We all know the drill. Before your next appointment you get an A1c test showing your average glucose over the last ninety days. Your doctor looks at it and says some version of: “Let’s see how we can get that lower.”

Naturally, as we all know, elevated glucose levels, which A1c tests can estimate, are associated with all sorts of complications, especially long-term ones like kidney failure, heart disease, retinopathy, and more.

OK, you think. How do we lower A1c levels?

Several suggestions come up: Start using an automated insulin delivery (AID) system, adjust the basal rates and insulin-carb ratios to try to keep those excursions at bay, and/or try a low-carb diet.

This is the common narrative, and nearly all T1Ds experience it.

The problem with this story lies in a very fundamental principle of physiology that science has known for decades, but has not translated to T1D management. Yes, there is an association between elevated A1c levels and complications, but the processes that actually cause harm do not lie with glucose alone. Furthermore, while that association shows up in population-wide data, the rate of individual variability is so high as to make the connection nearly useless.

The good news is that you have far more control over those downstream processes than you do over reducing glucose levels. It’s not that you should ignore glucose levels at all, but you need to do so with techniques that don’t cause the underlying harm. In other words, modern T1D management is so hyperfocused on A1c levels, they forgot that the methods for reducing glucose levels may win the battle, but lose the war.

If this feels like a radical paradigm shift in T1D management, rest assured that what is not new is that the physiology associated with all these processes has been known for decades and published in high-impact peer-reviewed journals. By adding attention towards oxidative stress, we can both reduce the incidents of complications, while also slowing down those that are already underway.

And the evidence is right there in front of us: the long-haulers — T1Ds that have lived with the disease for anywhere from 50 to 80 years. We can see what their A1c levels are, their rate of complications, why there’s no correlation between A1c and complications to individuals, and what they have in common that allows them to live so long.

Because of the chemical and physiological complexity, I want to lead with an analogy. Let’s start with a bicycle.

T1D and the Art of Bicycle Maintenance

Let’s say you have a brand new bicycle. You’re riding around, and it happens to start raining. That rain is like glucose. It lands on the metal and sits there a bit, but if you wipe it off, the metal is dry again. That’s the goal of keeping your A1c levels lower. Keeps the metal dry.

But you can’t get all of it because the water also seeps into the seams where it’s hard to get to. The air also has ambient moisture. And diabetes is like living near the ocean, where there’s not just a lot of air, but salty air. You can wipe down the metal, but the ambient moisture gets into those nooks and crannies that you can’t reach. And that’s where things matter., because over time, it starts to rust.

Those are three stages: water on the surface, water in the seams, rust.

That last stage is the oxidation process, because the chemicals in the metal mutate into something new and different: rust. And once in that state, it’s permanent. It’s no longer water that you can wipe away. Worse, it spreads. That chemical process turns into a feedback loop, eating all the metal next to it. That’s the insidious nature of spreading oxidation.

But it’s such a slow process that it takes decades before the bike starts falling apart. In many cases, you’ll ride your bike for decades before the seat stem breaks and you fall off. The rust destabilized the structure, but until it gives, you don’t know. And no one is measuring it, or teaching you how to avoid the rust.

The A1c test tells you how much water got into the seams, the second stage. It doesn’t tell you whether any of the harmful oxidation happened, nor does it tell you whether it will. In fact, what it measures isn’t permanent at all — water in the seams can still dry out. It’s the rust that’s permanent, and the rust is the one thing the A1c can’t see.

That may be useful in some minor ways for an individual, because it reveals how well you’re wiping the water away. But that’s just a short-term indicator, and it’s also deceptive because it turns out that the methods you’re using to try to keep the rain at bay — even avoiding being in the rain in the first place — may actually accelerate rusting. You may be pushing water into those seams by mere inattention.

You can see this rust in your connective tissue — the purest cases — like frozen shoulder, dupuytren’s contracture, trigger finger, carpal tunnel, stiff hands that won’t flatten against each other, plantar fasciitis, and achilles problems. That’s all rust. At least, the more visible part. It’s in invisible rust you can’t see that does the damage: vascularature, organs, other tissues.

Yes, those all happen to non-diabetics too. But it’s earlier, faster, and more aggressive with us because we’re living by the ocean with that constant, damp, salty air.

In our bodies, the process is called oxidative stress.

To be clear, oxidative stress is intrinsic to life. This is why people age, develop complications, and eventually, well, you know what comes next.

That oxidative stress is rooted in oxygen, but not the kind you breathe. It’s an unstable version of it, called reactive oxygen species, or ROS. They’re a normal byproduct of being alive, and that starts getting to the core of this topic. Oxidative stress is what we call it when there’s more ROS than your defenses can neutralize.

Your goal should be to neutralize the ROS, not amplify them, and the good news is that your body has ways to neutralize or clear ROS from your system. That’s what T1D management should focus on, but doesn’t. Instead, the management we’re using — more insulin, low-carb diets, pumps — either directly or indirectly add more ROS, overwhelming your defenses. Visceral fat, insulin resistance, and a whole series of downstream metabolic disorders that generate more oxidative stress are all building rust and allowing it to permeate our bodies.

You’re probably thinking, why not just take antioxidants? Vitamin C, vitamin E, the whole shelf at the pharmacy. It seems like it should work.

It doesn’t. A Cochrane review pooled 78 randomized trials covering 296,707 participants, and among the 56 trials with a low risk of bias, antioxidant supplements didn’t reduce mortality — they nudged it upward. People taking beta-carotene and vitamin E had roughly a 3-5% higher chance of dying during the study period. Vitamin C and selenium showed no effect at all, in either direction.

Those are small numbers, but notice that, among nearly 300,000 people, the best case for antioxidant supplements is that they do nothing.

The reason turns out to be that ROS aren’t only waste — they’re also a signal. When you exercise, your muscles generate a burst of them, and that burst is what tells your body to build more of its own defenses. Ristow and colleagues showed this directly: four weeks of training improved insulin sensitivity in men taking nothing, and didn’t in men taking vitamin C and E. The supplements blocked the rise in the body’s own antioxidant enzymes.

That’s the benefit of exercise: It reduces the oxidative stress that happens downstream of the A1c. Aerobic exercise in particular upregulates the body’s far more effective enzymes that stave off the rust.

What all the above have in common is insulin. After all, that’s the tool we use to manage glucose levels. Let’s take a closer look.

Insulin and the Goldilocks Effect

The first thing to know is that insulin is biphasic. That means it’s beneficial at physiologically balanced levels — too much or too little is bad. One of insulin’s many functions in the body is a vasodilator, which means it relaxes blood vessels, giving them plasticity.

A non-diabetic pancreas produces roughly 25 units a day for a physiologically weight-balanced person (not over or underweight). So 35–40 in a T1D is about 1.5× physiologic — which is what you’d expect given subcutaneous delivery, since you’re dosing to peripheral effect without the portal first-pass extraction a real pancreas gets.

One of the most important beneficiaries of insulin is the endothelium, the lining that protects your organs and arteries. It’s like the paint on the metal. Insulin stimulates nitric oxide (NO), which protects the endothelium, not just on the surface, but in the nooks and crannies — the parts you can’t wipe away.

Excess insulin becomes a vasoconstrictor, causing tissues to stiffen and weaken. It’s like the paint peeling off, exposing the metal to the elements.

When you take too little insulin, there’s not enough production of NO to achieve the vasodilation effect.

Either way, when the endothelium is affected, loss of paint allows the bad agents to enter. Hence, maintaining a healthy endothelium should be a priority, so balanced insulin levels should be prioritized. And no endo will ever explain or prescribe that.

Insulin Resistance: The Main Driver of Oxidative Stress

Another problem with excess insulin is that it’s a storage mediator, promoting fat accumulation and weight gain. But it’s not just that there’s a lot of fat. It’s that individual fat cells become engorged — a state called hypertrophied adipose tissue — and as they enlarge, they outgrow their own blood supply and go hypoxic.

Oxygen-starved tissue looks, chemically, like injured tissue. So the body responds the way it responds to injury: it recruits macrophages, the white blood cells that fight infection. They arrive and turn on the same respiratory-burst machinery they’d use against a bacterium — a deliberate burst of reactive oxygen species, or ROS, meant to destroy an invader.

The problem is there’s no invader. Nothing gets resolved, so the response never shuts off. It just runs, quietly, for years.

That’s the generalizable pattern, and it’s worth holding onto: any tissue running a chronic, low-grade immune response with nothing to resolve.

And here’s where it turns on you. The inflammatory signals those macrophages release interfere with insulin’s ability to communicate with the fat cell. Not a wall the cell puts up — the message simply stops getting through. That’s insulin resistance. The glucose stays in the bloodstream, so more insulin gets recruited to move it somewhere, anywhere that will take it.

Which promotes more storage, more hypertrophy, more hypoxia, more macrophages, more ROS. The loop feeds itself, and the oxidative stress compounds year over year. Add a damaged endothelium and you have the conditions underlying nearly every diabetic complication.

That’s the generalizable pattern: any tissue running a chronic, low-grade immune response with no infection to resolve.

And yet, your endo wants that perfect A1c number, unaware of the oxidative stress being imposed on you in order to achieve it.

With that conceptual overview, let’s look at all this from the clinical level and understand how we got here, what we can learn from it, and what we can do to fix it, beginning with the A1c test itself.

What the A1c Reveals and What it Hides

The NIH paper, HbA1c Standardisation: History, Science and Politics, explains how the A1c test was developed around 1977, but it sat for about a decade until the procedure was standardized in the mid-1980s.

The key to understanding the A1c is that it doesn’t measure your blood sugar. It measures how much of your hemoglobin — the protein inside red blood cells that carries oxygen — has glucose stuck to it. That’s called glycation, and everything that follows starts there. That’s the water in the seams. It’s not rust yet.

But note: this glycation is on hemoglobin, and hemoglobin turns over every few months. Glycated hemoglobin doesn’t harm you. What harms you is glycation on tissues that turn over slowly or not at all — collagen in your joints, your skin, your artery walls. The A1c doesn’t tell you about any of that.

Nevertheless, there’s a reasonable inference to draw: if your A1c is consistently high over time, glycation is presumably happening everywhere else too. So researchers put that assumption to the test in the first trial ever to use the A1c — the Diabetes Control and Complications Trial (DCCT), from 1983–1993.

The findings established that high glucose levels (measured by A1c tests) directly correlated to microvascular complications, such as retinopathy (leading to blindness), kidney disease, and neuropathy. As a group, people with lower A1c suffered fewer of these complications, and the group with higher A1c levels had worse outcomes. A follow-up analysis — the DCCT/EDIC — showed that aggressive insulin therapy produced a 42% reduction in cardiovascular events, 57% for major ones, and the benefit still held thirty years out.

The data was the first time a test could be used to actually set guidelines for T1D management. Since the cohort in the DCCT intensive-therapy arm achieved ~7.2%, the American Diabetes Association (ADA) recommended an A1c of 7% as the target with aggressive insulin therapy as the prescribed path. Follow-on guidelines would later expand to include the 50/50 rule for basal dosing, low-carb diets and other interventions.

But it came with costs that wouldn’t be considered seriously for decades: roughly triple the rate of severe hypoglycemia, and substantial weight gain. Participants who gained the most weight developed the lipid and blood pressure profile of those with insulin resistance syndrome, one of the largest contributors to oxidative stress.

Despite the fact that all this was eventually learned, it was too late: that the association between A1c and complications was baked into the guidelines.

So, if complications aren’t being driven by glucose alone, then what?

Advanced Glycation End-products (AGEs)

Remember the bicycle and the three stages of rust. Water, seepage, rust. Same thing with glucose.

The first stage is glycation, which happens when glucose attaches loosely to proteins. This is called the Schiff base. Then it rearranges into stable early stage complexes called Amadori products — this is HbA1c. Then, over months and years, a fraction of those mature into irreversible cross-linked Advanced Glycation End-products, or AGEs.

But that only happens where there’s oxidative stress to drive the conversion. That’s the step that turns the water lying in the seams into rust, and it’s the step that varies enormously from person to person.

The fact that there’s a fraction of a fraction from what the A1c measures isn’t as important as the fact that each fraction is highly variable and inconsistent because they only happen when exposed to particular conditions, some of which are actionable. That’s the part where inventions can and should happen.

Read that carefully: Your A1c could measure X number of glycated proteins, but your lifestyle choices and other interventions might keep many of them from becoming oxidized AGEs. Or, worse, you could engage in things that make more of those oxidized AGEs.

Accordingly, you could have two individuals with the same A1c, and yet, their management protocols with T1D could yield vastly different risk profiles.

Now look at these effects as a group: excess weight, cardiovascular disease, and eventually, kidney damage. Those aren’t three separate problems that happen to co-occur. They’re three expressions of insulin resistance and the sedentary metabolism that produces it. That means there’s potentially ONE intervention that can stave off AGE production, and we’ll see them in four separate biomarkers.

And the perfect place to look is in people who’ve had T1D for 50 to 80 years.

Long Term Correlation between A1c and Complications

The Golden Years Cohort, a group of 400 T1Ds in the UK who’ve had the disease for over 50 years, had a mean A1c of 7.6% (± 1.4). The Joslin Medalists, American T1Ds who’ve also had the disease for over 50 years — with some now passing 80 years — had an average current A1c of 7.3%. In fact, a 73-person longitudinal subgroup followed for 15 years had an A1c of 7.7%.

Across the full cohort, individual A1c values ranged from 5.0% to 14.0%.

An A1c of 14%! Imagine what your endo would tell you if they saw that.

Now let’s look at who got complications.

The paper, Protection from retinopathy and other complications in patients with type 1 diabetes of extreme duration: the Joslin 50-year medalist study (Sun et al., Diabetes Care 2011;34(4):968–974), examined 351 Medalists, mean age 67.5, mean duration 56.5 years, HbA1c 7.3 ± 1.0%.

After more than fifty years with T1D, here's how many of them had no sign of each complication:

• Kidney disease: 87% had none

• Cardiovascular disease: 51%

• Proliferative retinopathy: 43%

• Neuropathy: 39%

Yeah, there are complications, but what’s striking is that glycemic control didn’t predict who got the complications. The researchers also measured AGE levels directly — in blood samples from the same people, at the same time — and found that A1c predicted nothing, while AGE levels predicted a seven-fold difference in the odds of having complications.

Set that against what happens to everyone else. Among Americans 65 and older, 34% have chronic kidney disease. Among the Medalists who, after fifty years of type 1 diabetes, only 13% had nephropathy. Their kidneys were in better shape than their non-diabetic peers.

Nerves tell a more complicated story. Around 27% of adults over 70 have measurable peripheral neuropathy; among the Medalists it was about 61%. Elevated, but roughly one and a half times the background rate — not the catastrophe five decades of diabetes might predict.

We can’t compare diabetic nephropathy and chronic kidney disease to the general population because they’re diagnosed differently, and cardiovascular disease is defined so many ways the numbers stop meaning anything. But the shape is clear enough: these people did not simply survive their diabetes. Many of them aged better than the population around them.

And glycemic control didn’t explain which ones.

This brings us back to what’s under your control.

We spoke about the biphasic nature of insulin — too much or too little is bad — so now let’s look at how that translates to complications.

Adiposity and Insulin Resistance

As noted earlier, insulin drives fat storage. If you’re taking too much insulin to drive down A1c, you’re likely consuming a great deal more carbs than you want to either avoid or treat hypoglycemia. That fat storage builds weight gain around the middle, particularly visceral adiposity — the fat packed around your organs, not the fat you can see or pinch. That’s what leads to insulin resistance (IR). It’s metabolically active tissue that generates a continuous oxidative and inflammatory signal. There’s your villain right there.

IR is a classic symptom of type 2 diabetes, so when it also happens to type 1s, it’s called double diabetes.

In a meta-analysis titled, “Obesity in Type 1 Diabetes: Pathophysiology, Clinical Impact, and Mechanisms,” obesity in adults with T1D rose from 3.4% at baseline (1986 to 1988, prior to the DCCT) to 22.7% in 2004 to 2007, the direct result of the aggressive insulin guidelines that resulted from the DCCT.

By March 2023, a study titled, Prevalence and Management of Obesity in U.S. Adults With Type 1 Diabetes (Johns Hopkins/CDC NHANES data) found that 62% of U.S. adults with T1D were overweight or obese.

Weight gain is a direct result of overinsulinization, and according to section 9.27 of the ADA’s 2025 guidelines, the primary cause of that is overbasalization — people setting their basal rates too high on pumps, or taking too much basal insulin. This is a long and deep subject that I cover more thoroughly in a four-part series starting with Basal Insulin and its Effect on T1D Management and Long Term Complications.

These all lead again to the goal of chasing down A1c levels.

A Second Road to the Same Place

Everything above blames insulin — too much of it drives fat storage. But there’s another factor on weight management that is an accelerant that you can have greater control over if you know about it: fructose.

Table sugar is half fructose, and so is the high-fructose corn syrup in soda and juice. Unlike glucose, fructose goes almost entirely to the liver, where it’s processed without the regulatory brake that governs glucose. What the liver mostly does with the result is make fat. Unlike glucose, which goes through a number of pathways before it becomes fat, fructose takes the shortcut.

In a ten-week trial, overweight adults drank either fructose-sweetened or glucose-sweetened beverages providing a quarter of their daily calories. Both groups gained about the same weight — but visceral fat increased only in the fructose group, along with their insulin resistance and their oxidized LDL.

Think about that for a moment: the same number of carbs, but one of those carbs is converting to fat more than the other.

Think about the proverbial juice box that many T1Ds use to treat or avert a hypo. Apple juice — the one in most boxes — is roughly 70% fructose once the sucrose splits. A 200 mL box labeled 23 grams of carb hands you about 7 grams of glucose that treats the low, and 16 grams of fructose that doesn’t.

The fructose isn’t gone, though. It goes to your liver, and about 40% of it comes back out as glucose over the next three to six hours. So the shape of the thing is: too little up front, then a slow release you never bolused for, arriving long after you stopped paying attention.

So when you give your kid (or yourself) that juice box — or most any fruit juice or a product with high fructose corn syrup — you’ll notice the box doesn’t work fast enough. You drink a second one. Maybe a third. Your glucose levels rise again, but hours later, you’re at 220 and you blame yourself for over-treating. Know that good part of was from the first box, still being converted. Depending on how many boxes you drank, you’re going to get a stack of glucose spikes along the way.

Depending on how many lows you’re treating, you’re not just absorbing a lot of carbs, you’re taking in carbs that convert to fat at a much higher rate. This is the fast-lane to insulin resistance.

None of this means treat lows less aggressively. Treat the low, always. It means glucose tabs or dextrose do the job faster, dose more predictably, and skip the liver entirely.

So that’s insulin resistance. Now let’s talk about your arteries, the source of cardiovascular disease.

Lipids and Insulin

In my article, Extending T1D Longevity: Balancing Lipids, Insulin, A1c, I cite the primary cause of death for T1Ds as being Major Adverse Cardiovascular Events — or MACE. These include heart attacks and strokes, triggered by vulnerable plaques in artery walls that have built up over time from high LDL cholesterol.

LDL is a low density lipoprotein and its job is to deliver cholesterol to cells throughout the body. Because LDL is a protein, you know what sticks to it: glucose. The more LDLs there are, and the longer they circulate, the more they oxidize. And you know what that is: AGEs.

Now put the two together: dysfunctional endothelium allows those oxidized glycated LDL particles to attach to the arteries. Cue the inflammatory immune response, and now you have the “plaque” that narrows artery walls.

You have the explosive and you have the fuse. Now you just need a spark, which can be an acute hypoglycemic event. Hypoglycemia isn’t killing T1Ds by starving the brain of glucose; autopsies don’t show that. What it does is trigger the counter-regulatory cascade — adrenaline, cortisol, a spike in blood pressure and heart rate — and that surge is what can dislodge vulnerable plaque that’s already been built up over decades from high LDL and chronic insulin exposure.

Think about how often T1Ds get acute hypos, what’s called Level 2 events. A six-trial analysis of 1,433 T1D participants found median time below 54 mg/dL to be 1.5 to 10 minutes a day. Over a year: 9 to 60 hours below 54. Other studies show similar findings.

That’s a lot of opportunities to die of a stroke or heart attack if you’re a T1D with elevated LDLs and endothelial dysfunction, not to mention oxidative stress from IR.

It’s that amplification that the medical system doesn’t treat properly.

Two people with LDLs of 80 mg/dL aren’t running the same risk if one of them has spent thirty years with elevated glucose levels.

I’m going to step out of character here in order to emphasize what I believe is the most important thing a T1D can do at their next endo appointment.

Ok, I’m back to my calm and collected self.

Statins are the foundation of lipid therapy, blocking cholesterol synthesis in the liver and typically reducing LDL by 30-50%. They’re well-tolerated, inexpensive ($4-20/month generic), and have decades of safety data.

There’s a lot of misinformation about statins online, and the muscle-pain question deserves a direct answer. The SAMSON trial enrolled 60 people who had already quit statins because of side effects, and gave some a placebo and others the statin. By the end, placebo didn’t show a meaningful difference. When those very participants were shown their own data at the end, half went back on a statin.

PCSK9 inhibitors (evolocumab/Repatha, alirocumab/Praluent) are the most powerful option, reducing LDL by an additional 50-60% on top of statins. The FOURIER trial followed 27,564 patients who achieved median LDL levels of 30 mg/dL, translating to a 15% reduction in major cardiovascular events and, critically, continued benefit down to LDL levels as low as 10 mg/dL with no safety concerns. The VESALIUS-CV trial in NEJM studied 12,257 randomized participants, producing a 25% reduction in major cardiovascular events and a stunning 36% reduction in myocardial infarction.

The benefit was consistent across all subgroups, including those with diabetes.

Clinical guidelines for diabetics should be far more aggressive with lipid-lowering medications than non-diabetics. When it comes to LDLs and AGEs, there is no safe level.

Some might be thinking that you can reduce LDLs using a plant-based diet. Mechanistically, yes. Saturated fat suppresses the liver’s LDL receptors so that less LDL gets cleared from the blood. Therefore, replacing animal protein with plant protein lowers LDL. Eating more beans, oats, and vegetables and less processed meat is a good idea for reasons that have nothing to do with this article.

The question is not whether diet lowers LDL. It’s whether diet lowers LDL enough given what our LDL is doing. Remember, the amplification of damage from each LDL particle means we need vast reduction, not just the modest amount that diets can provide. For extensive detail on plant-based diets and their effect on LDLs for T1Ds, see my aforementioned article, Extending T1D Longevity: Balancing Lipids, Insulin, A1c.

You might be reading this and thinking: low insulin levels, low body weight, and low glucose levels… I know the answer: low-carb diets! Surely, they’d do the trick!

Nope. But there is a fascinating nuance worth observing.

Low-Carb Diets

As it happens, low-carb diets can actually make the problem worse, but from a different physiological angle. When carbohydrate intake drops too low, the body mobilizes stored fat for energy. Whatever it can’t burn gets repackaged into VLDL particles that get shipped back out into the bloodstream. VLDL delivers its cargo, shrinks, and becomes the LDL we’ve been discussing.

You just accelerated and amplified the same LDL problem as before.

Not only do you have far more LDL to deal with, you have the insulin problem as well, but this time, in reverse. Insulin relaxes blood vessels (vasodilation) at physiologically balanced levels. But if you don’t get enough insulin, you have vasoconstriction, which leads to the same endothelial dysfunction that we saw from insulin resistance.

The protective paint falls off.

In the literature review, Low-carbohydrate diets and all-cause mortality: a systematic review and meta-analysis of observational studies, the highest-restriction quartile in these cohorts — still well above Bernstein levels (<30g/day) — had the highest risk of overall mortality (32% increase), cardiovascular disease mortality (50% increase), and cerebrovascular mortality (51% increase).

Now, this is not to suggest that low-carb diets are bad, per se. They are effective for weight loss, which is incredibly important — adiposity and insulin resistance are long-term, deleterious conditions. More importantly, diets should be temporary. There’s a distinction between using low-carb to lose weight versus keeping A1c levels low. One can save you, the other can harm you.

I mentioned that time-limited actions are a reasonable trade-off, but let’s widen the lens here. What kind of timelines are involved in all these processes?

Time Horizons

Building up AGEs in your body takes decades, not months, or even a few years. Your body can tolerate several years of elevated glucose levels, lipids, low-carb diets, and many other assaults to a point. This is not permission to do it, but it is important not to get too caught up in the risk profiles and induce psychological stress or other anxieties (which can also increase AGE production). But of course, there’s also what glucose itself can do that we need to recognize.

A1c levels above roughly 9% carries its own risks independently of AGEs: DKA, opportunistic infections, and poor wound healing. At this level, bringing glucose down is the priority, even by accepting more insulin to do it. Getting glucose levels out of the severe danger zone is a reasonable trade, but you should aim for this to be like weight-loss: a temporary measure with the goal of tapering off.

A1c between 7–8% flips the picture. Glucose is present but not overwhelming, and what separates good outcomes from bad ones is almost entirely the AGEs process: oxidative burden, insulin resistance, lipids, fitness, smoking. If you engage in these other factors, then you’ll confer both protection, and the likely byproduct of your A1c falling. But this time, in healthy ways.

A1c below 6.5%. The question stops being your number and becomes what else is going on. If you’re exercising, and your body weight and insulin levels are physiologically healthy, you’re far better off than someone with the same A1c, but has high insulin levels, is overweight, or is on carbohydrate restriction.

As a general rule, intermittent stretches of elevated A1c won’t do much harm. Two years at 8% during college, against sixty years of exposure, is a small fraction of your lifetime exposure.

And a few days here and there where your glucose spikes to 400 may have some immediate repercussions, but you’re not going to suddenly raise your risk of a heart attack or neuropathy due to a sudden growth of AGEs.

One factor that is not part of any health guideline at all, and which addresses all of these things in the healthiest and easiest way is improving your metabolic health, which you may recall was the first of the items in the list of four commonalities that the long-haulers had in common.

Metabolic Health

Metabolic health is a catch-all category that includes a number of different biomarkers that affect the body’s ability to convert fuel into energy. The full picture is beyond the scope of this article, but the lever that moves it is exercise.

Of the many different kinds of exercise, we’re primarily interested in cardiorespiratory fitness, which involves high aerobic activity, like running, cycling, rowing, and more. As stated earlier, this level of activity helps reduce the oxidative stress that contributes to the production of AGEs. That is, aerobic fitness interrupts the transition from stage 2 to stage 3 AGE production.

And again, we can see this in longitudinal data. A paper from Mandsager involved 122,007 adults between 1991 and 2015. They found that the more physically fit people were, the longer their lifespan, even in the presence of other comorbidities, such as hypertension, coronary artery disease, or even smoking and diabetes.

In fact, the positive effects from good cardiorespiratory fitness were so disproportionate after factoring in these comorbidities, they found there was no greater predictor and protector of all cause mortality than cardiorespiratory fitness.

Two people with the same A1c can have vastly different risk profiles for complications if one of them is sedentary and the other has high cardiorespiratory fitness.

The devastating part is not just that it’s another reason why A1c is a poor predictor of complications, but the hyperfocused drive towards lowering A1c puts T1Ds at risk because they don’t know how to exercise and manage glucose levels at the same time. In fact, fear of hypoglycemia (FOH) is the leading reason why most T1Ds don’t exercise, according to the literature review paper, Fear of hypoglycemia, a game changer during physical activity in T1DM. So, they choose not to exercise.

As you can see, this is not only complicated, it’s highly individualized, and often is contrary to conventional T1D management.

Note: it may not necessarily be a good idea to immediately jump off the couch and start running, especially if you’re overweight and have high LDL levels. Going from a sedentary lifestyle to full-on aerobic exercise requires managing the drawdown of weight, and incremental building muscle mass and learning proper insulin dosing.

But a very good place to start for anyone is the simplest of tasks: a 15-minute walk once or twice a day, preferably after meals. That begins two critical things: habit-building, and a tapered reduction of post-meal insulin dosing, a technique that you need to learn to master manually.

The simple act of walking can reduce risk of complications considerably. According to the paper, “Walking for Exercise,” from Harvard’s School for Public Health, this simple process can improve your lifespan by several years (not to mention quality of life).

So, why is the medical system so hyperfocused on A1c, despite all this evidence?

Actually, there’s a sad, but understandable reason: because it’s easy.

Current Guidelines

In my article, Standard of Care: Who Defines it, How, and Why it Matters, I cite a report by The American Diabetes Association that comes down to this: physicians don’t have the time, expertise, or training to teach people how to manage all the various aspects to the disease.

I can’t sugarcoat this (er, so to speak) to say there’s an easy solution. As I used to teach my students, “if it were easy, everyone would do it.” As I said at the top, there are two separate, orthogonal objectives: teaching how to manage glucose levels, which is insanely complex, and teaching metabolic health, which is also complex. And both of those are highly individualized — you cannot and should not adopt population-wide methods to individuals.

That’s puts the medical system in a bind. The ADA sets guidelines for those very physicians, and the nature of the disease is complex. But the one thing abou the A1c target is that it’s easy. It’s a simple number, it’s a target, and the complexity can be somewhat alleviated by automated insulin delivery (AID) systems — insulin pumps that do all that work.

The insidious problem with that is similarly less obvious. In my article, Medical Literature Analysis: The Performance Paradox of AID Systems, I cite medical literature that finds that, while many people who are unable or unwilling to care for themselves do very well with automated systems (compared to how they could do on their own), otherwise capable adults suffered from these problems: weight gain, loss of agency, de-skilling of self-management, limitations on efficacy, quality of life, economic impact, and the “attractive nuisance” of disengagement from self-management.

We already know the problem with weight gain. But worse, the de-skilling of self-management skills and the loss of agency handicap these people because they lost the skills to know how to wean themselves off of automation to self-manage properly again (assuming they ever knew). They’re stuck in that system.

And worse, T1Ds typically adjust their lifestyles to meet the expectations of pump algorithms. A 2022 review of AID’s psychological implications — Nefs, Frontiers in Clinical Diabetes and Healthcare 2022 — documented exactly these dynamics.

Sedentary and overweight — the perfect storm.

Calgon, take me away!

GLP-1s and T1D

Lastly, people are certainly aware that GLP-1s help with insulin resistance and insulin requirements. Lots of data coming in already on how it helps lower weight, reduce insulin resistance and improve A1c. But they can also make exercise harder to manage.

These two points are essential and they go together. Yes, they reduce insulin resistance and insulin requirements, but it’s only one arm of the problem — reducing some sources of oxidative stress. You also need to address the stresses that already exist, and which continue to occur as a natural process of metabolism.

That’s where exercise is essential: it makes your body build its own antioxidant defenses, and that adaptation is the thing nothing else replicates.

But exercise also builds muscle mass, bone density, balance, aerobic capacity — the things that no drug will deliver. If GLP-1s become the next AID system, insofar as how they may be perceived as a game-changer, it may also follow the same path of creating more harm over time than intended.

I’ll write about that in ten years, when the trial data comes in. And I’ll be here to do it because... well, I exercise.

To Be Healthy, Get Back to Basics

We covered a lot here, but the irony to this story is that what works is getting back to basics. We long-haulers couldn’t test glucose levels or use any modern technology at all. We simply adhered to the basics of good health: eat well, get good sleep, go outside and play, and keep weight down.

I’m not aiming to oversimplify to tell a story; that’s just the truth that we had at that time. Today, we have more and better technologies — both devices and drugs — but the principle of physiology haven’t changed. It’s how you use them that matters.

In my article, The Four Habits of Healthy T1Ds, the first thing I explain is that, before you can become healthy, it requires a frame-of-mind: engage with your diabetes, don’t put it in the background. You don’t need goals or targets. Targets are distracting at best and dispiriting at worst, and either way they pull attention away from the things that work:

• Familiarize yourself with your glucose patterns.

• Split your boluses, and act on what you see.

• Exercise.

• Optional but useful: log what you eat and do.

Add lipid-lowering medication, and you’ve addressed the factors that can help extend your lifespan. Best of all, they also happen to be the easiest to control.

Share

Subscribe now

If you follow type 1 diabetes closely, you’ve probably heard that something sudden happened at BreakthroughT1D this summer. The T1D Fund, a subsidiary of BreakthroughT1D that invests in outside companies working toward a cure, had its board dismissed without notice. Governing documents were rewritten overnight. The largest donors in the field — families who have each given north of $100 million — found out after the fact, and at least one resigned in protest. There are open letters circulating, and a watchdog organization publishing on it, and a lot of speculation about who did what to whom.

To distill it as succinctly as possible, two things went wrong at once. The first is that the Fund drew enough internal concern that the above alterations were made (confirmed by the organization itself). And the second is that these actions caused major donors to walk.

We don’t know what those concerns were. What’s on the record is only what BT1D did about them. (BT1D is how BreakthroughT1D refers to itself. It should not be confused with BT1 — Beyond Type 1 — a separate organization with nothing to do with any of this.)

That said, the Fund itself, and the money it has access to, remains intact (at the time of this writing). As such, the story is not over.

As for the donors that walked, the most visible is David Panzirer, a main trustee of The Helmsley Charitable Trust, a now-former board member of the T1D Fund and one of its original founders.

On June 18, 2026, Panzirer resigned from The T1D Fund. In his open letter, which I have seen (though is not on a public website that I’ve been able to find), Panzirer stated: “What Breakthrough did by firing the Fund Board with zero notice or transparency has alienated the 4 largest families in T1D all within 48 hours. Each family has committed well over $100MM to the cause. NOTHING will happen if we don’t all stand up and demand transparency.”

An open letter that asks people to stand up and demand transparency invites outsiders like me to look at what’s available in the public record. The firings happened, the resignations happened, the documents were rewritten — but that narrative doesn’t tell you why any of it happened, or what comes next.

That’s the hull breach. If it isn’t repaired, the shakeup, and the alienation of key donors, stands to take the whole organization down. Without major funders, BT1D can’t survive. And those donors set the barometer for others that follow. The whole ship could sink. On the other hand, the reason the organization had to take action must have been important enough to take that risk.

My aim here is to focus on the Fund’s investments because that was its only function: investing to accelerate the cure. Whatever the problem was with the Fund, or between the Fund and BT1D, it starts there. Along the way, we’re going to learn about the economics of funding a cure that has virtually no financial upside for commercial enterprises to own. Participate, yes. But, own? No.

What the T1D Fund is

Structurally, the T1D Fund is a separate, wholly owned subsidiary of BT1D. Launched in 2016, it’s an LLC with BT1D as the sole member. The Fund files no return of its own, so its investments sit on BT1D’s balance sheet. Its managing directors are BT1D employees.

Its purpose is that of venture philanthropy, a means by which a charity can make investments in companies that serve the mission and turn a profit that can then be used for more investments. The model was originally conceived in 1969 by John D. Rockefeller III, one of the largest philanthropists of his generation. It has been adopted by many charities, ranging from school programs to other medical conditions. It is an excellent way for charities to fund startup companies that develop technologies in keeping with their mission.

However, it’s not universally a good fit, particularly when trying to find a cure for type 1 diabetes.

To be sure, that was not always the case. When the Fund was originally formed, it invested in companies that were strictly focused on T1D, and not only did those companies succeed, but the Fund returned over $100M back into its coffers, ready to be invested again.

But those companies were not focused on a cure — they were generalized for T1D management, such as insulin pumps and the like. While there were curative therapies in the portfolio, these were singularly focused on certain phases of T1D development.

The problem emerged when the Fund shifted its focus on technologies that were focused on a cure. And that domain changed. Curing T1D involves technologies (and companies) that fall into general categories, like immune tolerance and stem-cell derived tissues and manufacturing. All of these are essential for a cure, but none of them are focused on T1D.

The distinction here matters because of what “investment” does. Investing in T1D-specific companies like insulin pumps engaged the company to focus on the disease by definition. The investment helped fund a company that might not have a market outside of T1D, therefore, no investors. But that’s not the case for curing the disease. The act of purchasing shares in larger companies whose focus is elsewhere is not effective in getting them to focus on it. They have other, more lucrative targets in mind.

Making matters worse, because the companies are not focused on T1D, it can be argued that the Fund is not permitted to invest in them. This kind of investment vehicle (the Fund) requires companies to be directly “on mission”, or the tax-exempt status is compromised.

The drift in investment focus from daily management to a cure meant there was a shift in the types of companies receiving the money, which meant that the money being invested wasn’t really advancing anything towards T1D, per se. That is, the technologies those companies are producing may well lead to a cure, which is obviously great. But this would happen simply due to the nature of the technology. The Fund’s money is playing no role at all.

Whether this was as clear to the parties inside the organization as it is here is unknown. There may have been a dispute about this assertion. There may have been a dispute about whether the investments put the non-profit status at risk. Whatever discussions may have happened, it didn’t result in an agreement. And, as of this writing, there is no clear message from anyone about what comes next.

But one doesn’t need to speculate on what’s possible, and indeed, one of the oldest techniques in business: contracts.

Since the real goal is to get these companies to actually do T1D-specific activities, don’t invest in the companies. Just pay them to do the work you want done. Same money, different incentive structure. All of it on-charter.

But of course, you don’t want to just throw money around without a plan. There needs to be coherence—a strategy. This is where BT1D should act like a general contractor, by taking direct ownership of the whole pipeline and outsourcing each of the parts in a coherent roadmap towards a cure.

All this is what makes our particular disease unique from many others and why venture philanthropy doesn’t work here.

To be clear, two things can be true at once: no one at the Fund did anything wrong; it was a slow creep of technology drift that made “investments” less effective as a means to generate outcomes. At the same time, those investments may have been seen as a tax risk, due to their off-charter profiles. In short, the Fund may have just outlived its utility, and the money it holds may be better used elsewhere. That doesn’t attribute fault, though it could certainly explain disagreement. Perhaps vigorous disagreement.

Whatever prompted the concerns, we don’t know. But the way BT1D acted on those concerns may have made things worse, and it’s important to include this in the story because, without the investors that have walked, the ship may not actually reach its destination.

Finding a cure to the breach is essential. And to do that, we need to build upon what was just described above. Once we find a model that works, we can then use that as a blueprint for both the donors and the organization to consider.

The Hull Breach

For its first years, the Fund invested in entities squarely aligned with curing T1D, and the record shows it: Semma (islet replacement), Provention (teplizumab, which became the first FDA-approved drug to delay T1D onset), Pandion (immune modulation), and a row of T1D devices — Bigfoot, Biolinq, Capillary, Diasome, Protomer.

As the Fund’s page stated, it has “realized nearly $100 million in profits from portfolio returns” since inception, and that “every dollar has been reinvested to support future investments in promising T1D therapies and cures.”

As noted earlier, the currents changed. The technology worth investing and re-investing was no longer about T1D specifically. It broadened into immune tolerance, which isn’t just a T1D problem. So, the Fund’s investments went where the science went. Companies whose sole focus was T1D were becoming fewer and less relevant, and the money followed. Hence, the drift.

But one can say that these are essential to a cure, therefore, they’re related enough. Yes, but the distinction matters, and its subtlety may explain the murky waters of the disagreement.

A mechanism is how a drug works. An indication is what it’s approved to treat — the specific disease named on the FDA label, which a company only gets by running the trials to prove it. That technicality changes what the Fund can invest in, and also whether that investment actually motivates the company.

A company organized around mechanism says, “we suppress T cell activity, and we’d be delighted if type 1 diabetes turned out to be a beneficiary.”

A company organized around indication says, “we are proving this works in type 1 diabetes, and we’re spending the money to put it on the label.”

The Fund’s own portfolio page organizes its holdings into strategic areas, and the largest single bucket is “Immunotherapies” — a mechanism, not an indication. Three samples among the entire cohort of 25+ companies include:

• COUR Pharmaceuticals — nanoparticle therapies delivering disease-specific antigen or allergen to antigen-presenting cells in the spleen and liver

• Jaguar Gene Therapy — filed under Beta Cell Therapies, with no beta cell in the description

• Eledon — clinical-stage transplantation company. No mention of diabetes at all.

Not one of those names an indication. And the Fund’s homepage puts a number on the whole operation: 25+ companies, over $250 million in assets under management, most of them are mechanisms, not indications.

Here’s the part that bites: Each company got money from the Fund, but is under no obligation to produce anything. In the original companies, that wasn’t a problem because T1D was the only purpose of the company. That’s the difference.

So, put it all together: the Fund was investing in companies it might not be allowed to invest in, and the companies were not obligated, let alone motivated, to advance towards a cure.

If the sole purpose of the Fund is to make investments, these two problems suggest the Fund may have outlived its utility.

Now, this doesn’t preclude BT1D from working with these companies. Indeed, this is what contracts do: they pay the company to perform a specific service. But that’s a separate pot of money. In short, BT1D would have spent less and gotten the same work if they didn’t invest in equity, but instead just engaged in contracts. And the money they didn’t spend on an investment could have paid for another contract, such as a clinical trial.

If it was deemed that the Fund was not only unnecessary and a potential tax liability, it only amplifies the problem because there’s $250M allocated to it, either invested in companies doing nothing, or is waiting to be invested that do nothing.

Consider Eledon. The company has stated formally that it has no interest in T1D at all. They’re interested in whole organ transplants — kidney, liver, pig organs and the like — with the primary aim of supplanting tacrolimus as the immunosuppressant drug, which currently commands a $7B annual revenue stream.

BT1D owns stock in Eledon. BT1D also funded a clinical trial that happens to use Eledon’s drug, tegoprubart: an investigator-initiated islet transplant trial at the University of Chicago. Those are two separate pots of money doing two entirely different things.

The trial money bought an activity — T1D patients receiving islet transplants under a specific protocol, with measurable outcomes. That’s exactly what a charity should be paying for.

The stock bought nothing. It didn’t cause the trial. It doesn’t obligate Eledon to run another one, or to pursue an FDA indication in T1D, or to keep the drug available if the program gets deprioritized. Eledon owns the molecule, owns the regulatory path, and owns whatever comes next. BT1D owns some shares. They have zero control.

Run that across the portfolio and the pattern holds. Investments don’t compel companies to do mission-related activities. Only directly funding the specific activity does that.

As for my references to whether the investments were permissible, this is where BT1D’s “purpose clause” comes in. It’s stated in its articles of incorporation, and corroborated in its IRS Form 990: “improving lives today and tomorrow by accelerating life-changing breakthroughs to cure, prevent and treat T1D and its complications.”

Every word of that is bounded by a single disease. Not autoimmunity in general, not immune tolerance as a science, not cell therapy as a platform — type 1 diabetes. Investments made by a charity are supposed to serve that purpose.

The defense would be that the companies are “mission-related” — that a tolerance platform which could someday serve T1D is close enough. And that argument isn’t frivolous. But look at how it works in the Eledon case. BT1D paid for the trial that created Eledon’s T1D association, and that association is what makes the investment look mission-related. The nexus was manufactured by the investor.

No tax authority has ruled on any of it, and we don’t know if this may or may not have come up at a board meeting. That aspect notwithstanding, the greater point is that investing is not strategic. Even if every position were bulletproof on charter, the money still wouldn’t be buying activity. The tax concern may have been real, but it’s also beside the point to the larger picture.

Cue the iceberg.

$250 million, sitting in companies BT1D doesn’t control, doing work BT1D can’t direct. Major donors have either left or threatened to leave. Metaphorically, the ship was taking on water. Not from a single catastrophic breach, but from a long scrape that took years to develop, and that nobody could see from the deck. Until someone looked.

It’s time to patch the $250M scrape in the hull.

BT1D Needs to Behave Like a Product Company

We’ve established that the cure will come from larger companies’ technologies, and we can write contracts with those companies to do activities. Some manufacture cells, others develop immune protection, others focus on the delivery method. But no one owns the assembled protocol, and none of them has any reason to build it — the eligible population is too small to justify the work.

That’s the vacancy BT1D should fill. It has the money, the convening power, and the patient constituency. It’s the only entity on the field that could own the integration. It needs to act like a general contractor.

And here’s a critical aspect to that: a general contractor works in the interests of the people it’s building for. It also has to be indifferent among its suppliers at the start — otherwise it isn’t running a procurement, it’s ratifying a choice already made by someone else. That’s where owning stock in the suppliers becomes a problem. Not because anyone is acting improperly, but because it removes the one party whose job is to ask whether a different supplier would do better.

The thing about contracts is more than just funding the activity, it also sets terms, and those terms need to be sure of the optics of conflicts of interest. This is not easy, but it’s doable with the proper disclosures and more critically, internal due diligence of BT1D advisors and board members who themselves should be bound by conflicts of interests and firewalls between divisions. (I address this concern about BT1D in my article, Standard of Care: Who Defines it, How, and Why it Matters.)

For example, Eledon’s tegoprubart isn’t alone in its drug class. The following are in the same class.

• dazodalibep — Amgen (developed at Viela Bio, acquired by Horizon Therapeutics, then Amgen). Two Phase 3 studies in Sjögren’s disease, completion expected H2 2026.

• TNX-1500 — Tonix Pharmaceuticals (Nasdaq: TNXP). In April 2025 Tonix partnered with Makana Therapeutics on xenotransplantation.

• frexalimab — Sanofi. Sanofi has projected peak sales above €5B/year based on Phase 2 MS data.

• dapirolizumab pegol — UCB and Biogen, co-developed, in systemic lupus.

So with four other molecules in the same class, why tegoprubart? Possibly because it’s the best one — it has properties that reduce the clotting risk that plagued earlier CD40L drugs, and that’s a real differentiator. Though the competitors are possibly working on that. Probably.

But there’s a lot of “possibly” there. Why not the others as well? Were they thoroughly investigated? Did anyone engage in negotiation? Who ultimately made the decision?

No one from BT1D did; it was Dr. Piotr Witkowski, head of the UChicago program. He chose it. Why? We’d like to think it was the best choice. And maybe so.

But, there’s optics: In a January 2025 Frontiers in Transplantation article he discloses consulting relationships with Sernova, Eledon, Vertex, and Seraxis, a seat on a Vertex trial steering committee, and equity interests in Eledon and Sana Biotechnology.

None of the four alternative-molecule companies appears on that list. No Amgen, no Tonix, no Sanofi, no UCB or Biogen. Of the five molecules in the class, Witkowski holds a disclosed equity position in exactly one — the one under test.

Oh, and so does BT1D.

Maybe the other four are genuinely inferior for this application. Maybe they were never available — big pharma has little reason to supply a drug to a small islet trial. Both are plausible, and neither has been written down anywhere.

That’s the gap. There shouldn’t be maybe’s and probably’s. A general contractor would want its non-conflicted scientific committee to produce a report saying not just why tego is preferred, but why the alternatives were ruled out. Nobody in this chain had a reason to ask that question — which is different from saying anyone avoided it.

To reiterate: nothing’s inherently wrong or improper about these relationships. Financial entanglements are unavoidable in a field this small. But BT1D could be more self-aware of the optics when independent observers see the connections. Scientific researchers that sit on BT1D committees can be thorough in their assessments about what to fund and publish a competitive analysis when making those recommendations. And they themselves disclose their conflicts.

Not owning stock may improve the optics of course, but it also makes the optics of drug recommendations not only more transparent, but scientifically more rigorous.

As it happens, it’s what the old JDRF used to do. The Industry Discovery & Development Partnerships (IDDP) program made funding available to for-profit companies worldwide, publicly or privately held, focused on a priority area within JDRF’s mission, structured to take promising T1D research through discovery and development toward commercialization. This is what funded the advancement of automated pumps and other technologies used in daily management. And independent analysts were there the whole time.

These contracts also came with terms. In 2013, JDRF committed up to $3 million to Tandem Diabetes for a dual-chamber infusion pump, paid against development milestones. The program was terminated in 2016 after $0.7 million had been disbursed — and under the contract, Tandem repaid all of it. The money came back. That’s what a term does that a share of stock cannot.

There are two more advantages to a contract that BT1D has never leveraged. The first is a royalty back to BT1D on any product that comes out of work it funded. The second is a say in what that product costs the patient. Neither is available through an equity investment at any price. And crucially, can be set up to avoid the conflict-of-interest problem.

Royalties and Affordability

Funding trials is important. The problem is funding them as grants with no strings attached, while separately buying stock at the same time. This strategy yields no leverage of any kind — the kind that will actually matter later, especially if the trial succeeds and there’s a genuine pathway to a cure.

A better approach is to use the same money that went to both the trial and the investment, and bundle it all into a contract that has terms. Two of them, and the first is simple: BT1D funds the trials, builds the regulatory pathway, and delivers an FDA-approved T1D indication the company was never going to pursue on its own — in exchange for a royalty on commercial sales.

Yes, royalties. Again, standard practice for a set of conditions like this.

The company gets a free label expansion and some revenue for work they get paid to do that requires no effort on their part. Once the product hits the market through an FDA-approved protocol, the company receives revenue, and BT1D takes a cut of it in the form of royalties.

Which brings us to the second term, and it’s the one nobody thinks to write. Royalties can be lucrative for everyone, but they’re a double-edged sword for a patient organization. Making money is fine. Making it on your own constituents’ backs is not. So when the protocol is eventually approved by the FDA, what will the price be to the patient?

A curative therapy for a small population is exactly the kind of drug that ends up costing millions to develop — the orphan-drug trap, where a tiny patient pool has to carry all of it. For BT1D, that would be a disaster. The last thing BT1D wants to announce is that the protocol for islet transplants is FDA-approved, but it’ll cost $500,000 per patient. They want that price to be low, even though it would cut into their own royalty stream.

And that is where the last nail in the coffin goes for venture philanthropy for T1D. To keep the price to the patient low, the company that sells the protocol will not make enough money to justify their own investment. No for-profit commercial investor would do this unless there was a profit margin. A BIG profit margin. And that can only happen if the price were astronomical. It has to be because the number of patients who’d receive it is very low.

The canonical case is the Cystic Fibrosis Foundation’s relationship with Vertex Pharmaceuticals. Beginning in the late 1990s, the Foundation committed roughly $150 million to Vertex’s modulator program — work the venture market then considered too early and too risky — in return for royalty rights on anything that resulted.

Yes, they negotiated royalties, but that’s it. They didn’t anticipate what it would cost patients. Vertex charged patients $300,000 a year. And the royalty did something worse than fail to protect patients. It put the foundation on the wrong side of the price. CFF’s royalty was 10%, so the non-profit organization earned $30K for each patient that paid $300K for the treatment. In other words, the organization that existed to serve CF patients had a financial interest in the number that kept some of them out.

That’s the mistake BT1D does not want to repeat. It’s here where BT1D would want to get its own price for the drug, separate from list price, so that when the protocol becomes available in an FDA-approved therapy, the company doesn’t charge prohibitively expensive amounts. The fix isn’t a list-price discount—it’s an indication-specific net price capped via foundation-administered rebates, regardless of the sticker price. The patient pays $10,000 instead of $500,000, not because the drug is cheaper, but because BT1D bought that difference when it was the only buyer in the room.

The Fund is never going to be able to invest in a company that would do that because no commercial company would.

The next question is how to unload the water in the ship and keep it from sinking. What will BT1D do with all the equity that it owns?

Unwinding Existing Assets

As noted earlier, there is a very large amount of money sitting still, some of it locked up in stock that isn’t producing any work, and the rest is cash, designated for buying more stock that also wouldn’t produce any work. None of it is paying anyone to do anything. And every dollar of it could be funding a component on the roadmap starting tomorrow.

An obvious objection: isn’t this money restricted? Donors gave to the Fund, so doesn’t redirecting it require their consent? Mostly, no. The FY25 audit puts donor-restricted net assets for the T1D Fund at $3.2 million — about 1.6% of the total. The rest is board-designated, which means the board that designated it can un-designate it.

So the question isn’t how to clean up a compliance problem. It’s how to get the money moving.

The Fund’s own page claims over $250 million under management. The audited number tells you where it actually is: $65.5M in programmatic investments plus $5.8M in programmatic notes — $71.3M at fair value, spread across roughly thirty companies. The rest is either already exited or hasn’t been deployed at all, which turns out to matter enormously.

The composition determines the options. Of the $65.5M, only $8.8M is Level 1 — publicly traded, sellable Monday morning. The other $56.7M is Level 3 preferred stock in private companies, which is exactly as stuck as it sounds. So the clean exit covers about twelve percent of the portfolio, and the hard problem is the remaining $62M.

But the largest number in this story isn’t a position at all. The same audit shows $124.1M of BT1D’s operating investments designated as Fund-related and unavailable for general expenditure — undeployed capital, reserved for equity investing, never put into anything. There’s no counterparty, no secondary market, no discount, no negotiation.

Sit with that one. It’s not stuck in a company. It’s not illiquid. It’s not waiting on anyone’s signature but BT1D’s own board, which designated it and can un-designate it. It is nearly twice the size of the entire deployed portfolio, and it has been sitting there — earmarked to buy stock that produces no activity — while other components of a cure go unfunded.

That reframes the exercise. The hard part is $62M of illiquid preferred. Everything else is a vote.

For the assets in the portfolio, holding the positions and documenting a T1D rationale for each amounts to hoping the aggregate can be argued into the charter. Where mission-relatedness can be confidently argued, that’s another matter — and those positions should be identified. For the rest, two options.

Option 1 — Convert equity to contract. Go back to each privately held portfolio company and renegotiate the equity into a fee-for-service or milestone structure tied to a T1D program. BT1D surrenders the shares; the company commits to the work. This is mutually beneficial, because BT1D gets a contract that assures the work, and for the company, cancelling BT1D’s position is reverse-dilution — everyone else’s stake goes up without anyone writing a check.

Which makes this the most useful option, because it’s also a test. A company already running a T1D program signs, because it’s committing to what it’s already doing. A company that won’t sign is telling you there was no T1D program to commit to. Either the conversion works and the money starts buying activity, or the refusal tells you the position was never doing anything.

And for companies that refuse, you move to Option 2.

Option 2 — Sell the positions. Clean, and the only option that converts dead capital into spendable money. Divest and redeploy the proceeds inside the charter via the contract model.

Publicly traded companies are easiest. Eledon, for example, is publicly traded, and those shares can just be sold. It doesn’t have any effect on the company’s balance sheet (it’s not buying equity back), nor does it have to affect the relationship it has with BT1D. In fact, BT1D can use those very proceeds to engage in the T1D-specific activities that they arguably should have done in the first place.

For privately held companies, this isn’t as clean, because private positions are illiquid — you’d have to sell into a secondary market that knows you’re a forced seller, at a discount. And the markdown would land on positions already being written down. Selling into that means realizing a loss on top of a loss.

But there’s a negative side to the company for the same reason. That new highly-discounted trade sets a reference price that follows the company into its next round. If a company was worth $10M, but BT1D’s sale values it at $2M because they had no choice but to liquidate, that downgrade hits all shareholders the same, and makes it hard to raise money at even the previous valuation. If that’s not bad enough, the company may also have an unfriendly shareholder on the books, which could be devastating.

So for private companies, both BT1D and the company have a strong interest in option 1.

None of this is exotic. These are standard instruments, and the lawyers on both sides have done conversions like this before. What’s needed is the decision, not the invention.

What about the donors? What about the families?

Right now, nothing is happening. The Fund remains intact, and no money has moved. The governance fight is not worth arguing about. So, what’s next?

BT1D lost its largest funders, and that can have a devastating effect, not just financially, but on its own good will. It’d be like a startup company’s largest investor selling its own position for pennies on the dollar. No one wants to touch it.

And the families and other donors are distancing themselves from the one and only vehicle (BT1D) that has the infrastructure, scientific and political connections, and academic resources that can move the right technologies towards a cure. Sure, they could try to rebuild a new institution, but that’s a lot of money, time and work that could just as well be implemented in BT1D given the right planning and incentive structure.

The right path forward between them is not dissimilar from what we’ve learned about equity investing: you don’t just give someone money and hope. You write contracts with terms — money attached to specified work, with obligations, milestones, and consequences for non-delivery.

The donors should do exactly that, one level up. Not a demand for transparency, which is a request. Not a threat to withhold, which burns the thing everyone claims to want. A gift agreement with terms: fund the roadmap, name the deliverables, require the reporting, require conflict of interest boundaries (not just in funding grants and contracts, but throughout the whole organization), and specify what happens if the work doesn’t get done.

At its core, this is actually a collaborative, outcome-focused act. It’s the same discipline BT1D should be applying to its own suppliers as well. If the model is right for one, it’s right for both. If BT1D can’t be held to deliverables, then it has told you something about whether it intended to deliver.

Nobody has to lose for this to work. BT1D gets a better instrument, and a plan it can show donors. The donors get accountability without a courtroom. The companies get funded work on an indication they’d abandoned. And the patients get someone finally building the thing instead of buying pieces of the people who might.

Share

Subscribe now

When I was diagnosed with T1D in 1973, this was the whole protocol: take one long acting basal insulin in the morning, and go live your life. Maybe a bolus of Regular for a meal or two. No meter. No A1c. No number to hit or miss, and nothing much for the doctor to do other than write prescriptions and then treat you when your organs started failing.

But I was assured that a cure was only five years away. “By 1980 at the latest!”, my endo assured me.

The absurdity from today’s view is sobering, and it has nothing to do with a cure. It has to do with basic daily T1D management. There was no way to test blood glucose levels. It’s not because the technology didn’t exist. It did. It’s that it wasn’t commercialized and deployed.

Instead, it sat in hospital labs from 1970 until around 1980, when the first meters for home use finally reached patients. Even then, it wasn’t taken up for another ten years due to poor demand and lack of patient education. It took that long because the ADA didn’t push it, and the JDRF never had it on the radar. The most important number in the disease was measurable for years, and no one with the power to move it acted.

The same story with glycated hemoglobin — the HbA1c that we know and love today. The NIH paper, HbA1c Standardisation: History, Science and Politics, explains how the A1c test was developed around 1977, but it sat for more than a decade until the procedure was standardized in the mid-1980s. And yet even then, only 50% of labs were using it in 1993. It wasn’t until 2004 before 99% of laboratories were actually reporting it.

What does all this have to do with our drive towards a cure?

Everything. If you’re going to cure T1D, you need to either suppress or get around the immune system. And if you can’t monitor the immune system’s activity, you’re making the same wild, uninformed guesses that we T1Ds were doing in the 1970s with insulin dosing. You need a meter. And just as glucose testing technology was available, but unused for decades, the technology to monitor autoimmune activity exists today, and it’s not being used by the same companies that are developing cures.

I’ve stated this before in prior articles, but I didn’t emphasize it strongly enough. This article aims to clarify this assertion.

Autoimmunity is not on/off — It Throttles

In each of my prior articles on the path towards a cure for T1D, I present a different angle. One is about immune suppression, another is about stem-cell derived islets, and others talk about the various components and hurdles that need to be overcome.

The throughline among all of them is that the central barrier to a durable cure is the immune system, and yet, we’re treating everyone exactly the same: that we all have the same level of hyperactive autoimmunity.

Yes, it’s true that for many of us, our immune system killed the beta cells in our pancreas. We constantly have to explain that to people who conflate it with type 2 diabetes. No, we’re not them. Moreover, once beta cells are gone and exogenous insulin is required, it no longer matters how you got the disease.

But, as all my articles have stated, the immune system is not a binary on/off state. It throttles. And if you’re going to develop a cure, you must take this into account. Indeed, many T1Ds are entirely non-autoimmune. I provide the greatest number of citations on this in my article, To Cure T1D, Start with the Easiest Cases, Not the Hardest, where I quote the surprising statistic that nearly 22% of people diagnosed with T1D show no detectable autoimmune activity — their diabetes arises through non-autoimmune mechanisms entirely.

These statistics come from a number of sources, including a 2022 global scoping review by Ross and colleagues at the Harvard T.H. Chan School of Public Health, which pooled 125 studies across 48 countries. In the aggregate, they found that islet-autoantibody positivity varies enormously by region.

Keep in mind, these are autoantibody negative assays. Beta cells are killed by T cells. Autoantibodies are the byproduct of a destruction that has already taken place (or is in the process). They’re like the bullet casings at the scene of a crime. You know it happened, but you have no idea whether the killer is still there, or will come back. In other words, these antibody assays are imperfect. People who test negative may well have autoimmunity, but the immunity has waned such that there are no more antibodies left to detect. But that itself is worth noting: if autoimmunity wanes, then aren’t these people “functionally autoimmune” for transplant purposes?

These are great questions that open even more questions: if autoimmunity wanes, does it come back? Would these people be candidates for different kinds of curative therapies than those with higher levels of autoimmunity? Think of LADA patients — those who develop T1D gradually in adulthood. Estimates put LADA at 2–12% of all adult-onset diabetes, but what matters here is tempo: LADA progresses slowly enough that patients go years before presenting, and are routinely misdiagnosed as T2D along the way.

Testing for autoimmune T cell activity — not autoantibodies — is far more informative, because it reveals not just who is non-autoimmune, but how intense the autoimmune activity is. Just as T1D management is hit-and-miss without the ability to test for glucose in the bloodstream, a cure will be similarly erratic until we test for the presence and magnitude of a person’s autoimmunity. And unlike a bolusing error, a therapeutic drug that may restrict your immune system’s response to cancer and other opportunistic infections is a lot worse than just a hypo.

This was the mechanism underneath my “start with the easiest cases” argument. That piece made a sequencing case — build from the base of the pyramid upward — and stratification is what makes the sequence possible. You cannot start with the easiest cases if you have no way to identify them.

Despite the field knowing that autoimmunity throttles, the tools to measure it in real time are not being deployed. They’re sitting in a research lab in Australia, where the researcher is struggling to get funding. Not only are organizations hardly paying attention to this, they’re withdrawing funds from this research.

And the worst part is that building this assay that detects autoimmunity thresholds is the fastest, least-expensive, lowest-barrier technology in the entire field of T1D science. There is no area of research in T1D that does not benefit from this assay. The cost and resources to develop it to the point of widespread commercial deployment are infinitesimal compared to the time, cost and lives we’re investing today.

And to pour salt on the wound — or pour glucose into a DKA patient — all those other technologies need to assess autoimmune T cell activity if those are even going to work.

So, what is this assay?

BASTA — A First Generation Deployable Assay

BASTA — the Beta cell Antigen Specific T cell Assay — comes out of Stuart Mannering’s lab at St. Vincent’s Institute of Medical Research in Melbourne. He published his findings in Science Translational Medicine in 2025, and what makes it matter isn’t sophistication; it’s the simplicity.

To explain how it works, we start with what we ultimately need from it, which is embarrassingly simple. Draw a small tube of blood from the patient—about the same amount you’d take for any traditional lab work—and add the proteins that beta cells make. These are the same proteins the immune system attacks in type 1 diabetes. If the person’s autoimmunity is active, the T cells in that test tube of blood will recognize those proteins and react. If they don’t react, nothing happens and we know the person is non-autoimmune. But if there is a reaction, we want to know more than whether a reaction occurred, we want to know how strong it is, how much of the attack is underway.

This is not technically hard, per se, but the historical methods to do it have been intensive and equipment-heavy. Prior to BASTA, a very large blood draw was required because the abundance of these cells we’re looking for is very rare. The lower the abundance, the higher the resolution you need to detect them—hence, a larger volume of blood. In current protocols, research-grade equipment is needed that only a few labs have. In short, a huge amount of blood, a very long turnaround time to results, and extremely expensive equipment and radioactive processes.

In the BASTA assay, if a person has autoimmunity, their T cell receptors engage the antigens in the tube, and that readout is a protein called IL-2. It acts as a proliferation signal, telling the T cell and its neighbors to multiply into a larger population. It therefore acts both as the announcement that a T cell has recognized its target, and the fuel that expands the response.

BASTA listens for that signal, which happens within hours of the T cell recognizing its target. Catching that early signal is what makes the test both fast and sensitive — it reads a reaction in a day instead of a week, and it picks up autoreactive T cells too scarce for the older methods to detect. And because the amount of IL-2 climbs with the strength of the response, BASTA answers the “how much” directly: a bigger signal means a larger share of the immune system is engaged in the attack.

The practical advantages compound:

• 24-hour turnaround versus five to seven days. Clinically actionable in a way the prior assays never were.

• Plasma can be frozen after the culture step and shipped to a central lab for batch MSD analysis. This is the logistical breakthrough. It breaks the fresh-blood-only constraint that has confined every prior assay to specialized centers. A trial site anywhere in the world can collect blood, set up the 24-hour culture, freeze the plasma, and ship it. The measurement infrastructure lives at one central location.

• 0.5 mL of blood per antigen treatment. Compatible with pediatric populations, with frequent serial sampling, and with blood draws that are already occurring as part of trial protocols.

• Quantitative magnitude, not binary. Returns a continuous IL-2 signal reflecting real-time intensity of antigen-specific T cell activity, tracking exactly how autoimmune a patient is over time, against which antigens, and whether that activity is increasing or decreasing in response to treatment.

• No radioactivity. Standard biosafety requirements only.

The published performance backs this up. Asked to separate people with new-onset T1D from people without it, the test was accurate about 86% of the time using a single beta-cell antigen, and about 93% using a small panel of insulin-derived peptides in adults — strong for a test this simple. It also distinguished stages of risk that antibody tests blur together: among at-risk children carrying multiple diabetes-related antibodies, the T-cell response was markedly higher than in children carrying only one, meaning BASTA can see who is further down the road, not just who is on it.

And when people with established T1D were tested four times over several weeks, their results held steady — the signal is stable enough to track a patient over time rather than giving a different answer at every visit.

The assay isn’t perfect yet. It still faces engineering challenges: the test-to-test variability needs to be tightened for individual tracking, children occasionally show baseline immune background noise that blurs early detection, and the full panel of target proteins is still being validated. These are standard development hurdles, not dealbreakers—the kind of issues that get ironed out once an assay is run at volume.

These are solvable problems. They are the development agenda, not dealbreakers. And they are the kind of problem that gets solved by use — assays tighten when they are run at volume, on real cohorts, against real endpoints.

Therapies that require autoimmune T cell testing

Now we can talk about curing T1D.

The problem with how clinical trials are run today is that they are looking for different endpoints: A1c levels, time in range, C-peptide, and other biomarkers. When you put new islets into a person, investigators track each of these endpoints because they show whether the islets are producing their own insulin.

The problem is, whatever the intervention is, those very same endpoints could be affected by the person’s autoimmunity. If they’re totally non-autoimmune, the new islets might be working independently of the drug.

Transplanted islets face two different attacks: Alloimmunity is rejection of foreign tissue — the recipient recognizing donor cells as not-self. Autoimmunity is the original disease, aimed at beta-cell proteins regardless of whose cells they are. And both are very different in their response mechanisms.

Nearly everything the field has built addresses the alloimmunity with the expectation that it’ll capture autoimmunity at the same time. That is, suppress T cell activity.

But it’s not that simple. Alloimmune rejection does not produce a beta-cell-antigen-specific T cell response. Recurrent autoimmunity does. So when a graft is lost, a rising BASTA signal points at the disease coming back, and a flat one points somewhere else — rejection, drug toxicity, or an islet that never engrafted properly. Today all three arrive at the same place: a graft that stopped working, and no way to say which thing killed it.

For example, consider the cell-replacement therapies at center stage. In December 2024, a single patient at Uppsala University Hospital received Sana Biotechnology’s hypoimmune-engineered islets transplanted into the forearm without immunosuppression. The term is worth pinning down, because it's routinely misread: hypoimmune describes the gene editing, not the cell source. UP421 used primary islets recovered from a deceased donor, then edited. Sana's stem-cell-derived version of the same platform — SC451, built from iPSCs carrying identical edits — has not yet entered the clinic, though Sana expects to begin a Phase 1 trial in 2026. (As of Sana's Q1 2026 report (May 11), Sana is conducting nonclinical testing, manufacturing transfer to contract manufacturers, and clinical trial preparation, and expects to file an IND and begin a Phase 1 trial for SC451 this year. The July 13 NEJM follow-on release repeats the same expectation.)

The results from UP421, published in the New England Journal of Medicine and celebrated widely, showed that, at 14 months, the beta cells were still producing insulin with no detectable immune response.

But wait, there are multiple problems here. First, hypoimmune editing strips the markers that flag a cell as foreign — which is an alloimmune solution. Recurrent autoimmunity recognizes the beta cell by what it makes, not by whose it is, and no amount of editing changes that.

So, we have absolutely no idea why this single patient’s islets kept working because their autoimmune profile was never characterized. Whether the hypoimmune engineering worked, or whether the patient’s own attenuated (or even entirely absent) autoimmunity contributed to the outcome, is unknown.

Moreover, the islet mass was intentionally very low, because the goal was to see if there was an immune response (not whether it yielded glycemic control). With low islet mass and no measurement of autoimmune activity, there is nothing about this trial that can be generalizable. And even if this patient carried high autoimmune activity, the deliberately sub-therapeutic islet mass may not have provided sufficient antigenic stimulus to trigger a measurable response — leaving open the question of whether a full therapeutic dose would fare the same.

Next, consider Vertex’s Zimislecel trial — now in Phase 3, enrolling approximately 50 patients, where the same immunosuppressive regimen is applied to every participant. Will grafts fail regardless of the drug, because their pre-existing T cell autoreactivity will overwhelm the graft before immunosuppression can establish control? Might others succeed with less intervention because they simply don’t need it (because their immune response is attenuated)?

Now add Eledon’s tegoprubart trial, where 12 patients are showing promising outcomes. The same question arise, but in reverse: are those results evidence that tegoprubart works, or evidence that some subset of those patients had low baseline autoimmunity? Without pre-transplant immune profiling, the two explanations are indistinguishable — and the dosing implications of each are entirely different.

A further question cuts across all three trials: will the introduction of new islets trigger a resurgence of T cell activity in patients whose autoimmunity might appear to have waned? If so, by how much? We simply don’t know.

As it happens, there are two trials that took place in 2008 and 2009 — nearly twenty years ago — that demonstrated exactly this.

Huurman et al., PLOS ONE, 2008, enrolled 21 long-duration T1D patients receiving cultured cadaveric islet cell grafts using the traditional protocol. Before transplantation, and at regular intervals for one year afterward, the investigators measured cellular autoreactivity against islet antigens using LST — a Lymphocyte Stimulation Test — which is one of those older assays described earlier that BASTA replaces.

The findings were stark. Seven of eight patients without pre-existing T cell autoreactivity became insulin independent. None of the four patients reactive to both GAD and IA-2 before transplantation achieved insulin independence.

Crucially, autoantibody levels showed no such association — confirming that what mattered was not past immune activity, but the T cell activity itself, measured in real time.

Hilbrands et al., Diabetes, 2009, extended these findings in a cohort of 30 consecutively transplanted recipients under the same protocol. The results reinforced and expanded what Huurman had shown: of nine patients without baseline T cell autoreactivity, seven became insulin independent. Of 18 who tested positive, only six did. Combining T cell autoreactivity with B cell count above the median, the discrimination became even sharper: seven of eight patients with low immune burden became insulin independent; only one of eleven with high immune burden did.

The field has been generating this kind of noise for twenty years. The Bedrat et al. paper published in Diabetologia in July 2026 analyzed 560 TrialNet participants across six immunotherapy trials and made a striking finding that few have noticed. Two immunologically distinct subgroups with different therapeutic responses had been enrolled together in every trial, treated identically, and only identified retrospectively through transcriptomic analysis: subgroup one showed better responses to anti-CD20 therapy; subgroup two showed better responses to CTLA4-Ig.

The same patients, the same trials, opposite therapeutic implications — depending on which subgroup they belonged to. Nobody knew which subgroup any individual patient was in, because nobody measured it prospectively. The variable participant responses that have confounded trial after trial were not random. They were structured heterogeneity that no one thought to characterize before enrollment.

This is the cost of not stratifying.

Together these papers established three things that remain true today:

• First, the immune profile of the recipient at the time of transplantation is a more powerful predictor of graft survival than any metabolic or graft-related variable studied. Beta cell mass matters; immune burden matters more.

• Second, the relevant signal is T cell autoreactivity, not autoantibodies. Autoantibody status showed no significant association with outcome in either cohort. The field’s standard pre-transplant immune workup — which measures antibodies — is measuring the wrong thing.

• Third, T cell measurement changes what’s possible clinically. Across both cohorts, patients with no detectable T cell autoreactivity achieved insulin independence roughly 80% of the time (14 of 17). In Huurman’s cohort, none of the four patients reactive against both GAD and IA-2 achieved it. A patient with autoreactivity against two antigens has a near-zero chance under the same protocol. These are not marginal differences. They are the difference between a successful transplant and a failed one — and they were knowable before the graft was placed.

And that was in 2008 and 2009.

No matter what technology you’re going to use to try to cure T1D, all of it funnels down to monitoring autoreactive T cell activity before, during, and after treatment. And frankly, forever.

This isn’t just a good idea; it’s inevitable. Just as we’d never have managed T1D without testing for glucose, we’ll never cure T1D without testing for autoimmune activity. Twenty years from now, we’re either going to still be throwing billions of dollars towards a cure because we haven’t widely deployed these assays, or we’re going to look back on this with the same wincing we do now for ignoring commercial home blood glucose testing in the 1970s.

Epidemiology, Genetics, and Environmental Factors

The utility of this assay would address many other questions beyond just finding a cure. If this assay were as routine as the A1c test, we could characterize autoimmunity profiles across the full living T1D population, at population scale, longitudinally. We can look for trends. Associate profiles with phenotypes, genetics, and environmental triggers.

We can use those profiles to stratify patients before transplant (or other intervention), calibrate immunosuppression to individual immune burden, monitor whether interventions are working, and adjust dosing as the immune landscape changes over time.

For example, can BASTA predict T1D years earlier than current methods?

Consider a 2026 study in the journal Diabetes, where researchers used machine learning to spot a metabolic inflection point a year or two before clinical diagnosis. It’s a genuine advance, but it is still a metabolic measurement tracking a fire that the immune system started long ago. We don’t want to wait for the smoke; we want to catch the T-cells holding the matches.

But it is still a metabolic measurement that sits downstream of the immune event that caused it. That immune system was attacking the cells long before this OGTT method could work. What we really want to detect is theT cell activity that is driving that decline before any metabolic signal is visible at all.

Put most succinctly, the OGTT technique is a lagging indicator. BASTA may be a leading indicator, making it possible to detect onset far sooner. How soon, we can’t know because the assay isn’t being used.

That distinction matters enormously for intervention timing. Teplizumab — the drug that’s being used to delay T1D onset — its eligibility today runs on autoantibodies plus dysglycemia, both of which are downstream of the autoimmune attack the drug exists to interrupt. That’s way too late.

Every study of teplizumab, abatacept, and other immune modulators shows that earlier intervention produces better outcomes — which means the earlier the immune attack is detected, the larger the population that could benefit from intervention before substantial destruction has occurred.

Every question being asked in T1D research today would be better answered with this data. Every clinical trial currently underway would generate more interpretable results with it. Every therapeutic modality — from cadaveric islets to autologous SC-derived cells to gene-edited hypoimmune islets — would be better matched to the patients most likely to benefit from it.

There is one last critical key to all this: standardization.

The Standardization Imperative

The vision described in this article — population-scale autoimmunity profiling, longitudinal tracking, data that can be compared across trials, institutions, and countries — depends on something that sounds mundane but is historically one of the hardest problems in clinical medicine: standardization.

The field has been here before with the HbA1c test, as explained at the top of this article, and we cannot afford a twenty-year standardization gap for an assay like BASTA.

The core technical parameters that would need to be standardized for a T cell autoreactivity assay are already partially understood from BASTA’s published validation work and the legacy LST literature. At minimum they include: the antigen panel — which beta cell proteins are used to stimulate the T cell response, at what concentrations, and in what combinations. The current evidence supports at minimum C-peptide, GAD-65, and IA-2 as discriminating antigens in adults, with ZnT8 and IGRP under evaluation.

The detection method — whether IL-2, another cytokine, or a proliferation readout — determines sensitivity and the equipment required.

The sample handling protocol — time from blood draw to culture setup, temperature, freeze-thaw conditions for plasma — drives the inter-assay variability that currently sits too high for reliable longitudinal monitoring at the individual patient level.

And the reporting format — whether results are expressed as a stimulation index, a raw IL-2 concentration, or a normalized score relative to a reference population — determines whether data from different labs can ever be compared.

Each of these parameters, left unstandardized, produces the same problem HbA1c had in 1985: real data that can’t be compared, real clinical decisions that can’t be coordinated, and a measurement whose full value can’t be realized because no two labs are measuring quite the same thing.

The solution is an open standard — a publicly defined protocol, a reference antigen panel, a reference detection method, and a certification process for laboratories that want to run the assay and have their results be comparable to everyone else’s. This is exactly what the NGSP built for HbA1c, and it is exactly what will be needed here.

The Immunology of Diabetes Society is the natural candidate to lead this effort; it already runs inter-laboratory standardization workshops for T cell assays, and Mannering’s group has participated in those efforts. The IDS does not have the resources or mandate of the AACC, but it has the scientific credibility and the right community relationships to convene the process.

What it would need is a funding commitment from an organization with the reach and the interest to see this happen, which brings us to the organizations that are supposedly dedicated to finding a cure, and the question of whether population-scale immune monitoring registers as the kind of infrastructure investment worth making.

The home glucose tests and the HbA1c stories end well, as each became one of the most reliable and widely used tests in all of medicine, and it did so because the field eventually decided the stakes were high enough to do the coordination work. It’s just a shame it had to take as long as it did to get there.

Call to Action: The Field Now Has to Act

The evidence for T1D heterogeneity is no longer in dispute. The Huurman and Hilbrands papers proved two decades ago that immune profile predicts graft survival — and that the field’s standard pre-transplant workup is measuring the wrong thing. The Bedrat et al. paper in Diabetologia showed that immunologically distinct subgroups with opposite therapeutic responses have been hiding inside every TrialNet immunotherapy trial ever run, enrolled together and treated identically, their differences invisible because nobody measured them prospectively.

The same blind spot darkens other major plays in the field, like Vertex’s Phase 3 trials or Eledon’s promising antibody treatments. In every case, the same blanket protocol is applied to everyone. Are the successes due to the drugs, or did those patients simply have quiet immune systems at baseline? Are the failures a flaw in the medicine, or did a raging, undetected T-cell attack overwhelm the therapy before it stood a chance? Without baseline profiling, we are completely blind.

Hence, the callback to my article, To Cure T1D, Start with the Easiest Cases, Not the Hardest, where I argued that we need to stratify patients according to their autoimmune profiles. Those without any autoimmunity could receive autologous stem-cell islets — those grown from their own tissues — which could potentially involve very low-dose immunosuppressant drugs, or potentially, none at all. (The article gets into more nuance about that variability: the manufacturing process may introduce neoantigens that trigger an alloimmune response, but this is currently theoretical and has never been tested, let alone confirmed.)

Those with low-level autoimmunity might be able to tolerate hypoimmune stem-cell islets with mild immunosuppressants. And those with hyperresponsive autoimmunity might need the most aggressive treatment.

The science has been pointing in the same direction for twenty years. And now, the first generation of an assay that can measure real-time autoimmune activity exists in published, peer-reviewed form.

The question is whether funders, trial sponsors, and the institutions organizing T1D research are willing to prioritize the tools that let them act on that heterogeneity — or whether they will continue generating variable, uninterpretable results from unstratified trials, and wondering why the average effect is always modest and the individual responses always confounding.

The cost-benefit here is so glaringly obvious that Breakthrough T1D, Helmsley, and the other advocacy organizations should be doing all of the following:

• Convene the standardization through the body that already exists. The Immunology of Diabetes Society has run a T-Cell Workshop since 2000 and published harmonization guidelines as recently as 2022. Nothing new needs to be built. What it has never had is a candidate assay simple enough to run outside a specialist lab.

• Fund reference materials and a common protocol. Standard samples containing defined numbers of antigen-specific T cells, distributed blind-coded across labs. This is precisely what the field did for autoantibodies — the WHO adopted serum reference standards for GAD and IA-2 assays decades ago. Nobody ever did it for the T cell arm.

• Build the central-lab architecture, not thirty local ones. The IDS’s own multi-centre work found reproducibility was far better in a single laboratory. BASTA is designed for exactly that: culture the blood at any trial site, freeze the plasma, ship it, read it in one place.

• Start with samples that already exist. Banked specimens from completed trials, where outcomes are already known, would answer whether baseline autoimmune status predicts graft survival — retrospectively, cheaply, without enrolling a single new patient.

• Add it to trials already running. The blood draws are happening. This is an add-on, not a redesign.

• Open the regulatory path. If a fall in T cell activity predicts later C-peptide preservation, it becomes a surrogate endpoint — which compresses the timeline for every trial that follows.

• Then require it. Once the assay is qualified and standardized, funding for clinical trials should require baseline autoimmune stratification as part of the proposed protocol. Not as a courtesy to science, but because a trial that cannot interpret its own primary outcome is not a good use of anyone’s money.

Twenty years from now, we don’t want our children asking why we’re still taking insulin — or why they are. We want to tell them stories of yore, where smart people made the right calls back in the mid-2020s that ultimately led to real, durable cures. While we’re at it, we can tell them about the charmingly primitive things that were once in vogue: AI in our glasses, chatbots that did our thinking, and closed-loop insulin pumps — now behind glass at the T1D Museum that also holds other artifacts of a disease that has since been cured. It will likely be in the building that used to be the Joslin Diabetes Center.

Subscribe now

Share

We all know the famous balcony scene from Shakespeare’s Romeo and Juliet, where Juliet asks, “wherefore art thou, Oh Romeo?” She isn’t asking where Romeo is. She’s asking why — why does he have to be a Montague, the one identity that makes him impossible. The feud between their families might mean death. So, she asks, why are you the thing you are, when everything else about you is exactly what I want?

She was fourteen years old, and she had seen him exactly once — across a crowded room, at a party, for approximately four minutes. By the end of the evening she was prepared to renounce her family, her name, and her future for him. By the end of the week she was dead.

We don’t mock Juliet for this. We recognize her. That particular human experience — wanting something so completely, so urgently, that the wanting itself becomes the argument for having it. But she was naive, uninformed about the ways of relationships. Her love for Romeo was a fantasy, one that can be easily had because those around her had what she had not.

And yet, she also knew there was danger. But she pursued him anyway, because desire at that intensity doesn’t consult wisdom. It overrules it.

The T1D community’s relationship with islet transplantation has the same architecture. A cure exists — or something close enough to it that the distinction feels cruel. People have heard about it. They want what others around them have—a life without diabetes. The wanting has been building for decades, with institutional promises that a cure is five years away! And by now it has the force of something that feels not just desired but owed.

And surrounding that desire is a constellation of obstacles — insurance, regulation, access, bureaucracy — that together feel like the Montague-Capulet feud: arbitrary, structural, unjust, and entirely removable if the right people would simply decide to remove them.

In other words, the “villain” in this story is some mysterious actor that needs to be eliminated. But which one? Who is in the way? Is it the institutions? Insurance? A company that has exclusivity and is throttling access? Nearly everyone that gets near islet transplants is a suspect.

The ISLET Act, in this reading, is the rope ladder Romeo throws up to the balcony that just escapes the villain, whoever it might be. Finally. A way through. Pass the act, reclassify the islets, unlock the coverage, and the thing everyone wants becomes the thing everyone can have.

In my last article, Assembling the Components to a Cure, I examined why the ISLET Act doesn’t accomplish what people believe it will, and why pushing resources toward it is pulling attention away from the approaches that might actually produce a cure worth having.

Yet, the pursuit of the villain continues, but not with the circumspection that Juliet is aware of. The T1D’s passionate drive towards that cure is undaunted.

I’m reminded of the line from the play where Friar Lawrence warned, “these violent delights have violent ends.”

And so the passionate momentum behind the ISLET Act persists. In my “Assembling the Cure” article, I explained the rational reason why the Act is dangerous, but I didn’t address the passion about it. It’s here that I delve more deeply into what many believe are institutional barriers. Namely, insurance and access.

That may not sound as romantic as Romeo and Juliet, but oh, dear reader, it is. It’s rich with intrigue, good intentions gone awry. It’s the hunt for the villain, all wrapped up in a tidy little play of our own.

Let’s start with the biggest misconception about islet transplants: the number of people who’ve actually gotten them.

Islet Transplants by the Numbers

The Collaborative Islet Transplant Registry (CITR) was established in 2001 by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) and has provided the most comprehensive longitudinal dataset on human islet transplantation worldwide. It covers North American, European, and Australian programs. Participation is voluntary.

From 1999 through 2023, CITR collected data on 1,477 allogeneic islet transplant recipients — 1,134 islet transplant alone and 260 islet after kidney — who received 2,947 infusions from 3,442 donors. North American sites contributed 46% of recipients; Eurasian and Australian sites contributed 54%. (The link provided has extensive tables showing all the sources for where it collects data.)

That’s 25 years, 40 programs across 10+ countries. Spread out, it’s roughly 60 transplants per year globally. In North America alone, it’s closer to 30 per year — at peak. Yes, at the peak.

This dispels several of the rumors people have asserted, that the US does the least number per year (it doesn’t), and that other countries do thousands per year (they don’t). Historically, North American sites have contributed more recipients than any other region — though that lead has eroded sharply over the past decade, as it has not been evenly distributed over time. There was a global decline in new recipients between 2011 and 2023, where there was exactly one procedure in North America.

One transplant. One center. In all of North America. In 2023.

People look at that and shake their heads in disbelief. That’s what starts the conspiracy framing: it’s got to be insurance. It’s got to be regulatory. It’s got to be CellTrans/Lantidra — after all, they got FDA approval in 2023 and the exclusive right to market this product, so they must be deliberately throttling supply to protect their market position, which the ISLET Act would correct by reclassifying islets as organs.

Consider the questionable logic there: If you have a monopoly on a highly profitable procedure, why would you throttle access?

The conspiracy thinking follows that logic that since all T1Ds would love to have this procedure and if only one was done in 2023, someone or something is in the way!

The hunt is now on that villain. Surely, they hide amongst us!

Let’s start with insurance. Surely, they’re responsible. Those darn Montagues are at it again!

Insurance Coverage

In my article, I said that private insurers have written coverage policies for Lantidra, which is true. But let’s explore this ecosystem more carefully.

First, the private insurers that cover islet transplantation include Centene/WellCare, BCBS Michigan, Florida Blue, and others. But, the Cigna policy document contains the sentence that corrects the broader claim: “In the USA, islet transplantation is not yet approved for clinical use and reimbursement.” That’s Cigna’s own language, written after Lantidra’s FDA approval. Here’s what that reveals about the actual coverage landscape.

Private insurer coverage policies from a handful of payers were written to establish medical necessity requirements given the absence of coverage criteria provided by CMS and applicable Medicare Advantage Contractors.

That clause is the tell. The policies exist precisely because CMS hasn’t acted. This is not the fault of private insurers — they will cover the procedure, though it requires a prior authorization to get it. But the real barrier is a CMS National Coverage Determination (NCD) or Local Coverage Determination (LCD) for Lantidra. Without that, Medicare and Medicaid — which cover a disproportionate share of brittle T1D patients given their age and comorbidity profiles — don’t have a formal payment pathway.

Patients on Medicare or Medicaid can’t reliably access it even if a private payer would cover it.

Whether Lantidra itself was covered as a drug or an organ, the transplant procedure — the portal vein infusion, the hospitalization, the islet processing facility costs — require separate facility and procedural reimbursement codes. Those haven’t been standardized either.

So the more precise framing for this situation is that some private insurers have written Lantidra coverage policies, but there is no CMS determination, no Medicare coverage, and no standardized procedural billing pathway.

Ok, so insurance isn’t the villain. It’s CMS. Yeah, that’s it! It’s the original misclassification of islets! Therefore, the rope ladder of the ISLET Act’s organ-reclassification mechanism is an attempt to bring islets under the UNOS/transplant reimbursement framework rather than the drug/biologic framework. Eureka! Villain found!

Right?

Does organ reclassification automatically trigger UNOS/transplant reimbursement?

No — and this is the critical gap in the ISLET Act advocacy argument. Reclassification under NOTA (the National Organ Transplant Act) would bring islets under OPTN/UNOS governance for procurement and allocation, but it doesn’t automatically generate a CMS payment determination. Those are separate statutory and regulatory tracks.

What reclassification does:

• Removes the prohibition on valuable consideration for islets (currently they’re a biologic, not an organ, so donor compensation rules are different)

• Brings islet procurement under the OPTN network, potentially expanding the number of procurement centers

• Creates the predicate for CMS to write an NCD treating islets like other transplanted organs (kidney, liver, heart all have established transplant DRGs)

What it doesn’t do:

• Force CMS to act, or set a timeline for doing so

• Establish procedural billing codes for the infusion, processing facility, or post-transplant management

• Guarantee any specific reimbursement rate

Therefore, reclassification is not sufficient for solving the reimbursement problem. It opens the door; it doesn’t walk through it.

Ok, so the villain isn’t reclassification. But surely, there are other institutional actors that could make this happen. They are the villains, right?

Is legislation required, or is there another path?

Three alternative mechanisms exist, none of them clean:

CMS Coverage Determination (NCD/LCD) — CMS could issue a National Coverage Determination for islet transplantation as a procedure independent of the ISLET Act. This has been discussed for over a decade and hasn’t happened. The barrier isn’t statutory — it’s evidentiary. CMS requires randomized controlled trial data or substantial comparative effectiveness evidence before writing an NCD for a high-cost procedure. The Lantidra approval was based on 30 patients in two non-randomized single-arm studies, which FDA accepted but which CMS has historically not considered sufficient for coverage.

Transplant DRG extension — Congress or CMS could simply extend existing organ transplant DRG codes to cover islet infusion procedures administratively, without reclassifying islets as organs. This would be a narrower, more targeted fix that bypasses the NOTA reclassification debate entirely. It’s less politically legible than the ISLET Act but potentially more achievable.

Private payer pressure cascade — If enough commercial payers establish Lantidra coverage policies (which is already happening slowly), Medicare Advantage plans — which follow commercial payer precedent more readily than traditional Medicare — could begin covering it without an NCD. This is how several other biologics have achieved de facto Medicare coverage before a formal determination. It’s slow and uneven, but it doesn’t require legislation.

Keep in mind again that the whole aim of doing all this — the entire circus of shenanigans we’re discussing — is only useful if (and ONLY if) tego could be part of that islet transplantation. Because otherwise, this is even more much ado about nothing than it already was.

So, even if all this were to pass, tegoprubart still has to go through the clinical trials so it can be used here.

And even then, there’s that requirement: substantial comparative effectiveness evidence. There has to be evidence that this procedure is better than alternative interventions—existing insulin therapy—to make sure that the cure is not worse than the disease. Merely wanting to be cured of T1D is like Juliet fantasizing about Romeo because she sees happy couples all around her.

To some, that’s not good enough. The villain is still out there! Who is it?

To figure the real reason why no one’s getting transplants, let’s do a thought exercise by waving the magic wand and saying anyone, anywhere could just get it today. No barriers whatsoever. How many people would actually get transplants?

On demand: Is it Knowable?

The aggregate “how many T1D patients would get an islet transplant” question is noisy because it conflates actual need and eligibility with uninformed preference, and most T1Ds are uninformed. We have regulatory protections to keep people from putting themselves at risk. So, let’s look at who is actually eligible.

The brittle T1D / severe hypoglycemia population is actually well-characterized, where the health risks of continuing to live with T1D are actually worse than the islet transplants. The Lantidra indication — unable to approach target HbA1c due to repeated severe hypoglycemic events despite intensive management — maps to a defined clinical subgroup. Estimates put severe hypoglycemia unawareness at roughly 25–30% of long-duration T1D patients, though only a subset have the frequency and severity that meets transplant eligibility criteria.

A commonly cited figure in the islet transplant literature is approximately 10,000–15,000 patients in the US meet the rigorous eligibility criteria for allogeneic islet transplantation under current protocols.

If that number looks familiar, it’s because this is the number that advocates say would get it today. Yes, they could, and that’s the passion. That’s Juliet seeing Romeo from across the room.

But to all the tens of thousands of Juliets out there: would you get that transplant? The answer is something most T1Ds wouldn’t expect.

Plot Twist: The Real Villain is the Cure Itself.

The reason why people aren’t getting an islet transplant: no one actually wants it.

And when I say “no one”, I’m talking about the informed T1D, not the Juliet fantasizing about the boy with whom she exchanged glances from across the room for four minutes. The Juliet who thinks that the cure requires simply walking into a lab, getting an infusion that lasts 20 minutes, and walking home insulin-free. That’s the Juliet that dies at the end of the story.

The mature, informed Juliet has been around the block a while and can spot a highly toxic, and physically and emotionally abusive relationship a mile away. That Juliet knows better, and is also the informed T1D.

Remember, this procedure is available in Canada, Australia and other countries, completely covered by government healthcare, and carry very few limitations on eligibility. All the hand-wringing about America’s regulatory environment, CMS, eligibility requirements — all of that — doesn’t exist in other countries, and no one is getting it there either.

Here’s the informed protocol — the one that’s offered worldwide, that globally, has only seen 60 transplants a year at its peak, and has subsided to nearly none:

• A 90% serious adverse event rate — that’s in the Lantidra package insert, cited in the BCBS Michigan policy.

• It requires portal vein infusion, with documented risks of liver laceration, hemorrhage, portal hypertension, and portal vein thrombosis.

• It produces insulin independence that is temporary in most patients — some for one to five years, some for over five years, many never achieved it at all.

• And requires immunosuppression with a drug that is highly toxic.

Note that that drug—tacrolimus—can be replaced by tegoprubart, a far-less toxic immunosuppressant that, advocates believe, suddenly make the existing protocol viable for tens of thousands of people. Why wait for future developments, they argue, if the ISLET Act can lift the regulatory barriers and people can get it done with tego?

We’ve already established that the regulatory barriers aren’t really barriers at all; it’s the procedure. And the procedure isn’t just because of tacrolimus’ toxicity; it’s everything else about it that causes informed patients to opt out.

And tego is still an immune suppression drug. Once started, there are numerous other risks associated with opportunistic infections, cancer risks and so on. Furthermore, the person may now be “HLA-sensitized”, which makes them harder to match for any future organ transplant including kidney. This is a big deal, since many uncontrolled T1Ds often need kidney transplants. Getting islet transplants and reverting to normal glucose levels doesn’t suddenly reverse that risk.

Besides, the 90% adverse event rate for portal vein delivery is substantial, not to mention the fact that insulin independence is temporary.

And since the ISLET Act only affects cadaveric islets, which would still be required for this kind of transplant (ie., the act has no effect on stem-cell islets), there is no other approved (or tested) protocol for placing cadaveric islets anywhere other than the liver. So, even with the best outcomes of tego, this procedure is locked in place.

No informed patient with access to a competent endocrinologist is choosing that trade-off, which was the thesis of my article, T1D: The Cure/Disease Trade-off. If you want a real-world understanding of what the risks are, not to mention the incredibly high daily burden of being on any kind of immunosuppression, that’s the article to read.

It is for this reasons that, in the European JDRF Center for Beta Cell Therapy in Diabetes, immunosuppressive medication is routinely tapered off in recipients starting at least one year after transplantation on the grounds that the disadvantages of continuous immunosuppressive therapy may eventually be more detrimental to the patient’s condition and quality of life than regular intensive insulin treatment.

My closing argument in that article was simple: Your health should never get to a point where you would ever want to be cured. This is why most informed T1Ds don’t get islet transplants, except for that one person in 2023.

None of this is to dispel the emotional element, the one that still pines for Romeo. That’s real.

But there’s a lot more to take away from this besides just your own desire for a cure.

When I said that the ISLET Act is much ado about nothing, I was referring to its potential benefits, where there are none. What must also be considered is the potential harms: it’s diverting attention away from other approaches to a cure where the ISLET Act has absolutely no effect.

The Solution: Evolve the Cure.

As we found, the real barrier to accessing the cure is the nature of the cure. It’s just old technology that was a step in the evolutionary process of science. Iterative steps are necessary to move from early protocols to newer ones. The so-called “cure” that people are excited about now is not viable, but that doesn’t mean new ones won’t be.

To know where we’re going, it’s important to understand how we got here, both to understand where to move on from, but to avoid having history repeat itself.

To explain this, let’s go back to 1993. Whole organ transplants were common, but there were new therapies in development that are more complex than simply moving an organ from one person to another. New techniques in “gene therapy” were evolving that involved some kind of processing. This could include cells that are altered in some way, including stem-cell derived products. The FDA statement (Notice 58 FR 53248) defined any product that needs this kind of processing as something that would be altered in biological characteristics ex vivo to be administered to humans.

In plain English, that means that extracting islets from the donor pancreas is not a simple harvest. It involves enzymatic digestion of the donor pancreas using collagenase, density gradient purification, culture periods, and quality testing. The FDA looked at that process and said: this is not like transplanting a kidney, where the organ goes from donor to recipient with minimal manipulation. This is a manufacturing process that produces a cell product. Therefore, there should be quality control associated with them to assure safety in manufacturing.

In other words, this belonged in a different category — biologics — and those fall under Section 351 of the Public Health Service Act, alongside stem cells, cancer vaccines, activated T or B lymphocytes, chondrocytes for cartilage repair, keratinocytes, fibroblasts, and hepatocytes.

This wasn’t necessarily seen at the time as problematic. After all, islet transplants didn’t happen.

Until they did.

When Edmonton’s results arrived in 2000 and the first T1Ds were “cured” for the first time — data that looked initially exactly like the enthusiasm around the tegoprubart trial today — the field experienced its first genuine surge of optimism. This is when the CITR data started to show an increase in transplants, and this is when the real rub was happening with the FDA and insurance. CMS, Medicare and Medicaid, private insurance, and all the other problems listed earlier are there because the biologics classification was locked in.

The momentum behind trying to get islets reclassified started almost immediately, but the effort was bogged down in red tape, bureaucracy, and politics. In the end, there was one pathway that would do it: getting a Biologics License Application (BLA). And the only way to get that is for a commercial company to apply for it. That’s why CellTrans was founded, to start the regulatory pathway.

Next, because islets fell under the Orphan Drug Designation because fewer than 200,000 people in the US would be eligible for this specific indication, this status triggers a seven-year market exclusivity automatically upon approval. CellTrans didn’t lobby for this as a competitive weapon — it’s a standard incentive Congress created to encourage companies to develop therapies for small patient populations that wouldn’t otherwise be commercially viable. The company had no choice but to take it.

That’s why the ISLET Act exists: reclassifying islets as organs under NOTA, bypassing FDA’s somatic cell therapy framework entirely — is the only path that doesn’t require the FDA to admit it got it wrong in 1993 and create regulatory uncertainty across the entire cell therapy space.

And now comes the deeper irony might even make Shakespeare green with envy: the regulatory friction that blocked access for thirty years — the very machinery everyone complained about — may have been protecting patients from a procedure they would have chosen because the harms from it were not yet known.

Now that time has passed, and the dangers are known, we can see it in the CITR numbers: no one wants to do it. Not just here in the United States but in the countries where access was never blocked — Canada, Australia, France — where islet transplantation has been reimbursed standard-of-care for years. No regulatory barrier. No insurance obstacle. No CellTrans exclusivity standing in the way.

Again, the ISLET Act only affects cadaveric islets delivered into the liver, a procedure that no one wants to get, and it only affects cadaveric islets, not stem-cell derived islets,. Since this is where the future of cure therapy is headed, let’s look at that.

Sana Biotechnology’s SC451 are stem cell-derived islets that have been gene-edited to evade the immune system entirely, with an IND filing targeted for 2026 and a goal of normal blood glucose with no insulin and no immunosuppression.

NewcelX’s NCEL-101 is explicitly pairing with tegoprubart rather than tacrolimus, making it the only SC-islet program currently designed around a calcineurin inhibitor-free protocol.

These programs are unaffected by the ISLET Act., insofar as their own regulatory and scientific timelines in the biologics framework. But they are also commercially developed by larger companies that can afford it. The momentum, financial resources, and other initiatives that could move similar efforts by smaller companies that need funding are not getting the attention they need because of the movement pushing the ISLET Act.

Shakespeare’s Narratives

Romeo and Juliet survives centuries of scrutiny because it perfectly captures this exact friction. On one level, it is a timeless warning against extreme, unchecked passion. Even when pure, blind desire can distort reality, push reason aside, and lead straight to destruction.

And we also learned of the systemic failure of institutional guides. The star-crossed lovers were uninformed, but the authority figures around them—the parents nursing ancient, self-serving grudges, and the advisors engineering bizarre, high-stakes schemes to cover their own tracks—consistently prioritized status, pride, and personal agendas over actual safety.

When we look at the legislative theater surrounding islet reclassification, the parallel is striking. Political grandstanding and bureaucratic maneuvering offer a flashy, emotionally satisfying narrative for organizations to rally around, but they ignore the cold reality of clinical trial data and bedside risk-benefit calculations.

The T1D community deserves better from its institutional guides than a legally symbolic rope ladder that still drops patients into a 90% serious adverse event rate. They should not want to present us with the idea that they are streamlining access to a curative therapy that informed patients don’t actually want.

Type 1 Diabetes is a tragedy that doesn’t have just one clean cause, so the cure won’t either. If we are to move forward, the community must be driven by informed medicine rather than short-term fantasies, and our institutions must exhibit the humility and patience required to fund the right target. We have the pieces of a real cure on the horizon; we just have to be wise enough not to squander the momentum it takes to reach it.

Or, as Shakespeare might have put it:

The state of our affliction hath no single cause,
And thus, no single law shall break our chains.
Be driven by the light of sober truth,
And leave the tragic balcony to rest.

(Exit, extinguishing a lantern.)

Share

Subscribe now

One of the most essential books I used in my entrepreneurship curriculum is “The Mythical Man-Month” (1975), by Frederick P. Brooks Jr. (1931–2022). He was a computer scientist and software engineer at IBM, best known for managing the development of OS/360. Brooks shattered a fundamental assumption that the best way to complete a project was to throw more people at it. It is here where he coined the phrase, “You can’t make a baby in one month by getting nine women pregnant.”

How apropos it is that he cited biological processes, because when it comes to medical interventions, the same principle applies. And nowhere is this more evident than what’s brewing in the “cure industry” for type 1 diabetes.

Most of my readers are well adept at knowing how babies are made; you can do it yourself using tools you already have lying around the house: The seeds, the environment to grow them, and a way to keep the immune system from killing the fetus. Doing any one of these incorrectly will not result in a baby.

Curing T1D involves the same principles: you need islets, a place to put them in the body, and a way to prevent the immune system from killing them. Doing any one of these incorrectly will not result in a cure — at least, not one worth pursuing.

What they both have in common is time to develop, though at different time scales. As for babies, the clinical trials took millions of years with billions of willing volunteers looking to advance science—sometimes frequently—purely for altruistic reasons. If you’re reading this, you can affirm that the procedure is pretty well established.

But T1D? It’s not that simple.

And yet, there’s been a massive push for Congress to pass the ISLET Act to “make a cure for T1D available sooner”. These initiatives are misdirected because the assumptions are based on false information—the same kinds of mistakes that Brooks wrote about. But in the era of social media influencers, the enthusiasm is trickling upstream to the funding agencies that invest in and promote this pipeline, creating a circular feedback loop of misdirected groupthink.

In the same way nine women can’t make a baby in a month, no amount of urgency to accelerate the path towards a cure will compress the clinical trials necessary to see that outcome.

The aims here are not just to dispel the misunderstandings of the ISLET Act, but to paint a realistic pathway towards a cure. By the end of this article, you should be informed enough to realize that, not only are there clearly better paths, but those won’t take any longer than the expectations for the ISLET Act. Readers may already know that I’ve already published other articles on this topic, but none have painted the roadmap precisely, explaining each stem along the way.

The Spark that Lit the Fire

This all started from the hype following a trial involving Eledon Pharmaceutical’s new drug, tegoprubart, which meets the “immune system” arm of the three-part path towards a cure. As of the June 8, 2026 ADA presentation, all 12 patients in the study achieved insulin independence, no longer requiring exogenous insulin therapy, and achieving an A1c average of 5.4%. Participants showed no evidence of nephrotoxicity, hypertension, or neurotoxicity — the toxicities commonly associated with tacrolimus-based regimens, which were historically used.

(I wrote extensively about this trial and the drug in my article, Eledon and the T1D Cure: Whose Breakthrough Is It, Anyway?, where I described the drug’s best potential use is for whole organ transplants, particularly kidney and liver, where they have the opportunity to displace tacrolimus, which currently commands a >$7B annual revenue stream.)

This doesn’t take away from tego representing a seismic change in immunosuppression technology because of this lower toxicity. But as applied to T1D, the excitement has outrun the evidence, and the gap is filling with claims, projections, forecasts and other statements that simply don’t hold up.

Mind you, this is not about tegoprubart. The drug is great. It’s the protocol that’s wrong—the other two things that are required for a cure to work. And even beyond the protocol itself, we also have no idea about the individual patients’ pre-existing magnitude of autoimmunity. Turns out, that’s a bigger variable than all this. But let’s step through this one by one.

In short, the protocol used in the Eledon trial is a very old, outdated methodology dating back to 2000, which itself finds its roots in the 1970s. It works, yes, but is not a long-term solution. It’s like using an old flip phone instead of a smartphone. It’s like using a dialup modem instead of broadband internet. It’s like listening to vinyl when digital streaming is available. Ok, wait. That one’s pretty good. Scratch that. But you get the point.

The protocol used in the Eledon trial uses cadaveric (donor) islets, and is placing those islets in the liver. As we’ll explain in detail, both of those will technically work, but are not viable strategies moving forward. They are what’s called “investigational methods”, which are akin to using prototypes to demonstrate that foundational principles work, but not necessarily intended to be the final product.

Supporters of the ISLET Act say that, even though this may be old methods, it will at least make these crude cures available to people sooner while those other technologies develop. And that is is simply wrong.

It will take exactly the same amount of time to go through all these clinical trials — including tego within the existing, archaic methods — as it will take to press forward with those other, newer approaches. And we’ll develop that more in this article.

If it’s really just the same timeline, then what’s the harm?

The problem with the ISLET Act is partly the misinformation around it, which itself redirects attention, money and resources away from those newer and arguably far more essential approaches. Ironically, if the same people pushing for the ISLET Act were instead focused on these other approaches, they’d get exactly what they’re hoping for: a viable, sustainable cure sooner.

There’s a lot to this, so let’s pick it apart and understand the fuller picture.

The Backstory: The First Islet Transplant

In 1977, at the University of Minnesota, the first TPIAT procedure was performed. Total Pancreatectomy with Islet AutoTransplantation. A patient’s pancreas was removed (due to pancreatitis), their islets were isolated from the resected pancreas, infused into the portal vein to be delivered into the liver. No immunosuppression was needed because the islets were autologous—the patient’s own cells. The liver was chosen because of portal drainage, where the liver gets “first pass” at insulin and glucagon because that’s the way the body actually works.

Note that placing the islets into the liver was never intended to be a long-term solution. In fact, it’s not a very good location at all, but it was a viable choice at the time. As it happens, we’ve been locked into that ever since.

Fast-forward to the year 2000 and the Edmonton Protocol, the first replicable procedure for transplanting islets into type 1 diabetics. The procedure was essentially identical to TPIAT, except for the source of the islets. Here, cadaveric (donor) islets were used, since T1Ds don’t have functional islets to use. And because cadaver islets are foreign tissues, an immunosuppressant was required. They used a steroid-free regimen of sirolimus + low-dose tacrolimus + daclizumab (anti-IL-2R).

In essence, tacrolimus was the toxic component.

The outcomes were unprecedented: non-diabetic glycemic control, with many patients being insulin free for years. Many hailed it as a cure and immediately wanted to see it rolled out to the masses. (Sound familiar?)

But, as time went on and patients returned to insulin dependence, enthusiasm faded. More troubling, tacrolimus was highly toxic, often making the cure worse than the disease for anyone other than those whose disease state was mortally brittle. And indeed, the procedure is still done today, but highly restricted to those patients whose disease state is nearly fatal.

We’ve been stuck with this protocol ever since; a procedure where all three components needed for a cure are wrong. But, because it was something rather than nothing, it has persisted. Besides, as an investigational method, new things could still be learned.

Fast-forward to 2023, when the company CellTrans, was granted FDA approval for Lantidra, which is the same protocol as Edmonton back in 2000, except they’re using tacrolimus instead of the cocktail they used before. Other than that, the islets and the liver are required elements in that protocol, and the company was granted seven-year exclusivity to market the product.

But now that Tegoprubart has shown to be highly effective as the immunosuppressant to replace tacrolimus, the community is rallying around the ISLET Act to lift this monopoly to make the procedure more generally available by multiple centers, and for the FDA to approve an accelerated pathway for tegoprubart to replace tacrolimus.

The rationale is that thousands of people could get access to a cure pronto.

And that’s not true at multiple levels. First, tego needs to go through clinical trials to show efficacy and safety, and those trials won’t start for another year or more. There’s nothing for the FDA to fast track.

Besides, even if tego were eventually used, the other two elements of the procedure — the cadaveric islets and the placement in the liver — make the entire protocol a physiological dead end, which will be explained here. Just because early participants are raving about how great it is to be free from insulin, no one sees the risks, and that’s what trials are for.

Nevertheless, the heckler’s veto is having its say, and public pressure is on Congress to pass the ISLET Act. So, let’s take a look at what it actually does and doesn’t do.

The ISLET Act

The ISLET Act — formally the “Increase Support for Life-saving Endocrine Transplantation Act” (S. 3105 in the Senate, H.R. 8018 in the House) — has a deceptively simple mechanism: it amends the Public Health Service Act to add “human cadaveric islets” to the definition of organs, and simultaneously clarifies that cadaveric islets are not a drug or biological product under the Federal Food, Drug, and Cosmetic Act or the Public Health Service Act.

In short, it moves regulatory oversight from the FDA’s drug pathway to HRSA’s organ transplant system — effectively putting cadaveric islets in the same category as kidneys and livers.

The Senate bill was introduced November 5, 2025 by Mike Lee, with Ted Budd as co-sponsor, and is a partisan Republican bill currently referred to the Senate HELP Committee. The House companion bill (H.R. 8018) was referred to the House Committee on Energy and Commerce in March 2026.

The Core Arguments For the ISLET Act

1. Access and equity. Currently, classifying islets as drugs creates logistical and financial barriers that make widespread access extremely difficult. The ISLET Act would align islet transplantation with how kidneys, livers, and pancreases are regulated.

2. International alignment. The ISLET Act would allow open access to islet transplantation for the most disadvantaged patients with T1D in the US, adopting the same approach already proven safe and effective in Canada, the EU, Australia, and Japan.

3. Commercialization ethics. Many argue that organ donation is an altruistic act, and commercialization of donated tissues is ethically problematic. A for-profit company controlling distribution of material derived from human donors strikes many transplant physicians as a category error.

4. Competition and innovation. The organ designation would promote healthy competition between transplant centers, cost-effectiveness, innovation, and progress toward reducing immunosuppression — and ultimately, a cure for diabetes.

The Core Arguments Against (or Complicating Factors)

1. The supply problem doesn’t go away. Advocates completely overestimate how accessible islets actually are.

According to the OPTN/SRTR 2024 Annual Data Report, there were 16,989 deceased donors in the United States in 2024 — but the vast majority of their pancreata never enter the islet pipeline at all. Of those donors, only 878 pancreata were recovered for whole-organ transplant, and 3,243 were recovered for research purposes, of which 2,010 were accepted for islet cell research.

The pancreas is the most self-digesting organ in the body, so even among those accepted for islet research, a large share of isolations fail to yield a viable therapeutic product. The organ that looked fine on procurement frequently isn’t.

Whole-organ pancreas transplant patients get first priority. Whatever remains for islet isolation is further constrained by the multi-donor requirement: a single patient achieving insulin independence generally requires islets from two to three deceased donors — sometimes four — because even a successful isolation recovers only a fraction of a pancreas’s total islet mass.

As for the islet research pipeline in the US, the ceiling is roughly 2,000 pancreata per year enter, per the OPTN/SRTR Deceased Organ Donation chapter. Active areas of development — improved OPO procurement rates, better isolation protocols for marginal donors, and expanded use of DCD pancreata with normothermic regional perfusion — could expand that number meaningfully over time. But even under optimistic assumptions, divide the realistic near-term yield by two to three donors per recipient and the ceiling isn’t thousands of patients a year. It’s 600–700. Reclassifying islets as organs doesn’t change that arithmetic.

The only thing that changes it is eliminating the multi-donor requirement — which is precisely what tegoprubart’s improved islet engraftment efficiency promises, and what stem cell-derived islets eliminate entirely by removing the donor equation from the calculation altogether.

2. Insurance coverage is entirely unaffected. The Act is also said to unlock insurance coverage — but coverage already exists. Lantidra is FDA-approved, and approved islet transplant is already covered by many payers for the narrow group it’s indicated for: people with severe, hard-to-control hypoglycemia. What limits access isn’t a coverage door waiting on Congress; it’s who qualifies and how few procedures are physically possible. The Act could change how the costs are accounted for — organ procurement is reimbursed differently than a biologic drug — but rearranging the cost structure doesn’t widen the eligibility criteria or add a single pancreas. You can’t improve the economics of a procedure that you can only perform a few hundred times a year.

3. The post-BLA reality is messier than either side admits. First, the BLA (Biologics License Application) is the FDA approval pathway for products derived from living material, governed by Section 351 of the Public Health Service Act. Its purpose is to prove three things about a biological product: safety, purity, and potency. That’s very hard to do for cadaver islets, given that their sources are, by definition, dead people, where we don’t know the variability of donor history, cold ischemia times, and isolation yields.

In other words, the ISLET Act is effectively much ado about nothing for actual patients, because, well, nothing would actually change in whether islets are available to them, how and where they get surgeries, or their insurance coverage.

But, the lack of actual benefit to patients is missing the real, tangible harm they will suffer. The ISLET Act is misdirected momentum: the advocacy energy and funding gravity it concentrates around—the cadaveric+liver+tego combination—is what crowds out far better pathways to an arguably more scalable cure. The danger isn’t the bill; it’s what the field stops looking at while chasing it.

This raises the obvious question: what are better paths towards a cure, and what role does tegoprubart play?

We begin with one of the other arms of the cure: islet placement within the body.

Location, location, location…

Remember again that there are three elements to islet transplantation: the cell source, the placement, and the immunosuppressant. Tego is only the latter. For this section, we’re going to focus on location for implanting the islets, which involve three criteria:

First-tier, non-negotiable: islet survival — oxygen, rapid revascularization, geometry to spread the mass, mechanical stability. Most of the body fails here. This is what kills the naked subcutaneous sites, encapsulation techniques, and especially the liver. Islets just don’t survive in the liver well, but other locations are viable.

Second-tier: monitorability and retrievability. We need to be able to see, monitor and (if necessary) extract the islets. (Or, potentially put more in.) This is another point where the liver fails and other sites win.

Third-tier, the long-game optimization: drainage route. The liver wins when it comes to ideal glucagon and insulin efficiency. It not only yields excellent glycemic control, but the beta cells’ direct deposit into the liver means one avoids hyperinsulinemia.

This is a bit technical, but when insulin travels to the peripheral tissues first, it gets soaked up before the liver gets its pass. So, to get more insulin into the liver, a great deal more insulin has to be generated, which is responsible for endothelial dysfunction, the precursor to heart attack and stroke. We see this in both type 1 and type 2 diabetes. (See my article, Extending T1D Longevity: Balancing Lipids, Insulin, A1c.)

Draining into the portal vein is highly efficient, but isn’t required for good glycemic control — systemic-draining sites achieve excellent A1c and time-in-range, as we’ll soon see. It’s a trade-off to be sure, and it’s a shame because the liver is unsuitable because it’ll eventually kill the islets.

But there are other health consequences to delivery into the liver as well. The first is Instant Blood-Mediated Inflammatory Reaction, or IBMIR, and it starts the moment the islets hit the bloodstream when infusion starts. Coagulation and complement cascades fire, platelets swarm and stick, white cells move in, and the islets end up wrapped in clots and infiltrated by the very inflammation that’s supposed to protect us from invaders. Up to 60% of the transplanted mass can be destroyed within hours to days. The little guys are dying in the doorway before they ever find a place to live.

The islets that do survive engraft somewhere they can’t be watched or recovered. Scattered through the hepatic sinusoids, they’re unimageable, un-biopsiable, and impossible to remove if something goes wrong.

The liver isn’t hostile because of the immune system. It’s hostile because it’s the wrong physical place to put islets, which means placement and immunity are two different problems, and the field has been spending its effort on the second while the first quietly eats half the graft.

Even among TPIAT patients, their own cells don’t survive this procedure, many of whom become insulin-dependent type 1 diabetics because of islet cell loss. In such cases, they become effectively identical to non-autoimmune T1Ds, a whole category of T1D that we haven’t even touched on yet. Hint: there are a lot of them.

If the field actually succeeds in their stated goals of passing the ISLET Act and getting tego approved for this existing procedure, that will virtually ensure that we’ll be locked into a protocol that won’t help the majority of T1Ds, while potentially harming those who do.

Moreover, it makes it harder to get funding for either research or trials for other locations to place islets or other islet sources (stem-cell derived). This could add yet another item to the lengthening list of avoidable medical errors that Marty Makary — himself an islet transplant surgeon and former head of the FDA — wrote about in his book, Blind Spots: When Medicine Gets It Wrong, and What It Means for Our Health (Bloomsbury, 2024).

(I wrote about this in my article, Marty Makary’s “Groupthink” and the Medical Errors in the T1D Ecosystem.)

Changing the infusion site from the liver to another location is the ultimate goal, and that by itself isn’t a labeling “tweak” — it would be a new indication requiring new clinical data, essentially a new BLA. Hence, a whole new and different set of clinical trials—expensive, time-consuming and competing with the cadaver-liver-tego route that everyone’s aiming for.

This section was necessary to affirm that we need to get out of the liver. And we’ll talk about the omentum and the ARS later, but first, we need to see how immunosuppression plays into all this, because it is the binding glue that leads into the stem-cell derived islets, and that’s when we can talk about the other placement sites.

What we don’t yet know about tegoprubart

The excitement around tego rests on one word: non-toxic. And on its own terms that’s earned — no nephrotoxicity, no neurotoxicity, none of the islet-poisoning that makes tacrolimus self-defeating in a procedure meant to keep islets alive. But “less toxic” is not the same claim as “less risky,” and the gap between those two is why the years of clinical trials are important.

Tego is a costimulation blocker. It interrupts one specific handshake — the CD40-CD40L signal — that a T cell needs to commit to attacking. That’s a precise, elegant mechanism, and it has a precise, known weakness: it works best against naïve T cells seeing an antigen for the first time. Memory T cells, the ones that have already learned a target, lean on that handshake far less.

This is why tego cannot stand alone — the trial pairs it with mycophenolate, an old antiproliferative that doesn’t care how a cell got activated and simply caps how fast any lymphocyte can multiply. (That is, it slows down the proliferation of those T cells, easing off on the accelerator. In practice the regimen starts even earlier, with an induction phase that depletes the existing T-cell pool outright before tego and mycophenolate take over maintenance.)

Tego prevents the activation; mycophenolate cleans up what slips past. So the “tego cures it” story is, at minimum, a tego-plus story, and the no-rejection results belong to the combination, not the headline drug.

That same memory-sparing trait raises the question nobody is funding the time to answer: what does this do to ordinary immune defense? Tacrolimus suppresses everything bluntly — bad, but predictable. Tego suppresses selectively, which should spare your existing immunity to pathogens you’ve already met while leaving you exposed when you meet something new. That’s what is not yet known, and you can’t know that by watching 12 patients for two years. You need to watch thousands of patients for ten years. Big difference.

Of course, what we’re really talking about is cancer. Every chronic immunosuppressant carries a malignancy cost, because tumor surveillance is one of the jobs you’ve switched off. Maybe tego’s profile is better than tacrolimus here. Maybe it’s just differently shaped. Twelve patients over two years cannot tell you, because the signal takes five to ten years to appear.

And finally, dosing. Tego isn’t dialed to disease severity; it’s a fixed weight-based dose calibrated to saturate the CD40L pathway. Past saturation, more drug buys no extra graft protection — it only ramps up the cancer and infection risk. So the knobs that actually matter aren’t milligrams; they’re how long you block the pathway and how much mycophenolate rides alongside. And right now there’s no framework for setting either, patient by patient. The default for an unproven population is maximum coverage for maximum duration — which is exactly how you load infection and cancer risk onto the patients who needed the drug least.

All of this talk of dosing will get even more nuanced once we talk about stem-cell derived islets, especially those that are gene-edited to evade the immune system. Their immune cloak may not be perfect, thereby potentially requiring some immune suppression, but it certainly won’t be the maximal suppression that tego is currently using for cadaveric islets.

None of these are reasons tego won’t work. They’re reasons you cannot know it works — safely, durably, at the right dose, in the right people — without Phase 2 and Phase 3 trials that take up to ten years. That is not bureaucratic foot-dragging. That is the irreducible runtime of the experiment.

Do we really want to invest 10 years of clinical trials on a protocol that we are so convinced should be fast-tracked if we can’t even evaluate whether this protocol is ultimately the best one? We already know the liver is not the ultimate destination, and we also know we need more data on tego’s long term effects (particularly the dosing calculations). And that leads to the stem cell derived islets, so let’s take a look.

Stem-cell Derived Islets

With stem-cell derived islets now under development, the supply problem may be lifted sooner than later, thereby making a procedure scalable. But it won’t be the Lantidra product, or frankly, any other existing protocol. Entirely new protocols will have to be developed, and there’s no reason tego can’t ride alongside them all.

Vertex, Sana (SC451/UP421), and others are all pursuing allogeneic (foreign cell line) stem-cell derived islets under FDA’s biologics framework — the very framework the ISLET Act explicitly leaves intact.

Each of these could be instrumental, not just in their ability to scale up the supply of islets, but hypoimmunity, where the islets themselves will be less provocative. This would certainly affect the dosing protocol for tego.

Another critical development is the ability to make autologous islets, where the cells are made from the individual’s own tissues rather than from a foreign cell line.

Wait, you’re thinking, if T1D is autoimmune, then islets made from my own cells will just trigger that same autoimmunity, right?

Again, it’s not that simple. As we’ll show later, many T1Ds don’t even have autoimmunity, and research shows that people’s immune response wanes over time, while some patients (LADA) have very low autoimmunity in the first place. For these cases, autologous islets would be ideal, possibly paired with low doses of tego.

But even all that notwithstanding, there’s evidence that autologous islets work even in fully developed autoimmune T1Ds.

Wang and colleagues, from Hongkui Deng’s group, published their findings in Cell 187, 6152–6164 (2024), Transplantation of chemically induced pluripotent stem-cell-derived islets under abdominal anterior rectus sheath in a type 1 diabetes patient. Here, autologous chemically-induced pluripotent stem-cell-derived islets were transplanted back into the same patient, achieved insulin independence at ~75 days post-transplant, A1c dropped from 7.6% to ~5% range, and a TIR >98% at one-year follow-up.

Let that sink in: the same patient’s own stem cells were used to make new islets, and achieved ideal glycemic control. And that woman was a fully maximal autoimmune patient. But she was chosen because she was already on immunosuppressants for two whole organ transplants earlier. The trial was intended to see whether the autologous islets worked.

Not only did they work well, they were not placed in the liver, but in the anterior rectus sheath (ARS), a layer in the abdominal wall beneath the fibrous sheath that wraps the abdominal muscles, reached through the belly wall. This is far easier to access than threading a catheter deep into the liver’s plumbing.

That alone changes the character of the procedure. Where intraportal infusion is an interventional-radiology affair carrying all the clotting-and-bleeding hazards above, getting islets into the ARS is comparatively minor surgical access — closer to routine abdominal-wall work than to cannulating a major vein.

Two advantages stand out, and the liver can offer neither. First, you can see it: the ARS graft can be imaged on demand by ultrasound and MRI, so the cells can be monitored instead of vanishing into the sinusoids. Second — the big one for an experimental therapy — the site supports graft retrieval. If the islets misbehave, you can get them back out. For a cell therapy where no one yet knows the decade-out behavior, “watchable and removable” isn’t a convenience; it’s a safety property the portal route forfeits completely.

Interestingly, you could also put more islets in if needed. Imagine that.

The human ARS evidence is still thin — a single patient, using her own reprogrammed cells. So the claim isn’t that the ARS is established. It’s that on every axis that matters — access, monitoring, retrievability, ideal glycemic control and early survival — it points somewhere the liver can’t follow. This is a place where new funding dollars and enthusiasm should be directed.

Another site worth naming is the omentum, the apron of vascularized tissue draped over the abdominal organs. On paper it may be the most physiologically attractive option of all, because unlike the abdominal wall it drains into the portal vein — recapitulating the first-pass delivery to the liver that native islets have, the one genuine advantage the liver site offers, but without dumping islets into flowing blood. It’s accessible, richly vascularized, monitorable, retrievable, and large enough to hold a stem-cell-derived graft; portal drainage may even confer a modest immune benefit.

The catch is that the clinical attempts to use the omentum haven’t yet matched the physiology — early human trials produced insulin independence that faded within a year, but they didn’t establish why. (Hint: not studied or funded enough.)

The omentum belongs on the list as a strong candidate on first principles, not as a proven destination.

All of these should be studied with tegoprubart as a potential immunosuppressant, and those trials would also provide opportunity for a more fine-tune dosing protocol. We’re no longer talking about cadaveric islets, where tego is the primary protector from T cell activity. Using these new locations, along with stem-cell derived islets, the whole dosing mechanism would need refactoring.

And that’s a good thing, because it leads to the next aspect of the immune system that I alluded to earlier: autoimmunity. It’s not what we thought.

The Role of Autoimmunity is Shifting

We’re all taught that T1D is an autoimmune disease. And for most, it is. But it’s not that simple. Autoimmunity is not a binary switch—it throttles. In fact, as alluded to earlier, there’s a sizable number of T1Ds that show no autoimmunity at all. Even though some have had their beta cells destroyed, it’s not always due to autoimmunity. This can include viral infections, checkpoint inhibitors (cancer therapy), and many other causes.

In my article, To Cure T1D, Start with the Easiest Cases, Not the Hardest, I cite statistics gathered by the ADA and international diabetes organizations that estimates the number of non-autoimmune T1Ds is roughly 22% of all T1Ds, which itself may be an undercount due to the imprecision of the assays used to assess autoimmunity. Furthermore, recent research shows that autoimmunity wanes, with many becoming “functionally” non-autoimmune, insofar as how much immunosuppression they would actually need.

Even among those with more active autoimmunity, the level of intensity is not constant. LADA patients, for example, have much lower T cell intensity than children who are newly diagnosed.

What we really need, independently of all this, are assays that assess actual T cell activity. One such test is called — BASTA the Beta cell Antigen Specific T cell Assay. It comes out of Stuart Mannering’s lab at St. Vincent’s Institute of Medical Research in Melbourne. He published his findings in Science Translational Medicine in 2025, and what makes it matter isn’t sophistication; it’s the simplicity.

Draw a small tube of blood from the patient—about the same amount you’d take for any traditional lab work—and add the proteins that beta cells make. These are the same proteins the immune system attacks in type 1 diabetes. If the person’s autoimmunity is active, the T cells in that test tube of blood will recognize those proteins and react. If they don’t react, nothing happens and we know the person is non-autoimmune.

Here’s the best part: if there is a reaction, we want to know more than whether a reaction occurred, we want to know how strong it is. In other words, it return a magnitude — how much autoreactivity, and against how many targets.

I have a fully-dedicated article to this thesis called, Autoimmunity Stratification: The Greatest Blind Spot in T1D Cure Research, which argues that T1D autoimmunity is an essential tool — a “meter” if you will — that’s as essential to understanding how to dose drugs like tegoprubart as a glucometer is for detecting glucose to know how much insulin to take.

If someone said, “you’re a diabetic, therefore, we’re going to give you 10u of insulin right now,” that would be insanely dangerous. But of course, there might be random outliers where that is luckily the right amount. The way clinical trials are run right now, we’re giving people drugs that suppress their immune system, and we don’t know how how active the immune response actually is. This is why trial results are all over the board.

For example, back in 2008, a Belgian group (Huurman et al., with Roep and Pipeleers) ran a trial involving 21 established T1D patients where each received cadaveric islets through the portal vein under full immunosuppression. Standard protocol; nothing about the technique was new.

Because they banked the blood samples, they found that, of the patients with no pre-existing T-cell autoreactivity, seven of eight became insulin-independent; of the four reactive to both GAD and IA-2, none did.

And that was in 2008. The field has been ignoring this science for too long.

In short, autoimmunity throttles, and that presents enormous opportunities and risks for immunosuppressant therapies. The risk of not knowing autoimme activity is that we overdose the patient. The opportunity of testing ahead of time is that we not only dose well, but the graft survives—along with the patient—for much longer.

This is the trick that pregnancy already pulls off — not a blunt immune shutdown but a calibrated stand-down, dialed precisely enough to tolerate foreign tissue without leaving the body defenseless. That’s the target: dose to the patient’s actual immune intensity, not to a worst-case maximum.

Put it all together, and you can see the trajectory: the moment you move toward the pathways that actually scale — autologous and stem-cell-derived islets — the allo problem shrinks or disappears, and autoimmunity becomes the whole game.

If future trials do not know where each patient sits on that spectrum, the results will scatter and you can’t assign causality between the dosing and the outcome.

My article on stratification makes all this abundantly clear, and calls for all trials involving immunosuppressants to identify patients’ autoimmune thresholds before, during and after clinical trials.

Without doing any of the above, the real cost of the rush to approve tegoprubart prematurely is a waste of time and money, not to mention the timeline to a cure.

Groupthink? Or good strategy?

Reviewing it all: autologous islets, hypoimmune islets, alternative placement sites, and the stratified dosing of (yes, likely) tegoprubart. That’s how you get to a cure that can meet the needs of most people.

The ISLET Act doesn’t address any of this. And the social media misinformation engine is making things worse. It’s not just noise, it’s pressure. When patients and influencers on podcasts insist the cure is being withheld, that insistence flows uphill — to advocates, to funders, to the FDA — and every institution feels that pressure.

This is an example of what I’ve called uncoordinated groupthink, where each player drifts toward the prevailing current because that’s where everyone else seems to be going. This drags along with it the funding, the attention, and the goodwill. In that ecosystem, going along is individually rewarded. Eledon benefits when its drug is the headline. Advocacy groups benefit when donors feel a cure is near. The investigators benefit from enrollment and momentum. None of it requires a meeting.

But there’s no one at the table questioning whether this approach makes sense. As I always say, “When everyone at the table agrees, someone’s got it wrong.”

Tego is fine. What we really need, in order, are assays that assess autoimmune activity, not just for these trials, but globally. We need to better characterize T1D as a disease. Next, we need to fund clinical trials that include the ARS and the omentum. Procedurally, these will be easier, less dangerous, and less expensive. Last, we need to include stem-cell derived islets, both autologous and gene-edited cell lines in funding grants. I’ve already written about this.

Look, we all understand. We all want a cure. Pronto. When you’re pregnant, you really want that baby to come out. But you also know that doing it too soon carries risks. It’s worth the wait to do it right. There’s no easy way out of this, but when it comes, you treasure it.

Share

Subscribe now

In the classic Saturday Night Live sketch, “Theodoric of York, Medieval Barber“ (April 22, 1978), Steve Martin plays Theodoric, a confidently ignorant medical practitioner in 1303 A.D. who dismisses anything resembling actual science in favor of bloodletting. After mercilessly torturing patients using the technique, the sketch reaches its comedic peak when, after a patient dies and her mother accuses Theodoric of being a charlatan, Theodoric delivers a soaring monologue questioning his ancient superstitions. He passionately envisions a future where medicine relies on observation, experimentation, and a scientific method that could lead humanity into a standard-bearing age of rebirth—a “Renaissance.”

Then, after a brief beat, he looks away and simply says: “Naaaaaah.”

The humor is two-fold. First, there’s the comedic look back on bloodletting, how it was actually taken seriously to cure diseases for nearly 1500 years. In The History of Diabetes Treatment: How We Got Here, Aretaeus of Cappadocia (1st–2nd century A.D.), who gave diabetes its name (derived from the Greek word for “siphon”), recommended bloodletting to treat it.

Clearly, it didn’t work.

But the second comedic look is funny in a more sardonic way. The skit mocks how science was dismissed, despite its obvious advantages, and we think that it was so archaic, because we did go through a renaissance, and we do follow the scientific method.

Uhhh… right?

Unfortunately, despite the fact that science does, in fact, demonstrate many a tipping point where the evidence for a given medical claim is abundantly clear that it either does or doesn’t work, we cannot necessarily be assured the medical community resolves to adhere to the science. There will always be a Theodoric of York in our midst.

Such is the topic of Marty Makary’s book, Blind Spots: When Medicine Gets It Wrong, and What It Means for Our Health (Bloomsbury, 2024). Its thesis is simple: even here in the modern era, the field repeatedly commits to wrong answers, defends them long after the evidence has turned, and takes a generation to admit the mistake. Makary builds the case through a series of well-chosen disasters.

And though Makary doesn’t talk about type 1 diabetes in his book, his prominence in the T1D ecosystem is well-established. At Johns Hopkins he was a hepatobiliary and pancreatic surgical oncologist and, more to the point for us, the chief of Islet Transplant Surgery—the man who literally transplanted islets into people with type 1 diabetes to free them from insulin. He won the National Pancreas Foundation’s Nobility in Science Award, and has spent a career in and around the organ that defines our disease. He was the Commissioner of the FDA in the midst of a tumultuous political environment.

It’s ironic that Makary’s background is in T1D, since the arguments he makes in his book should be held up like a mirror to ourselves. Yes, there’s strongly established science that contradicts the very practices we engage in today in the T1D ecosystem, from daily management to interventions, technologies, and of course, the drive towards a cure. The medical literature is unambiguous about many of these things, and yet, we continue to engage in practices and technologies that have shown to be either archaic, or genuinely harmful.

The real questions are why and what can we do about it?

That’s what we’re here to explore, and odd as it may sound, I’m going to do so by taking a closer look at Makary’s main thesis itself. The “blind spots” that he’s referring to are actually multi-layered, and they apply differently to different actors in the T1D ecosystem. As we explain these, we’ll cite T1D-specific “medical errors” that are victim to these.

Background

I’m going to begin by changing hats. In my substack, I am a medical researcher and analyst in the T1D space. If you’re reading this, you already know that. But this hasn’t been my whole career.

In 2011, I founded the Center for Entrepreneurship at UC Santa Cruz after a long career in starting technology companies since 1990. My role here was to help the university’s researchers bring deep science to market — to help professors and graduate students translate their core science into products and intellectual property that could be used in the real world. (My proposal was approved by the UC Regents in 2012.)

The most viable patents in the portfolio were from the life sciences—namely, genomics, bioengineering, among others—which is where I cut my teeth on truly vetting scientific claims, validating medical literature, pressure-testing authors on their claims, and ultimately, understanding how a given IP can be commercialized into real-world settings.

But perhaps the most beneficial learning experience for me was sitting across the table from executives, chief scientists, and MDs of major pharma companies who were highly interested in core research and patents filed by UCSC faculty and researchers.

Because they were eager to get first looks at our IP. To help researchers understand their market needs, they shared with me highly detailed and confidential information on how their companies worked in the emerging areas they wanted to pursue. I had signed more NDAs than I can remember.

I used the fundamentals I learned from this process to build the academic curriculum for the courses I taught. This was not just how to start tech companies, but how organizations work. How technology goes through life cycles. And sadly, how “medical errors” are uncomfortably common.

Medicine is evolving and imprecise. To gain a competitive advantage in the marketplace, the marketing emphasizes what it needs to. But that’s where we get to tipping points, where evidence just points the other way.

Makary’s “blind spots” is a manifestation where any one (or more) of the basic principles of business, science or organizational psychology can fail. We begin with Makary’s basic thesis.

Makary’s Blindspot: Groupthink

Makary opens his book with peanut allergies—the cleanest case, and the one he documents most thoroughly. In 2000, the American Academy of Pediatrics told parents to keep peanuts away from at-risk children until age three. The recommendation rested on weak evidence and ran against known immunology that states that early exposure builds tolerance, and withholding it does the opposite. This recommendation caused a massive spike in peanut allergies in American children for the next decade and a half, and the guideline wasn’t fully reversed until 2017. Seventeen years of medical literature pointed to it, but it took that long to correct it.

Makary then moves through hormone replacement therapy, where a non-significant statistical association was used to convince the establishment that HRT caused breast cancer—costing a generation of women the well-documented benefits of treatment, with no reduction in breast cancer to show for it.

Through all these examples, he names the cause: groupthink. In Makary’s view, cognitive dissonance keeps doctors from absorbing information that contradicts what they were trained to do. Conformity keeps the consensus intact, and inertia prolongs the practice long after it should have collapsed. That’s the blind spot the title refers to—not that doctors can’t see, but that a whole field can be looking right at something and miss it because everyone around them is acting the same way.

There’s even a number that puts a clock on it: 17 years on average. A 2025 analysis in Frontiers in Health Services breaks it down further: about 6.3 years for evidence to reach reviews and textbooks, and another 9.3 years to actually change what happens at the bedside.

Makary’s framework holds that groupthink itself is essentially passive. People conform. People resist new information. Institutions are slow. Change is hard. In this telling, the actors are basically innocent—well-meaning physicians and administrators caught in the gravity of consensus, going along because everyone goes along. The 17-year lag becomes a kind of weather: regrettable, impersonal, nobody’s fault in particular.

But this thesis starts to feel incomplete as his narrative moves forward. When he covers OxyContin and the decades of assurance that opioids weren’t addictive, we start seeing something more than just passive. The Sackler family engineered a body of reassuring literature that its pain killer wasn’t addictive, despite enormous evidence to the contrary, and a sales apparatus to keep a profitable falsehood in circulation. They paid doctors to keep it going. They created an ecosystem where people were actively coerced in one form or another.

He covers the HIV-tainted blood supply, the demonization of dietary fat, the reflexive overuse of antibiotics. Different organs, different decades, same shape.

The pattern he describes is real, and the human cost is enormous, but Makary collapses all these into the same bucket of “passive conformity”, which is where we need to dig deeper. There’s more nuance to this than he portrays.

The term “groupthink” was coined by William H. Whyte Jr. in a March 1952 Fortune article, riffing on Orwell’s “doublethink.” Later, in 1972, Irving Janis, a Yale psychologist, used that term to describe cohesive groups suppressing dissent, which he describes in his book, Victims of Groupthink: A Psychological Study of Foreign-Policy Decisions and Fiascoes.

It’s that very notion of “cohesive groups” that starts to bring us closer to how groupthink works, and how it takes different shapes in different groups, a concept that exceeds Makary’s simplistic framework. If we want to correct bad medical practices, it’s not just about being less passive and paying attention. There are organizations that shape public opinion outside of the medical establishment, which comprise the “cohesive groups” that Makary doesn’t address. In the following sections, we’ll delineate those groups, of which there are four, beginning with Makary’s thesis: Passive Conformity.

Groupthink #1: Passive Conformity

Makary’s definition of groupthink being largely a consequence of passive conformity accounts for most of the medical errors we see in the T1D ecosystem today. Here are a few examples that I’ve written about over the course of my substack.

Heart disease, not hypoglycemia, is the #1 killer of T1Ds. As the article in that link explains, the fear of dying from a low is a myth that just doesn’t, well, die. It has gripped the T1D ecosystem for decades, far out of proportion to the actual risk. The “medical error” here is that the true cause of death is not getting attention or treatment. Hypos trigger a counter-regulatory response that ruptures vulnerable plaque in the arteries, causing a cardiac event. People think the patient dies from the hypo, unaware that it was cardiac. T1Ds are highly vulnerable to cardiovascular disease from a lifetime of elevated insulin and LDL levels, which are the primary drivers of vascular disease. (Elevated glucose levels can act as an accelerant.) What should be surfaced, but isn’t, is lowering LDL through medical interventions like statins and PCSK9 inhibitors, which can add years of life to T1Ds.

DKA is far rarer than commonly stated, and fearing it makes people sicker. That article explains that, while over 40% of T1D onsets are preceded by the horrifying experience of DKA — one that stays etched in the mind — the phenomenon is otherwise extremely rare outside certain vulnerable groups (those with eating disorders and pump mechanical failures). The medical error, outside of the unfounded fear, is that the act of trying to prevent DKA itself drives unhealthy practices in three critical ways.

• First, the current “remedy” to avoid DKA is to take insulin when ketone levels are elevated. But that action is almost universally unnecessary; DKA does not just spring up surprisingly, the way a hypo might. DKA requires about ten hours of zero insulin on board. Not just “too little” insulin. None. And the symptoms leading up to DKA are intolerable—nausea, vomiting, acute pain. No one gets DKA by surprise, and there’s plenty of time and motivation to prevent it.

• The second problem with the fear of DKA is that the next generation of CGMs will test for ketones, so now people will see their ketone levels constantly. When people see high ketone levels, which are often physiologically healthy, they will not just take more insulin, but they will actually stop doing the healthiest thing a T1D can do: exercise. Ketone levels rise with exercise because ketones is a byproduct of converting fat into glucose, which is what happens during exercise. So, people will start to do activity, see ketone levels rise, and then stop.

• The third thing is the last nail in the coffin: trying to lose weight. Losing weight, by definition, means the body uses fat as energy—hence, the weight loss. But, like exercise, losing weight produces ketones. So, T1Ds will naturally take insulin believing they’re staving off DKA, which then turns into fat again.

In short, misinformation about DKA risk is about to rise significantly once the marketing materials start getting released by the CGM manufacturers. They might as well just recommend bloodletting in the instructions, as it’d likely just as effective.

All this discussion about DKA now leads to the next big medical error: the recommendation for basal insulin ratios. Yes, people take far too much basal insulin. As the article explains, this is largely because of a faulty paper published in the 1980s that was based on T2Ds with very high insulin resistance. This created the false belief that the liver is a leaky faucet, constantly dripping glucose that needs its own steady drip of insulin separate from food insulin. None of that is true, and the myth leaves people with far more insulin than they need — and the hypoglycemia and weight gain that follow. My article cites literature showing that ~60% of T1Ds are either overweight or obese due to overbasalization. The ADA’s 2025 guidelines finally corrected this, nearly thirty years after the original 1988 guidance, but the correction hasn’t permeated the T1D ecosystem. Doctor’s still prescribe a 50/50 basal ratio, which is slowly killing T1Ds through metabolic disease and cardiovascular disease.

And this leads to another medical error that was popularized by the late Richard Bernstein: Ultra-low-carb diets. As this article explains, these diets are a leading accelerant of heart disease because it asks the body to utilize nothing but fats as a fuel source. When the body has no access to usual carbohydrates in the diet, the metabolism has to go into hyperdrive to convert all the fats in the diet, which is like trying to get a VW bug to pull a massive trailer up a hill. That conversion process at this level of intensity generates massive amounts of LDLs, accelerating cardiovascular disease.

The medical literature has long shown this, and fortunately, most clinicians don’t recommend it, but they also don’t warn against it. That Richard Bernstein was lucky enough to reach his 90s is one reason followers are so passionate. (As my article points out, there’s evidence to show that some people are just genetically tolerant of ultra-low-carb diets, much the same way some people are genetically tolerant of obesity. Genetics can often work for you as well as against you.)

The list above and more errors all lead to the single most problematic narrative within the T1D ecosystem: the hyperfocus on A1c levels as the target of T1D management. That singular focus is the fuel that justifies the acts listed above (except for hypos and heart disease). In my article, A1c: The Good, The Bad and the Ugly, I explain how medical literature points to longevity in T1D requires normal weight, low insulin levels, and crucially, low lipid levels. These are the common denominators among T1Ds that have lived 50+ years with the disease. Nowhere is A1c listed there.

The Golden Years Cohort, over 400 British T1Ds who’ve had it >50 years, had a mean A1c of 7.6% (± 1.4), and The Joslin Medalists, had an average 7.7%, with some going to 14.0%.

No, this does not mean you can ignore A1c levels, but it does mean that you need to keep weight, insulin and LDLs levels LOW, and most of that is best achieved with exercise, not low-carb diets, insulin pumps, or any other drugs (other than statins for lipid management).

Medical errors in T1D are driven by exactly what Makary describes as groupthink, and you can hear them repeated constantly in podcasts, social media sites, advertising, and yes, even from many doctors and other practitioners. They speak authoritatively, and it’s easy for the audience and non-technical podcast hosts to not know the difference. When sound bites are uttered without being challenged by knowledgeable and informed hosts, the ecosystem of poor information perpetuates the groupthink.

Nearly all of these are passive. Few are actively promoting these with malintent.

But sometimes, there is intentional coercion, which brings us to the next stage of groupthink. To illustrate, let’s return to the fun history of bloodletting.

Groupthink #2: Coercive Groupthink

In the 1700s, the main proponent of bloodletting was Dr. Benjamin Rush — one of the most prominent physicians of his day, a signer of the Declaration of Independence, and a major figure in politics and public health. Rush wasn’t merely swept along by prevailing wisdom. His almost messianic message led many doctors to adopt his bloodletting methods, including those who treated President Washington during his final hours.

There was a dissenting voice — William Cobbett, who called Rush “the Bleeding Physician of Philadelphia” in his Porcupine’s Gazette. Cobbett investigated the city’s mortality statistics and determined that deaths had actually increased after Rush began his aggressive bloodletting campaign — using data, whereas Rush had used only his own experience and anecdotes. This is a 1790s data-vs.-anecdote argument. The data guy lost — Rush sued Cobbett for libel and won.

Rush himself, consistent to the end, had himself bled twice during his own final illness.

This is not passive groupthink. Rush was a man with institutional authority, professional prestige, political standing, a professorship at the University of Pennsylvania medical school, and the social capital of a Founding Father. Disagreeing with him wasn’t an intellectual exercise — it cost you professionally and, as Cobbett discovered, legally.

Rush could beat Cobbett in court. He could not beat Mother Nature — the patients kept dying, and the mortality figures Cobbett compiled didn’t care who won the lawsuit.

Dr. Rush is an example of coercive groupthink, which is a dramatic departure from Makary’s more sedate “passive conformity”.

These kinds of cases have been said to exist in the T1D ecosystem as well, but to assert that here would be to accuse a person or group of committing fraud, which is out of context. Such accusations would have to be backed up by documentation, followed by a trial and an adjudication by a judge (and possibly a jury).

For purposes of this article, we’re going to assume that other kinds of medical errors are not fraudulent, per se, but more uncoordinated actions among groups, where they all know that a given error exists, but quietly just go along with it. This is not illegal, but nevertheless more insidious because there is no single bad actor or conspiracy to point to. It’s a set of conditions where there’s an unspoken agreement to just play along. I call this “uncoordinated coercion”.

Groupthink #3: Uncoordinated Coercion

Uncoordinated coercion involves people who shape public perceptions and guidelines do so, but without formally organizing it. Because they’re the insiders of a loose network, they don’t have to. Instead, they do this through a process that cognitive psychologist Steven Pinker calls common knowledge.

In his 2025 book, When Everyone Knows That Everyone Knows… Common Knowledge and the Mysteries of Money, Power, and Everyday Life (Scribner, 2025), Pinker draws the distinction between everyone privately knowing a thing and everyone knowing that everyone else knows it, creating a recursive loop of awareness. The business continues so long as this knowledge remains confined to insiders. The goal here is to keep the consumer from knowing what the insiders know.

I initially wrote about this in my article, Dexcom’s Stock Performance: A Business Masterclass in “Common Knowledge”, where I document how the medical evidence for the performance of CGMs had peaked nearly ten years ago, but CGM companies continue to game the system by releasing new products that are either no different than prior ones, or potentially worse—and yet, promoting them as “more accurate”. This is because they’re conducting trials that artificially yield improved MARD ratings—the metric that determines how “accurate” a CGM is compared to a blood glucose test.

This is an example of uncoordinated agreements among the manufacturers, but it is not an example of a medical error that harms patients (though Dexcom has been sued for fatalities and injuries from their G7, which I document here).

An example that illustrates uncoordinated groupthink that has harmed patients is found in automated insulin delivery (AID) systems. In my article, Medical Literature Analysis: The Performance Paradox of AID Systems, I cited over ten years of medical literature from mostly ADA journals — including one from the ADA itself — that pointed to an emerging trend about who benefits and who doesn’t from these systems. I used the term “The Barbell Distribution of Clinical Benefit”.

In distilling what the medical literature finds, I said:

The true beneficiaries of AID systems are the far outliers in T1D demographic profiles—those who are either entirely unable to manage themselves, or are hyper-engaged in self-management—leaving those in the middle as either not realizing any material benefit from automation, or may be potentially harmed by it.

Those “outlier” groups include children, teens, and those with cognitive disabilities—these people greatly benefit from AID systems. These are lifesavers for parents of these children, and those parents become highly engaged activists for these devices.

But over time, data has been accumulating that long-term use of these systems leads to weight gain, loss of agency, de-skilling of self-management, limitations on efficacy, quality of life, economic impact, and the “attractive nuisance” of disengagement from self-management.

The net effect is disengagement. When a person outsources their T1D management to an algorithm — an imperfect one at that — the best way for it to “work” is a sedentary lifestyle. The algorithms cannot handle exercise well at all, and they can be particularly harmful as exertion levels increase—the very kind of exercise that leads to good health, especially insulin sensitivity and weight management. When users modify their behaviors to fit what the algorithm expects, the lifestyle leads to a cascading series of metabolic disorders, typically starting with obesity.

The ADA published in its own position statement with its European counterpart that echoed nearly everything mine did (though mine delved further into more domains). Their joint article is Automated insulin delivery: benefits, challenges, and recommendations. A Consensus Report of the Joint Diabetes Technology Working Group of the European Association for the Study of Diabetes and the American Diabetes Association.

Making matters even more intriguing is the fact that these systems have never improved since they were first brought to market. In my article, Why AID Algorithms Don’t Improve: The Category Error, I cited the outcomes from when Medtronic published the results of the first clinical trial that used a hybrid closed loop (HCL) automated insulin delivery (AID) system: A1c was ~7%.

Ten years later, with multiple revisions of the algorithms, and five commercially available AID systems on the market, independent researchers have conducted trials to compare how all five perform: A comparative study from Ohio State (2024), the CLEAR trial at Cambridge, a crossover trial at McGill, and a January 2025 overview in diaTribe. They converge on the same conclusions: Ten years. No change in A1c or TIR.

Given all this, how/why would the ADA “recommend” AID systems?

As it happens, they don’t. At least, not in the way people think they do.

In my article, Standard of Care: Who Defines it, How, and Why it Matters, I dive into the ecosystem of “recommendations” within the diabetes community at large. In reality, the ADA does not recommend automated systems causally. Rather, they recommend that caregivers should consider many key caveats before recommending AID systems. This is not reflected in the uniform way companies promote it in their advertising, and no one is putting pressure on the FDA to tighten up these marketing tactics.

Again, no coordinated conspiracy — at least, not one that I aim to investigate — but clearly there are strong interests at play engaged in uncoordinated coercive groupthink. Every pump manufacturer knows this — they’re the ones running the trials. It’s common knowledge, and as long as the consumer doesn’t know, or there’s no one else to surface the problems or for the FDA to enforce restrictions on marketing practices, the practice continues.

With no one protecting the practice, the last stop is patient advocacy organizations. This is what leads to the next and final type of groupthink, and one that Makary doesn’t touch at all.

Groupthink #4: Bureaucracy

In an ecosystem like T1D, we rely on organizations like Breakthrough T1D and many others to advocate for patients’ interests. Currently, they raise funds for research, push for policy, and interface with the FDA on regulatory matters, and build an expansive ground operation to recruit more T1Ds and provide support.

These are all good and effective acts. But when it comes to correcting medical practices now known to be harmful, they could contribute more, and one cause that may be standing in their way is a phenomenon posited by Jerry Pournelle, an operations-research scientist and one of the first bloggers. He called this phenomenon the Iron Law of Bureaucracy:

“In any bureaucracy, there will be two kinds of people: those who work to further the actual goals of the organization, and those who work for the organization itself. […] In all cases, the second type of person will always gain control of the organization, and will always write the rules under which the organization functions, while those dedicated to the goals that the bureaucracy is supposed to accomplish have less and less influence, and sometimes are eliminated entirely.”

The arc is familiar: an organization begins with mission-driven founders, and as it grows, the people who sustain it — fundraisers, financiers, fiduciary stewards — become indispensable, and indispensability becomes control.

Breakthrough T1D is arguably the oldest body in this space, and you can read the arc in its board. Among the 16 governing members, only 2 carry any scientific or medical credential; 14 do not — and one of the two is Claudia Graham, a former Dexcom executive (VP of Marketing, then SVP of Global Access). All the rest have a background in finance. Yes, finance.

This matters because the organization is built for two functions — fundraising capacity and fiduciary oversight. There’s nothing wrong with either in itself. The problem is who’s missing: no independent scientist, no clinician, no patient-as-expert, no dissenting voice positioned to question a device strategy or flag an outdated practice.

When everyone around the table agrees, someone’s got it wrong.

Critically, no one is there saying, “Hey, these medical errors are harming our constituents. Shouldn’t we be doing something about it?”

Device recalls? Safety alerts? Even the FDA puts out notifications. But BreakthroughT1D is silent.

When the board becomes calcified with homogeneity, it foretells what the Iron Law predicts: the mission people end up with “less and less influence.” The institution survives, not because it’s meeting its mission, it’s because the people who control it are able to raise money and protect it from outsiders.

This is not cynicism. Correcting known medical errors that harm patients — to shave off any number of years from the 17-year average — is a mission. For many institutions, that job is on no one’s agenda.

And that brings us to the most important mission of all: the pursuit of a cure. Various organizations fund good science, but the strategy guiding what that science should do has fallen behind the literature. BreakthroughT1D’s curative programs are an example, as they state plainly that they favor “treating large numbers of people, rather than pursuing personalized therapies.”

Personalized therapies: That’s the tell.

In two recent pieces — To Cure T1D, Start with the Easiest Cases, Not the Hardest and There Will Never Be a Single Cure for T1D. There Will Be Several — I lay out why that’s backwards, and it points to physiological realities that aren’t making it to the decision-makers.

An estimated 22% of T1Ds have no autoimmunity at all, and the autologous stem-cell-grown islets that are the “personalized therapy” the T1D Fund deprioritizes, are already within reach. And for the rest of us with autoimmunity, it’s not a hard, red line. Autoimmunity has variability, and it wanes over time. “Personalized therapies” is precisely where the future of a cure lies, and to deny it is to deny modern scientific knowledge and research.

Making it even more solid is that the personalized (autologous islets) approach is less expensive over the course of the patient’s life. Yes, the upfront costs are higher because the cost of creating the islets on a per-person basis is high, but the downstream costs are far lower than the patient having to be on a lifetime of toxic immunosuppressants.

Shockingly, the companies and the tech that can do personalized therapies are already sitting in its own portfolio, but they’re not allowed to pursue them. James Shapiro, who proved islet transplants possible with the Edmonton Protocol in 2000, has written about this. In the 2025 ADA article, Advances in Cell Replacement Therapies for Diabetes, seventeen luminaries in the field summarized the technologies currently underway for curing T1D, where they also added this:

“The autologous concept theoretically allows for successful transplantation without need for immunosuppression in patients without autoimmunity, substantially altering the risk-benefit for patients with diabetes in favor of transplantation.”

Groupthink within the bureaucracy is responsible for not seeing or supporting this.

So, what can be done?

Makary’s Prescription

Here is where Makary offers his prescription for avoiding groupthink: humility. Hold our beliefs more loosely. Apologize when we’re wrong. Make room for the contrarian who turns out to be right. All true, and all necessary. You can picture him typing it into his computer like Theodoric mid-epiphany, delivering a soaring speech to the medical community at large.

We, the readers, love it.

But the people who actually need to do these things — to be publicly, consequentially humble — aren’t us. They’re the people who run the corporations, the societies, the advocacy groups… the FDA. These are the individuals with the most on the line. They know exactly what happens to a career that says the words “we were wrong.” So they’ll do what Theodoric did. Think about it for a beat. And say:

Naaaaah.

Rush was never going to concede bloodletting — the admission would have cost him everything his stature was built on, and belief or not, he wasn’t going to pay it. The dissenters, the William Cobbetts, can only do so much.

I’m one of them. I write the articles that say the quiet part out loud — the kid pointing out that the emperor has no clothes. But I’m not inventing anything. I’m communicating complex medical literature to a lay audience. My ideas aren’t unique — they come from hundreds of researchers, clinicians, and ethicists who publish the material I cite. The material that lies in the blind spots of the organizations that suffer from groupthink.

And that’s the trap in Makary’s prescription, including during his tenure at the FDA. The people who most need to heed it are the ones it can never reach, because humility, for them, is a career killer, or a product killer, or a fundraising killer. “Hold your beliefs loosely” is a lovely thing to say to a room. It’s a terrible thing to ask of someone whose job depends on having been right.

So, again, what can we do?

Leadership

Let’s go back to that scene I described earlier, where I was sitting across the table from executives, chief scientists, and MDs of major pharma companies, learning what actually moves them. I wasn’t alone — I had the professors, post-docs, and other innovators pitching their IP. But I didn’t let them go in unprepared. Here’s what I told them.

If you want an incumbent on your side, start from where they’re standing. Their reputation and their job are on the line, and the fastest way to lose them is to present your idea as a replacement for the thing they already built and bet their career on. The moment your technology competes with their life’s work, you’re not an opportunity — you’re a threat.

So don’t replace. Augment. Tell them your idea makes what they built better. Maybe it’s the next generation. It’s an add-on. There are many narratives that can work, but it’s not about changing direction. They don’t have to concede that anything before was wrong. You then need to help them tell that story inside their own organization. It’s all about messaging.

The same is true across the T1D ecosystem — the companies, the ADA, the advocacy groups. The instant the literature threatens a stated position, public humility becomes a career killer, and no one stands up to “make room for contrarians.” But framed as addition rather than admission, the same change becomes survivable — even attractive.

That advice worked for the meetings I described… because we were there. We were invited.

The difference between that scene and the organizations I’m referring to here is that leaders are reluctant to be challenged. They’re not inviting outsiders in to pitch new ideas, at least, not in a way that works.

Even internal committees are subject to the same phenomenon because they have similar carrots and sticks. The leaders themselves need board members with diverse academic and career backgrounds, and who are willing to play devil’s advocate.

In one of my classes, I taught a section on leadership, where I cited a few meaningful studies. The first is Hewlett, Marshall & Sherbin, “How Diversity Can Drive Innovation” (HBR, Dec 2013), and the second is from 2016, “Diverse Teams Feel Less Comfortable — and That’s Why They Perform Better,” which itself cites a 2011 study finding that management teams with a wider range of educational and work backgrounds produced more-innovative products.

I then had students form teams to come up with a hypothetical product design, do the research on what it would take to bring it to market, including market analysis, production, corporate logos and the whole nine yards. You’re a real company.

I then did my own A/B testing: For a few teams, I required them to have at least one student from each of three majors: Economics, Arts, and one of the sciences that isn’t associated with the main technology being promoted as the “product”. The other teams formed teams of whatever composition they wanted. As expected, all the engineers chose other engineers—and friends, no less.

At the end of the year, the teams with diverse majors (who also didn’t know each other beforehand) won all the top slots by independent judges (VCs and tech execs from Silicon Valley).

In a successful organization, whether a non-profit or a for-profit company, if the board doesn’t have its own adversarial perspectives, there’s no healthy conversation. Diversity in education and work backgrounds is, as the papers showed—and my own business competition results—may feel uncomfortable to the leaders, but an effective leader can tolerate that discomfort.

A retort may be “we disagree all the time!” Sure — on internal matters, maybe even strategy, but not in the way it really matters for the ultimate constituents. So long as that long list of medical errors continues, the core mission isn’t being met.

This is where bold leadership matters, because once calcification sets in, there’s no one — inside or out — sitting across the table to explain that the calcification is why the errors persist. Leaders have to return to the fundamentals of what makes an organization effective and self-correcting.

So, finally, again, what can we do?

Mother Nature Always Wins

Mother Nature is the great arbiter of human activities. She is indifferent to what we want—she will always do what biology demands. You can perform bloodletting for 1500 years, and patients will always die. You can build devices, you can promote diets, you can try to coax the immune system to shut down, but Mother Nature will tilt the board in her favor. You cannot game the system for long.

Groupthink is psychology—people use techniques to persuade others, usually in an attempt to game the system, to fool Mother Nature. Therefore, to answer, “what can we do?”, we have to use psychology to fight back.

But we have to be mindful of one last thing: Rush believed in bloodletting. He believe it so much, he wanted it used on himself when he was dying. People can convince themselves of anything when they surround themselves with people who just agree with them. And when they’re powerful, they shut down dissent. This is when people are afraid to call out that the emperor isn’t wearing any clothes. They don’t want the emperor or his guards to show up at their house one night.

In my articles, I’ve called out CGM manufacturers for gaming the MARD data; I’ve pointed out the harms of AID systems; I’m making it clear now that bureaucracies have groupthink. And when I’ve published those, I get email from those in the ecosystem that rely on those companies and the FDA and other orgs, and love what I say. But when I ask them to publicly support my posts on social media, everyone says they can’t—their businesses would be hurt. License agreements would be at risk. Funding would vanish.

The kid who calls out that the emperor isn’t wearing clothes gets away with it because he doesn’t have anything to lose. I have no affiliation with anyone, and I don’t need sponsorship to do my research. I’m entirely independent. What we need are leaders who are themselves the kid in the crowd that says, “we can do better”, but that can only happen among those leaders who themselves are insulated. Dr. Rush believed his story, but it might have been possible for the right person to incentivize him with positive rewards.

There’s an old saying that the greatest human emotions are fear and greed. Fear usually wins when it comes to emotions (which explains negative political messaging) but greed also wins, especially in the business arena. This is where the “yes, and” psychology can be effective.

• AID systems are remarkable for many people — and we’re going to help T1Ds build self-management skills, including for those on MDI, so everyone can get healthier with modalities that work for them.

• A1c matters — and so do lipids, metabolic fitness, and other biomarkers. We’re going to build guidelines and education that put the whole picture together, leading to better health and lower stress.

• Curative therapies are in development — and in addition to targeting those with maximum autoimmunity, we’re going to also pursue techniques that span the full spectrum of immune profiles, helping more people sooner while informing the harder cases.

Change doesn’t filter up from the literature; it filters down from leaders at the top, whose mindset becomes everyone else’s. You don’t move them by demanding they admit they were wrong. You move them by handing them a narrative where the next chapter makes the last one a stepping stone instead of a mistake — where they don’t have to say “we were wrong,” and instead say, “we’re making it better.” Mother Nature shows this works every time.

Get that right, and Theodoric looks up from his speech, thinks about it for a beat —

and this time says:

Yeahhhhh.

Subscribe now

Share

In my recent piece, To Cure T1D, Start with the Easiest Cases, Not the Hardest, I delved into a provocative idea put forward in an article published by the ADA in 2025 titled, Advances in Cell Replacement Therapies for Diabetes. Seventeen luminaries in the field summarized the technologies currently underway for curing T1D.

Buried within the text, there’s this:

“The autologous concept theoretically allows for successful transplantation without need for immunosuppression in patients without autoimmunity, substantially altering the risk-benefit for patients with diabetes in favor of transplantation.”

“The autologous concept” they’re referring to is creating islets directly from the patient’s own tissues, not some other cell line that might prompt an immune response. This idea was counter to where everyone else in the field is headed: Trying to avoid the immune system. What intrigued me was that the very people who built islet transplantation are the ones proposing the autologous islet concept for non-autoimmune T1Ds.

The initial shocker that most people feel is that “non-autoimmune T1D” is even a thing. We’ve all been taught that type 1 diabetes is because our own immune system killed our beta cells. If we don’t have autoimmunity, what killed the beta cells?

The details from my article explained that the non-autoimmune T1D population is large, and that autoimmunity is not a static state for those of us who are autoimmune. Instead, autoimmunity throttles. And therefore, my article suggests, not every T1D would receive the same type of cure. Instead, we should stratify people according to their autoimmune activity. Some, like the non-autoimmune T1Ds, would be the “easiest” patients for the reasons stated in the above quote: just make new islets from their own stem cells and reinsert them. No immune suppression required.

For everyone else, the approach employs items from a menu of options ranging from hypoimmune (gene-edited) islets to islet transplants that may take varying doses of immunosuppressive drugs that are comparable to their autoimmune activity.

What I didn’t realize at the time was that the technologies that can do this come from companies that are already backed by none other than BreakthroughT1D.

This raises the obvious question: Why isn’t it happening?

That’s the aim of this article.

Subscribe now

BreakthroughT1D and Helmsley Charitable Trust: Funders of T1D Cure Research

We start with BreakthroughT1D and the organization’s flagship cell-therapy effort, Project ACT, launched in 2024 as the biggest coordinated push yet to get cell therapies to market, explicitly the kind that don’t require immunosuppression. Their strategy is primarily focused on allogeneic cells — donor-derived islets from a single cell line. There are also hypoimmune islets, gene-edited in order to evade the immune system. I explain these in my earlier article.

BreakthroughT1D’s open funding calls reiterate that “immune tolerance in islet replacement” is aimed at beating the immune barrier. But here’s the key: BreakthroughT1D states plainly that its global cell-therapy strategy is focused on allogeneic cells to treat large numbers of people, rather than pursuing personalized therapies.

That’s the pivotal phrase right there: personalized therapies.

This implicitly excludes autologous cells because, by definition, each islet is created from each individual’s own stem cells. That’s the quintessential definition of “personalized therapy”.

The thinking seems rational on its face: creating individualized islets for each person is expensive, slow and laborious. With current technologies, this can run $100K per person. That’s not a scalable solution in the organization’s eyes.

The institutional worry is infrastructure. Making individualized islets demands rigorous manual quality control, introducing dozens of potential failure points that a single, mass-manufactured, gene-edited cell line bypasses entirely.

But these reasons are short-sighted. Regarding cost, yes, making stem-cell derived tissues is currently very expensive, and $100K per patient is hard to digest. But it’s also a nascent technology; it’s unlikely to remain expensive in 10, 15, and 20 years.

As for quality control and individualized manufacturing, it presumes that the pipeline won’t build in efficiencies over time, which is unrealistic. Stem-cell derived tissues are seen as a major growth sector throughout medicine writ large. Streamlining per-patient cell-lines is the industrial strategy for the stem-cell manufacturing trade.

Lastly—and this is what it really comes down to—the cost of manufacturing is only one component in the entire curative process. For a non-autoimmune patient, that’s the only cost. For maximally autoimmune patients (where the current strategy is targeted now), allogeneic islets are, by definition, someone else’s cell. Foreign tissue is the entire reason these cells have to be wrapped in a device, gene-edited to hide, or defended with immunosuppressive drugs. And all of that is expensive, complex, and riddled with confounders that account for why a cure has been so hard to achieve after all this time.

So the comparison isn’t islet against islet, or even the up-front costs. It’s the full cost of a cure against the full cost of a cure. Against the approaches that aim to evade the immune system, the autologous route drops the cost of hardware, the implant surgery, the replacements when the device fibroses or fails, immunosuppression, a lifetime of drugs, the monitoring around them, the infections, cancers, and kidney damage that come with chronically suppressing an immune system.

The autologous islet’s manufacturing premium is paid once. The immune-management layer is paid for years. Over a patient’s lifetime, the gap doesn’t merely narrow — against immunosuppression, it can reverse.

Even setting costs aside, or even whether autologous islets would ever work, there’s a greater reason to conduct autologous clinical trials: what we learn from them pays for itself insofar as it informs all other therapies moving forward. It doesn’t have to be seen as a long-term solution any more than the original Edmonton Protocol in 2000 that used cadaveric islets and heavy immunosuppressants. That trial could never be scaled up because of the islet source itself, but it had to be done to advance the science. (That said, autologous islets for the non-autoimmune cohort could be a viable solution.)

One or more of these reasons likely explains BreakthroughT1D’s restriction on ‘no personalized therapies.’ But the retorts above suggest that position should be reconsidered.

As it happens, there is one grant associated with the autologous islets question, and the recipient is islet-transplant pioneer James Shapiro. His goal is framed around whether recurrent autoimmunity would still demand immunosuppression. His research currently does not involve non-autoimmune patients, but of course, it wouldn’t take much to include them. It’s a small line item that informs decades of future work, much as the Edmonton Protocol did 26 years ago.

As for Helmsley, they pour money into trials, which is essential here. One example is $52M to the GPPAD network and its POInT trial, testing whether oral insulin in infancy can head off the autoimmune response, plus newborn genetic screening across five countries. Funding autologous islets in non-autoimmune patients is comparatively far less expensive than all the other trials requiring devices, immunosuppressants, and yes, hypoimmune gene-edited islets.

This now brings us back to the beginning; BreakthroughT1D has already funded the components that could run a trial today using autologous stem-cell derived islets on non-autoimmune T1Ds. All we need to do is assemble the parts. So, let’s examine some of the tech they’ve funded that directly pertain to this goal.

Tech for the Non-Autoimmune T1D

BreakthroughT1D’s investment fund portfolio includes:

• Treg engagers: Mozart Therapeutics — Treg bispecific engagers for autoimmune indications.

• Antigen-specific tolerance: COUR Pharmaceuticals — antigen-specific immune tolerance that generates disease-specific Tregs, explicitly to avoid general immune suppression.

• Beta cell regeneration: DiogenX — regenerating insulin-producing beta cells.

• Immunotherapies: the broad bucket — peptide desensitization, soluble-TCR immunomodulation, and similar.

As stated earlier, these technologies are currently looking at solving, evading, or preventing the immune attack. But these technologies can also be leveraged towards curative treatments for those without autoimmunity as well as many with low-grade autoimmune activity. That leads to the first task: Stratifying autoimmunity.

Assay Development to Assess Autoimmunity Thresholds

In my prior article, I cited a new generation of tests — BASTA (Science Translational Medicine, 19 Mar 2025) among them — that detect beta-cell-specific CD4⁺ T cells, which drive the attack — but the direct killing is done by CD8⁺ cytotoxic cells. The assay is designed to be deployable at a scale a clinic could use. Better still, these assays do not return a yes or a no. The goal is to return a magnitude — how much autoreactivity, and against how many targets.

The value here is that the assays are essential for everything in the T1D ecosystem, even for the existing strategy of full autoimmunity. As it applies to our goals here, the assays can identify those with no autoimmunity at all—the ones standing at the first door, nearly open.

An important addition: what is most needed is an assay that can be deployed across clinics worldwide to map and stratify autoimmunity by disease duration, age of onset, geographic regions, and other population-level factors. It would show how much autoreactivity exists and where it has waned, making it a far more effective sorting tool across all of T1D research, not just towards a cure. These could help improve understanding about the nature of autoimmunity, from onset to old age.

Helmsley is a primary funder of Blue Circle Health, a non-profit healthcare platform that provides free patient care to T1Ds. This organization would be an excellent launching platform for this assay. Helmsley also is a primary funder of T1D Exchange, a survey and data analysis non-profit that could also gather this data and provide clean insights about T1D autoimmunity thresholds.

This bears directly on the 22% from my first article. That number came from autoantibody testing, which — as the first piece argues — undercounts the functional population. Antibody testing flags virally triggered cases as autoimmune even when they aren’t durably so. An example is molecular mimicry, where a virus resembles a beta-cell protein closely enough that the immune system attacks both. That’s a real autoimmune attack, but a one-time, externally triggered one, not the self-renewing kind driven by genetics. The assay can discriminate between these patients.

For transplant purposes, the above scenario describes a cohort who lack the genetic program to mount a fresh attack on new islets, making them “functionally non-autoimmune”. And because antibodies fade whether or not the T cells behind them have, an antibody-negative result increasingly marks a response that’s wound down on both arms — functionally absent, which is what a transplant cares about.

The assay helps identify which door a patient is standing in front of. The next step is which curative approach is best suited for each door.

Growing new native beta cells?

This brings us to another company in T1D Fund’s portfolio, DiogenX. Their platform involves a regenerative protein (DGX-01, a Wnt/β-catenin modulator) meant to coax the patient’s own beta-cell capacity back rather than replace it, with no transplant and no immunosuppression. They talk about “modifying the course of diabetes,” “prevention and reversion,” and harnessing the patient’s “remaining endogenous beta cells.” That reads more like active, new-onset T1A — a disease with a course still running and residual mass left to rescue.

But imagine running trials on the non-autoimmune T1Ds: would they potentially grow new beta cells?

To understand this reasoning: In a T1A patient, the immune system systematically hunted down and obliterated the cellular architecture. In a non-autoimmune T1B/f-T1B patient, the original decline may have left behind a more intact progenitor microenvironment or a higher baseline of residual, quiet beta cells that are structurally primed to receive a regenerative signal. That’s where DiogenX’s product could work.

To be clear, the tech is still preclinical — animal proof-of-concept and an effect on human beta-cell mass in the lab, but no human trial yet — so it’s the earliest-stage bet here. And yet, hiding in plain sight are those who are confirmed non-autoimmune T1D. Once the drug is further along, one could imagine the simplest route might be to skip the transplant entirely and just give the patient a drug cocktail that allows their own bodies to regenerate new beta cells.

And the approach reaches past the strictly non-autoimmune into that second door: those who sit low on the spectrum but not at zero. These are the functionally non-autoimmune, which can also include LADA patients. For this group, it might be possible to use DiogenX, or even autologous islet transplants to get new islets, but if there’s low-intensity autoimmunity, this is where Mozart’s Treg engagers, or COUR Pharmaceuticals or AnTolRx’s antigen-specific tolerance drugs might come in. They quiet whatever residual autoimmunity the assay flagged, potentially allowing the new islets to survive, but requiring much lower doses than a maximally autoimmune patient would need.

Depending on how clinical trials go, this combination of technologies might even work with those at the third door: people who used to have strong autoimmunity, but have faded due to older age or other as-yet discovered reasons.

So far, we’ve already identified three high-value applications for three companies in the fund portfolio that are not being leveraged as well as they could. And so far, none of them require surgeries.

Once you get into islet replacement, the technologies shine even brighter. The assays can also help in further development of hypoimmune SC-islets, cells that are edited to dodge alloimmunity. This is the fourth door. What would need to be discovered in those trials is whether the editing also blocks recurrent autoimmunity—that’s not yet settled (and something that Shapiro’s group is pursuing). When a graft fails, a magnitude readout from this assay tells you which barrier broke. Yet another important reason for an assay like BASTA to come to market.

Here’s where we can get to new clinical trials that have not yet been done, and should be.

The Autologous SC-Islet Trial for the Non-Autoimmune Patient

This is the intervention everything else has been clearing the way for — and the one the ADA’s seventeen authors were pointing at when they wrote that autologous transplantation could work “without need for immunosuppression in patients without autoimmunity.” It’s the most direct expression of the whole thesis: not engineering a cell to hide from the immune system, but returning the patient’s own cells to a body that was never going to attack them.

Because the islets are the patient’s own cells, there’s no alloimmune barrier — and none of the immunosuppression, encapsulation, or gene-editing the rest of the field is built around. The published autologous case has already shown the cells can yield insulin independence with 98% time-in-range.

While glycemic control was achieved, the one real unknown stands out: iPSC reprogramming can introduce neoantigens — novel antigens the immune system has never seen — and whether they provoke a response in humans, and how much, has never been measured. (This was discussed at length in my prior article.) That is precisely what the trial would answer, and if it turns out to matter, the same tolerance tools already in the portfolio are on hand to dial it back. But the expected case, in a confirmed non-autoimmune patient, is the cleanest cure on the board: their own cells, returned, no drugs.

And this needn’t wait on regeneration or compete with it. The same screened population feeds both — an autologous graft, which can be tested now, and DiogenX’s regenerative protein once it reaches the clinic. They’re two shots at the same goal in the same patients, not a sequence; one may prove simpler, the other more durable, and running them in parallel is how you’d learn which — on exactly the people the rest of the field is overlooking.

There’s a second payoff that has nothing to do with curing those particular patients. Every cell-therapy program today reads graft survival through three confounds at once — allorejection, recurrent autoimmunity, drug toxicity — and can’t cleanly separate them. A non-autoimmune autologous trial strips all three away and exposes the floor: how much graft loss is just engraftment, vascularization, and manufacturing, the problems that have nothing to do with the immune system. That baseline would tell every hypoimmune and encapsulation program what it’s up against, because right now none of them can tell a failure of their immune trick from a failure of the islet itself. And the neoantigen readout would, for the first time, de-risk — or flag — every iPSC-derived product in the pipeline, the allogeneic ones included.

Summary

The shape of it is simple: a large, overlooked population — the non-autoimmune, plus the larger “functionally non-autoimmune” group — have the potential to be cured the most direct way there is: with their own beta cells, returned, and none of the engineering the rest of the field is built around. None of that requires inventing a new technology. Every piece already exists, and nearly all of it is already funded, sitting in the T1D Fund’s portfolio. The only thing missing is the initiative.

BASTA represents the potential for a deployable assay that sorts patients by the magnitude of their autoimmunity, one of the field’s hardest problems, but it could be used to tell a hypoimmune program whether a failed graft died of alloimmunity or recurrent autoimmunity, and letting immunosuppression be dosed to the threat a patient carries instead of by blanket protocol.

Mozart’s and COUR Pharmaceuticals’ tolerance drugs become low-dose brakes for the middle of the spectrum. DiogenX’s regeneration, pitched at active disease, works cleanest in exactly the non-autoimmune patient it isn’t aimed at.

Which leaves the one thing that would tie it together and has never been done: a trial of autologous SC-islets in a confirmed non-autoimmune patient, with no immunosuppression. It’s the cleanest potential cure on the board. Of course, the issue with the neoantigen has yet to be resolved: whether it’s actually a concern, and the degree in which it needs additional research has yet to be established. That’s what trials reveal.

Whatever headwinds that neoantigen may present, it’s a single problem to deal with, which is far more manageable than the multitude of problems with more complex interventions. It is, quite literally, the wide-open door from the cartoon at the top — the one with the low-hanging fruit and the cobweb, because no one has attempted to walk through it.

BreakthroughT1D and Hemsley already know how to stand up a trial. Assembling what they’ve already paid for would take a single sentence in a request for proposals. Ok, a sentence that may have a lot of commas, semicolons, bullet lists and dollar amounts, but a single sentence nonetheless. Who knows, it could even win a Nobel Prize for literature. Science knows no bounds.

Subscribe now

Share

Building things usually works best when you start with the foundation and build upward. When the problem is complex, science adds a second rule: start with the simplest version of the problem, test it, learn from it, and then advance by controlled increments of complexity. That is how we built the pyramids, learned to fly, transplanted organs, and made machines that now fit in our pockets. Start with the base case. Prove what works. Then climb.

So why aren’t we applying that same logic to the quest to cure T1D?

A cure for T1D is one of the most technically complex problems in medicine because it requires solving several hard problems at once: making new insulin-producing islets, and preventing the immune system from destroying them. The field has made extraordinary progress on the first problem. The hard part now is the immune system.

For a review of these technologies and approaches, the ADA published an article in 2025 titled, Advances in Cell Replacement Therapies for Diabetes. Seventeen luminaries in the field explain the actual technologies associated with each of these challenges. Authors include James Shapiro (who developed the Edmonton Protocol, the first reproducibly successful islet transplant in 2000), Camillo Ricordi, Bernhard Hering, Michael Rickels, and most of the others who built the field of clinical islet transplantation.

The bulk of the article explains each of four different approaches currently under investigation to tackling the immune problem:

• wrap the islets in a membrane the immune system can’t cross (encapsulation)

• place them where the immune system patrols least (immune-privileged sites)

• edit the cells so the immune system can’t recognize them (hypoimmune islets)

• retrain the immune system itself to stand down (tolerance induction)

This is why the authors claimed the immune system as “the next frontier”. And they’re right. It’s the top of the pyramid. If that frontier is crossed, the rest of the cure begins to look achievable.

But what they don’t discuss is how to get there. The current technologies are being built for the hardest cases: patients whose immune systems are fully armed, fully autoimmune, and ready to attack anything that looks like a new beta cell. By engineering against maximum hostility first, the field is trying to place the capstone before proving the foundation beneath it.

But that might not be the best way to solve the problem.

In the section of the review article dealing with stem-cell grown islets, the authors plant the idea of growing islets from the patient’s own tissues — autologous stem-cell grown islets — and just putting them back into the patient. No encapsulation, no immunosuppressants, no gene-editing. It’s not a straightforward procedure in reality, but the concept is succinct. They said:

“The autologous concept theoretically allows for successful transplantation without need for immunosuppression in patients without autoimmunity, substantially altering the risk-benefit for patients with diabetes in favor of transplantation.”

This would easily go unappreciated because of its simplicity. And naturally, this would only work for patients that have no autoimmunity.

Wait, T1D without autoimmunity?

Yes, that fact slides by quickly, and it’s a fact about T1D that few are aware of. Consider this striking statistic: Roughly 22% of T1Ds don’t even have autoimmunity. Accordingly, those patients don’t need most of the technologies being developed to evade the immune system. (I know that figure is surprising, but we’ll get back to it.)

The review article doesn’t talk about that—they only mention it because they wrote about two trials that attempted it. These trials involved involving making autologous SC-islets and putting them back into patients. The results: ideal glycemic control and the islets survived.

But the problem was that the subjects were already on immunosuppression (from organ transplant surgeries), and the patients were not actually determined to be non-autoimmune. The trials were essentially an attempt just to see if autologous islets would work. They did.

The paper then moves on.

But, let’s pause and consider that autologous SC-islet idea more closely. This isn’t just a curiosity — when we dive into the science and the medical literature, we find there’s a great deal more to this idea. And the more you look, the more that’s revealed.

This is the base of the pyramid. The easy cases that have never, ever been tested.

Why not? Three reasons, the first of which comes from the surprise you’re feeling about that 22% figure — one that is derived from multiple sources and based on testing criteria that is inconsistently applied (which we’ll dive into later). Most people think that non-autoimmune T1D is both small in number, and as statistical and medical anomalies. Freaks of nature, whose reason for even acquiring the disease could be due to any number of causes that might also have to be dealt with that may only complicate things further. This is why this subgroup is formally called, “idiopathic diabetes” — meaning, “we don’t know.”

The second reason follows closely: If there’s not that many, and they’re medical anomalies, the institutional incentives, historical precedents, funding sources are less interested. It’s not that they’re unaware, of course—they know the cohort exists—but it leads directly to the third reason: Autologous islets are “personalized” — you have to make a new set of islets for each individual, which is perceived as “not scalable”. The institutions want to fund efforts that can cure the most number of people, and autologous islets can only cure one-at-a-time.

At least, that’s how it looks. But as usual, it’s not that simple. In fact, my article, There Will Never Be a Single Cure for T1D. There Will Be Several, explains that BreakthroughT1D has already funded all of the technologies needed; the pieces just haven’t been assembled together. That article picks up from here and is an essential add-on to this one.

This article focuses on just the technical requirements, and we do so in three discrete steps.

First is the idea that non-autoimmune T1Ds are outliers. As we’ll show soon, these are not that rare, but their prevalence is not evenly distributed worldwide, and most of them are not in the United States or Europe. That’s part of why we don’t see them.

But it turns out that autoimmunity is not an on/off toggle—it throttles, especially over time. There are also those who acquire the disease later in life (LADA), who exhibit low-intensity autoimmunity, making this cohort potentially viable candidates for autologous islet transplants (alongside mild immunosuppressants that are far less toxic and more tolerable).

This fact has long been known — and in some cases were used to actually predict graft survival among those getting islet transplants — but has historically been overlooked by the field. These assays, if standardized, would change our entire perception of T1D as an autoimmune disease, allowing for treatments that can be stratified according to immune profiles.

That stratification reveals a more complex picture. It’ may well be that it’s only when you get to those with the most aggressive immune response, typically the ones who are within 5-10 years of diagnosis or those in stages 1 and 2 (before islets are destroyed), do you get the most difficult cases. And as it happens, those are the people the industry gets pressure to cure: The children.

When you consider the entire population of T1Ds, those with the maximum immune response represent the fewest, hardest, and most expensive population to treat.

It’s not that attention or resources should be shifted away. Instead, it’s more that the cure should not be seen as a one-size-fits-all intervention. If we look at autoimmunity as stratified, not only does it become possible to cure each group according to what is medically appropriate for them, but the process of doing so yields better knowledge about the nature of each therapy.

This kind of stratification is also more aligned with the scientific method: start with the simplest case, remove all the confounders, isolate conditions to their most rudimentary components, and see what happens. This allows for problems to be solved one by one as we work through the complex biology of islet placement, types of immune responses, and more. Let each new discovery inform the next tier of complexity as we move up the immune-severity ladder. Along the way, we may develop curative protocols that can be applied today to certain populations whose immune responses can tolerate them.

Early trials expose failure modes, laying the foundational work needed to solve the rest of the immune system challenges as we climb the pyramid.

This article explores this approach in these discrete steps:

• What non-autoimmune T1D actually is, how to identify it, and why the immune system turns out to be a spectrum rather than a switch — including how large that population really is. (Two Types, Determining f-T1B, and the Numbers section)

• What the existing trials and literature already show — and exactly where they stop short of the experiment that matters. (Autologous vs. Hypoimmune)

• The economic consideration — not just in the cost of trials, but the savings in R&D and other costs that can be avoided (An Economic Consideration)

We begin by understanding more about non-autoimmune Type 1 Diabetes.

Two Types of T1D and the Meandering Dividing Line

The American Diabetes Association has long divided Type 1 diabetes into a number of subtypes. Type 1A (T1A) is the autoimmune form — the kind nearly everyone means when they say “Type 1.” It is marked by detectable autoantibodies against the beta cell’s own proteins (most commonly GAD65, IA-2, ZnT8, and insulin itself), by high-risk HLA genes, and by the full cellular machinery of autoimmunity: T cells, B cells, and antigen-presenting cells coordinated against the patient’s own islets.

Type 1B (T1B) is the idiopathic form. The patient is insulin-dependent and ketosis-prone like any Type 1, but shows no sign of autoimmunity — no antibodies, no autoreactive markers. The beta cells were destroyed by something other than the immune system. “Idiopathic” is the formal name, and it carries its usual medical meaning: we don’t know why.

True idiopathic T1B is a diagnosis of exclusion — no autoimmunity, and no other cause anyone can name. But it sits within a wider family of non-autoimmune, insulin-deficient diabetes: cases in which the beta cells were lost to an identifiable, non-immune process. Most of these are formally classified elsewhere — often as “other specific types” — but they share idiopathic T1B’s one defining feature, the absence of autoimmunity. They include:

• direct viral destruction of beta cells (as opposed to a virus that triggers autoimmunity)

• toxins or drugs that are directly toxic to beta cells

• diseases of the exocrine pancreas — acute or chronic pancreatitis, cystic-fibrosis-related diabetes (CFRD), and the iron overload of hemochromatosis — sometimes grouped together as “Type 3c”

• and the truly idiopathic cases: the wastebasket where, by definition, no one knows

This may feel like a true dividing line, but here’s where it starts getting complicated. There are two additional causes that often get swept into this group, which include:

• Diabetes from molecular mimicry, where a virus resembles a beta-cell protein closely enough that the immune system attacks both.

• Diabetes triggered by cancer checkpoint inhibitors, drugs that work by releasing the immune system’s brakes.

These two sit awkwardly in this camp because neither is non-autoimmune, and neither belongs on the first list above. That is, there was an autoimmune process that killed the beta cells, but it was not genetically driven. The difference is that genetic autoimmunity is durable and self-renewing — it will destroy fresh beta cells even when those cells are the patient’s own, as identical-twin transplants showed decades ago in Sutherland, Sibley, Xu et al.

By contrast, a one-time, externally triggered attack — such as molecular mimicry and checkpoint inhibitors — in someone without that genetic program may not persist at all. Same etiological bucket, opposite implications for a transplant — which is the first sign that the cause of T1D is the wrong axis to be sorting on.

Remember, the question we care about is not how a patient lost their beta cells. It is whether their immune system is, right now, capable of mounting an attack that would destroy a transplanted islet. That is a functional question, not a diagnostic one. A patient whose autoimmunity never existed and a patient whose autoimmunity has waned below the level of doing harm are, at the moment of transplant, in nearly the same position: neither has an active attack to launch, and a new islet would survive. Nearly — because a faded autoimmunity is not a vanished one, a distinction the next section has to take seriously.

We can extend this further by noting that many people develop “operational tolerance” to tissues that might have otherwise provoked an immune response. The medical literature is replete with cases where organ transplant patients terminate their immunosuppressants (largely due to the side effects becoming too burdensome) and yet, their organs remain functioning. There’s evidence of this in T1D islet transplant patients too.

Put all of these non-autoimmune (and waned autoimmunity) together, and we come up with a a group that I call functional T1B, or f-T1B. That is, they may well have no autoimmune activity at all, or they may have a low, weakened immunity, or may be likely to develop operational tolerance.

The label is deliberately functional rather than diagnostic: it gathers everyone whose islet autoimmunity is operationally absent at the time of transplant, however they arrived there. Idiopathic T1B belongs to it. So do the non-autoimmune secondary forms above. And — as the next section will argue — so do some patients formally diagnosed as T1A, whose autoimmunity has worn down over time.

When we take this spectrum into account, it can help shape how the technologies evolve and are applied. Identify the easiest cases first, and build the simplest technology necessary to “cure” them. That is the foundation layer: the patients whose transplant problem contains the fewest immune variables. If the cure cannot be made to work there, it is unlikely to work higher up the structure. If it does work there, every failure and every success tells us what the next layer requires. And people are cured along the way.

From here on, f-T1B means this functional group. The formal idiopathic subtype, Type 1B, I’ll name in full on the few occasions I need it on its own.

Determining Who is “f-T1B”

To determine whether someone is an f-T1B, we need an assay for it. The standard test looks for autoantibodies — but antibodies tell you that autoimmunity has been present, not whether it is still actively attacking. It may also be that the antibodies have waned while the T cells are still at the ready — or that they've waned because the whole response, T cells included, has wound down. An antibody titer can't tell those apart. It’s like finding bullet casings on the ground: there was once a murder, but we don’t know if they’re still there, or whether they’ll come back.

Only a direct measure of current T-cell activity can, and there are assays that can do so, but they are largely in research labs, not yet capable of scaling up for commercial use, but that is beginning to change.

A new generation of tests — BASTA, the Beta cell Antigen Specific T cell Assay — comes out of Stuart Mannering’s lab at St. Vincent’s Institute of Medical Research in Melbourne. He published his findings in Science Translational Medicine in 2025, and what makes it matter isn’t sophistication; it’s the simplicity: Draw a small tube of blood from the patient—about the same amount you’d take for any traditional lab work—and add the proteins that beta cells make, and wait for beta-cell-specific T cells to attack the targets. The assay measures the magnitude of attack: how much autoreactivity, and against how many targets. (A longer analysis of this approach can be found in my article, Autoimmunity Stratification: The Greatest Blind Spot in T1D Cure Research.)

The concept of throttling autoimmunity is not novel at all. The paper, Immune-evasive beta cells in type 1 diabetes: innovations in genetic engineering, biomaterials, and computational modeling, reveals some remarkably important detail.

• C-peptide responders were far fewer among people with longstanding T1D than new-onset — empirical support for autoreactivity waning with duration.

• In children, even autoantibody-negative non-diabetic kids gave responses similar to T1D kids, so specificity varied sharply by age and antigen — exactly the uncharted structure that broad deployment would map.

The point being that autoimmunity is not a switch but a throttle.

I can’t overstate just how important this is, and therefore, the importance of more widely utilizing the BASTA assay for other purposes beyond looking at cures. This simple assay allows detection of early autoimmune activity far sooner than what other assays or technologies can possibly do.

An OGTT (oral glucose tolerance test) can measure C-peptide activity, and autoantibodies can detect onset, but these are lagging indicators, long after T cell activity has already begun. Detecting T cell activity could theoretically allow new therapies to prevent or delay onset for much longer if the immune activity can be modulated earlier.

But since this article focuses on islet transplantation, the BASTA assays could have a huge impact on eligibility criteria for different kinds of therapies. If autoreactivity comes in degrees, then the question stops being “is this patient autoimmune?” and becomes “is this patient’s autoimmunity below the line where a graft can live without immunosuppression?” And if suppression is needed, can we dose for the profile of the patient, rather than a one-dose-fits-all protocol?

Testing for a person’s degree of immunity in this fashion has precedent, and it’s been in plain view since 2008. A Belgian group (Huurman et al., with Roep and Pipeleers) ran a trial involving 21 established T1D patients — all T1A — where each received cadaveric islets through the portal vein under full immunosuppression. Standard protocol; nothing about the technique was new.

This trial was repeated in Hilbrands et al., Diabetes 2009 — Differences in baseline lymphocyte counts and autoreactivity are associated with differences in outcome of islet cell transplantation.

In both cases, the association between pre-transplant autoimmunity and outcome was established; Hilbrands showed that baseline lymphocyte counts and autoimmune profiles could be used prospectively to predict who would do well.

What was new, and unexpected even to the investigators, was a correlation they found after the fact: of the patients with no pre-existing T-cell autoreactivity, seven of eight became insulin-independent; of the four reactive to both GAD and IA-2, none did. Antibody levels — the markers we actually use to sort T1A from T1B — showed no association with outcome at all.

That is not a finding about transplant technique. It is a finding about patient heterogeneity: the T1A label conceals a sizable subset whose autoimmunity is functionally absent. That scientific fact dserves sitting with.

Operative T-cell autoreactivity predicted who kept the graft; antibody status predicted nothing. That is the throttle, measured — operative autoimmunity varying patient to patient among people who all carry the same diagnosis.

That this has sat in the literature since 2008 reveals more about how much more there is in the literature that we can learn from—and act upon. If clinical trials for a given technique show efficacy for patients with particular autoimmune profiles, then the assay can be used to screen them. As the tech improves, the screening score moves with it.

In other words, the architectural plans on how to build the pyramid has already been done. We just weren’t looking at the blueprints.

With that, we have a working definition of the population and a way to identify them.

Now, given the effort, time and expense for pursuing this theory, how do we know the pyramid of recipients is big enough to justify the time, expense and effort? I mentioned that 22% of T1Ds are non-autoimmune, so let’s break down where those numbers come from.

T1B by the Numbers

Assessing the potential number of candidates that might qualify as f-T1B for immunosuppression-free islet transplantation is not yet possible because such a definition never even existed till now.

Instead, we can look at conventional T1B status that’s a subset of f-T1B. Even by conventional measurements, the figures surprise most people. Besides, our primary goal is not to get precise figures, but to validate that the population size justifies the effort, especially if the aggregate complexity, cost and ultimate deployment of this base-case “cure” are markedly improved over the “cure everyone” approach the field is currently pursuing.

This is what funding organizations will be looking for.

We begin with two anchors that we know are solid. For total prevalence, the IDF Diabetes Atlas, 11th Edition / T1D Index v3.0 (2025) estimates 9.5 million people living with T1D globally — up 13% from 8.4 million in 2021. For autoantibody prevalence, the most comprehensive synthesis is a 2022 global scoping review by Ross and colleagues at the Harvard T.H. Chan School of Public Health, which pooled 125 studies across 48 countries and found that islet-autoantibody positivity varies enormously by region — with IA-2A prevalence, for example, ranging from roughly 75% in Europe down to about 31% in Africa.

We need to pause for a moment and point out a key factor that drives these numbers. They are derived from autoantibody testing, and antibodies record that autoimmunity happened, not whether it is still running. That cuts both ways. Some share of that 22% are antibody-negative because their autoimmunity faded, not because it never existed — which makes 22% too generous as a count of true non-autoimmune T1D.

But the functional population, everyone whose islet autoimmunity is operationally absent at the moment of transplant, is larger than 22%, because it also includes patients still carrying antibodies whose T cells have gone quiet. The antibody assay is wrong in both directions at once, and only a T cell assay resolves which patient is which. That is the argument, not a footnote to it.

We don’t know if they even cancel each other out because we don’t have widely deployed assays that detect real time T cell activity. But that doesn’t negate the premise, or the fact that the prevalence is not to be ignored. Overlaying autoantibody-negativity rates onto regional T1D totals produces a rough global picture.

The percentages below are estimates — directionally meaningful, not precise — but they make the central point unmistakable: non-autoimmune T1D is not a rounding error, and it is far more concentrated outside of European-ancestry populations.

On these estimates, roughly 2 million people worldwide are living with non-autoimmune T1D — and that’s only the fraction we’d capture with today’s incomplete testing. To put 2 million in context, the IDF ranks the largest national T1D populations, in order, as the U.S., India, China, Brazil, the U.K., Germany, Russia, and Canada.

The regional skew is noteworthy. The European and North American figures — the populations most T1D research is built on — are exactly where T1B is rarest. Step outside those populations and the picture inverts. A 2025 study in The Lancet Diabetes & Endocrinology — the Young-Onset Diabetes in sub-Saharan Africa (YODA) study, run across Cameroon, Uganda, and South Africa — found that roughly 65% of young-onset, insulin-treated, clinically diagnosed T1D patients of Black African ancestry were autoantibody-negative, and, crucially, also had a low genetic risk score for autoimmune T1D.

The same pattern, attenuated, shows up in the United States. Comparing the African cohorts against the U.S. SEARCH for Diabetes in Youth data, the researchers found that about 15% of Black Americans diagnosed with T1D showed the same non-autoimmune signature — antibody-negative and low genetic risk.

Put the demographics together and we see that non-autoimmune T1D isn’t a curiosity at the margins. It’s a large, real, and globally uneven population. And it’s highly likely to be undercounted.

Now add the f-T1Bs to the candidate list and the total population could be millions more, and that makes the review article’s premise of autologous SC-islets even more interesting: could we actually cure millions more people than we imaged?

We don’t know, because it’s never been tried. But while we consider that, there’s another approach also cited in the review article: hypoimmune islets. These are gene-edited cell lines designed to evade the immune system entirely. Odd as it may sound, the two serve different purposes and have trade-offs between them, but could be combined in exactly the way this article proposes: stratefication. So, let’s review both.

Autologous vs. Hypoimmune Islets

Autologous and hypoimmune islets are often discussed as competing technologies, but in this framework they sit at different levels of the pyramid. Autologous islets for the f-T1B population are the foundation case: can a patient’s own manufactured islets survive when the immune problem is minimised? Hypoimmune islets aim at one or two levels up the pyramid: can one engineered product survive across many immune layers?

Let’s start with autologous islets.

Since autologous islets are grown from the patient’s own cells, they carry the patient’s own surface markers, fully exposed to whatever autoreactivity exists. If we’re considering true T1Bs who have no autoimmunity at all, one assumes that there’s nothing to reject, no autoimmunity to reignite, no edits, no device, no drugs. Nothing triggers the immune system.

But is it even possible to grow such cells that are clean of autoimmune response? And will they actually yield glycemic control?

The ADA review article cited two trials that grew autologous SC-islets and put them into patients. One is from Wang and colleagues, from Hongkui Deng’s group. In their paper published in Cell 187, 6152–6164 (2024), Transplantation of chemically induced pluripotent stem-cell-derived islets under abdominal anterior rectus sheath in a type 1 diabetes patient, autologous chemically-induced pluripotent stem-cell-derived islets were transplanted back into the same patient, achieved insulin independence at ~75 days post-transplant, A1c dropped from 7.6% to ~5% range, TIR >98% at one-year follow-up.

The same patient’s own stem cells were used to make new islets, and achieved ideal glycemic control. That alone is worth a moment to ponder: Stem-cell grown islets from the patient’s own tissues physically worked. What we don’t know is if this could be deployed to f-T1Bs because the trial didn’t test for that. In fact, the patient was on immunosuppressants because of two liver transplants for cryptogenic cirrhosis and a whole-pancreas transplant that failed from graft thrombosis.

It’d be nice to see a follow-up trial on a T1B, of course. But we do know there would be at least one main hurdle: the manufacturing process itself. That is, even though the researchers created new islets from the patient’s own tissues, the manufacturing process might have added non-self proteins.

This is exactly the thesis of a 2019 paper from Deuse, Hu, Agbor-Enoh and colleagues (Nature Biotechnology) that showed that reprogramming an ordinary cell into a stem cell, then expanding and maturing it back into an islet cell, accumulates mutations in its mitochondrial DNA (or, mtDNA) — the small, separate genome inside the cell’s energy factories — and some of those mutations make new proteins the patient’s immune system has never encountered. These are neoantigens, which are new molecular flags for the immune system.

In other words, the manufacturing process itself made the product partly foreign. So even a perfectly autologous islet is not immunologically silent — it is “self” at the level of sourcing but not-quite-self at the level of the finished protein.

However, there has yet to be a trial to test the Deuse assertions. What we’d be looking for include:

• The effect size in humans. The mechanism and the measurable immune response are real, but no one has shown this is what killed grafts in earlier trials, such as Deng’s group, where immunosuppression was used.

• The magnitude next to autoimmune recurrence. A handful of new flags is a bounded, tunable problem — not the primed immune army that has hunted beta cells for decades.

• Mitigation and whether it’s tractable. The mutation load tracks with how the cells were reprogrammed, how long they were grown in culture, and the donor’s age; cleaner, shorter manufacturing lowers it. This is engineering, not biology.

• Whether an SC-cell line can be screened for these mutations before use — though more of them accumulate as the cells mature into islets, after the screening point, so screening contains the problem rather than erasing it.

These would tell us how much further work is needed, if any.

This is what brings us to an alternative to autologous islets, which happens to be where the field is moving already: hypoimmune islets. This is a single cell line, rather than a per-person line, where the cells are gene-edited to “hide” from the immune system by stripping the surface flags (MHC) that let T cells inspect what’s inside. (If they can’t see inside, they don’t know that it’s a foreign body that needs to be attacked.) Actually, it’s two-for-one against allo and auto (because both are MHC-restricted), so the idea is that these hypoimmune islets would be built once and produced for everyone — drug-free because the edits do the foreign-tissue evasion that the autologous route accomplishes. It could/should work for both the T1A and T1B populations.

A demonstration of the gene edits can be seen in Sana’s UP421, which involves hypoimmune-edited islets placed in a long-standing T1D patient with no immunosuppression. After fourteen months, the islets were still producing insulin and surviving.

But, like the Deng group’s trial, UP421 just missed telling us what we wanted to know because they didn’t use SC-islets — they edited cadaveric islets (just to isolate the gene editing aspect). In effect, neither trial gets us what we want to know:

• In the Deng group, they used autologous stem-cell grown islets along with immunosuppression.

• In UP421, they didn’t use immunosuppressants, but they also didn’t use stem-cell islets; they used cadaveric donor islets (which they edited).

And what both trials don’t know is the autoimmunity status of the patients. They could be non-autoimmune, they could have low-intensity autoimmunity, they could have high-intensity, hyperreactive T cell activity. We just don’t know. And because we don’t know, these trials don’t really reveal the basic science we truly need to know.

Let’s assume for the sake of the narrative that, yes, their edits work because they strip off the molecular ID badges — the MHC proteins — that every cell shows so the immune system can inspect it. Here, a patrolling immune cell finds nothing to read and moves on. But the immune system can still poison the islets from a distance with inflammatory signals (the cytokine arm); and because one cell line would be used for everyone, antibodies can latch onto leftover surface proteins and call in complement, an antibody-driven (humoral) attack the badge-stripping never touches.

There’s even a twist the edits create rather than fix: natural killer cells are built to destroy any cell that isn’t showing an ID badge — a trigger known as “missing self” — so removing the badge makes them more suspicious, not less. Anticipating that, the designers answer with yet another edit — a “don’t-eat-me” signal called CD47.

And that’s the pattern in miniature: each fix closes one door and tends to open the next. Lather. Rinse. Repeat.

Whether the whole stack holds up inside a living human immune system is exactly what no one has shown.

These and other concerns make hypoimmune islets just as much a thorny theoretical approach as autologous islets. If they work, they’ll take care of the autologous SC-islet track, and T1A and T1B can be cured by the same technology. Unless and until they do, the autologous SC-islets are easier, less complicated, and can serve a substantial population right away.

This leads directly into pragmatism. In other words, money.

An Economic Consideration

Clinical trials are expensive. Measured against the trials the field actually runs, an autologous-T1B pilot would be far leaner. For comparison:

• Organ acquisition alone for a single patient runs $50,000–90,000, because each patient typically needs islets from 2–3 deceased-donor pancreata. Counting pancreata that are processed but fail release criteria, this rises to as much as $180,000 per patient.

• A U.S. analysis put the fully-loaded cost of islet transplantation alone (ITA) at $138,872 (Moassesfar et al., American Journal of Transplantation, 2016). Those per-patient costs are dominated by the two cost centers an autologous-T1B trial removes — organ acquisition and the immunosuppression management donor cells require.

• A trial testing a new immunosuppression regimen carries that donor cost plus the drug arm and the extended safety monitoring that immunosuppression demands.

• A trial testing an encapsulation device carries the donor cost plus the device.

Because of their complexity, these trials cost vastly more than what an autologous SC-islet T1B trial would by stripping those variables out: no donor organs to procure (the cells are the patient’s own), no immunosuppression arm (there’s no immunosuppression), no device (the cells go in unencapsulated).

What remains is a structurally simpler trial — fewer variables, cleaner monitoring, healthier subjects, and none of the multi-donor organ logistics that have constrained islet trials for thirty years. For scale, the NIH’s Clinical Islet Transplantation Consortium ran its Phase 2/3 protocols across roughly 100 patients at multiple centers; a T1B pilot needs only 10–15 to establish baseline feasibility before any comparative trial.

That covers the trial. The larger expense sits upstream in the R&D and infrastructure needed to have a cell product to go to trial in the first place. Consider that Vertex paid $950 million to acquire Semma Therapeutics in 2019 and an additional $320 million to acquire ViaCyte in 2022 — $1.27 billion in acquisitions alone, before the years of internal R&D and manufacturing buildout that followed. Vertex’s three parallel programs all exist to solve one problem: getting cells past a host immune system that wants to destroy them. VX-880/zimislecel does it with lifelong immunosuppression, VX-264 by sealing the cells in an encapsulation device, and the hypoimmune line by gene-editing the cells to hide.

Compare all that to the easier autologous. The patient’s own cells, in a host that isn’t attacking them, and that requires no drugs, no devices, and no edits.

Autologous iPSC-derived islet manufacturing is feasible today. Deng’s group in Tianjin has produced clinical-grade autologous islets for individual patients. And yes, it’s nascent and expensive — roughly $100K per patient in cell production alone, before surgery and follow-up.

Scalability is a potential yellow flag of course, but the alternative is hypoimmune SC-islets. While the hypoimmune, single cell line could scale better from a manufacturing perspective, it has a much harder uphill road for technically working, especially as the target patient goes beyond the T1B population. This doesn’t make it worse, nor does it imply that it’s one or the other. But it is definitely too soon to pretend we have a crystal ball on either of these concerns to make confident forecasts.

Nevertheless, the benefit of treating the easiest patients first allows for an acceleration of the process in ways that have both scientific and economic benefits that the structural advantages (cheaper, cleaner, faster) start accruing from day one relative to the comparators, and can benefit millions of people.

The economics are so favorable, and the potential result so widely beneficial, that the low barrier to entry makes this an attractive value proposition for a small startup to bring to investors, funding organizations, and granting agencies.

Building from the Ground Up

This is where we loop back to the ADA review we began with. Its authors are right that the immune system is the next frontier, and right to lay out the four ways across it. What the ground-up approach adds isn’t a fifth way across — it’s the map.

Each of those four is being engineered for one abstract enemy: the maximally hostile immune system, imagined rather than measured. The ground-up approach measures it, and the foundational trials it produces are exactly what will benefit those other technologies.

Begin with the patient who has no immune fight at all—non-autoimmune—no door is needed, just a refinement of the basic autologous SC-islet manufacturing process. Insert enough autologous islet mass to test for an immune response, not for glycemic control. If the graft draws an immune response because of that neoantigen the Deuse paper suggests might happen, experiment with some of the new milder immunosuppressants at low doses. Eledon’s tegoprubart comes to mind. Low dose, low duration.

As the immune response arm is addressed, also test for glycemic control by increasing islet mass and experimenting with different site locations — the anterior rectus sheath and the omentum are currently interesting candidates.

As results progress, recruit f-T1Bs with low grade autoimmunity. As the autoimmune signal climbs, the data says what the minimum sufficient answer is — a gene edit, a nudge toward tolerance, a slighting longer duration of immunosuppressants. Nothing toxic and not life-long.

As we iterate through the various degrees of autoimmunity, we may eventually find that we’ve cured a large share of the T1D population, leaving only the most acutely affected for last. These are the people where the most complex and expensive interventions might have to be used.

The idea that there will be one technology the will cure everyone is the perception, even if that’s not necessarily how insiders feel. Either way, by not trying to address the patient whose immune system poses the least resistance, we’re missing out on more than just the opportunity to cure a large subset of people, but the scientific knowledge we gain from the experience helps inform the higher tiers.

Subscribe now

Here’s a reminder to read the next article in this series, There Will Never Be a Single Cure for T1D. There Will Be Several, which explains that the technologies needed to go this path already exist. They just need to be assembled.

The old stories all warned us about this. The genie, the wishing well, the deal made in haste — every culture has its version of the cautionary tale, and the moral is always the same: be careful what you wish for. The horror in those stories is never that the wish gets denied. It’s that the wish gets granted, exactly as asked, and the wisher discovers the consequences too late.

It’s one of humanity’s oldest insights, captured in another well-worn phrase: the cure can be worse than the disease.

Both come to mind when T1Ds think about “the cure” that always seems to be just on the horizon. It’s the kind of fantasy the old cautionary tales were built around — wished for, never seen, and impossible to evaluate until it’s too late. For many, we can imagine life without T1D because we have memories of what it was like before it happened to us. That’s the fantasy of a cure. But the reality of a cure requires interventions that we cannot possibly understand because too few T1Ds have experienced it.

In my article, Eledon and the T1D Cure: Whose Breakthrough Is It, Anyway?, I explain the technical reality behind the science of a drug, tegoprubart, that has been promoted as being a very big step towards a cure. And while it’s a major advancement for organ transplants, including kidneys, heart — and yes, pancreatic islets for T1Ds — it’s not the end-all “cure” that many T1Ds believe (largely due to oversensationalizing from the press).

And this is where I get comments and questions that, on the face of it, are understandable:

“I would go for that! It doesn’t need to be better! You’re always telling us that T1D is a disease that carries serious risks and burdens as well. So even if tegoprubart isn’t a cure in the strict sense, it could be a meaningful clinical improvement compared to having T1D.”

Fortunately, the researchers, the FDA, and everyone else with access to the long-term data — the data this article is about to reveal — already know that you should not want this cure for yourself or even most patients.

Breakthrough T1D’s official framing of who’s eligible for any islet transplant is very specific: “These transplants are only approved for adults whose risk from severe hypoglycemia outweighs the risks associated with immunosuppression.” That’s the formal risk-benefit threshold the field uses. It’s not “T1Ds who want to be cured”, because it would be unethical to include anyone whose current condition is better than what the immunosuppression regimen would impose on them.

So, let’s talk about what the researchers and the science community knows that is not presented in press releases or discussed in public discourse.

Since there are not enough T1Ds currently living decades with new islets on lifelong immunosuppressants, researchers look at other diseases for long-term data. Kidney transplant is an excellent stand-in, since the same drugs are used here. Below are stats known from those taking a general class of these drugs, with the caveat that we do not yet know to what degree tegoprubart will share some of these risks and side effects. The expectation is certainly less severe, but how much is entirely speculative.

Cancer and infection risks on lifelong immunosuppression

Cancer risk overall. Risk accumulates with duration of exposure rather than arriving all at once. In a cohort of kidney recipients whose grafts survived beyond twenty years, overall cancer incidence reached 4.4% at ten years, 14.6% at twenty, and 33.2% at thirty (Schachtner et al., Clinical Kidney Journal). Non-melanoma skin cancer followed a steeper curve — 10.3%, 33.5%, and 76.8% across the same intervals. An earlier single-center series found total malignancy incidence of 36% at twenty-five years (Kidney International, 2004). The pattern across studies is consistent: duration and cumulative dose of immunosuppression matter more than which agent is used.

Specific cancer risks. A commonly cited lifetime risk of squamous cell carcinoma among kidney transplant recipients is 65-200x higher than the general population, but pools a lot of people from around the world. A 2024 review reports standardized incidence ratios for NMSC ranging from 7 to 121 depending on country, with cumulative incidence highest in Australia and New Zealand at 39%, versus 12% in Europe and 1% in the Middle East.

Specific high-risk cancers. Post-transplant lymphoproliferative disorder (PTLD), Kaposi’s sarcoma, cervical cancer, and other virus-associated malignancies are dramatically elevated because immunosuppression reduces immune surveillance against oncogenic viruses (EBV, HPV, HHV-8).

Beyond cancer. Chronic immunosuppression also elevates risks of opportunistic infections, cardiovascular disease, kidney damage (especially with calcineurin inhibitors like tacrolimus, which tegoprubart is replacing), bone loss, and cognitive effects. And, ironically, diabetes.

Yes, there’s yet another kind of diabetes that can happen post-transplant, called NODAT — New-Onset Diabetes After Transplantation — which is primarily driven by steroids (prednisone) and to a lesser degree by calcineurin inhibitors and mTOR inhibitors, which directly impair beta-cell function and induce insulin resistance. The newer regimens being used with tegoprubart are largely steroid-free, which reduces but doesn’t eliminate the risk. That said, early indications seem to look very promising that tegoprubart is not as susceptible to this.

An important context for the above cases is that they are among the general population—people that do not have T1D. And that’s particularly bad for T1Ds, because decades of both excess glucose and insulin exposure exacerbates all these risk factors beyond that of the general population. In other words, for each of those risk profiles above, now multiply them by, say, 2x. Maybe 3x. We don’t know because no T1D has ever been on these drugs long enough because the islets don’t survive long enough. But we do know that T1Ds already have elevated risk in each of these areas anyway, so the effects are multiplicative.

This is particularly the case for cardiovascular risk, where MACE (Major Adverse Cardiovascular Events) is the leading cause of death among T1Ds, accounting for a life expectancy reduction of anywhere from 11-15 years. Since our risk of MACE is significantly higher than the general population, if you add on top of it even the most mild immunosuppressant drugs, then the risk profile is unthinkable.

(Reducing cholesterol levels—particularly, LDL—has been shown to be the primary intervention that has extended the lifespan of T1Ds since the 1990s. For more, see my article, Extending T1D Longevity: Balancing Lipids, Insulin, A1c.)

Those are the risk profiles. Now let’s examine what daily life would be like.

The lifestyle reality on lifelong immunosuppression

Those who absolutely abhor the daily burden and struggle of daily T1D management may feel there’s relief in all that suddenly going away. There’s a lot to say about that, to be sure. Compared to T1D daily management, taking daily pills feels relatively easy. All those health risks about cancer and viral infections may well be horrible, but at least, the daily burden and burnout of T1D is relieved.

Unfortunately, the data doesn’t bear that out.

One of the things that makes T1D so agonizing is compliance—taking insulin when you should. Pre-bolusing for meals, correction boluses, exercise, and attention to diet. Perhaps most infuriating of all is the unpredictable volatility. You never know what’s going to happen next.

This embodies most of what T1Ds hate so much about their disease. It’s so stressful for some that even the word “compliance” can be triggering.

And then there’s the injustice of it: you’re doing what you’re supposed to, yet the outcomes don’t match. Effort ≠ Outcome.

When you’re on immunosuppressants, none of that goes away — it just changes. Dosing must be rigorous and the tolerance for patterns of noncompliance is far lower than with standard T1D self-management. And it’s not just about dosing regularity; it’s the entire set of restrictions and limitations that come with it—restrictions that can make T1D compliance a breeze by comparison.

Drug levels need periodic monitoring (blood draws). Every other medication, supplement, and over-the-counter drug has to be checked against the regimen — grapefruit juice is famously off-limits with calcineurin inhibitors. Tegoprubart specifically requires IV infusion every three weeks in addition to oral medications.

Infection precautions. Transplant patients are advised to avoid crowds, sick contacts, and various foods (raw fish, undercooked meats, unpasteurized dairy, certain cheeses). Some restrictions ease over time but never fully lift. Travel becomes more complicated — vaccination requirements differ for immunosuppressed travelers, and some destinations (areas with endemic diseases) become inadvisable.

Sun exposure. Given the 65-200x skin cancer rates, transplant patients are advised to avoid significant sun exposure for life. Daily sunscreen, protective clothing, regular dermatological surveillance. For someone whose lifestyle involves outdoor activities, this is a meaningful constraint.

Exercise and weight management. The paper from Janse et al. (2024, in The Lancet Healthy Longevity) is sobering. Kidney transplant recipients commonly experience post-transplant weight gain (especially fat mass), reduced physical functioning, and elevated cardiometabolic risk. Transplant patients can’t necessarily exercise their way to good outcomes, and exercise is perhaps the most important health factor associated with T1D. The immunosuppression-driven cardiometabolic dysfunction is harder to overcome with lifestyle than the T1D-driven dysfunction is, because the drug effects persist independent of the patient’s behavior.

Sleep and fatigue. Persistent issues for kidney transplant recipients, not fully responsive to lifestyle intervention. A 2024 study using the same ACT trial data found exercise and exercise+diet interventions didn’t significantly improve sleep or fatigue trajectories compared to usual care.

Mental health. Depression and anxiety rates are elevated in transplant recipients vs. the general population. Some of this is attributable to the drugs (steroids historically, though that’s less relevant for newer regimens), some to the chronic vigilance required.

One might wonder how many of these transplant patients can put up with that.

The answer: not many.

Non-adherence rates among kidney transplant recipients run between 28% and 52% across 38 studies from this literature review, with a 2019 cohort where 71 of 182 patients (39%) were not consistently taking their medications. And they stop in full knowledge of their health risks—losing the kidney they received and keeping them alive.

They stop because of the burden of daily management and the restrictions on lifestyle.

The further patients get from the acute experience of needing the transplant, the more likely they are to drift into non-compliance. This speaks directly to the phenomenon of T1D “burnout”. We all experience it: When the burden of daily management and the suffering is so overwhelming, you just want to quit T1D. We all joke that we’re just going to give it up—go cold turkey. But you know you can’t because the moment you do, you feel it. Your glucose spikes, you’re on the roller coaster. This may cause high levels of distress, but that pain is a constant reminder to do something. Take insulin or eat. One or the other. Both. Whatever. The point is, the pain keeps you going.

Contrast this with immunosuppression: You don’t feel it. There’s no more “acute” experience of non-compliance. In fact, you feel better when you’re non-compliant. You no longer feel the anxiety from the side effects of the drugs. The entire motivation for staying engaged is gone. This is why the compliance rate for kidney patients is so poor. The fear that drove them to accept lifelong immunosuppression fades. The daily burden does not.

The consequences are real. The Sellares study shows 315 recipients that were followed roughly three years: 47% of the 50 allografts that failed were due to antibody-mediated rejection, 32% of patients were non-adherent, and half of all AMRs were attributable to non-adherence — with non-adherence 10 times more frequent among patients with graft failure.

One paradox of curing diabetes is that the cure only works if you’re compliant with the new restrictions that may actually be more burdensome than what you’re currently going through. Which makes it all the more ironic to say: Your health should never get to a point where you would ever want to be cured. This is the potential future the magic genie doesn’t tell you about when you make your wish to be “cured” of T1D — at least, using today’s technologies.

Given all this, it should come as no surprise that, in the European JDRF Center for Beta Cell Therapy in Diabetes, immunosuppressive medication is routinely tapered off in recipients starting at least one year after transplantation on the grounds that the disadvantages of continuous immunosuppressive therapy may eventually be more detrimental to the patient's condition and quality of life than regular intensive insulin treatment.

A Thought Experiment: What if YOU were “cured”?

What the announcement about tegoprubart has shown is that the health risks may well be further reduced. That’s the headline. More T1Ds could be eligible because of that. Rather than those who are near death because of their acutely brittle T1D, we might be able to dial it back somewhat to those who are at risk of slightly less, but still urgent risks—amputations, for example.

But let’s play this out a bit further by running an interesting (or troubling) philosophical thought exercise. Let’s say the ultimate cure is found that does not involve immunosuppressants—you just go into the doctor’s office, get an injection of new islets, and a few weeks later, they graft to wherever they end up going, and you’re now fully cured. What would you do then? Not how would you feel, but what would your daily life look like?

If that day were to ever come, you’d then join the rest of the population that needs to watch their weight, exercise, and try to avoid becoming a T2D, a problem that’s now at epidemic levels. The 150 million Americans who are either T2D or “prediabetic” are also rubbing the magic lamp, hoping that GLP-1s or other interventions will help them lose weight and to “cure” their T2D.

There’s always a genie willing to help, and that genie isn’t going to be honest about the wish people are asking it to grant. Sure, here’s your GLP-1 drug! Look how amazing it works. The headlines call it a miracle drug. T2D and obesity can be eliminated. It also has shown to relieve food and alcohol noise, psychological effects, and more.

And yet…

A Cleveland Clinic study (Gasoyan et al., Obesity, June 2025) analyzing electronic health records from 7,881 patients on semaglutide or tirzepatide shows that 22% of semaglutide users and 16% of tirzepatide users stopped treatment within the FIRST THREE MONTHS.

In a literature review published in Diabetes, Obesity and Metabolism, 2025, 20-50% of GLP-1 users discontinue within the first year. For obesity-specific (rather than diabetes) use, the numbers are starker — only 27% adherence and 32% persistence at one year, dropping to 15% persistence at two years in a 2023 Prime Therapeutics study.

Among those who quit, there’s the rebound effect. Weight returns with a vengeance. A1c levels rise higher. All of it worse than if they didn’t start.

The pattern isn’t disease-specific — it’s a feature of how humans relate to ongoing pharmacological interventions where the burden of compliance accumulates over time.

The burden of compliance. Sounds like a good title for a novel.

T1Ds that experience burnout are just being human. We all feel it. There’s no way around that. Interventions of any kind are burdensome. And if I haven’t used the word paradox enough, here’s one more to consider: The paradox of T1D is that the behaviors that actually make you healthier—exercise and lipid management—are exactly the same that makes everyone healthier. You’d have to do it whether you’re cured or not.

Odd as it may sound, those two interventions can allow T1Ds to live long, healthy lives, reaching and exceeding the lifespan of non-diabetics. What’s magic about exercise is that it reduces insulin requirements, improves glucose absorption, builds mitochondrial health, improves the body’s ability to fight infection, improves psychological health, and keeps weight off.

And the medical literature supports this.

The Golden Years Cohort, a group of 400 T1Ds in the UK who’ve had the disease for over 50 years, had a mean A1c of 7.6% (± 1.4). The Joslin Medalists, American T1Ds who’ve also had the disease for over 50 years, had an average current A1c of 7.3%, and 7.7% among the 73-person longitudinal subgroup followed for 15 years. Across the full cohort, individual A1c values ranged from 5.0% to 14.0% — and current/longitudinal glycemic control was not related to complications.

All of them (including me) grew up without being able to test glucose levels, no A1c tests, no CGMs, and just two kinds of insulin, Fast and Slow, administered 2-3 times daily. What all long-haulers have in common is good metabolic health: normal weight, low insulin requirements, and crucially, low lipid levels. It’s not low-carb diets, insulin pumps, or any other drugs (other than statins for lipid management).

I am by no means suggesting it’s easy. For many, it’s not. And there’s definitely a cohort of T1Ds that feels like those behaviors would be far easier without having T1D. Still more simply cannot do so with T1D. All fair, but this has to be put into context of the fantasy of a “cure”. There’s no easy way out of being healthy unless you’re compliant with something.

In closing, none of this is suggesting that a cure will never happen, or that it won’t benefit many people, or that we should stop trying. The aim here is to provide a more realistic perspective on what’s actually involved in a cure and what it requires from you. False hopes eventually fade because of unrealistic expectations, and this can lead to heavy despondency. Reshaping your expectations towards a more realistic one should help put things into perspective, so you can make sound choices, and reconsider your relationship to your disease.

As for daily management that leads to being healthier, I discuss a different approach in my article, The Four Habits of Healthy T1Ds, where I explain how building small, simple, step-wise habits yields the best outcomes and reduces mental overload. No goals. No targets. Just small tasks that you can do to improve glycemic control and a sense of self-identity.

Share

Subscribe now

While we wait for that elusive cure for T1D, the World Health Organization has been busy solving a much more pressing issue: we ran out of drug names ending in ‘-mab.’

Odd as it may sound, all those strange, unpronounceable words that companies use to name their drugs are part of a bizarrely formal system. The Nonproprietary Names (INN) program is essentially a linguistic construction site with an interesting, but important primary objective: to help companies come up with distinctive drug names while avoiding trademark conflicts. The big crisis? They literally ran out of distinct names for monoclonal antibodies. The WHO decided on a new suffix: –bart.

Yes, “bart”. One of the people at the WHO must have been a fan of the Simpsons.

One of the first users of this naming scheme is also creating a huge buzz as a potential cure for type 1 diabetes: Eledon Pharmaceuticals’ new drug called tegoprubart. The company’s stock is up 149% on the hype, though investors have a hard time saying the name.

Now that you know that “-bart” is a suffix for monoclonal antibodies, you’re probably already way ahead of me as I’m about to explain what the drug is, how it works, and why it has very little to do with T1D.

In short, the drug is an immunosuppressant that specifically targets the pathway that rejects transplanted tissues. This can include any tissue, from whole organs all the way down to pancreatic islets. Need a new kidney? Tegoprubart would be used to allow that. New heart? Same deal. The list goes on.

Naturally, since T1D is characterized by the immune system destroying beta cells, the road to a cure involves transplanting new islets (that contain beta cells) alongside an immunosuppressant to keep the immune system from attacking them. The excitement about tegoprubart is that it can do that.

And yet, it turns out that T1D may be one of the last of many diseases to benefit from it once you understand what the drug actually does and what the true barriers for a T1D cure actually are. As this article will explain, there’s a misalignment that’s not spelled out by the popular press that needs scrutiny. The fact that the trials using this drug to help T1D subjects achieve non-diabetic A1c levels is what makes it ironic that they are not likely to benefit from this drug anytime soon.

To unpack all this, let’s delve into what the drug does, what it doesn’t do, who benefits, and who doesn’t.

• Tegopurbart is an immunosuppressive drug, particularly alloimmunity, which targets foreign tissues. If you transplant an organ from a cadaver (donor), tegu is a far better drug to use than the prior generation of drugs.

• The T1Ds in this trial received cadaveric islets—hence, the foreign body element that tego addresses. But these patients still had their autoimmune machinery intact ---they were still targeting beta-cell antigens, which tego doesn’t touch. This is why additional immunosuppressants were required, and that’s the part no one talks about. Those additional drugs are not viable as long-term uses in patients. The cure is worse than the disease.

• The primary beneficiaries for tego are whole organ transplant patients, not T1Ds.

• Why test the drug on T1Ds if they can’t benefit? Cadaveric islets are deliberately used as a disposable proof-of-concept tool to test immunosuppressants for other diseases. Islets are much easier, cheaper, streamlined, and definitive to test than patients with whole organ transplants. If the drug works on the alloimmunity part, it’s a win for those other beneficiaries. It has nothing to do with autoimmunity.

In summary, most immunotherapy drug trials start by testing T1Ds (first mice, then people) because our disease is a perfect testing ground. This is why we constantly hear about mice getting cured, and occasionally people (like this trial). Everyone in the science community knows they're not a viable cure. So the trial design is essentially: use something we know won't scale (cadaveric islets) to test something we want to evaluate (tego's alloimmune suppression), in a patient population where the endpoint is easy to measure (insulin independence). T1Ds are the test bench, not the beneficiary.

Long story short, Tego represents a critical step forward in immunosuppression, and it’s worth learning about it. If you think it’s a cure because of all the hype, you need to read this article.

The 25-year detour

Tegoprubart is a lab-made antibody that achieves something that’s been an essential roadblock to addressing many diseases that involve organ transplants: the immune system. In this case, there’s a single communication pathway called CD40-CD40L (also written CD40-CD154). If you’re a T1D, you will eventually have the word “CD40” etched into your brain, because that’s the bad guy that is going to reject any kind of islet transplants, because that’s what activates T cells to attack a transplanted organ or tissue. Block the signal, and you can prevent transplant rejection.

For the past 25 years, the two primary problems among drugs blocking that pathway have been highly toxic, and they are broad-spectrum. In this latter case, they prevent the T cell from fending off common viruses and other pathogens, not to mention tougher things like cancer. The two problems together often meant that the cure was worse than the disease. Prior drugs created dangerous blood clots, killing many patients. As a 2024 review in the American Journal of Transplantation put it, clinical development of anti-CD40L antibodies was halted because of thromboembolic complications that animal models had not predicted. The entire drug class went into the freezer.

This is why tegoprubart, along with two competitors (dazodalibep and TNX-1500), are getting so much attention—they are the next generation antibodies that have been engineered to remove the platelet-crosslinking problem while preserving the immune-blocking activity. The early results suggest the engineering worked.

The next question is how well it works, and for whom.

What the T1D trial shows (and doesn’t)

The trial generating most of the publicity is an investigator-initiated study at the University of Chicago, led by Dr. Piotr Witkowski, funded by Breakthrough T1D and The Cure Alliance. Twelve adults with long-standing brittle T1D received transplants of insulin-producing islet cells from deceased donors, with tegoprubart replacing tacrolimus, the standard immunosuppressant that’s been used in organ transplantation for decades. This isn’t the one causing blood clots, but its toxicity has been severe enough that many patients who could benefit from transplants haven’t been eligible for them. The magic of tegoprubart is that it’s far less toxic, opening up the opportunity for transplants to a much larger population.

It’s worth pausing on what “replacing tacrolimus” actually means here, because there are two headlines at once. First, tacrolimus is a roughly $7 billion-a-year global market, so if there’s a newer, less toxic drug that can replace it, it grows the potential market by orders of magnitude, allowing more patients to get transplants that might not otherwise be eligible. That news alone will move a tiny little startup into the big leagues.

The second headline is the clinical trial involving T1Ds that shows that it “works”, which were presented at ATTD in Barcelona in March 2026. In that trial, all 12 patients received pancreatic islets from donated cadavers, and after four weeks post-transplant, they achieved insulin independence, with mean A1c levels around 5.35%. No rejection episodes, and no detectable kidney, neurological, or blood-pressure toxicity of the kind that tacrolimus typically causes.

As Bart Simpson would say, “Cowabunga!”

That’s your double-header, and why there’s all the excitement. The next question for everyone (not just investors, but patients too) is who benefits? And that’s where the shine starts dimming for T1D patients.

In the T1D trial, patients receive an induction package at the time of transplant — anti-thymocyte globulin (a powerful lymphocyte-depleting antibody) over several days, plus etanercept to control the inflammatory response when the new islets engraft.

After that initial period, the maintenance regimen settles into tegoprubart by IV infusion every three weeks, indefinitely, alongside daily oral mycophenolate (CellCept) — also indefinitely.

Read that again. Indefinitely. Two drugs as ongoing maintenance, plus the one-time induction cocktail you accepted at the start. The “less toxic immunosuppression” story is real, but it’s about replacing the most toxic drug in the regimen (tacrolimus) with a less toxic one. It’s not a story about eliminating immunosuppression. Patients remain on systemic immune-suppressing therapy for life, with all of the considerations that come with it.

Now, the drug is too new to know how differently those restrictions are compared to traditional therapies, which have been historically highly toxic. Tego is considerably less so, but it’s too soon to say the degree of improvement. So far, what we know is that the side effect profile is genuinely better than tacrolimus, which is part of why the trial is generating excitement. Tegoprubart doesn’t appear to damage kidneys, doesn’t cause tremor or hypertension, doesn’t induce post-transplant diabetes (yes, a completely different form of diabetes), and has far fewer drug interactions. That’s real.

But mycophenolate (the daily oral companion drug) carries its own GI and blood-count issues, the IV-every-3-weeks schedule is itself a lifestyle imposition, and the long-term cancer profile of CD40-CD40L blockade is genuinely unknown — the first patients have been on the drug for less than two years, and meaningful cancer-incidence data require decades, not months.

As for durability, the earliest patient, Marlaina Goedel, was transplanted July 17, 2024, and is documented as still insulin-independent as of late 2025 / early 2026. The full cohort of 12 patients past the four-week mark all have A1c below 6.0%, with mean A1c around 5.35%. That’s genuinely encouraging early durability evidence, and not something to dismiss.

But Edmonton-protocol patients also looked great early on. The decline became visible around year 2 and accelerated through year 5. The Eledon trial enters that danger zone starting now, and the next 2-3 years will tell whether tegoprubart-based regimens hold where tacrolimus-based regimens did not. Durability remains the open question, just with a longer baseline than the trial’s original four-week reports suggested.

Autoimmunity

Perhaps the most sobering reality for T1Ds is that our disease is primarily driven by autoimmunity, a whole different and far more complex layer of immune activity than merely rejection of foreign bodies. Sure, you can transplant a kidney, heart, or other organs, but in most T1Ds, it’s the autoimmunity that matters. Tegoprubart does not address that at all.

In other words,

• Alloimmunity attacks the cell because it's foreign. This is what Tego protects.

• Autoimmunity attacks the cell because it's a beta cell. Tego does not protect.

That’s why the other drugs had to be used in this trial. For T1Ds, Tego is nearly useless for any kind of durable, real-world cure for T1D because it doesn’t protect against the autoimmunity part.

Put another way, Tego is fine demonstration purposes like this trial because cadaveric islets are not seen as viable cures by anyone. They’re used solely to test and evaluate drugs like Tego that attack foreign tissues. But for any real-world cure, you wouldn’t use these islets, you’d have to create new islets from stem cells, and those are going to include beta cells, and Tego will not protect the immune system from attacking them because they’re beta cells (regardless of whether they are foreign).

Long story short: if you’re transplanting anything that’s not a beta cell, tego is great! This is why everyone (else) in the science community is excited: there’s a new immunosuppression drug that addresses the rejection of a foreign body. Kidney transplant patients have every reason to celebrate.

For T1Ds, you can now hear Homer Simpson yelling at his son, “Bart!”

To cure T1D, new islets have to be made, and there may be a use for tego in some cases, but it won’t be a prominent player. There are several ways to deal with this that I cover in two separate articles.

The first is To Cure T1D, Start with the Easiest Cases, Not the Hardest, which describes the variability of autoimmunity itself. Roughly 22% of T1Ds aren’t autoimmune at all, while the rest of us have variability in our autoimmunity. Studies show that it wanes over time, so islet transplants could, theoretically, start with growing our own islets from our own stem cells. Here, you wouldn’t need the immunosuppressants for foreign tissues, but you would need some other way to suppress the autoimmune portion, depending on severity. The article gets into technical detail on that.

The second article, There Will Never Be a Single Cure for T1D. There Will Be Several, covers the same ground, but looking at existing technologies that could be used, precisely because of the variability among individuals.

Who Benefits from Tegoprubart?

The primary beneficiaries are solid-organ transplant patients. Eledon’s Phase 2 trial in kidney transplantation, called BESTOW, reported mean kidney filtration rate of 69 mL/min/1.73m² at 12 months in 51 patients — what the company believes is the highest reported in larger kidney-transplant trials evaluating rejection prevention — with rejection rates that matched tacrolimus. Phase 3 is the next step, pending FDA guidance later this year.

Kidney transplant patients reading this story have arguably more to celebrate near-term than T1Ds do. The infrastructure already exists at scale. There’s no donor-cell-supply bottleneck the way there is in islet transplantation. And the patients themselves are a much larger population. If Phase 3 lands, the kidney is where tegoprubart becomes a real product first.

Next up, tegoprubart was used as part of the immunosuppression regimen in the first pig-kidney xenotransplant and the second pig-heart xenotransplant. It’s also been tested in a Phase 2A safety trial in ALS, a neurodegenerative disease where the rationale is that CD40-CD40L signaling contributes to inflammation in the brain and spinal cord. The ALS results, published in PLOS Medicine, were primarily about safety rather than disease modification, but they add to the picture of a drug that could matter across a wide range of indications.

Why use T1Ds in the Tegoprubart trials?

So why run a T1D trial at all if the real target is somewhere else?

Because T1D is uniquely valuable as a readout system, not necessarily as a sequential stepping stone. In a T1D patient, beta cells are gone — destroyed. If you give the patient donor cells (which are known to work) and add your investigational immunosuppressant, you can measure the drug’s effectiveness by whether the cells continue to function. The signal is clean. There’s no partial-organ-function ambiguity, no slow-decline endpoint, no subtle inflammatory marker to track. The cells either work or they don’t, and the patient’s A1c (or C-peptide levels) tells you which. Or, if patients’ A1c or C-peptide levels vary in the first several months when donor cells are known to perform their best, then you know something’s going awry with the drug, not the islets. That kind of clean signal is much harder (and prohibitively expensive) in whole organ transplant trials.

This is why the T1D literature has so many stories about T1Ds that are cured — especially mice, which are cured on a routine basis. It’s “easy” to give T1Ds cadaver islets and have them achieve excellent glycemic control temporarily while some investigative immunotherapy is given. The goal is not to cure T1D at all, but to examine the performance of the islets to measure the performance of the drug.

Consider the study published in the Journal of Clinical Investigation from Stanford Medicine titled, Prevention and reversal of autoimmune diabetes by mixed chimerism and islet transplantation, which reported that “19 out of 19 mice were prevented from developing T1D, and 9 out of 9 mice with established diabetes were cured. No immunosuppressive drugs required afterward. No graft-versus-host disease.”

This made headlines because it was funded by BreakthroughT1D, despite the fact that the researchers themselves specifically stated in their own paper that they have no intention of curing T1D, but instead, have their eyes on rheumatoid arthritis. Naturally, they said that T1D could be a candidate if someone were ever able to grow those islets from stem cells, as if that’s a minor problem in the grand scheme of things. (Clearly, the investigators had no idea about the problem with donor islets and growing stem-cell islets.)

Nevertheless, there’s a nod to that Stanford study simply because it genuinely is a much more palatable approach to autoimmunity than many others, and perhaps it could one day be used in the broader toolbelt, but certainly not anytime soon (or in conjunction with any other known technologies).

Does it make it wrong for T1D advocates to promote this as being more than it really is? That’s a tricky answer because adding more tools to the toolbelt is important because you never know whether a future discovery in some other area might be able to leverage it. Same thing with tegobrubart.

Hence, the pragmatic reality when it comes to T1D research, funding, and advocacy: T1D trials generate headlines that kidney trials don’t. “Functional cure” stories move stocks and motivate patients; “mean eGFR of 69 at 12 months” stories don’t. The T1D trial has produced disproportionately more press than the much larger BESTOW kidney trial, where the benefits are real, tangible, and can be deployed quicker. BreakthroughT1D’s partial funding makes them look great, even if the general public is not fully informed.

The Road to a Cure

The road to a cure is not a direct line, but a series of tangential innovations that stem from prior work—the toolbelt. Now that tegoprubart is in the mix, it’s being used to further advance new research.

Sernova has begun a clinical-trial collaboration with Eledon to test tegoprubart with their cell-pouch device. Eledon also has a collaboration with NewcelX for stem-cell-derived beta cells, with the first clinical trial targeted for 2027.

Those trials have their own challenges, of course, but they’re still important.

“Beta cell only” transplants cannot work—you need the entire islet because glycemic control relies on the entire package of cells contained within pancreatic islets. Beta cells do more than just secrete insulin; they communicate with alpha cells to signal glucagon secretion and delta cells in the same process. You need those cells next to the beta cells, or glycemic control doesn’t happen.

While the trial using beta cells can still reveal useful scientific information, and they’re easier and cheaper because you don’t have the complexity of full islets, just make sure you’re not clinking champagne glasses when they publish glowing results.

As for the pouches and other encapsulation techniques, those are not long-term viable solutions either.

The current literature and successful trials show that islets perform best (defined as yielding ideal glycemic control) when they reside in the liver, but that doesn’t mean they’ll last long. The liver is incredibly hostile and suffers from other shortcomings.

The liver’s main advantage is first-pass exposure: portal drainage delivers insulin to the liver at high concentration before it reaches the rest of the body, the way native islets do, which keeps peripheral insulin levels from having to run high to compensate. That’s why donor islets were placed in the liver in this trial, and why subjects achieved an A1c of 5.4%.

But there are problems with the liver that make it highly unlikely to be a long-term solution: First, it’s hard to get into. Portal vein infusion carries risks of thrombosis and internal bleeding. Inflammatory response destroying up to 50% of the graft within hours.

Also, if something goes wrong — like a cancerous growth — retrieval is impossible because the islets scatter through the hepatic sinusoids as thousands of microscopic deposits. There’s no discrete graft to remove. It’s also why you can’t monitor them, which is a separate problem from retrieval and matters just as much once the cells are stem-cell derived.

This is why companies like Sernova argue that a “pre-vascularized” pouch (like the Cell Pouch) creates a “natural” organ-like environment that avoids these problems entirely, potentially allowing for better long-term cell survival even if the immediate insulin kinetics are less efficient.

Other locations have been considered, which I discuss in more detail in my article, Assembling the Components to a Cure, such as the anterior rectus sheath and the omentum, both of which are near the stomach, so there’s better access to nutrients.

To sum it up: For a “cure” to happen, there must be reliable and sustainable stem cell-derived islets that do not grow into other tissues or become cancerous; autoimmune suppression that does not affect the other functions of the immune system (or the immune system is reprogrammed from the ground up); and the islets must be placed in a location that can both survive and yield acceptable glycemic control.

That’s a big deal. As Bart Simpson would say, “”¡Ay, caramba!”

Once again, the “Assembling” article is essential reading (and picks up a lot of what’s discussed here).

The right read on tegoprubart is neither “cure” nor “nothing to see here.” It’s that the immune-protection side of several different problems just got materially better, and we’ll see over the next few years which diseases are positioned to actually use it. But the T1D community needs to be more circumspect about clinking champagne glasses.

Share

In December, 2025, a woman told me about a coma she experienced from an acute hypoglycemic event; her husband had to revive her. I asked to see her CGM data from the night of her event to see why her Omnipod-5 automated dosing algorithm might have miscalculated her dose. The moment I saw her CGM chart, I recognized a pattern immediately: a series of chaotic glucose readings, zig-zagging around in seemingly incoherent ways, all clustered together.

I recognized them because I wrote about them in my 2023 article, Continuous Glucose Monitors: Does Better Accuracy Mean Better Glycemic Control? In that article I wrote about my experiment with the new Dexcom G7 at the time, where I wore it alongside the G6 for a month to see how they compared. The left-hand chart below is from that article; on the right are the CGM readings from the woman’s Dexcom data. I refer to the woman as Jane Doe to protect her identity.

The patterns were similar. And our experiences were also similar. She suffered a coma and I came pretty close to it. What the two graphs have in common is what I wrote in my article: sudden, spontaneous and physiologically implausible glucose movements that I predicted would harm users. I wrote:

Turns out, the G7’s data shot up to 270. If this really was my real glucose level, the stacked boluses I’d taken would have perfectly corrected these readings, and I would have had a soft landing. But, as the insulin started to kick in, my glucose levels plummeted to 49. And I was being conservative! Had I taken just one more unit, I may well have gone into hypoglycemic coma, and I might not be here to be writing this article.

My immediate speculation was that Jane Doe’s Omnipod-5 algorithm saw the same noisy data, and didn’t discriminate it from the rest of the readings and just dosed too much insulin. The same mistake I made in 2023.

It turns out, this is hardly an outlier—it happens very frequently, and the severity is increasing. An article published by Hunterbrook Media titled Dexcom’s Fatal Flaws cited 60 claimants alleged hospitalization and deaths as of September, 2025. In late October, they published a follow-up article reporting three more deaths. A number of individual and class-action lawsuits are in progress. All tied to the same G7 “accuracy” problems.

But to put this into perspective, while these deaths and injuries are only alleged in litigation, not yet adjudicated, they represent an implied mortality rate of roughly 1 in 2,600 to 1 in 3,000 among US AID users.

And we know that not all the adverse events associated with the sensors are captured in this data. Jane Doe is not represented here, nor are similar accounts found in Facebook and Reddit discussion groups.

To put this into perspective, when the FDA moved to restrict the Johnson & Johnson COVID-19 vaccine in May 2022, the agency cited 9 confirmed deaths attributable to a rare clotting disorder across 18.7 million doses administered — a death rate of roughly 1 in 2 million. The agency had actually moved faster than that: the initial pause came after just 6 cases and 1 death among 7 million doses, a rate of approximately 1 in 7 million.

Against that backdrop, just the known adverse events with the G7 is somewhere between 700 and 2,700 times the threshold that triggered the FDA’s J&J pause.

Could it be that these clusters of zig-zagging glucose readings might be responsible? I pose the question this way because of what I observed and wrote about in my 2023 article: Those are physiologically implausible glucose readings. Glucose doesn’t move like that in the body. If that’s the case, then we can presume that Jane’s real glucose levels were unknowable. The numbers are as good as random. The Omnipod-5 algorithm probably assumed that her real glucose levels were somewhere in the middle, so it just averaged the numbers together. That may work. But not always. And that’s the risk.

Clusters of physiologically implausible glucose readings are like Russian Roulette: It’s good to be lucky, but you shouldn’t be playing the game. It only takes one to cause real harm. The more clusters there are, the more bullets you load into the empty chambers in the gun—your risk profile rises.

The real issue is that they should never happen in the first place.

We’ll tackle that another day. What we’re faced with now is a different problem: Can we detect these clusters of incoherent data, and if so, what do we do? What we’re talking about here is CGM sensor integrity. To be clear, fault detection methods for CGM data have been explored in academic and artificial pancreas research contexts: Facchinetti et al. (2013) and Mahmoudi et al. (2016) describe real-time detection of CGM failures — spikes, drift, pressure artifacts — but in controlled research settings where the raw materials are artifacts of the CGMs themselves rather than the resulting readings they produce.

What I’m presenting in this article is different: The characterization of spontaneous clusters in real-world commercial CGM exports and their correlation with adverse insulin dosing events, has not, to my knowledge, been addressed in the published literature or in any public tool.

To that end, I wrote an algorithm that detects these patterns in real-world glucose readings produced by any CGM with two objectives: first, to assess the runtime “health” of a sensor; and second, to correlate unreliable glucose readings to adverse events. In the interest of general public health, I am making this algorithm and the package freely available for continued research and development. More on that later.

From Jane Doe’s Coma to an Algorithm

Once I made the connection between my data and Jane’s, I spent the next few months developing an algorithm to detect these patterns. Using statistical analyses described in medical literature (much of which I cited in my 2023 article), I developed an algorithm that detects when reported glucose readings become less probabilistically likely to follow one another. When you put them in sequence, they tend to collect into clusters that can be categorized into degrees of severity.

I call my algorithm a sensor integrity detector (SID). As its name implies, it detects clusters of physiologically implausible glucose readings (PIGRs).

Below is an analysis of Jane’s CGM readings over her 24-hour period, interpreted by my SID algorithm. As you can see, it identified many clusters of PIGRs from that single day where she suffered from the event.

The red boxes are “critical” clusters and the orange boxes are “elevated” clusters. We’ll get into the details of what those mean later. The “1” and “2” labels identify hypo events that come after a cluster. Black represents a nocturnal hypo; yellow represents hypos that happen during the day. Jane’s coma happened at the “1”.

NOTICE THE BLACK CIRCLES. These represent manual BGM readings that Jane did. Many times, they matched the CGM readings. These are the fortunate events. But you can see at 3:30am, she had several readings that were over 100 mg/dL lower than what the CGM said, and her pump algorithm never knew it. That’s why it kept dosing insulin, leading to her coma.

If the algorithm recognized that the clusters were unreliable, it would not have dosed anything. Therefore, her event was preventable.

This struck me as so obvious, that I couldn’t believe I was the only one to discover it. I searched online for any discussion of this sort of thing, in T1D discussion boards, AID development sites, and people I knew.

Crickets.

It was only when I started to describe the aims of why this algorithm would work that I discovered one citation: a patent filed by Dexcom on exactly this phenomenon.

Dexcom Patent for Detecting Unreliable CGM Data

Dexcom filed a series of patents that precisely characterize the methods for detecting the same noisy glucose patterns that my algorithm just did. The first one was filed in 2013, WO2014107275A1, Outlier detection for analyte sensors, US 12,354,030 B2 (granted July 2025), which focuses on detecting unreliable CGM data in real time. The language in the patent includes this line:

“One goal is to prevent closed-loop systems from incorrect dosing based on end-of-life noise, and to suspend or modify dosing when EOL signatures are detected.”

The patent then goes on to describe what should happen when these criteria are met:

1. Display instructions to change the sensor.

2. Transmit data to connected devices (pumps) indicating degraded data quality.

3. Prevent automated delivery systems from dosing based on erratic readings.

Despite the existence of these patents and the methods they describe, I have found no public evidence that they are being deployed in a way that warns connected systems before automated insulin dosing occurs.

That changes the entire framing of how the adverse events associated with the aforementioned lawsuits would be interpreted.

To understand why, let’s look at the lawsuits currently filed against Dexcom.

The Dexcom Lawsuits

As noted by the Hunterbrook Media article, the lawsuits are generally citing “the G7’s inaccuracy”, without much more detail than that. To support their claims, the claimants point to an FDA warning letter that revealed Dexcom made an unauthorized design change to a key component of the G7 that internal studies showed was inferior by “every accuracy metric.”

But this could be a red herring. Dexcom stands by its accuracy claims, asserting that the G7’s MARD value is a standard throughout the industry and has remained the same ever since the sensor was released in 2023. (MARD stands for Mean Absolute Relative Difference, where the CGM readings are compared against a blood glucose analyzer. The lower the MARD rating, the closer the CGM data matches the blood analyzer data.)

Indeed, my 2023 analysis (and a new 2026 review of the latest G7 algorithm covered later) show that the G7 has always performed largely on par with the G6 and almost every other CGM on the market—in the aggregate. That is, when you calculate longer periods of time.

The following chart from my 2023 article shows how the differences in actual readings between the two sensors over days and weeks are largely the same with the G7 reporting slightly lower values on average:

The G6 averaged 121 mg/dL, versus the G7’s 116, and the standard deviations (SD) were 33 vs. 34, respectively. By the traditional metrics of measuring a CGM’s “accuracy”, the G7 and G6 are nearly indistinguishable.

So, again, in the aggregate, the G7 is just as accurate as the G6.

As for the victims that suffered adverse events from these faulty sensors, Dexcom can say they already acknowledged and demonstrated that some sensors can, indeed, fail, and they showed it in the clinical trial they conducted to gain FDA approval for the G7.

In their article, Accuracy and Safety of Dexcom G7 Continuous Glucose Monitoring in Adults with Diabetes, the results showed that roughly 2% of sensors can have MARD values northward of 20%, with an additional 2-3% having MARD ratings exceeding 14%. But they also have no idea why; as they stated in their filing, “the basis of anomalously poor accuracy in a small proportion of sensors is unknown.”

Below is a figure included in that report, representing the MARD values for 619 sensors placed on the arm or abdomen:

Dexcom’s defense could be that those users were unfortunate enough to have encountered one of the outlier sensors, and that the company had already disclosed it in clinical testing, and that whatever the FDA letter means, the G7’s overall accuracy profile has not fundamentally changed.

There’s a lot of “ands” in that sentence.

Therefore, attributing the G7’s known ratio of inaccuracies doesn’t change the reality that these are statistically consistent with what’s always been known about the G7. We’re sorry for your loss, but medtech is an imprecise business — or so the argument would go.

A more consequential framing is whether Dexcom could have detected these bad sensors while they were in use and, if so, warned either the user or the AID system that the readings were unreliable for dosing. You know, exactly as described in their own patents. That new framing has never been proposed from my research.

Again, Dexcom could say that they are detecting noise in sensors, which explains the “sensor error” messages people get.

But is that sufficient? Dexcom may say “yes, it’s sufficient. And those erratic glucose readings you see, well, we believe they are still viable enough for dosing.”

That’s what they might say. Of course, we don’t know. But to consider it, we’d have to see the patent or its application in action.

This is why it matters that I wrote an algorithm that effectively performs the same class of detection and that I can directly correlate clusters of unreliable data to hypoglycemia. My SID algorithm can be used as a tool to test that assertion quantitatively and objectively—to determine the degree in which these detectable patterns arise.

But again, this is not unknown — Dexcom’s trial said that a small percentage of sensors can have MARD ratings north of 20%. But is it really? Now that I have my algorithm, I ran it against my entire dataset from my 2023 article, where I wore both the G6 and G7 at the same time. The results speak for themselves:

The G7 produced 24x more clusters of PIGRs than the G6. To validate it further, I ran the SID against CGM data from open source AID users from the OpenAPS Data Commons, which covered over 360,000 hours of sensor time. Below are the results: Each bar represents the percentage of time that each sensor spent producing unreliable data. The G7 performed far worse than all of their own sensors.

That’s what brings us back to Jane Doe’s event and the unreliable CGM readings that her Omnipod-5 algorithm interpreted as requiring an amount of insulin that caused her coma. The Omnipod’s algorithms are designed to be as conservative as possible; hypoglycemia is the primary risk to be avoided. If the dosing algorithm did its calculations right, this shouldn’t have happened. But if the data it was using was unreliable, then this might explain that mis-dosing calculation.

In a legal proceeding, if discovery were ever to show that these more severe signatures were detectable—especially those that resulted in death or injury—and that meaningful safeguards were left under-deployed (i.e., the algorithm didn’t flag the data as “unreliable enough” to trigger a “Sensor Error” fault), the case would begin to look more serious than ordinary negligence.

If this is the case, the reverberations would not stop with Dexcom.

Who Does This Affect?

The next obvious question is this: if Dexcom could have detected these unreliable events, and I was able to demonstrate a workable method as well, should pump algorithms have been able to detect them too? That is, why should pump algorithms have to rely on Dexcom if pump algorithms could also detect unreliable glucose readings?

If the answer is yes, then the liability landscape widens considerably to include any technology that uses CGM data to either administer insulin, or present glucose values to users in a way that would affect their in-the-moment dosing decisions.

This is where the duty-of-care question becomes harder to avoid: once a hazard is knowable, companies that act on the data cannot easily treat reliability as someone else’s problem. That, in turn, complicates some of the protections they might otherwise hope to invoke later, including PMA-related defenses.

If it was not common operating knowledge before, it is much harder to treat it that way after this. And one does not necessarily have to have access to some mysterious, proprietary, secret algorithm to detect these signatures. My code demonstrates it sufficiently. The patent’s importance here is not that it blocks inquiry, but that it shows the problem had already been worked up into something concrete. The important point is not patent doctrine in the abstract. It is that nothing about this problem makes it off-limits to other companies that rely on CGM data and have reason to take reliability seriously.

In fact, not doing so may pose its own risks: If someone is harmed in ways that can associate unreliable glucose readings to dangerous insulin dosing, it’s no longer possible to say, “We didn’t know”, or that “Dexcom didn’t tell us”, because the methods are there.

Now, just because pump companies may also be responsible for assuring glucose readings are reliable, this doesn’t alleviate Dexcom. The duty question does not end with the last actor in the chain.

The Problem is Getting Worse

Because Dexcom is the largest player in this market, there’s concern the problem could get worse due to three critical catalysts. The proverbial “perfect storm”.

The first is the rapid uptake of AID systems, increasing the number of users who delegate more of their daily management to algorithms and are therefore less likely to notice these errant glucose patterns in real time. (I discuss this and other moral hazards of automation in my article, Medical Literature Analysis: The Performance Paradox of AID Systems.) So long as AID algorithms do not have safeguards for these spurious CGM readings, the risk rises.

The second catalyst is unnecessarily high basal rates, which are the “background” insulin doses that pump algorithms infuse like a leaky faucet. The default settings are set so high that people have far more insulin onboard than is physiologically necessary, so they’re constantly teetering on the brink of hypoglycemia as it is. A simple miscalculated bolus from a cluster of unreliable CGM readings can quickly trigger an acute event.

(I discuss basal dosing at length in Basal Dosing: How Much Do We Need?, which cites DeFronzo, et al., in 1979, who found that in T1D, alpha cell dysregulation is so severe that there’s nearly an imperceptible amount of “background glucose production” that needs to be taken care of by basal insulin. But high basal rates from AID systems is pushing too much insulin into people, raising the risk of hypos. The problem has reached the point where section 9.27 of the ADA’s 2025 guidelines now recognizes overbasalization as a major health concern for T1Ds.)

So, with high basal rates and automated dosing calculators relying on potentially unreliable glucose readings, the risk elevates substantially.

Now comes the third catalyst: Dexcom announced discontinuation of the G6 as of July 1, 2026, requiring users to move over to the G7. And yet, Dexcom’s own study shows that approximately 26% of the new 15-day Dexcom G7 sensors are expected to fail or not last the full duration.

I can’t overstate how monumental this declaration is.

It’s not just a simple case that someone’s Dexcom app suddenly posts a short message: “Change your sensor”, and then they are sent a replacement. It’s not that simple.

Before a sensor technically fails, it goes through a period of gradual degradation. The days, and most importantly, the readings leading up to the ultimate failure are going to devolve into a mess of unreliable glucose readings. Data shown later in this article illustrates this perfectly, raising the pivotal question: How is this “replacement” protocol supposed to work? Users aren’t sophisticated enough to know that their sensors are producing unreliable data. Heck, until now, no one knew that such a thing could even happen.

And if users don’t know, will Dexcom tell them? What will the cutoff point be where Dexcom notifies the user that the sensor is eligible for replacement? Without guidance—let alone awareness—users are going to be more like Jane Doe and just accept whatever the system is doing. This raises the risk profile considerably.

Dexcom’s own patent has the ability to detect when a sensor is starting to degrade. When they choose to alert the user now becomes a policy decision, not a technical decision.

By taking the G6 offline—an otherwise very safe and effective sensor—users will be forced to use a risky sensor that Dexcom declares will fail at least 26% of the time, possibly more. And it’s not just everyday users that will be affected. The G6 is deeply integrated across the entire healthcare system, including hospitals, ICUs, and many other critical care settings. If the G6 has to be swapped out for the G7, this puts at risk the most vulnerable among us.

So, who’s going to come to the rescue? The FDA?

The regulatory framework that governs CGM approval has no requirement for manufacturers to characterize how often individual sensors produce physiologically incoherent readings, how severe those readings are, or whether real-time anomaly detection is functional and deployed. The only thing the FDA requires is that a CGM shows comparable performance against a reference platform, and as we’ve already seen, a sensor can easily appear to be “accurate” by these measures, while also producing exactly the kind of erratic clusters described in Dexcom’s own patents as end-of-life failure signatures.

The approval framework is silent on this. Until that gap is addressed, the question of what constitutes an acceptable failure rate remains entirely in the hands of the companies whose business interests are directly affected by the answer.

But there is one more pressure point.

Regulation Through Litigation

If there’s one thing Americans are really good at, it’s litigation. And that can be a very powerful tool for instituting change. The concept of Regulation Through Litigation describes changes — particularly those affecting industries — that were brought about by litigation rather than legislation or self-regulation. Tobacco being one of the textbook examples, but the list is extensive.

In the case of CGM data being unreliable for dosing a lethal drug, lawsuits are no longer hypothetical. The industry will be watching very closely how the Dexcom cases proceed. And more may well come. And that’s why the focus of this article is on both the Dexcom cases and the company’s patent. These may be the canary in the coal mine.

One thing to be on the watch for is how companies will seek protections under the auspices of PMA approval (“pre-market approval”). This is where companies often enjoy broad protection from liability claims after the FDA has approved a product.

But there are carve-outs for this. If a clinically important hazard is knowable, if it intersects with duties already embedded in postmarket surveillance, hazard analysis, warning, complaint handling, or risk management, then the inquiry does not stop at “the FDA approved it.”

So the inquiry does not necessarily end there. And as they do, we’re not necessarily looking for what a trial, judge or jury finds — although, that could be monumental — it’s what was found during the discovery process. That’s the work lawyers do to collect evidence leading up to a trial. That is often where the most consequential changes are made (whether or not the public is made aware of them).

During discovery, lawyers collect internal communications and documents surrounding a claim. And defendants are required to turn them over. And, as stated before, it’s not just Dexcom. It’s anyone that is part of a chain of technologies that led to a person suffering harm. It’s a classic case of the paper trail between what engineers knew, and what was implemented.

If discovery were to show that engineers understood this class of unreliability and failed to address it, PMA would become a less comfortable refuge. And it’s not just that they knew, but had every opportunity to act to mitigate the risk. And the existence of the patent and my independent implementation of detection makes this all the more apparent.

A practical fallout from this is straightforward: any company touching CGM data in a way that influences dosing or treatment decisions has reason to think seriously about data integrity. Third parties do not necessarily need to wait for Dexcom to deploy anything. If they have the data, they can begin testing for these signatures themselves.

With that accountability landscape in mind, the next question is what detection actually looks like in practice — and whether it’s as technically daunting as it might sound. It isn’t.

Detecting Physiologically Implausible Glucose Readings in CGM Data

As stated earlier, my SID algorithm detects PIGRs — Physiologically Implausible Glucose Readings — by tracking CGM readings, one by one, looking for stretches where the oscillation between them exceeds what physiology can explain. Once the pattern is detected, the severity of the cluster elevates till the pattern subsides. Obviously, a little jumpiness is tolerable, but clusters of increasingly larger amplitudes can actually be dangerous. This is where there’s a strong indication that something in the hardware is amiss, so there’s no way to know what the true glucose level could possibly be. This is where dosing decisions become dangerous. Not assuredly so. And, like the Russian Roulette analogy earlier, noisy data may very well be within range of real data. But this doesn’t mean we should be complacent with it.

With that in mind, my algorithm’s aim is mechanism, not policy. The parameters in my code are tunable. You can set how much amplitude is necessary to elevate a cluster’s severity, which will catch clusters that are, technically, physiologically implausible. Reasonable people can disagree as to when the noise is sufficiently bad for dosing decisions, and the ratio of Russian Roulette risk tolerance. This is where policy decisions affect health outcomes. This can and should be baked into minimum FDA standards, but those don’t exist. Therefore, it’s in the hands of companies and their business risk/reward analyses.

This article is not going to get into any technical detail about how the algorithm works. All that is laboriously explained in the documentation that accompanies the code, which is being made publicly available under the MIT license: Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software

The code can be downloaded from GitHub.

The distribution is a zip archive. When you download it to your computer, you then unzip it (usually, double-click on it in your file finder), and you’ll get four more zip archives and a README.md file:

After downloading the package, users are encouraged to run the code using an AI assistant, such as Claude, ChatGPT, Gemini, etc. It is so much easier than running analysis by hand.

NOTE: Some have reported that not all AI assistants will understand what to do with the code, despite the fact that there are clear instructions to the AI. In my own testing, the same AI agent will know what to do right away, and other times, it doesn’t. It’s not clear how pervasive this problem is. But I am working on a longer-term solution that I will post later.

Those with access to large datasets — OpenAPS, Nightscout, Sugarmate, Gluroo, etc. — are encouraged to use the associated API packages (fully documented) to run large analyses to get a sense of broad sensor health performance.

For individual users, to understand your own CGM’s health, you need to upload your CGM data, so to do that, you go to the website associated with your product and download your data. Dexcom’s Clarity patient portal is shown below. After you login, select a time span (I suggest 90 days), and choose the Export button—not the download button. You don’t want a PDF report, you want the raw data in CSV format.

Once you have your CGM data, open a browser tab to your preferred AI agent and upload both files—the SID distribution and your CGM data. The AI assistant will read the documentation and immediately know what to do. You can then ask it questions. If you’re lost, ask what to do. If you want new ideas, just ask. If you don’t even understand what any of this is about, ask. Those things are pretty smart that way. You can also point the AI to this article if you want to ask questions about the details.

Sample Reports

This section will show some sample reports that the SID package will produce.

The first sample is a daily report generated for March 3, 2023, the first day of the first G7 sensor I ever wore. This data comes from my 2023 article. Red boxes mark Critical clusters and orange boxes mark Elevated clusters.

All reports highlight two key datasets: The clusters themselves, and the number of hypos that happened after a cluster. The black numbered symbols represent nocturnal hypos that occur during or shortly after a cluster. These are guideposts to look for the insulin doses that may have led to the event. CGM data from Dexcom and Abbott do not contain insulin doses unless users manually add that information. And they rarely do, and AID systems don’t export their dosing data to anything.

The table at the bottom lists the time the cluster ended, how many minutes after the cluster it took before glucose levels dropped below 70 mg/dL, and the lowest BG value recorded while glucose levels remained below 70. (BG units will show mmol/L if that’s what’s used in the CGM data report.)

Now let’s look at a clean report: Below is a G6 chart for the same day.

As we can see, there are no clusters of PIGRs in the G6. But you can also see that the overall glucose readings are roughly similar to the G7. That’s what makes the G7’s “accuracy” claims feel plausible. We can see that by overlaying the charts. (This is also the leading image from my 2023 article.)

At a high level, the G6 and G7 run mostly in parity. The G7’s jumpy readings look odd, but still relatively close to the G6. True, but that’s not how dosing decisions are made. They are made, not just based on fixed glucose levels, but trajectory and rate of change. This is why the jumpiness is problematic.

If a cluster of readings breaches a threshold where they are now physiologically implausible, this could give the illusion of rapid glucose movements. If we zoom into the 4pm window, we can see this in stark detail, which brings us back to the original image at the beginning of this article.

When zoomed in like this, we can now see that these wild glucose readings are highly unreliable, and nowhere near where the G6 was—the “reliable” readings. That’s why I made my own dosing error and caused my glucose levels to drop to 49. You can imagine an AID system algorithm making the same decision. And once this becomes apparent, it’s much easier to see them.

And yet, Dexcom promoted the G7 as being “the most accurate CGM”. Technically, they were right. But that was based on easily manipulable metrics like MARD that the CGM industry has adopted and the FDA permitted. If the standard were based on sensor integrity, the CGM market would look very different, and manufacturers would have different incentives to aim for.

Now that we have an algorithm that detects clusters, let’s run it on the entire dataset where I wore both the G6 and G7 to get a dashboard of the whole month. In total, there were 19 Critical clusters, and 90 Elevated clusters. Each bar in the chart represents the number of elevated and critical clusters over the full day of readings. We start with the chart for the G7 and notice that there were 27 hypo events that succeeded a cluster.

Even for me, a particularly well-controlled T1D who’s pretty adept at dosing (multiple daily injections, or “MDI”), I can also be swayed by CGM readings that look like they require dosing.

Of course, that was in 2023 when the G7 first came out. I had no idea what was going on. The G6 was so reliable, and with the G7’s claim of being more accurate, I assumed the G7 was somehow seeing nuanced data that I’d never considered before. I thought I could get better control still. Turned out, these were entirely unreliable readings, and my experience confirmed it.

And we can see in the aggregate time-in-range (TIR) chart from 2023, just how different the outcomes were. When I used the G6 to make decisions, I achieved a TIR of >90%. When I used the G7, my TIR dropped to the ~70% range.

Let’s now focus on another part of that report: The sensor’s overall “health” as indicated by the purple bars that have the sensor numbers: S1, S2, and S3. This color is one of four that represents a scoring formula to determine how reliable the sensor is overall.

For you math nerds, I add up all the amount of time the sensor was in elevated and critical clusters, along with the amount of time the sensor produced clean data:

For comparison, below is the dashboard for the G6 that I wore at the same time: A total of 7 elevated clusters and only one critical one.

There are no hypos shown here because, even though I had them, none were associated with the G6 sensors I was wearing.

My sensor integrity detection algorithm can produce a variety of reports, including situations like my experiment wearing multiple sensors, but it can also produce reports from large datasets, indefinite time frames, and so on. Here’s a multi-sensor report showing both the G6 and G7 combined.

OpenAPS, Nightscout, Tidepool

If you use OpenAPS, Nightscout or Tidepool, your insulin dosing data is there, and my reports will graph it. Below is an example of an anonymous donor using an unknown CGM because this data is often not included in many people’s uploads. We can see this is a pretty noisy sensor—therefore, probably highly unreliable most of the time. The insulin dosing data includes both large doses as well as SMBs (super micro boluses) and are represented at the bottom of the CGM graph.

This particular user had three hypos that are directly associated with insulin dosing that appears inconsistent with the CGM readings at the time. But of course, there are a lot of unknowns here, such as which doses were automated versus user-initiated, nor whether the person ate food and bolused for it. This is why daytime data is hard to interpret—lots of confounding factors.

By contrast, nighttime data is easier because we can generally assume no manual bolusing or eating, so any insulin delivery is likely algorithm-driven.

Notice there’s one nocturnal post-cluster hypo, labeled “4”. That cluster started before midnight on Dec 15, but hypo nadir falls just after midnight on Dec 16, so the table correctly reports it (with the +1d marker), but the chart can’t render it because the scatter plot is bounded to the Dec 15 24-hour window.

The new Dexcom G7-15day Sensor

In December, 2025, Dexcom updated their G7 algorithm to smooth out some of the jumpiness in readings. The new G7-15day sensor inherits this new algorithm. To test it out, I wore the G6 and the G7-15day sensors at the same time, just as I did in 2023, but for 70 days. Below is a graph of the new G7-15day sensor for January 6, 2026. While the overall data are far better than those of the old G7 algorithm, the new sensor itself is not quite there yet.

Here’s a dashboard report for six sensors worn over the period. Here, we find there were 19 Critical level PIGR clusters throughout the period. In other words, the new G7 algorithm is clearly improved, but it is not yet as generally clean as would be ultimately needed.

It’s very important to notice one more thing in sensors 4 and 5: the number and severity of the clusters increases as the G7-15day progresses towards the end of its life cycle. This suggests that the hardware begins to degrade progressively, not all at once. This is the vulnerability users would need to be concerned about when I mentioned earlier that 26% of 15-day sensors won’t make it to the end. This gradual degradation is what happens as it approaches that endpoint, and this raises risk during those periods.

Now let’s talk about the G6.

Dexcom G6 can have bad data, but….

In my 2023 data, the G6 was almost entirely clean. Indeed, that’s almost always been the case for me and most others. The G6 is a well-established and reliable CGM. My analysis across years worth of G6 data rarely finds a problem.

But this doesn’t mean the G6 won’t have issues. It can, but this is where things are different from the G7.

The G7 is a self-contained unit — the sensor and transmitter are built together as one piece. In the G6, the transmitter is a separate unit that clips onto each new sensor. One transmitter typically lasts three months, during which you cycle through roughly nine 10-day sensors, attaching the same transmitter each time.

That design difference matters. Both the G6 and G7 transmitters use sealed batteries that can’t be replaced. As those batteries age, they don’t fail cleanly — they degrade unevenly. When that happens, the electrical signals the transmitter reads from the sensor become inconsistent, and the glucose values it reports start to drift and spike in ways that look a lot like bad sensor data. In the G6, this is easier to spot because the transmitter is separate: when one goes bad, every sensor you attach to it will produce unreliable data. In the G7, the transmitter and sensor are fused together, so when a unit goes bad, there’s no way to know whether the problem is the sensor, the transmitter, or both.

What we can say is that my algorithm detects the same signatures in both cases.

Below is a report of my G6 data from September through December 2025, covering four different G6 Transmitters. That’s a lot for a three month period, but you’ll understand why in a moment. Let’s examine the dashboard:

Notice the concentration of bad sensors (S1 - S3) in transmitter 89PKFK. The number and severity of these bad clusters made me acutely aware that the transmitter was failing. In the G6, when a series of sensors are bad, the problem is the transmitter, not the sensors.

So, I got a new transmitter (89W2NK), which started fine, but at the end of S2, again, the number of clusters were increasing, all bumped up against the end. Knowing the transmitter was failing again, I got yet another transmitter, 8QXYWL, which was largely clean. That is, there were minor clusters (that if you look at each individually, are comparatively small), but they were also not increasing in frequency the way the old G6 transmitters were.

This is very much the pattern Dexcom’s patents describe: as transmitters begin to fail, the data become progressively unreliable.

The problem with G6 transmitters is that they tend to fail pretty reliably as they get to the end of their 90-day period OR the end of the use-by date on the box. And, for some reason, whenever I get a new G6 transmitter from Dexcom or my pharmacy, that use-by date on the box is pretty close to the current date. If I don’t use it right away, that transmitter won’t last its entire 90-day period.

Now that we’re talking about bad transmitters, let’s return to Jane Doe.

Jane Doe: Mystery Solved

Jane was using a G6. And her problem wasn’t her sensors — it was her transmitter. One bad transmitter caused every sensor she used during that period to produce faulty data. This is not something generally known.

You can see in her dashboard that she kept swapping sensors, over and over, well before their normal expiration. She thought she was getting a run of bad sensors. The transmitter was the source, and it was affecting every sensor it worked with. (I changed the transmitter ID to protect her identity.)

Her acute hypoglycemic event happened on December 5th, but she had suffered 83 hypos tied to bad clusters during the entire 90-day period. Any one of those nocturnal events could have been lethal.

After her coma, she was told by her endo to change sensors. Again. It didn’t solve the problem.

Next Steps

With all we’ve covered here, restating my earlier warning may have a different resonance now: If these signatures are as detectable as the data suggest, then this begins to look less like unavoidable noise and more like a class of preventable harm. The saddest part is that this has been known for over 20 years when the first patents on detecting these patterns were filed.

Reasonable people can debate exactly where to draw the line between ‘noisy but usable’ and ‘don’t trust this data.’ That is the policy debate this kind of detector forces into the open. One could reasonably ask what the commercial consequences would be of deploying this detection more aggressively. That question is worth sitting with, while also understanding that this article is not the first to deliberate on it.

CGM and pump companies are aware that clusters of data can be unreliable, so they’ve had to debate the “noisy-but-usable” consequences long before now. A system that’s constantly bothering the user is unusable at best, and frightening at worst. That puts everyone in the ecosystem in a very uncomfortable position. The G7’s behavior caught everyone flat-footed.

And let’s not forget the Russian Roulette analogy. The likelihood that a cluster is still within the scope of a person’s real glucose levels is unknown, but likely. The fact that things work out OK anyway gives the false illusion that this is an acceptable problem.

But let’s pull the lens back and realize that this class of recurrent, transient, self-contained signal incoherence should never exist in the first place. It should not be the new normal. Before the G7, this problem was so rare that we didn’t think we needed it. And hopefully, the G7 will become noise-free. As the company works towards that, it would be wonderful to delay retiring the G6 until the new G7’s performance is as clean as the G6 with a good transmitter.

And we should also use this as a learning opportunity to make things safer throughout the industry.

One thing we need to be alert to is the industry response. The easy and dangerous way around this problem is not to solve it, but to smooth it away — to do exactly what current pump algorithms often do, averaging unreliable data until it merely looks usable. That would calcify the problem rather than fix it, much as CGM trials eventually learned how to optimize for flattering MARD values.

The article Limits to the Evaluation of the Accuracy of Continuous Glucose Monitoring Systems by Clinical Trials is a step in the right direction: it explains how MARD can be manipulated or misinterpreted by trial design, and it recommends better reference methods, weighted accuracy measures, and more transparent reporting across glucose ranges and wear periods.

But even that is not enough. What also needs to be measured is local sensor integrity — whether the trace itself becomes briefly incoherent in ways that can be clinically dangerous even when the average performance still looks respectable. To address that, my proposal is for trials to include using independent reference CGMs alongside the sensors being tested. This would catch attempts to conceal transient incoherent traces by averaging numbers together because those averages would be far out of sync with the reference sensors.

As for what those sensor-integrity standards should look like, this is where movements like #WeAreNotWaiting can help. The group has long pushed open interoperability across hardware and software platforms in much the same way open standards helped stabilize many early internet-era technologies. That same ethos could now help define practical standards for sensor-integrity analytics, using my reference platform as a working model. Working with industry, the open-source community could propose a formal standard for inclusion in future trial design and FDA review. At that point, “accuracy” in the MARD sense would no longer stand alone.

Until then, the broader ecosystem can help monitor sensor integrity even if manufacturers do not move fast enough. OpenAPS Commons, Nightscout, Tidepool, and other data-rich platforms are in a position to surface population-level performance patterns and make it harder for signal-quality problems to hide behind aggregate metrics.

Most urgently, downstream systems that act on CGM data need to begin checking its integrity for themselves. That includes apps, dosing calculators, hospitals, and closed-loop systems. If the data are incoherent, then the system should know it is acting under degraded conditions. And in some cases—especially where insulin dosing is involved—it should refuse to act at all. This is both for patient safety and for their own liability protection.

This is not as hard of an ask as it may sound because the code could be licensed or developed in-house. For example, the screenshot below are mockups showing how Gluroo and Sugarmate, two of the most popular T1D management apps, might represent unreliable CGM data.

AID systems software can monitor for sensor degradation and switch into manual mode until the data is clean again.

Dosing calculators can simply say that they can’t recommend dosing while the CGM’s data is unreliable. (As it happens, many insulin dosing calculators don’t even factor in current glucose readings in their formulas. They simply calculate dosing based on the person’s carb-to-insulin ratio and the number of carbs they’re entering into the calculator. This is notoriously dangerous, and is a separate matter entirely.)

Hospital systems that integrate CGM data into inpatient care — increasingly common in ICU settings — carry perhaps the highest duty of care of all, given the acuity of their patients and the directness of clinical action on that data.

Clinics and healthcare providers can identify when sensors or transmitters are suspect, so when patients call with problems, the HCP can be prepared to offer better advice. This could have been helpful to Jane Doe.

And, of course, there’s the CGM companies themselves. They could immediately deploy certain processes that require no FDA submission, no firmware update, and no action from any user at all. Their backend systems already aggregate data from every sensor in the field; running coherence detection centrally on that incoming data stream is a deployment that can be done within months. When a signature is detected, the system flags it. As they accumulate, the condition escalates automatically in the system. What happens next is not a technical mystery. It is a policy choice.

One option is to send automated messages to users’ phones, emails, care teams, and emergency contacts. Internal support systems would also be aware. All service agents identify the patient when they call. Imagine if Jane Doe called Dexcom to ask about her CGM: whoever answers that call would have their computer screen flash large red, all-capital letters: USER’S SENSOR SHOULD BE REPLACED IMMEDIATELY. Or a transmitter for the G6.

At this point, with what we all now know, pump manufacturers deserve much closer attention. When a closed-loop system receives incoherent sensor data and administers insulin anyway, the question is no longer just whether the system failed. It is whether that failure was tolerated as a matter of policy.

The technology exists. The methodology works. It’s time to deploy it.

Subscribe now

Share

Picture it: 1963. Arnold Kadish, a physician in Los Angeles, straps a backpack-sized machine to a patient. Tubes running in, tubes running out. It measures blood glucose intravenously every 15 seconds, and then delivers insulin and glucose or glucagon automatically to keep levels stable. The first closed-loop automated insulin delivery (AID) system. It works. It’s also completely impractical — the size of a Marine’s field pack, tethered to the patient, invasive, and useful only in a clinical setting, where a person is entirely immobile. It was only tested on one diabetic volunteer.

But the concept was proven: a machine can close the loop.

It would take nearly sixty years, over a billion dollars in research funding, and a cast of characters — a grad student from Cleveland, an Italian engineer, a Bulgarian mathematician, and a desperate father with a million-dollar check.

And that’s when the murders began.

Ok, just kidding. There were no murders. But the effort to build mathematical algorithms that would control glucose levels in real-life situations became torturous, expensive and ultimately futile, largely because of a category error of trying to engineer a solution to a problem that, by definition, is chaotic. The metabolic system is a complex adaptive system, not a clean, predictable, albeit “complicated” system.

It’s an understandable mistake; glucose management looks like a solvable system. After all, Kadish’s closed loop system was pretty straightforward. He measured glucose levels, and administered insulin or glucose as needed. Today, we have CGMs and insulin pumps—how is that any different? Why not put an algorithm inside the pump’s circuitry to measure those glucose levels and administer insulin as needed. Sort of like cruise control on a car.

It’s a fair question and a reasonable approach. So you write algorithms that would be embedded in the insulin pump. Now you need to test them. The FDA requires preclinical trials before you can test on humans, and the default path — animal trials — is expensive, slow, and unreliable because animal glucose metabolism isn’t like humans’.

But, what if you could simulate a person’s glucose metabolism? That would do the trick, right? Test it on the simulator, and if it worked, then go to human trials. Cut the long, expensive and metabolically mismatched animal trials out of the loop.

That’d be great! But, that’s where the category error lies. The highly restrained conditions that were required for Kadish’s system to work were misleading. In a real person under real world conditions, it’s anything but straightforward.

My first article in this series used a pendulum as an example. You can swing it and mathematically predict exactly its position anytime into the future. But if you add one more arm at the end of the pendulum, so that it dangles and swings along as well, it will exert forces back onto the first, and now you have a chaotic system. By definition, such systems cannot be predicted, and therefore, not controlled. You can model it all you like, and you can use those models to simulate exactly the same behavior. But that simulation cannot be used to predict any future state of a real double-armed pendulum. That’s a “complex” system.

And glucose metabolism has dozens of little arms that are jutting out from the main pendulum and each other, creating an enormously complex and chaotic system. Most of the arms aren’t even observable, let alone modelable — cortisol levels, glucagon secretion, and dozens more. And even if we could see them, their behaviors are chaotic—not predictable. And if you can’t predict, you can’t control.

But, we have closed-loop insulin delivery (AID) systems today. Yes, we do, but as the first article explained, they have reached an apex where they can’t get better because of this category error. In fact, the reason they work at all is the surprise ending of that article. If you haven’t read part 1 yet, now’s your chance.

And yet, the pursuit continues, all centered around the belief and faith that engineers would be able to program their way around this performance ceiling. As my advisor for my computer science thesis, David Huffman, used to tell me about data analysis, “if you torture the numbers enough, they’ll confess to whatever you want to hear.”

The AID systems we have today were ultimately designed to work with a specific simulator to get around the animal testing. And that simulator — its history, why it was designed, and what its aims were — is the innocent and brilliant bystander that got caught up in something it was never intended to do.

This is the story of how multiple groups came together: the brilliant scientists that designed a simulator to understand the progression of type 2 diabetes, the ambitious engineers that wanted to bring closed-loop AID system to market, a non-profit advocacy group that spearheaded the campaign, and the political pressure on the FDA to change its regulations to permit it.

And it begins, as most stories do, with good intentions.

A Young Scientist Asks a Question

It’s 1965. The Vietnam War is raging. Draft calls have hit levels not seen since Korea — 33,000 in a single month. The counter-culture movement hadn’t materialized yet. There weren’t protests, but lots of anxiety. No long hair and bell bottoms yet, it was still crew cuts and narrow ties. Across the country, college attendance slowly ticks upward as twenty-two-year-olds weigh their futures against casualty reports on the evening news.

One such graduate student was a young man named Richard Bergman. According to his published account, Origins and History of the Minimal Model of Glucose Regulation,” (Frontiers in Endocrinology, 2021), he’d come to Pittsburgh to study endocrine modeling under John Urquhart, at the suggestion of his uncle Oscar Hechter. But, after spending time with I. Arthur Mirsky — a towering figure in carbohydrate metabolism — Bergman became captivated by a deceptively simple question: why do some people lose the ability to regulate their blood sugar?

And that’s what led him to learn biology.

Type 2 diabetes was already a growing concern, but the underlying mechanics were poorly understood. Clinicians knew that some patients’ cells stopped responding normally to insulin — what would come to be called insulin resistance — but they had no way to measure it precisely, and no framework for understanding how insulin sensitivity and insulin secretion interacted to maintain glucose homeostasis, or to predict when that interaction would fail. Bergman decided his PhD thesis would focus not on the adrenal system he’d been recruited to study, but on the endocrine pancreas. When his advisor abruptly left the lab, the 22-year-old found himself an accidental independent investigator.

Over the next decade, Bergman built an experimental laboratory dedicated to an unfashionable idea: that mathematical models, rigorously tested against real physiological data, could reveal things about glucose metabolism that clinical observation alone could not. The endocrine community was skeptical — modeling had a reputation for abstraction disconnected from biology. Bergman was determined to change that.

The breakthrough came through an unlikely transatlantic connection. At an international endocrinology conference in Copenhagen, Bergman was approached by a colleague from the University of Padova who introduced him to Claudio Cobelli, a young biomedical engineer already making a name in the emerging field of systems physiology.

Both men had started as engineers, but their paths had diverged. Bergman had spent a decade building an experimental physiology laboratory — running animal studies, developing new protocols, doing wet-bench science. He used models as tools to interpret what he observed in living systems. For Bergman, the model served the experiment. For Cobelli, the simulation was the product.

Cobelli traveled to Bergman’s lab, and together they confronted a fundamental design problem: previous models of glucose regulation were either so complex their parameters couldn’t be estimated from real data, or so simple they couldn’t capture what the data actually showed. What was needed was a model just complex enough to explain the physiology — and no more.

The Minimal Model

Bergman set three requirements for any model he would build. First, it had to be grounded in known physiology — not an abstract engineering construct, but a representation of actual biological processes. Second, it had to be complex enough to account for what he observed in his experiments. Third, and crucially, it had to be simple enough that its parameters could be calculated from a single test in a single individual.

But simplicity presented an immediate problem. The beta cell’s response to glucose is nonlinear, multiphasic, and influenced by dozens of factors. A complex adaptive system.

Their breakthrough was to cut the loop. Rather than try to model insulin secretion, they treated plasma insulin concentration as an input — measured data from the patient — and plasma glucose as the output — what the model would predict. This exceedingly simple “partition analysis” allowed them to model just the insulin-sensitive tissues without having to model the black box of the pancreas.

After multiple attempts, their 6th model — the simplest representation that could account for what the data actually showed — became the one. Hence, they called it the Minimal Model.

In 1979, they published their answer in Quantitative estimation of insulin sensitivity (American Journal of Physiology. 1979;236(6):E667–E677). The model created a parameter they called SI — insulin sensitivity index. When they validated SI against the laborious glucose clamp method, it correlated beautifully, yielding a practical tool for quantifying insulin resistance. By looking at SI thresholds in an individual, clinicians could determine who would develop Type 2 diabetes. It earned Bergman the ADA’s Banting Medal — their highest scientific award — in 2006, and became one of the most cited frameworks in diabetes research.

After the 1979 publication, Claudio Cobelli returned to the University of Padova and continued to refine glucose-insulin modeling. Over the following decades, his group in the department of Information Engineering built increasingly sophisticated simulations, adding meal absorption kinetics, subcutaneous insulin dynamics, and glucagon counterregulation. Each time they opened a new box, they found more boxes inside. More evidence of the complexity of the metabolism.

Boris Kovatchev and the Mathematics of Risk

Meanwhile, at the University of Virginia, a mathematician named Boris Kovatchev had turned his attention to diabetes from an unexpected angle. Kovatchev had earned his PhD in mathematics in Bulgaria, came to UVA for computational neuroscience, and turned his attention to diabetes in 1996.

His foundational insight was deceptively simple but profoundly consequential: the blood glucose measurement scale is asymmetric in a way that distorts every statistical analysis performed on it. A glucose reading of 40 mg/dL represents a life-threatening emergency. A reading of 300 mg/dL is dangerous but not immediately fatal. Yet the standard measurement scale treats a 40-point deviation in either direction from normal as equivalent.

In a 1997 paper in Diabetes Care, Kovatchev proposed a logarithmic transformation that corrected this asymmetry, creating what he called the Low Blood Glucose Index (LBGI) — the first metric that could reliably quantify the risk of hypoglycemia from routine blood glucose data.

This was not a model of how glucose behaves. It was a model of how glucose data misbehaves when you try to analyze it — and it was exactly what the field needed. The LBGI became the standard tool for assessing hypoglycemia risk, and Kovatchev’s risk analysis framework became foundational to diabetes data science. It was a tool designed to help clinicians, not algorithms, make better decisions.

In 2003, Kovatchev founded the UVA Center for Diabetes Technology, creating the institutional home where the Padova modeling tradition and UVA’s risk quantification work would converge.

Then came JDRF.

The Big Beautiful Bet

In 2004, Jeffrey Brewer stood before the board of the Juvenile Diabetes Research Foundation and issued a challenge. His son had been diagnosed with Type 1 diabetes in 2002, and Brewer was “appalled” by how much of the disease management burden fell on patients — and how dangerous it was to stray from the routine. He offered $1 million, but only if JDRF committed to getting an artificial pancreas on the market.

At this point, one surely might wonder: He can’t be the first to think of it. Indeed not.

Brewer was already on the board of the JDRF, so he and everyone else were aware of efforts to bring automated insulin delivery to market. Hovorka’s ADICOL project in Cambridge (EU-funded, using model predictive control) was already producing human trial data.

But the elephant in the room was the work of Gary Steil, a biomedical engineer at Medtronic MiniMed, which used PID control (proportional-integral-derivative), a different control approach, modeled on beta-cell insulin response. Ten patients, ~30 hours, subcutaneous sensing and delivery. Results: mean glucose 133 mg/dL on closed-loop vs 133 mg/dL on open-loop pump therapy. TIR 70-180 was 75%. Numbers that would prove remarkably durable.

So Brewer’s proposal wasn’t what it appeared to be. He wasn’t proposing to build a system—it was to bring it to market. And, as everyone knew at the time, the bottleneck was the pipeline: The FDA required a traditional path through preclinical animal trials. But animal models of T1D are terrible — you induce diabetes in pigs or dogs with streptozotocin, and they’re now T1Ds. The problem is, they don’t behave like human T1D patients. Moreover, these trials were slow, expensive, and arguably uninformative.

Bringing any of the existing closed-loop systems under development to market required changing the FDA regulations to allow a simulator instead, but only one of the players was able to do it. Not Medtronic—they’re a regulated company, so they can’t publicly pressure the regulator who controls their product pipeline. Not academic researchers—they have no political infrastructure. JDRF was perfect: They had the brand, the congressional relationships, the fundraising machine, and — critically — the moral authority of “we’re doing this for the children.”

So, again, they needed the simulator to bypass the animal trials, but they didn’t need the algorithms. They already existed. This is where the UVA/Padova group comes in: they had the only simulation of human glucose-insulin dynamics sophisticated enough to potentially convince the FDA to accept it as a substitute for animal trials.

The JDRF Artificial Pancreas Project, once launched, funded the regulatory pathway. The JDRF would later change their name to BreakthroughT1D, and its own Chief Scientific Officer, Sanjoy Dutta, later said the simulator “saved at least five years of animal studies because we didn’t require an algorithm to be tested in an animal model to be deemed safe and effective before going into human studies. That entire chunk was eliminated.”

The FDA accepted the simulator by 2008. Specifically, they accepted the UVA/Padova simulator (FDA Device Master File MAF-1521) as a substitute for preclinical trials for closed-loop insulin algorithms. This had never been done before for a complex biological system.

As the old saying goes, be careful what you wish for, you might get it. In this case, they got a Master File, and that needs explaining.

What is a Master File?

This is really where it all matters, so sharpen your pencils and take notes.

Let’s say someone wants to be able to detect cancer in a blood sample in a doctor’s office using a new sensing device. They can test it against all sorts of cancers to know that it works. But getting samples of all the different types of cancers is expensive and laborious.

So, I see a barrier that I can lift: I create a simulator that generates a whole bunch of synthetic cancer-like DNA strands. My simulator would allow that diagnostic developer to avoid relying on actual samples, saving them a ton of time and expense. They can throw all these synthetic samples into their sensor and declare it ready for the real world: a product that can detect cancer in a doctor’s office.

My brilliant synthetic cancer simulator can be submitted to the FDA as a Master File. Anyone who wants to build a cancer detection device can reference my simulator in their FDA application — with my permission. Yes, I own the Master File, and it’s private and confidential. I’m the gatekeeper.

This is what the UVA/Padova simulator was—a device that can simulate a person. It isn’t a real person. So, if you want to make an algorithm that delivers insulin and you want to avoid animal trials, you could run your algorithm against that simulator, and it would spit out results: “yeah, glucose levels are under control now!” The FDA would then approve the algorithm.

(To be more accurate: the FDA doesn’t approve the algorithm in isolation. They approve the device system — pump + sensor + algorithm — through a PMA or De Novo submission, referencing the simulator’s Master File as part of the preclinical evidence.)

Just as my synthetic DNA strands I made were not reality, the only cancers this device is proven to detect are those that resemble the synthetic strands I designed — which capture what I think cancer looks like, not the full range of what it actually is.

The UVA/Padova simulator doesn’t interact with reality at all. It generates synthetic data from mathematical equations. When a glucose-control algorithm runs against it, the “results” are the simulator’s opinion of what would happen in a human. There’s no patient. The algorithm is being tested against a theory of a patient. And, perhaps most critically, you cannot verify it with some other system. With the cancer sensor, you could at least spot-check it against real tissue samples. With the simulator, there is no independent reference. The only way to know if its predictions match reality is to test on real humans — which is the step the simulator was accepted to replace.

Now that you understand that, consider the fact that the FDA accepted the simulator on a confidential filing with no public review. No one to point out the example I just provided above, something simple enough that anyone with a biology degree could spot. Just because you can simulate a system’s behavior doesn’t mean you can predict or control it. Simulation ≠ reality.

And that’s when the murders began.

I can’t help it. It’s such a good line.

What the simulator does is evaluate algorithms against its own assumptions about physiology. By approving the simulator, the FDA inadvertently created the conditions where algorithms that agree with it pass, whereas the algorithms that disagree fail, even if the disagreeing algorithm is closer to reality. So, the algorithms are no longer incentivized to be good—they’re incentivized to pass the test. This largely meant that there wasn’t going to be much variability in the performance of competing algorithms, since they all had to pass the same test.

And given that, the simulator’s accuracy is all the more critical, and it wasn’t ready. They never modeled real type 1 diabetics.

The Perils of Modeling T1D Physiology

Modeling non-diabetics had been routine for decades, but T1D physiology was too complex. Alpha cell dysregulation alone means they couldn’t reliably model glucagon secretion, which is notoriously stochastic in T1Ds. This is also the case with somatostatin and many other regulatory hormones involved in glucose management.

Nevertheless, the team made what seemed like straightforward modifications to model a T1D in a way that locked in all algorithms ever since: They set the basal rate too high. They published their approach in the paper GIM, Simulation Software of Meal Glucose-Insulin Model (Journal of Diabetes Science and Technology, 2007):

In order to simulate a type 1 diabetic subject, the insulin secretion module is substituted by a subcutaneous insulin infusion module. In order to account for the higher (vs normal) basal glucose, e.g., 180 mg/dl, the software introduces a higher endogenous glucose production, e.g., 2.4 mg/kg/min... Because the simulated subject is assumed in good control, all the other parameters are kept at values of the normal subject.

Read that carefully. They took a model of normal glucose-insulin dynamics, removed the pancreas module, added a subcutaneous insulin pump model, cranked up endogenous glucose production, and assumed everything else worked exactly like a non-diabetic.

So, now we have two foundational errors baked into the system.

First, the category error of attempting to turn a simulator into a predictor ignores what complex adaptive systems theory tells us about predictability. It looks real, but isn’t.

The second problem is that assumption about “higher endogenous glucose production”. This decision embedded a specific error that would reverberate through every algorithm and every diabetic that ever used an AID system. The model assumed the liver produces glucose at a constant, elevated background rate — essentially treating hepatic glucose output as a fixed parameter to be offset by basal insulin.

In fact, T1Ds do not experience hepatic glucose production like a leaky faucet, as the above model assumes. In fact, that production is shockingly low and sporadic. DeFronzo, et al. confirmed this in 1979, and published their findings in Glucose clamp technique: a method for quantifying insulin secretion and resistance (American Journal of Physiology-Endocrinology and Metabolism, 1979).

Because the simulator overrepresented hepatic glucose production, the algorithms trained on it defaulted to higher basal rates to compensate. As we just learned, if it’s in the simulator, it has to be in the algorithm. And all AID algorithms suffer from overbasalization, a topic covered in my four-part series, Ye Olde Basal Rate: Its Effect on T1D Management and Long Term Complications.

With that out of the way, the simulator was ready for prime time.

Just three months after the FDA accepted the simulator in January 2008, the first human closed-loop trials began — at the Universities of Virginia, Montpellier, and Padova — using control algorithms that had been validated entirely in silico. No animal testing. No years of preclinical work. From computer model to human subject in a single academic quarter.

Over the next seven years, the field moved from those supervised overnight studies to increasingly ambitious trials. Algorithms were tested in hotels, then diabetes camps, then home settings. Each step was enabled by the simulator: researchers could pre-screen scenarios computationally before exposing patients.

By 2015, Medtronic had developed its 670G system through this pipeline and enrolled 124 subjects across ten sites — nine in the U.S. and one in Israel — for its pivotal trial (Bergenstal et al., JAMA 2016; ClinicalTrials.gov NCT02463097). The results: Time-in-range: 72%.

Despite these very modest improvements, Medtronic’s chief medical officer declared that the system “should be safe and more effective than what the person with diabetes can do on their own.”

Let’s pause and consider that statement “more effective than what the person with diabetes can do on their own” — that was both patently bold, and patently wrong.

Within a year from when he made that statement, a review of 8 clinical trials, Optimal prandial timing of bolus insulin in diabetes management: a review, showing that MDI users on the effects of something as simple as pre-meal bolusing showed that the simple act of timing a bolus 15–20 minutes before eating — no algorithm, no automation, no hardware — reduces postprandial glucose excursions by nearly 30% and reduces hypoglycemia, figures that an AID algorithm has never been able to achieve.

In other words, this simple act of pre-meal bolusing achieves glucose improvements that exceed what the 670G trial demonstrated with a $7,000+ system.

But Medtronic, JDRF and the FDA still promoted the 670G as turning the tide in T1D management.

The Great Unraveling

The 670G’s pivotal trial in 2016 showed an A1c of ~7%.

By 2025, independently run clinical trials across all five commercially available AID systems showed the same thing: ~7%. Different hardware, different software, different companies. Same result. The algorithms hadn’t improved — they had converged on a ceiling that was visible from the very first system.

Interestingly, Dana Lewis and Scott Leibrand, complete outsiders with no experience in any of this, wrote code from scratch and started OpenAPS in 2014-2015, an open-source AID algorithm that users can download and use on many commercial systems. Neither was a control theorist. They didn’t build from the Minimal Model or the simulator. And their systems also perform largely the same.

Why all these systems achieve the same outcome should be evident by now: Complex systems can be modeled, but modeling doesn’t mean they can be predicted, so there was an upper limit on how well they could ever perform. But there was also a lower limit: suspension of insulin for hypoglycemia. The gap between the upper and lower limits is covered in my article, Why AID Algorithms Don’t Improve: The Category Error, which assembles everything discussed here: chaos theory, complex systems, category errors, and the ultimate algorithm design that leaves very little room for variability.

Despite the performance ceiling, it can still be seen as a major benefit to many users. The greatest beneficiaries are those who either cannot or will not self-manage their disease. These people have A1c levels at 9%, 10% and much higher. To bring theirs down to 8% is a remarkable achievement, to be sure.

But the story didn’t end there: Discontinuation rates were high. Alarm fatigue was pervasive. And among those who stayed on the systems, daily life gradually narrowed to accommodate what the algorithm expected. Users ate more predictably. They avoided unannounced meals. They reduced the kinds of physical activity that introduced glucose variability the system couldn’t handle.

The research documenting these behavioral shifts is detailed in my article, Medical Literature Analysis: The Performance Paradox of AID Systems, documenting these behavioral shifts and their consequences: weight gain, loss of agency, de-skilling of self-management, limitations on efficacy, quality of life, economic impact, and the “attractive nuisance” of disengagement from self-management.

Lost in Translation

So, what happened to all the characters in our story?

Bergman had never joined the effort at all. He is still at Cedars-Sinai, still studying insulin resistance, still doing the T2D diagnostic work he’d started in Pittsburgh forty years earlier. His Minimal Model was cited in the lineage of every AID algorithm on the market, but he had no hand in any of them. Bergman, the young scientist who started it all in Pittsburgh, managed to stay focused on biology and avoid the coming regulatory storm. The kid from Cleveland got out clean.

Cobelli stayed closer to the fire. His group at Padova maintained the simulator, publishing revisions in 2013 and 2017.

Kovatchev had crossed over entirely. As director of the UVA Center for Diabetes Technology, he is now embedded in the regulatory and commercial infrastructure — the institutional bridge between Padova’s models and the FDA’s approval pathway. The mathematician who’d built the risk framework was now part of the apparatus that depended on it.

But recently, his language changed.

In 2025, his group published a study in NPJ Digital Medicine describing what they called “Adaptive Biobehavioral Control” — a system that creates a personalized “digital twin” for each user, letting them simulate changes to their settings before applying them in real life. Time in range improved from 72% to 77%. Kovatchev called it “human-machine co-adaptation,” and the phrase was telling: it acknowledged, for the first time in the program’s history, that removing the human from the loop had been the wrong direction.

The user needed to stay engaged.

But there was a catch that the press coverage didn’t mention. The digital twin powering each patient’s simulation was still built on the UVA/Padova simulator — the same model Dalla Man had constructed from healthy subjects decades earlier, the same model the FDA had accepted as a substitute for preclinical trials in 2008, revised in 2013 and 2017 but never rebuilt from the ground up.

And the study’s own data told a story the framing obscured: patients who received traditional feedback — graphs, metrics, the usual dashboards — showed no improvement over standard AID alone. It was only the interactive simulation, the part that required active user participation, that moved the needle.

The technology had improved outcomes not by predicting glucose better, but by giving users a reason to pay attention. And the broader literature on digital twins in biological systems is not ambiguous.

A 2025 review in Nature Digital Medicine — De Domenico and colleagues, writing from the perspective of complex systems science — examined the theoretical foundations of the entire enterprise and concluded that:

… even in the ideal case that a perfectly functioning computational framework was technologically accessible, the nonlinear dynamics of interacting biological units lead to emergent phenomena that cannot be simply simulated or predicted.

The review was about all digital twins in precision medicine, but its conclusion applied with particular force to glucose homeostasis, where the system involves not just molecular and cellular interactions but the full weight of the human exposome — stress, sleep, diet, exercise, illness, emotion — factors the authors described as “almost impossible to replicate or reproduce.”

They weren’t alone. A 2023 National Academies workshop on digital twins in biomedical research reached a similar conclusion: the technology showed promise for narrow, constrained subsystems, but scaling to whole-body physiology remained out of reach. One participant urged researchers to deprioritize AI and big data in favor of mechanism-based models — which is precisely what the AID field has not done.

And in a 2024 paper in PNAS Nexus, Emmert-Streib and colleagues drew an explicit distinction: complex adaptive systems “provide primarily a qualitative understanding, rather than detailed quantitative predictions,” which “distinguishes them from digital twins” that require numerical accuracy to be actionable. The glucose-insulin system is a complex adaptive system. The simulator treats it as if it isn’t.

Meanwhile, Heman Shakeri — a complex systems researcher now on Kovatchev’s own faculty at the Center — had just received a $4.7 million grant from Breakthrough T1D and the Helmsley Charitable Trust to develop next-generation, fully closed-loop AID algorithms that eliminate the need for meal announcements entirely. They’re funding the same person who wrote, The Driver-Blindness Phenomenon: Why Deep Sequence Models Default to Autocorrelation in Blood Glucose Forecasting, where his team found that:

Deep learning models for glucose forecasting “consistently fail to leverage clinically informative drivers—insulin, meals, and activity—despite well-understood physiological mechanisms.” The gain from adding these variables is “typically near zero.” They attribute this partly to “physiological heterogeneity that undermines population-level models.“

The funding comes from the same organization that had redirected the field’s trajectory two decades earlier—BreakthroughT1D. And they are now funding a complexity-aware scientist to keep working within a paradigm that he himself said complexity theory says can’t work.

Epilogue

As for Jeffrey Brewer: he wasn’t wrong to be appalled.

If you’ve ever watched a child — your child — wake up shaking at 3 AM with a glucose of 42, or sit out of a birthday party because the timing doesn’t work with their insulin, you don’t think about complex adaptive systems theory. You think: there has to be a better way.

And if you’re an engineer, you think: I can build it.

That instinct isn’t foolish. It’s human. And, I can speak to this in two ways. When I was diagnosed in 1973, my mother cried for a week. She got involved with the JDF (Juvenile Diabetes Foundation), which is when we first heard those immortal words, “There will be a cure in five years”.

Later, when I received my degree in computer science with a focus on AI in 1985, I wanted to work in medicine one day. That day came when I founded a biotech company in 2011 that made a molecular sensing platform for a multitude of biomarkers. It was then that I learned about the complexity of biology, a completely different paradigm from software engineering—or any kind of engineering. My company had three arms: hardware, software, and biochem. The mental models of these three groups were so disparate it was like watching people who don’t speak the same language try to agree on where to go for lunch

So, I get it at all levels. As a parent, and as an engineer. But what most engineers who don’t learn and appreciate unless and until they learn biology, is that they see every problem as a deterministic, calculable, solvable problem. And biology—glucose metabolism—is not, and that category error is what lures so many well-intentioned technologists down a rabbit hole.

The pursuit of a system where a T1D never has to pay attention is the way engineers frame the problem—and what desperate parents hope for. This is why full page ads, congressional testimony and heart-breaking stories are so effective.

It also makes the far more modest, achievable and successful task of keeping the user in the loop increasingly more difficult. Every study, across every system, finds the same result — the more engaged users are with their own management, the better their outcomes. It doesn’t have to be overwhelming—even modest engagement outperforms automation every time. The technology works best as a tool you use, not a system you surrender to.

And that’s when the healing begins.

Subscribe now

This year, 2026, marks ten years since Medtronic published the results of the first clinical trial that used a hybrid closed loop (HCL) automated insulin delivery (AID) system, which was conducted in 2015. The “hybrid” part means that the user inputs data for meals. The group’s mean A1c was ~7%.

Medtronic’s chief medical officer declared that the system “should be safe and more effective than what the person with diabetes can do on their own.”

Ten years later, with multiple revisions of the algorithms, and five commercially available AID systems on the market, independent researchers have conducted trials to compare how all five perform: A comparative study from Ohio State (2024), the CLEAR trial at Cambridge, a crossover trial at McGill, and a January 2025 overview in diaTribe. They converge on the same conclusions:

• HCL (hybrid closed loop) with human input (70% TIR): ~7.0% A1c

• FCL (fully closed loop) without human input (50% TIR): ~8.0% A1c

(Note that TIR (time in range) maps to A1c levels via a formula that Beck et al. published in Diabetes Technology & Therapeutics, 2019, where they analyzed 545 adults with T1D across four randomized trials.)

Ten years. No progress.

Sure, but at least it must have been better than the old style of multiple daily injections, right?

Nope. In the 2025 paper, Shining the Spotlight on Multiple Daily Insulin Therapy: Real-World Evidence of the InPen Smart Insulin Pen (Diabetes Technology and Therapeutics), 3,793 T1D users on smart insulin pens with CGMs achieved 70.3% TIR — an A1c equivalent of 7.0% — when they dosed at least three times daily.

And InPen is owned by Medtronic. Yeah, them.

Perhaps unsurprisingly, it’s been this way since before automation as well. A research study by the T1D exchange from 2018, which compiled data from US-based clinics prior to the availability of automation, and found that users over the age of 26 saw nearly identical A1c levels between pump users and those on MDI: ~7.3%.

So, before automation: 7.3%. Ten years after automation: 7.0%.

But there is one thing that automation has done that’s worth recognizing. Look at the data from that 2018 study, and notice the A1c levels for those under the age of 26.

Those on the younger side had higher A1c values, and it’s exactly those people who benefit greatly from automation. People who either cannot or will not engage in self-management can see their A1c levels drop from 9%, 10%, and considerably higher, all the way down to 8%. It’s that subgroup of people who are the main beneficiaries of AID systems and that you see cited in most papers published by pump companies and others that promote these devices.

And this highlights something often hidden in the details of all these studies: the great variability among subgroups. Those who come from distressed socioeconomic backgrounds have less access to healthcare and resources, and less support from family/friends. People in this group are less informed about good self-management techniques, and are less engaged. So, they tend to have worse A1c and TIR percentages. Those who come from more stable environments, access to resources, and are educated about self-management are also more engaged with their disease. So, they tend to have better A1c and TIR numbers.

The point is that human engagement yields better numbers and healthier outcomes, regardless of method of insulin administration. This has been true since insulin was discovered in 1929.

But the promise of automation was supposed to break that barrier—that diabetes could be put in the background, that people could finally be relieved of the daily burden of self-management. Yet, after ten years of these systems, that hasn’t happened. WHY?

Perhaps we should start by asking a more fundamental question that no one in the development cycle ever asked: Is this problem even solvable in principle?

Complicated Systems

Before mathematicians attempt to solve a problem, they first have to determine what type of problem they’re trying to solve. This isn’t philosophical hand-wringing, it’s the foundation of serious scientific work. Some problems are merely difficult; others are impossible. Knowing the difference before you start saves decades and billions of dollars. And in this case, lives.

Since we’re talking about controlling glucose levels using a CGM and an insulin pump, we’re in a domain of science called control theory. That is, there’s a sensor that detects glucose levels, and there’s a controller (an algorithm) that delivers insulin. There’s a cause and effect at play, and that interplay is the core thesis of control theory. In fact, we engage with automation everyday.

When you’re driving a car and flip on the cruise control, an algorithm determines the speed of the car and the amount of gas being delivered. These are the sensor and the controller. By constantly monitoring the speed, the algorithm knows just how much to add or let up on the gas. The hills, valleys, and other road conditions may cause more or less gas to be applied (or the brakes), but this is the closed loop.

Now think of a fully automatic self-driving car. It has lots of sensors, lidar, cameras, sonar and other gadgets around it, constantly monitoring the environment. The controllers are not just gas and brakes, but the steering wheel.

Now think of an airplane, or the telephone system, or an oil refinery. Big and complicated. Indeed, they’re a system of independent parts. To build a controller, you have to start with statistical models from physical properties of the parts of the systems you’re observing and controlling. Those parameters follow what is called a lognormal distribution — a pattern where values can’t go below zero, most readings cluster in a typical range, and occasional high values stretch out in a long tail.

You know that you can control the systems if the behaviors of the parts are stable, stationary and independent of one another. Thus, engineers can characterize the system mathematically and build controllers with known performance bounds. It applies to the car, the airplane, telecommunications systems, oil refineries, and more.

The system’s statistical signature is a property of the system itself, not of its context. You can identify the transfer function, it stays identified, and your controller works tomorrow the same way it worked today.

With this, you can create a simulator that looks, acts, and can be reliably used to resemble your system. Indeed, these simulators are important because you want to test a jet engine before putting in a billion dollar airplane. Or putting living passengers in it.

These simulators exhibit a property of complicated systems: They’re predictable, which is also a way of saying they’re tractable. You can not only calculate each of the properties, those calculations allow you to predict future states of the system.

Many medical devices fall into this category. The canonical examples include pacemakers, cochlear implants, ventilators, and even the NeuroPace responsive neurostimulation system for epilepsy. In each case, the control task has the properties you’d expect of a complicated system: the relevant biomarkers are directly measurable with minimal lag, the input-output relationship is well-characterized, the physics are stable, and the “actuator” operates on the same timescale as the disturbance. A pacemaker detects an arrhythmia and delivers an electrical pulse within milliseconds. The system it’s controlling — cardiac electrical conduction — follows predictable electrophysiology.

Closed-loop AID algorithms were designed with the same engineering logic: If you look at CGM data, it sure looks like a lognormal distribution: Glucose values can’t go below zero, they cluster in a typical range — say, 100-200 mg/dL — and there’s a long tail that goes higher. And this is the basis for assuming it has the properties for control: A sensor, a controller, and a lognormal distribution of data. Indeed, we can simulate different physiologies from huge datasets.

But there’s one hiccup here: In glucose control, the “parts” are not independent. They’re integral to one another. These parts include regulatory and counter-regulatory hormones, cortisol, glucagon, and a litany of other pathways that affect glucose levels, and each other, which feeds back to glucose levels. The way each component behaves affects the others in ways that no longer make them “stable and predictable.” This is an entirely different category from the complicated system—it’s a complex system. (So, it’s less of a hiccup as it is a loud, audible belch.)

Complex Systems

When a complicated system turns into a complex system, it’s a new category of system, where it may well be calculable and simulatable, but not predictable, and therefore, not controllable. It is a central characteristic of chaos theory. Glucose homeostasis exhibits the defining properties of those systems—sensitive dependence on initial conditions, non-stationarity, path dependence, and unmeasurable state variables (all those hormones and other physiological properties that affect glucose levels). Therefore, the theoretical ceiling on prediction is a property of the system, not a limitation of technology.

To see how a complicated system becomes a complex system, consider the pendulum on a grandfather clock. Completely calculable and predictable. You can calculate its position at any point in the future. Engineers solved this in the 17th century.

Add a second pendulum at the bottom of the first pendulum, and watch what happens: As the main pendulum swings, it causes its extension to swing, and that swing feeds back onto the motion of the first pendulum. By definition, the integration of the way the parts affect one another turns it into a completely unpredictable, chaotic system.

Same physics, same materials, one additional hinge. Two double pendulums released from what looks like the same position will diverge within seconds into completely different paths. No equation can tell you where any part of the larger system will be seconds from now—not the tip, or the base joint, or anywhere in between. This type of system is intractable. And it’s entirely because the parts are not independent of one another. It’s the feedback loop that makes each part itself change its own behaviors.

Nothing got “harder” in the engineering sense. Nobody added a new force of nature. It’s the same gravity, same joints, same metal arms. What changed is the category of system—from tractable to intractable. And no amount of better measurement, faster computation, or more precise manufacturing will make it predictable again.

Here’s where things can get insidiously deceptive that engineers need to be highly aware of or they’ll fall into a trap: Because you know the physics of all the parts, you can still simulate the system. You can take a million videos of a double-pendulum, collect all the statistics—all of which will look lognormal in their distributions—thereby allowing you to write a simulator that models a double pendulum. More deceiving is that the simulator will indeed exhibit a behavior that could very well be one of an infinite number of possible behaviors that could happen.

And this is what it all comes down to—there is no chance that the simulation can be used to predict the behavior of a real double-pendulum that looks like it’s starting with the same initial conditions. You can read how to do it and watch animations of simulations on the Wikipedia page. The math is extraordinarily complex — almost worth looking at the page just to appreciate that — but the ability to simulate a system’s behavior has nothing to do with the ability to predict its outcomes. And if you can’t predict the states of the parts, you can’t control them.

The problem with trying to predict glucose—and this was the category error that started all this—is that it appears to have a lognormal distribution of data. So engineers treat it like a complicated, controllable system. The CGM is the speedometer, and insulin is the “controller” that acts on it. Everything else is treated like “disturbances” — not as integrated parts, but as separable parts. You can collect zillions of hours of CGM data and create simulations of what appear to be real humans, then try to map those simulations into an algorithm that controls glucose levels and you get… No difference. The noise in the chaotic system just overwhelms the signal.

So how about this idea: Rather than control the whole system, why not tune some small aspect of a subsystem. Maybe to reduce hypoglycemia at night, perhaps? At least many of the variables are more stable—you’re not eating, you’re not physically active, you’re basically in the most “stable” state you can be in.

But again, even that is not possible because, unlike a double pendulum where the parts are fully observable, glucose metabolism has parts you can’t even see, let alone measure and/or simulate, even at night: The body moves between the sympathetic to parasympathetic nervous system, muscles and other tissues take up glucose through non-insulin mediated pathways, glycogen stores are replenished (without insulin mediation), and dozens of other activities are still swirling around.

Collectively, the interdependent parts have such an enormously disproportionate effect on glucose trajectories and rates of change—whether awake or asleep—having an algorithm try to do anything sensible is like having cruise control keep your car at a fixed speed while a hurricane lifts your car and drops it into the next town.

If you’re not overwhelmed yet, there’s another aspect to human biology that takes it yet another step further into complexity: it adapts. And that leads to the grand theory that’s been established for decades in the field of biology: complex adaptive systems.

Complex Adaptive Systems (CAS) Theory

The science of complex adaptive systems (CAS) recognizes that biological systems self-organize and follow their own internal rules. Interventions, such as medications or even psychological awareness, can have effects that may meet some objectives, but not all, and not predictably, and not for everyone.

Most importantly, biological systems adapt—they behave differently as conditions change. When you’re sick, hormones react and respond differently; if you’re pregnant; when you grow from childhood to young adult to middle age and beyond. Your physiology is never constant long enough for the system to be “stable”.

Even day-to-day physiology varies dramatically.

A useful illustration comes from a Stanford group (Hall et al., PLOS Biology, 2018) that used CGMs to monitor glucose variability in response to different kinds of foods. In a standardized-meal sub-study, a subset of 30 participants ate three predefined breakfasts (including cornflakes-and-milk) on two separate days each. Even with that kind of controlled input, post-meal glucose excursions differed markedly across individuals — as well as within the same individual. Below is a figure from that study:

A larger Israeli study from the Weizmann Institute (Zeevi et al., Cell, 2015) pushed the same point, but at a larger scale. The investigators used week-long CGM data in an 800-person cohort and captured tens of thousands of meal responses (46,898). They found high inter-individual variability in postprandial glucose response—even to meals that were ostensibly identical.

Another study titled Imprecision nutrition? finds the same thing: Same meal, same person, two occasions. Within-subject correlations were only r = 0.46 (Abbott) and 0.45 (Dexcom), with intraclass correlations of 0.28 and 0.17 — the authors conclude responses to duplicate meals were highly variable and that CGM-based personalized advice requires aggregated repeated measurements.

You may be reading that and think that you’ve experienced the same thing. Heck, T1D is crazy, right?

Here’s the kicker you didn’t expect: ALL of the subjects in these studies were healthy individuals with no diabetes diagnosis. They were screened to have no metabolic disease, they had low A1c, no physical disabilities of any kind. They were, as much as the study could determine, the perfect specimens of human health.

You can pick up your jaw off the floor now.

Now let’s look at T1Ds:

• Children with T1D, six controlled standardized breakfasts — Diabetes Research and Clinical Practice (https://pubmed.ncbi.nlm.nih.gov/34836409/). Within-person variability in time to peak exceeded between-person variability in all three conditions: controlled breakfast 18.5 vs 38.9 min, free-living breakfast 14.1 vs 49.6, free-living dinner 5.7 vs 64.5. Same population, controlled meals, and the within-person SD is more than double the between-person SD.

• Bozzetto, 61 T1D subjects on pumps, same meal on three separate occasions. Reproducibility is good enough to characterize a person in aggregate and nowhere near good enough to dose a single event.

The takeaway is that glucose dynamics are more complex than ever assumed. The wild glucose excursions are the body itself—a healthy one at that—trying desperately to stabilize glucose levels. These are the properties of a complex adaptive system, and are the basis for CAS Theory, which has been explored in medical literature for decades. Examples include these citations:

• The challenge of complexity in health care (BMJ, 2001). Here, the authors state that the “behaviour of any complex system is fundamentally unpredictable”. I.e., one cannot model glucose as a superposition of linear transfer functions.

• From Crossing the Quality Chasm (Institute of Medicine/NCBI Bookshelf), the section Redesigning Health Care with Insights from the Science of Complex Adaptive Systems recognizes that biological systems are complex adaptive systems where detailed algorithmic prediction is “fundamentally not possible.”

• The Paradox of Intervening in Complex Adaptive Systems (PMC, 2018) directly addresses the paradox of trying to control systems that, by definition, resist external control.

Even data scientists are weighing in on this. A 2025 paper by Heman Shakeri, Assistant Professor of Data Science at University of Virginia, addresses this in his paper, The Driver-Blindness Phenomenon: Why Deep Sequence Models Default to Autocorrelation in Blood Glucose Forecasting:

Deep learning models for glucose forecasting “consistently fail to leverage clinically informative drivers—insulin, meals, and activity—despite well-understood physiological mechanisms.” The gain from adding these variables is “typically near zero.” They attribute this partly to “physiological heterogeneity that undermines population-level models.“

Unfortunately, the automated insulin delivery field has never engaged with this body of theory. Nowhere in the AID literature will you find serious discussion of complex adaptive systems, irreducible uncertainty, or the theoretical limits of algorithmic control over biological processes. Instead, you find engineering-style approaches, using control theory as a foundation for attempting to make these algorithms better.

The AID field has spent fifteen years and billions of dollars trying to build better grandfather clocks that have multiple pendulums, wondering why they can’t keep time.

And that’s why the algorithms never get better. The properties of a chaotic system will never permit it.

But it raises the obvious observation: If prediction doesn’t work, how are these systems working at all? How is it that people are ever in range? The algorithms must be doing something right. Right?

Indeed—let’s take a look under the hood at what the algorithms are actually doing.

Painted Into a Corner

If CAS theory tells us that glucose can’t be predicted or controlled, then what are these algorithms actually doing? And how is it that people are ever in range?

To answer that, it helps to understand just how constrained the engineering problem actually was — not because the engineers lacked skill, but because the system they were building left them almost no room to maneuver.

To illustrate, we’re going to use Tandem’s algorithms, but the same principles apply to all of them. And this is evident from the common A1c/TIR outcomes we see from clinical papers that publish these findings.

So, we begin at the low end of the glycemic range, where the floor was already high. Consider what happens with the simplest possible automated intervention: suspending insulin when glucose is dropping toward hypoglycemia. That’s all that Tandem’s Basal-IQ does — a process called Predictive Low-Glucose Suspend (PLGS). No dosing algorithms, no correction boluses, no meal detection. According to Tandem’s own study, users on Basal-IQ achieved a TIR of approximately ~64%. That single safety mechanism — stop delivering a dangerous drug when the patient is heading toward a dangerous low — captured the majority of the achievable benefit. Everything above that 64% floor is the territory where the actual algorithm operates.

And it feels like there’s a lot to go, right? Why not 90-100%? That’s where reality gets in the way.

First and foremost, SAFETY. This is where the ceiling is structurally low. To get tighter control, your model may suggest infusing insulin at times that may not be warranted, entirely because you can’t detect any of those pesky, unseen, unreliable hormones that may be affecting glucose levels. The algorithm might suggest a 5u bolus for dinner, but 30 minutes later, the person takes a post-dinner walk, and suddenly, they’re admitted to the ER. These unforeseeable aspects forced the potential for better control to be lowered by a lot. Add these up over time, and the ceiling on TIR drops.

Another aspect of safety is the reliance on CGM data, which is highly subject to unreliable variability. This is not because the CGM is inaccurate, per se. Glucose is highly volatile in the body, and detection comes with wide error bars. CGMs can be out of calibration, or just show readings that are dramatically out of alignment with physiological reality.—something that happens frequently when levels are rapidly moving up or down. An algorithm can’t just believe CGM data as gospel—it has to factor in that readings may be faulty, especially if the algorithm would indicate substantial correction doses. Indeed, overdosing insulin due to faulty high readings is not uncommon. These are all covered in detail in my article, Continuous Glucose Monitors: Does Better Accuracy Mean Better Glycemic Control?. Taking all this into account, and you shave off several more potential TIR percentages.

Safety demanded conservatism. Boris Kovatchev — the mathematician whose risk analysis framework underlies these systems — established decades ago that errors toward hypoglycemia are asymmetrically dangerous. A glucose of 40 mg/dL is a medical emergency; a glucose of 300 mg/dL is not. Any system that delivers insulin must be biased heavily toward under-dosing rather than over-dosing. This isn’t a design flaw — it’s the mathematically correct response to the risk landscape. But it necessarily caps how aggressively the algorithm can act, which translates to lower TIRs.

Between the high floor, the low ceiling, and the safety constraints, the operating space for any algorithm — no matter how sophisticated — was remarkably narrow. And, again, this applies to all algorithms.

This does not diminish the engineering that went into capturing those points. Kovatchev’s risk-asymmetric framework, individualized insulin sensitivity parameters, and a 30-minute prediction horizon based on CGM trend data are all the byproduct of years of simulation, clinical testing, and iterative refinement. The people who built it are serious scientists solving a hard problem with real ingenuity.

But here’s what the constraints actually produced. We can see it in the article, The Hybrid Closed-Loop System Tandem t:slim X2™ with Control-IQ Technology: Expert Recommendations for Better Management and Optimization:

• Below 70 predicted → suspend insulin

• Below 112.5 predicted → reduce basal

• Between 112.5 and 160 → deliver user’s programmed rate

• Above 160 → increase basal

• Above 180 → deliver 60% of a correction bolus, max once per hour

That’s what the algorithm reduces to once all the constraints bind. Not because the underlying engineering is simplistic, but because the operating space left almost no room for the sophistication to express itself. It’s as if you trained a concert pianist for twenty years and then asked them to perform in a room where the piano has five keys.

The 60% correction factor tells the story. If the system had genuine predictive confidence, why deliver only 60% of the calculated dose? Because the designers understood — correctly — that the prediction was unreliable, and the consequences of over-delivery were potentially fatal. That 40% haircut is Kovatchev’s risk framework in action: the system is deliberately conservative because the asymmetry of harm demands it. Tandem Diabetes Care itself describes this approach openly.

The results confirm the narrow operating range. In the study One Year Real-World Use of the Control-IQ Advanced Hybrid Closed-Loop Technology, approximately 9,000 users went from a baseline TIR of ~63% (on Basal-IQ) to about 74% on Control-IQ’s full system. Other studies converge on similar figures: roughly 70% TIR across all available AID systems. The entire contribution of the algorithm — all that engineering, all that science — accounts for roughly 8–12 percentage points of TIR improvement, depending on the population.

And notice how much of this is driven by the user, not the algorithm. In the paper Therapy Settings Associated with Optimal Outcomes for t:slim X2 with Control-IQ Technology in Real-World Clinical Care, researchers found that 67% of Control-IQ users had less aggressive settings than clinical guidelines recommend — meaning the system is already operating conservatively, and most users make it even more conservative. Nearly everything that determines how the system behaves — basal rates, correction factors, carb ratios — is user-settable, as it should be. The algorithm operates within those human-defined parameters, not instead of them.

Were the simulators necessary? Was any of this really necessary?

The safety concerns, plus the heuristics in the Control-IQ algorithm, prompts the obvious question: Was any of this necessary? Did we really need decades of research to come up with these algorithms? Could someone have arrived at essentially the same AID algorithm without any of the control theory scaffolding — without Bergman’s Minimal Model, the UVA/Padova simulator, Kovatchev’s MPC framework, any of it?

The answer is yes, and there’s direct evidence. Dana Lewis and Scott Leibrand started OpenAPS in 2014-2015. Neither was a control theorist. They didn’t build from the Minimal Model or the simulator. They started from the practical problem: “I have a CGM reading, I have a pump, what rules should govern insulin delivery?” From there, they built heuristic systems (oref0 and oref1) described explicitly as reference designs based on observable behavior, not physiological models. The outcomes are comparable to commercial systems. The Tidepool Loop FDA clearance used Control-IQ as its predicate device precisely because the FDA recognized functional equivalence.

The punch line sort of hurts: The entire intellectual apparatus — decades of physiological modeling, the simulator, the regulatory shortcut, hundreds of millions in R&D — produced algorithms that two people with no background in the underlying science replicated by empirical tuning. That’s not a knock on Lewis and Leibrand (it’s impressive). It’s evidence that the “algorithm” was never where the value lived. The operational space is so narrow that anyone reasoning from the same constraints — one input variable, one actuator, catastrophic downside risk — would land in roughly the same place.

The ceiling isn’t a function of which approach you take. It’s a function of what the system allows.

The Biggest and Best Variable: The User

This brings us to a painful irony. Any algorithm works best when the user conforms their lifestyle to meet the algorithm’s expectations: consistency and low volatility. And sadly, that’s a very unhealthy lifestyle.

A 2022 review of AID’s psychological implications — Nefs, Frontiers in Clinical Diabetes and Healthcare 2022 — documented exactly these dynamics. Users weren’t passively accepting the algorithm’s output. They were actively restructuring their lives to produce the numbers the algorithm rewarded: eating predictably, avoiding unannounced snacks, timing activities to coincide with low insulin-on-board, and — most consequentially — reducing or eliminating exercise that would introduce glucose variability the algorithm couldn’t handle.

Imagine a concerned parent telling their child when they can do certain things, like limiting physical activity or carbohydrate intake or just playing with their friends at a party — all with the aim of increasing time in range.

The behavioral adaptation produces TIR numbers that make the algorithm look effective, but the causality is partly inverted: the users are doing the work, and the algorithm is taking the credit. And the saddest part about this — the moral hazard — is that the user is doing the least healthy behaviors, all in the hopes of optimizing their TIR and A1c levels.

Shockingly enough, companies are starting to use this as a feature. In a press release, Diabeloop’s DBLG2 receives FDA and CE clearance for automated insulin delivery without meal input, the company says the quiet part out loud: removing meal announcements costs 3-6 percentage points of time-in-range, but they’re marketing it as a feature because it “reduces cognitive workload.” That’s behavioral conformity playing out in real time — except now the manufacturer is explicitly trading glycemic control for disengagement and calling it innovation. Diabeloop just got FDA clearance to remove the single most important point of user engagement: the meal bolus. And the FDA said yes.

The health consequences of this conformity have been appearing in medical literature since closed-loop systems were introduced. In my article, Medical Literature Analysis: The Performance Paradox of AID Systems, I cover ten years of research — much of it from journals published by the American Diabetes Association — showing that AID system users gradually develop most or all of the following: weight gain, loss of agency, de-skilling of self-management, limitations on efficacy, quality of life, economic impact, and the “attractive nuisance” of disengagement from self-management.

And if the user is entirely disengaged? TIRs hover around 50%, as the algorithm operates in what is effectively fully closed-loop mode — confirming that without human input, the system alone can’t get much past the safety floor.

Oh, and if we’re going to talk about engaged users, we have to recognize the DIY/looper community. They tend to do better than commercial AID systems because they can personally customize their algorithms to match their personal physiologies. This may work well for them, but are not generalizable to the broader population.

Personalizing your algorithm is the modern-day equivalent of MDI—you choose the dosing (basal and bolus) to match what your empirical experience has shown works. This is how T1Ds have always managed themselves, whether MDI or pumps.

The point is, the engaged user knows what works for them, and that should always be encouraged. T1Ds have always had the freedom to choose how much to dose at any given time, whether that’s done manually via MDI, or by personally tuning the algorithm. Whatever floats your boat.

Epilogue

This article took you on a very long journey, meandering through discussions of control theory, complicated vs. complex systems, and the rigorous nature of building physiological models that could simulate real diabetics.

And yet, two guys built mostly the same thing using simple heuristics.

So, why didn’t I just say that in the beginning and save you the extra fifteen minutes needed to learn about control theory and multi-arm pendulums?

I’m reminded of the scene in The Wizard of Oz, where the Scarecrow asks Glinda why she didn’t tell Dorothy that the ruby slippers would take her home when she arrived in Munchkinland in the first place. Instead, she went on this wild journey with flying monkeys, a dysfunctional Wizard, and a wicked witch that wouldn’t really be interesting until the sequel 75 years later. Why, Glinda, didn’t you just tell Dorothy the truth in the first place?

Glinda responds: “Because she wouldn’t have believed me. She had to learn it for herself.”

Part 2 of this series will cover how we got here: The story about how a young biomedical engineer who was looking to understand the progression of type 2 diabetes eventually found his models ending up in a system to control insulin delivery, a tool that it was not designed to do. And how the JDRF and Medtronic worked with the FDA to approve the first simulator that led to the first HCL. It was a big beautiful bet that was placed with good intentions, but contradicted the basic principles of biology and the medical system.

Share

Subscribe now

I recently came across the word, glossolalia, which means,” the phenomenon of speaking in an unknown language, especially in religious worship.” You can imagine someone using ancient Latin, Hebrew or Arabic in religious contexts, and yet, not really knowing the literal translation of the phrases.

This term came to me when I observed how the phrase “Standard of Care” is used in advertising and other promotional material in the realm of type 1 diabetes. It turns out to be far different than conventional understanding. And the use of it is increasingly becoming more of a marketing term, much like how MARD has become for CGMs.

The original intent of MARD was to measure the “accuracy” of a glucose meter, but because it’s so easily manipulated—and decidedly does not correlate in any way to actual clinical outcomes—it has not only lost its actual value as a measuring instrument, it’s become a marketing tool that brings no value to those with diabetes. (I cover this extensively here and here.)

“Standard of Care” has become a victim of this as well, but the effects go way beyond just technologies; it applies to all sorts of recommendations, from diets to diagnostic measurements, like A1c levels, lipid measurements, and more.

This article aims to address:

• What “Standard of Care” actually means

• Who gets to declare it

• The factors that go into the rating system

• Who the intended audience is

• What are the legal ramifications (and legally, the term carries very different meaning and weight)

Throughout this article, it’s important to understand the most important truism of medicine: Every single medical technology, drug, diet, and other interventions will affect people differently. There will always be those who benefit and those who don’t.

“Standard of Care” aims to take this complexity and nuance into account by weighing the pros and cons of benefits and risks derived from credible studies and trials published in peer-reviewed medical journals. Because there are those who benefit (or not) differently, the “rating system” may seem inappropriate. If there are nuances, why give something a specific rating?

The short answer is that the audience is not your average consumer. It’s clinicians. The intention is that clinicians treat ratings as a starting point, then consider the exceptions and carve-outs to determine whether the intervention is appropriate for a given patient. With very few exceptions, the term “standard of care” is rarely ever a simple statement of support with no carve-outs.

Most people don’t know this, and as such, have very different—and often dangerously misleading—understanding of whether an intervention comes with potential risk. This article is intended to help you navigate this tumultuous process.

We’re going to use automated insulin delivery (AID) systems as an example because it has become a very hot topic within the T1D community, especially as multiple forces are emerging in both the technical and medical realms. But as you read this, be aware that it also applies to everything else you think is “established fact” about T1D management, whether it’s diets, insulin dosing methods, dosing calculators, and so on.

AID Systems and “Standard of Care”

As noted above, “standards of care” aims to weigh the pros and cons of benefits and risks derived from credible medical journals. As such, I recently posted an article that did just that for AID systems. It was titled, Medical Literature Analysis: The Performance Paradox of AID Systems. The article cited over ten years of medical literature that pointed to an emerging trend about who benefits and who doesn’t from these systems. I used the term “The Barbell Distribution of Clinical Benefit”. I said:

The true beneficiaries of AID systems are the far outliers in T1D demographic profiles—those who are either entirely unable to manage themselves, or are hyper-engaged in self-management—leaving those in the middle as either not realizing any material benefit from automation, or may be potentially harmed by it.

This drew responses from some who said, “But AID systems are Standard of Care”, and they cite two sources:

• Medtronic’s website states, “the American Diabetes Association calls automated insulin delivery systems the standard of care for people with type 1 diabetes.”

• BreakthroughT1D also makes the claim explicitly on seven discrete pages on their website, but also in congressional testimony, annual reports, and advocacy materials. Notably, none of these cite the ADA as the source—they simply declare it as fact.

This is a classic case of glossolalia. Both institutions are using a term without proper context or appropriate attribution.

Yes, the ADA does recommend AID systems, but with very specific carve-outs that have important notations for clinicians about the appropriateness for different people and conditions. It’s those exceptions—and there are a lot—that my article highlights: weight gain, loss of agency, de-skilling of self-management, limitations on efficacy, quality of life, economic impact, and the “attractive nuisance” of disengagement from self-management. The literature also points to long-term effects, such as a sedentary lifestyle, which leads to a cascading series of metabolic disorders.

For this research, my citations included studies, trials, and meta-analyses from the following journals:

• ADA journals (Diabetes Care, Diabetes, Clinical Diabetes): Citations: 9

• Diabetes Technology & Therapeutics (Mary Ann Liebert, not ADA but diabetes-focused): Citations: 4

• Lancet Diabetes & Endocrinology: Citations: 2

• Diabetologia (EASD journal): Citations: 1

• Non-Diabetes Journals: Citations: 10

  • Endocrine (Springer)

  • Social Science & Medicine

  • JAMA Internal Medicine

  • PMC ethics reviews

  • General medical/health economics journals

And many of those articles are “literature reviews”, which are single articles that examine hundreds of other papers and distill their findings into an overview paper. In all, I covered medical literature that spans over ten years, covering hundreds of thousands of T1D patients.

Perhaps most importantly, which I even stated in the article, virtually everything outlined in my article is not just published in ADA journals and others, but the ADA itself published in its own position statement with its European counterpart: Automated insulin delivery: benefits, challenges, and recommendations. A Consensus Report of the Joint Diabetes Technology Working Group of the European Association for the Study of Diabetes and the American Diabetes Association.

In a way, I didn’t really need to publish my article at all—the above document comes to the same conclusions. The difference, however, is that their document is a highly dense tome of clinical language that is tough to read, at least, not without a glass of cognac and a roaring fire. It’s a foreign language to the lay audience.

My target audience is not steeped in the genre of complex medical literature. Hence, my independent analysis.

The natural question then becomes: If the ADA recognizes all these problems, how/why would they “recommend” AID systems?

And what about BreakthoughT1D’s claims? And Medtronic?

Let’s start with the ADA.

Understanding What the ADA Actually Says

The ADA’s Standards of Care is a 350+ page document updated annually, covering everything from diagnosis to complications. For technology recommendations, Section 7 (Diabetes Technology) and Section 9 (Pharmacologic Approaches) are most relevant to the discussion on AID systems. Each carries a letter grade indicating the strength of evidence behind it:

• A = Clear evidence from well-conducted, adequately powered randomized controlled trials (RCTs).

• B = Supportive evidence from well-conducted cohort studies.

• C = Supportive evidence from poorly controlled or uncontrolled studies.

• E = direct evidence is lacking, so clinical Experience weighs in.

These grades are explicitly about strength of evidence, not a universal instruction. They inform judgment; they do not replace it. In fact, the ADA is very explicit that their ratings are based on the strength of the evidence, not the strength of the recommendation. An A-rating may have multiple rigorous trials that support it, but there are other considerations that should be taken into account.

Here’s where it gets complicated for AID systems. The ADA makes several related but distinct statements:

Section 7.8 covers “...early initiation, including at diagnosis, of CGM, CSII, and AID depending on a person’s or caregiver’s needs and preferences.” This carries a C rating—meaning the evidence for early initiation comes from weaker or poorly controlled studies, not large RCTs.

Here, the “C” rating is the lowest rating that has studies associated with it, so a physician needs to be diligent. Recently diagnosed patients can have very different profiles: A three year old child, or a 45 year old LADA (late-onset autoimmune diabetes) who has a medical degree, or a 20 year old college student, or a 60 year old priest. These are all distinctly different people with different aptitudes for adopting an AID system. Hence, the need for medical literature to tease out what is actually appropriate. It’s more nuanced than just an ADA rating.

Moving onto section 7.25a, it says “AID systems are the preferred insulin delivery method over MDI, CSII, and sensor-augmented pumps in people with type 1 diabetes.” This carries an A rating—meaning rigorous trials clearly show AID systems achieve better glycemic metrics than alternatives. This is the part that most people (Medtronic, BreakthroughT1D) focus on—at least, in spirit.

But wait. Let’s zoom in on that statement: “perform better on glycemic outcomes” — yes, but better than WHAT?

Here, we look at the medical literature for context: Hundreds of studies that show that AID systems perform “better” when the same person’s A1c > 8%. For them—and there are a lot of these people—AID systems get an A rating.

But, the studies also show that for individuals whose A1c <8% before using the system, they see no observed clinical benefit. In fact, if their A1c is lower than 7.5%, they tend to drift upward to 7.5% (+/- .5%), and even drift higher. That’s why the clinician should use their “Experience” as to whether the device is appropriate for any given patient. It’s not a universal “you should prescribe an AID system” proclamation.

Compare all this language to the rating for CGMs, where Section 9 states flatly: “Its use is now considered Standard of Care for most people with type 1 diabetes”—there are no equivalent qualifiers about capability, carve-outs, or even ratings. No A, B, C, or E. It’s just declared “Standard of Care”. Yes, that’s the one rare exception.

The distinction matters. When Medtronic claims “the ADA calls AID standard of care” or when BreakthroughT1D states “AID systems are now standard of care” without qualification, they’re stripping away the individualization that the ADA explicitly pairs with its ratings. They take the A-rated finding (“AID performs better in trials”) and present it as if the ADA endorsed universal adoption—which the C-rated and E-rated guidance expressly does not support.

Ok, that’s the ADA. But what about BreakthroughT1D? They’re a well-recognized organization. Why can’t BreakthroughT1D act as an authoritative voice on this?

This is an important question, and that leads to comparing them with the ADA in their structure and other aspects that lead to what carries authoritative weight.

What is the ADA’s Authority?

The ADA (founded 1940) is a professional medical association whose members are physicians who pay dues. Those dues-paying members are the intended recipients of the ADA’s communications. The ADA publishes peer-reviewed journals (Diabetes Care, Diabetes, Clinical Diabetes, and others), and has established credentialing within the broader medical establishment.

The ADA’s primary function, as it were, is to address public policy in a manner that its members can use broadly and generally to address the entire population of people who have one of a variety of forms of diabetes. This is a highly complex disease, as readers of this article are sure to understand, so the challenge is to educate, inform, and otherwise set general guidelines for those members.

The ADA derives its influence from a combination of factors rather than any legal charter:

Professional legitimacy: Their Standards of Care document is developed through a Professional Practice Committee using systematic evidence review, graded recommendations (A through E based on evidence quality), and annual revision cycles. This methodological rigor creates credibility within the medical profession.

Institutional adoption: The real power comes from downstream adoption. When CMS references ADA standards for insurance coverage determinations, when hospital credentialing committees cite them, when medical boards use them to evaluate physician conduct, and when malpractice attorneys invoke them as the benchmark for reasonable care—the standards acquire functional legal force without ever being codified into law.

Payer alignment: Insurance companies routinely tie coverage decisions and quality metrics to ADA standards, creating financial incentives for adherence. In fact, this is exactly why insurance companies, from Medicare to private companies, don’t reimburse for AID systems unless the person would actually benefit from them. (As per my “paradox” article, Medicare and most insurers require an A1c > 7%, with some cases even higher. Remember, these devices are really expensive, especially compared to MDI, where many patients achieve the same outcomes, or even outperform them.)

There’s another very important aspect that brings even greater context:

• The target audience is health care providers (HCPs), who are dues-paying members of the ADA. Not the general public.

• Most HCPs are either not trained in diabetes management, or (even if they are) don’t have the time and/or experience to give focused, dedicated attention to individuals, due to their vast overload of patients under their care.

Wait, what? That last point seems important.

According to a report by The American Diabetes Association, “the vast majority of individuals with diabetes receive care in primary care settings—not endocrinologists—including physician offices, community health centers, clinics, and pharmacies.” A report from the National Institutes of Health shows that Primary Care Providers (PCP) expressed low self-confidence in managing T1D. The result is that more than 85% of insulin initiation recommendations originated from PCPs, according to a report by ​ScienceDirect.

And this problem is getting worse. As I explain in detail in my article on America’s healthcare system, the number of endocrinologists is at an all-time low and dropping fast, as medical students are not going into this field. According to Stability Health (2022), “34.2 million Americans live with diabetes, but there are fewer than 8,000 endocrinologists to treat them.” Of those, the 2026 data shows 2.1 million have T1D, according to the CDC.

The map below shows that over two-thirds of U.S. counties lack endocrinologists, according to aggregated data from GoodRx’s article titled, “Endocrinologist Deserts: A Critical Healthcare Gap for Millions in the U.S..”

Of those endocrinologists who do treat diabetes, the overwhelming majority treat Type 2 diabetes—which represents approximately 91% of all diabetes cases versus just 6% for Type 1. Given that T2D prevalence is roughly 20-25 times that of T1D, this leaves perhaps 15-25% of endocrinologists with meaningful T1D caseloads.

Put it all together: It’s a reasonable inference that 5-10% of all endocrinologists are truly proficient in modern T1D management (AID systems, CGM interpretation, exercise physiology, carb-ratio optimization).

With 2.1 million T1Ds in the US, chances are pretty high that your doctor is not in that 5-10% of endos that are truly experienced in T1D management.

This is why the ADA finds themselves in a tough spot. Their guidelines are addressing a problem that’s more akin to public policy: They are trying to provide a safety net at the bottom end of the health spectrum by streamlining healthcare guidance to physicians that are largely ill-equipped and inexperienced at T1D management.

Automated insulin pumps are not necessarily seen as the best way to manage T1D—as the evidence shows—but they do accomplish a very urgent need: Provide a quick and easy solution for medical professionals under stressed conditions, and who do not have the time or resources to learn the true nuances of management that they can teach to patients.

That brings us to BreakthroughT1D, who purports to be advocates for T1Ds. Shouldn’t they provide that additional education and support for T1Ds so they can learn how to better manage their disease? To be less dependent on automation that literature shows may not be suitable for everyone?

BreakthroughT1D (JDRF)

BreakthroughT1D (formerly JDRF) is structurally different from the ADA—it’s a research funding and patient advocacy organization, not a professional medical association. Their stated mission is to “improve the lives of people living with type 1 diabetes,” and they’ve done genuinely important work: approximately $2.5 billion in research funding since their founding in 1970, significant contributions to CGM development, and tireless awareness campaigns that have shaped public understanding of T1D.

But advocacy organizations and clinical standard-setting bodies serve different functions. BreakthroughT1D doesn’t convene panels of practicing clinicians to issue clinical practice guidelines through peer-reviewed processes. They don’t grade evidence on A-through-E scales. They don’t publish journals that require methodological transparency. When they make proclamations about “standard of care”, they’re engaging in advocacy, not guideline-setting.

There’s nothing necessarily wrong with advocacy of this nature, but patients need to be mindful of the source and strength of recommendation. Its Industry Advisory Panel—comprising Abbott, Dexcom, Insulet, Medtronic, Tandem, and others—explicitly shapes the organization’s “advocacy and research priorities.” According to its website, the IAP’s purpose is explicitly to shape advocacy priorities:

Breakthrough T1D’s Industry Advisory Panel (IAP) is a forum whose purpose is to develop an open dialogue and strengthen relationships with our industry partners. This includes sharing our development and organizational strategies, advocacy and research priorities, and marketing initiatives. In turn, the IAP provides members an opportunity to provide feedback on these updates as part of a collaborative and synergistic relationship with each IAP member.

The question is what countervailing process exists.

This raised a question I hadn’t planned to explore, but which is relevant to the larger point: what internal mechanisms does an advocacy organization have for stress-testing its public claims against the evolving clinical evidence, especially when its advisory structure includes manufacturers of the technologies it publicly advocates for?

Searching for context, I came across a 2016 article by Rebecca Robbins, who was a reporter for Stat News at the time (now at the NYTimes). The article, The risky bet behind the first ‘artificial pancreas’ for diabetes patients, chronicled how the JDRF (formerly known as the Juvenile Diabetes Research Foundation) went on a costly, and risky, campaign to enlist academic researchers, global companies, members of Congress, and even federal regulators to embrace the concept of a device that could take over much of the process of regulating blood sugar in patients with diabetes.

As the article stated, JDRF “spent roughly $16 million more to fund companies working on the technology,” while also receiving funding from the same companies, “including more than $5 million from Medtronic” (per the 2016 Stat News investigation). Current figures are not publicly itemized, and sponsorship “levels” (which are tiered by dollar amounts) are not the only way to contribute money. From that article,

Those close ties with industry have sparked some criticism — mostly out of concern that advocacy groups won’t be able to freely fight for patients if their finances are tied to the fate of a corporate partner.

Ten years later, as longer-horizon evidence accumulates, the same governance question becomes more important: how does an advocacy organization keep public-facing claims synchronized with the evolving clinical literature when its advisory structure includes manufacturers? There does not appear to be a formal, independent clinical-evidence review process comparable to a guideline body, and I’ve found no publicly documented independent clinical review board to evaluate the claims being made.

The concern isn’t that BreakthroughT1D receives money from device companies or that BreakthroughT1D doesn’t actually believe the narrative themselves. These are unrelated. Many organizations, including the ADA, also receive money from corporate sponsors, and at much greater levels. In fact, there has been considerable concern about this for years.

This is not an assertion of wrong-doing, nor that BreakthroughT1D is villainous. Advocacy organizations are supposed to advocate. They’re supposed to push for broader access to technologies they believe help patients. The problem arises when advocacy statements get confused with clinical guidance. It’s one thing to say, “we believe everyone should have access to AID systems” but it’s another thing entirely to say “AID systems are standard of care for everyone.”

The ADA, despite its own industry entanglements, maintains structural firewalls: a Professional Practice Committee with conflict-of-interest disclosures, an evidence grading system that distinguishes strength of evidence from strength of recommendation, and peer-reviewed journals that create accountability. When the ADA rates AID systems, it also adds caveats about individual circumstances—caveats that are documented in its own peer-reviewed medical journals and get cited in public discourse.

One more note worth considering: The FDA requires extensive disclosures when drugs are advertised—those rapid-fire lists of side effects you hear voice-overs say during a commercial. Drugs must also demonstrate efficacy in large-scale clinical trials before approval, and the FDA increasingly encourages (though doesn’t always require) comparative data showing how a drug performs against existing treatments. Medical devices like AID systems face a lower bar: they only require demonstration of safety and efficacy for their intended use.

There is no patient advocacy organization that is advocating for these changes, to educate patients on a variety of self-management methods, or to clarify the nuances that the ADA does.

Legal Considerations

Perhaps the most underappreciated aspect of the phrase “standard of care” is that it carries implied legal and institutional weight, even when it is used casually.

To put this into context, the lawsuits that Dexcom is facing for wrongful deaths are based on an assertion that Dexcom changed a component in the manufacturing process—a change the FDA later discovered—which affected the G7's accuracy. The claims against Dexcom are not about misrepresentation.

That’s distinctly different from representing a product as “standard of care” because it implies representing something about professional consensus, as if it were a settled, universally applicable conclusion. When public-facing materials fail to explain the caveats clinicians are supposed to apply, the phrase can function less like guidance and more like persuasion.

Promotional health claims in general are expected to avoid material misrepresentation, and the more a statement sounds like an authoritative medical consensus, the more scrutiny it can attract—informally (from clinicians and patients) or formally (from regulators or litigants). When broad public declarations shape patient expectations and clinical conversations, that can matter if disputes arise later.

The practical risk is often not a courtroom verdict; it is being named in a lawsuit and subjected to discovery—subpoenas, depositions, demands for internal communications about how the ‘standard of care’ claim was substantiated. Then comes the downstream cost of controversy, reputational damage, and time-consuming demands for substantiation.

The legal ambiguity of ‘standard of care’ is compounded by a historical pattern: the ADA’s own guidelines have reversed course when evidence accumulated that earlier recommendations caused harm, which are worth considering for a broader view.

Future Developments

It’s important to put this into a chronological perspective. When AID systems were introduced, they provided promise for T1Ds: To relieve the burden of self-management, to potentially automate a highly complex disease, and to achieve healthier glucose control than what the patient can do for themselves.

And there’s no question, that has been achieved for a large number of patients. But as time went on, it became evident that the beneficiaries were not everyone. That the systems’ performance seemed to reach an apex. That there were other unanticipated physical and mental health concerns. That the costs were much higher than comparable methods (MDI) for equivalent control. And more.

It would follow that the ADA is likely to continue to revise their rating systems and provide greater context in future documents. And this is very common, as we can see from some examples:

Aspirin for primary prevention: For decades, diabetics with cardiovascular risk factors were told to take daily aspirin. Following the ASCEND, ARRIVE, and ASPREE trials (2018), the ADA now recommends against aspirin for patients over 70—the risks outweigh the benefits.

Tight glycemic control: “Lower is better” was gospel until the ACCORD trial (2008) had to be halted early because patients aggressively targeting A1c <6% had 22% higher mortality. The ADA now emphasizes individualized targets.

The 50/50 basal-bolus rule: A mainstay of insulin education for decades, now flagged in the 2025 guidelines with warnings about “overbasalization.” It should be noted that most AID algorithms still require a minimum basal rate that exceeds physiological needs, a problem that the FDA needs to address, and is one of the leading concerns about AID systems. I cover the whole topic of overbasalization in a four-part series, Ye Olde Basal Rate: Its Effect on T1D Management and Long Term Complications.

Blood pressure targets: The once-standard <130/80 recommendation for all diabetics was relaxed to <140/90 for most patients after evidence showed aggressive targets didn’t improve outcomes. (Recent medical research appears to be heading back to the <130 target again. Stay tuned.)

The Low-Fat Diets (1980s-2000s): Dietary guidelines demonized fat, believing that dietary fat caused cardiovascular disease, leading to products that replaced fat with sugar. The result: population-wide weight gain and the Type 2 diabetes epidemic. Current guidelines now distinguish between healthy and unhealthy fats. I talk about this in my article, Extending T1D Longevity: Balancing Lipids, Insulin, A1c.

High-Carb Diets: For decades, diabetics were told to eat high-carb diets because it made glycemic management “easier” (especially before CGMs). The result: blood sugar volatility, increased insulin requirements, weight gain. The ADA then shifted to low-carb diets, but later found that, in excess, it leads to cardiovascular disease. The ADA now recommends carb and fat balancing that’s more physiologically balanced for exercise. I cover this in my article, The Paradox of Low-Carb Diets: A1c vs. Metabolic Health.

These reversals aren’t failures—they’re evidence that guidelines evolve as data accumulates. But they share a common pattern: confident recommendations based on the best available evidence at the time, followed by years of accumulating data that revealed unintended harms, followed by quiet revisions.

AID systems are on a similar trajectory. The harms aren’t acute—there’s no single adverse event that triggers an FDA recall. They’re diffuse and delayed: gradual weight gain from overinsulinization, metabolic dysfunction that compounds over years, disengagement that forecloses the development of self-management skills. These are the kinds of harms that only become visible at population scale, after enough time has passed.

Summary

The topic is how to better understand ADA “standards of care” claims, and it can apply to many things (as we just covered). We used AID systems as a case study of how the phrase “standard of care” is both misunderstood and misused in the T1D space by institutions. In a word, it’s glossolalia.

No one—patients or doctors—should look at “standard of care” attributions as though they are black and white, case settled, known benefit, little risk.

Remember the rule at the top of this article: Nothing in medicine is universally beneficial, and there are often hotly debated topics that are left unresolved. In November 2014, the FDA convened a public workshop on insulin bolus calculators that revealed deep divisions among experts. After lengthy debate, there was no consensus. The FDA ultimately issued guidance in 2015 requiring regulatory oversight for apps that “use patient-specific parameters and calculate dosage,” but the fundamental tension remains: these tools can cause as much harm as good, and reasonable experts disagree about the balance. (I discuss the risks of IOB and dosing calculators in my article, The Insulin Absorption Roller Coaster and What You Can Do.)

The takeaway isn’t that guidelines are too ambiguous to trust. It’s that “standard of care” is a phrase that deserves scrutiny, not reflexive acceptance. Who’s saying it? Based on what evidence? With what accountability? And—perhaps most importantly—does it apply to you?

The best defense against borrowed authority is informed skepticism. Understand the difference between advocacy and clinical recommendation guidelines. Read the primary sources. The good news is that literature is there. Yes it’s dense and often biased and hard to read. The aim of my substack is to bridge that gap by translating glossolalia into language that the non-technical, non-medical audience can understand. And like the ADA, when evidence reveals new information and new understandings, I revise articles to be up to date with current science.

Share

Subscribe now

Most readers of my substack know all too well that my articles can be a bit time-consuming to read. Many have asked me, why can’t I just distill a topic down to the basic facts?

If I did that, it would be too easy to dismiss them as “opinion”. To be credible, I cite articles and studies published in peer-reviewed medical journals, and where the principal investigators do not have a conflict of interest with companies. It is increasingly common for tech companies to fund and/or run their own clinical trials that produce outcomes favorable to them, flooding medical literature, thereby showing up more frequently in literature searches. Doing diligent research is not for the faint of heart.

But, for my readers, I understand: sometimes, you just want the basic facts, and move on.

With that in mind, let’s meet in the middle. The aim of this article is to review the top ten myths of T1D that are either half-truths, exaggerations, or completely false. For each item, I will present the topic in short form that makes it easy to understand, followed by a link to my long and arid technical review.

This may look like a long article, but think of it as ten very short articles!

Now, because they are short, one particular section may prompt you to say, “No! That’s not true!” (You see, this is why longer articles with lots of citations are better.) If you have such an urge, go to the article that I link to, read it carefully, and then submit a comment with your own arid response. (Or, just message me.) I will try to reply with a thoughtful and potentially less arid explanation.

Ok, on with the list! [ Drum roll sound effect here. ]

#1: HbA1c Levels and Complications

The T1D management guidelines we live by are aimed primarily at driving A1c levels lower — more insulin, tighter automation, carbohydrate restriction. They were built on a population-level finding that’s been misapplied to individuals.

Meanwhile, the underlying physiology that actually determines your outcome goes unmeasured, untracked, and unmentioned. Nobody is testing for it, nor are they telling you how to slow it down. The result is a silent corrosion that runs for decades before anything shows up on a chart.

The strange part? None of this is new. The researchers who ran the trial that gave us the 7% target went back and measured that corrosion directly — and found it predicted the next decade of complications better than the A1c itself.

That was 2005. Nothing changed.

When we look at health outcomes of those who’ve had T1D the longest, we see distinct characteristics. The Golden Years Cohort, a group of 400 T1Ds in the UK who’ve had the disease for over 50 years had mean HbA1c of 7.6% (± 1.4), with some levels as high as 8.5%-9%, and one individual as high as 14%. None had an HbA1c below 7%.

Another group is the Joslin Medalists, American T1Ds who’ve also had the disease for over 50 years. They had an average A1c of 7.7%, and also had low lipid levels and insulin requirements (0.46-0.52 u/kg/day).

What they shared was metabolic health: normal weight, low insulin requirements, and crucially, low lipid (LDL) levels. Neither group achieved this through perfect glucose control.

The article, A1c and T1D Complications: What It Reveals, What It Hides, and How the Damage Actually Happens, traces how we got here, what the number can and can’t see, and why the people who’ve lived with type 1 for 50 to 80 years didn’t get there on A1c alone - or at all. What they DID have in common is measurable, it’s actionable, and almost nobody is talking about it.

#2: The #1 Cause of Death for T1Ds is not “dead in bed syndrome”. It’s MACE.

The old myth about the “dead in bed syndrome” holds that hypoglycemia can deplete your brain of glucose, especially during sleep. And this keeps many T1Ds and their parents up at night.

But no autopsies have ever shown that the brain was deprived of glucose. It’s because an acute hypo can cause a heart attack or stroke—otherwise known as a Major Adverse Cardiovascular Event, or MACE.

And that can only happen if you have advanced cardiovascular disease (ASCVD), and that takes decades to develop. Stats show that T1Ds die 11-15 years sooner than the general population, and the number one cause of death is heart attack or stroke—by a huge margin.

The good news is that ASCVD can and should be treated using statins and PCSK9 inhibitors, which are cheap, easy, safe, and highly effective. Diabetics of all types are particularly vulnerable to ASCVD because of two key accelerants: a lifetime of excess insulin levels and LDL levels. LDL is the “bad cholesterol” that you often hear about.

Excess insulin levels harm your arteries’ endothelium (a protective lining), making it easier for LDL to embed onto artery walls, creating plaque. The stage is now set for a heart attack. What typically triggers it is an “acute hypo” event, which elevates both heart rate and blood pressure, which causes vulnerable plaque to rupture, and that’s it.

It can happen any time. But when it happens during sleep, that’s when people attribute the death to “dead in bed syndrome”.

If you keep your LDLs level low, your risk of ASCVD is low enough that you can live as long as the general population.

Indeed, studies show that the longest-lived T1Ds are so, not because of the myth that they have working beta cells, and not because they have excellent A1c’s (many are north of 8-9%), but because they all—yes, all—have low LDLs, have low daily insulin levels, and are not overweight. Why this is not part of standard T1D management care is another problem entirely.

For more, read Extending T1D Longevity: Balancing Lipids, Insulin, A1c.

#3: DKA is exceedingly rare

Many T1Ds are taught to fear Diabetic Ketoacidosis (DKA), but they are not informed about physiological ketosis, a healthy state that occurs during exercise, fasting, or weight loss. Ketones also provide anti-inflammatory benefits, and can be used as an energy substrate for exercise.

Because no one educates T1Ds about this—worse, they are told to constantly fear and test for ketones—patients not only live in a state of constant stress about DKA, they engage in highly unhealthy behaviors to avoid it: taking too much insulin, not losing weight, and not exercising, the latter of which are essential to good health and generate elevated ketones.

The truth is that DKA doesn’t just spring up out of nowhere. It is a vanishingly rare, pathological state that requires three simultaneous conditions: a total lack of insulin, high glucose (with some rare exceptions), and high ketone levels. It also takes 10 hours to set in, during which time one is extremely uncomfortable, making it painfully obvious that you should just take insulin. The majority of cases are those with severe eating disorders (where they withhold insulin to lose weight), with the remaining cases being mechanical pump failures or illness.

For more, see Ketones: The Unjustly Demonized Villain in T1D Management.

#4: Basal insulin: Actual metabolic need, the 50/50 rule, and other myths

Most T1Ds take far too much basal insulin — you know, “background insulin”. This has long been presumed to take care of basic metabolic needs, outside of your food intake. The theory stemmed from a flawed trial back in the 1980s, which mistakenly assumed the liver produces at a steady rate, like a “leaky faucet”. This led to the 50/50 rule, where 50% of your total daily dose (TDD) should be basal insulin.

But a pivotal study in 1979 established that the liver only produces glucose when there’s no food onboard for periods of time—mostly at night and for short bursts between meals. And that study has been replicated, even all the way up to today.

The only time the liver produces glucose (under normal physiological conditions) is to provide fuel for the body when there’s no glucose coming in from food. If you eat, your liver is quiet. So if you also have “background insulin” active, then it’s partially serving to address the meal’s carbs. If you bolus on top of that, it can be fine, but it’s very hard to calculate.

Depending on the amount of basal insulin you take, that calculation just gets even more complex.

If you reduced your basal dose or rate, your mealtime boluses would go up proportionally, but here’s the benefit: you’re moving insulin dosing around from basal to bolus, but you’d be doing so more efficiently—giving yourself the right amount of insulin when you need it, and not getting too much when you don’t.

For more, see these articles:

• Ye Olde Basal Rate: Its Effect on T1D Management and Long Term Complications

• Basal Dosing: How Much Do We Need?

• Basal Insulin Reduction: A How-To Guide using “Active Learning”

#5: Insulin and Food Absorption Variability: The myths of IOB and IOC Calculators

When you were first diagnosed, you likely learned about insulin-to-carb ratios, correction boluses, and how to estimate your “insulin onboard” (IOB) and “carb onboard” (COB) as part of your self-management protocol. Indeed, modern automated systems use these same ratios in order to make dosing decisions, too.

This is great for newly diagnosed patients, who don’t yet know how to manage themselves. A calculator is good for training—like training wheels on a bike.

But these estimates are gross generalizations based on formulas established in controlled clinical studies, where patients are measured in reclined, sedentary positions, where they measure the rates of glucose and/or insulin absorption. These are not real-world conditions, so as time goes on, one eventually learns that these estimates are not that accurate, and if you (or your pump) blindly trusts these numbers, things start to go awry.

In reality, food and insulin absorb at highly inconsistent rates, depending on things like physical activity, time of day, amount of sleep, and other routine things in one’s life.

Moreover, the longer one has T1D, factors like lipodystrophy (tissue damage at injection sites) and delayed gastric emptying (gastroparesis) make insulin and food absorption even more volatile and stochastic.

When you experience stubborn highs that don’t seem to come down no matter what you do, or binge eating food to bring up glucose levels that also don’t seem to budge, that’s just how it goes sometimes.

The short story is that IOB and IOC are great if you’re learning T1D, but are highly unreliable, especially as your duration with the disease passes around five years. Try to wean yourself off of them and learn techniques that can help navigate these turbulent periods. See these articles:

• The Insulin Absorption Roller Coaster and What You Can Do

• Food Absorption Variability and Treating Hypoglycemia

#6: CGM “Accuracy” (MARD) and other blood glucose myths

T1Ds are often misinformed that a lower MARD (Mean Absolute Relative Difference) rating makes a CGM superior—or even simply “more accurate”. But MARD is largely a marketing term that can be easily manipulated in clinical trials. Moreover, “accuracy” has no bearing on clinical outcomes.

This leads to other myths about blood glucose: that a finger-prick reading is “the gold standard” for accuracy, or that there’s a “lag” between CGM and blood glucose levels. Those happen to be true under controlled lab conditions where people are tested under sedentary conditions, but in real-world settings, glucose moves around the body into different “compartments” at different rates in different conditions. It is simply not possible to measure true systemic glucose levels.

That doesn’t mean CGMs or blood glucose readings aren’t useful—of course they are. But they are approximations that should be understood as “potentially good enough” for making critical dosing decisions. What’s more, the technology that measures interstitial glucose levels was established ten years ago, and the limitations of physiology are such that they can’t get better than what we see today.

The best way to use a CGM is not solely by individual readings and whether they are “accurate”, but to use them to observe directional movement (up/down/steady) and rate of change. In other words, you’re looking for trend lines and “glucose curves”.

Learning those patterns is the key to optimizing dosing decisions. The better you understand these concepts, the better you can observe and learn how your body reacts to insulin dosing, food types, exercise events, and other self-management essentials.

For more, see Continuous Glucose Monitors: Does Better Accuracy Mean Better Glycemic Control?

#6: Ultra-Low-Carb Diets

While ultra-low-carb diets (like the Bernstein diet) can produce impressive A1c numbers, and can be highly effective at weight loss over short periods of time, long-term exposure to insufficient carbohydrate consumption introduces severe metabolic stress and cardiovascular risk.

When carb intake is too low, the liver must package massive amounts of fat for energy, and the part they don’t use is sent back out into the bloodstream raising LDL and apoB levels, the primary drivers of heart disease. Studies spanning decades consistently show that T1Ds who adopt low-carb diets over long-periods of time had the highest risk of overall mortality (32% increase), cardiovascular disease mortality (50% increase), and cerebrovascular mortality (51% increase).

Furthermore, these diets can deplete glycogen stores, making intense exercise — the thin g that actually makes you healthier — dangerous and increasing risk for other metabolic disorders.

While low carb diets are essential for weight loss, one should balance energy intake with energy expenditure. In other words, your carb intake should match your physical demands (exercise), and since exercise is the healthiest thing you can do, carb up as necessary to optimize exercise. For more, see The Paradox of Low-Carb Diets: A1c vs. Metabolic Health.

#8: The “Standard of Care” Myth of AID Systems

There is a widespread belief that Automated Insulin Delivery (AID) systems are the “standard of care” for T1Ds, and therefore, inherently superior to Multiple Daily Injections (MDI).

Strictly speaking, the American Diabetes Association, which is the governing body for ‘standard of care’ has not designated AID systems as such. They have designated AID systems as the “preferred method of insulin delivery” under many circumstances, but doctors should consider a number of factors before recommending them to patients.

Those who benefit the most are those who cannot or will not self-manage their disease, which can include children, young adults, those who are recently diagnosed, those with certain cognitive disorders, and other conditions. For these people—and there’s a lot of them—AID systems are a miracle, and we’re lucky to have these systems.

But this is where things get murky. There’s an upper limit on performance, and population-wide data for adults over 26 shows no substantial difference in A1c outcomes between pump and MDI users (or non-automated pumps).

If a person is able to attain A1c levels <8% and TIR to 7% on their own, it’s almost assuredly because they are engaged users. In fact, studies show that the simple act of bolusing 20-30 minutes before meals usually gets someone’s A1c and TIR percentages into these ideal ranges, regardless of MDI or pump.

What studies have found about AID systems for these “capable users” is that the people become disengaged. They trust the system. They stop bolusing before meals, and suffer other effects, such as a loss of “agency”, frustration, and significant costs (over MDI). Studies also show that users can become complacent, leading to weight gain, and “de-skilling” of self-management.

Despite the ADA’s strong recommendation in favor of AID systems, both the ADA and the European and American Diabetes Associations put out a joint position statement on AID system adoption, which cites all of these same concerns, and more.

I delve into all of the nuances in my article, Medical Literature Analysis: The Performance Paradox of AID Systems, which comprehensively cites hundreds of studies—most of which are published in ADA journals—that speaks to each of the concerns about these systems.

It’s vitally important to also see the article, Standard of Care: Who Defines it, How, and Why it Matters, which highlights the highly complex ecosystem that establishes how T1D management guidelines are established.

#9: Everything you think you know about a “cure” for T1D

The running joke in the T1D community is that a “cure” is only five years away… and always will be.

This joke persists because of a fundamental disconnect between what science reveals, what researchers actually investigate, and how headlines translate that work. These are three vastly different things.

Scientific papers might report “improved C-peptide levels in recently-diagnosed patients over 12 months”, but by the time it passes through institutional PR departments, advocacy organizations, and science journalism, it becomes “NEW BREAKTHROUGH BRINGS CURE CLOSER.”

A true cure for established T1D requires solving two separate, enormous challenges: resetting the immune system to stop the autoimmune attack, and creating fully functional islets of Langerhans with proper architecture and vascularization.

Unless you have both, there is no cure. When the Edmonton Protocol used whole donor islets in 2000, patients achieved A1c levels of 5.8%, the feeling was that the cure was five years away.

The current strategies are using islets grown from stem cells, opening up new and exciting opportunities, but there are still many barriers ahead.

There’s a lot more to it, but it’s definitely worth learning more in my article, Assembling the Components to a Cure

#10. “Exercise makes glucose control harder” (or “Exercise is dangerous for T1Ds”)

Many T1Ds avoid exercise entirely because of hypo fear, or have been taught that the unpredictable glucose swings make it “too risky.” And yes, exercise does cause glucose volatility—sometimes it spikes, sometimes it crashes, sometimes both in the same session. But avoiding exercise because of this is clipping years off your lifespan.

Exercise is the single most powerful intervention for metabolic health and longevity. (And if you have low lipid levels too, you can live to infinity and beyond.)

Exercise builds mitochondrial fitness, which improves how efficiently your body uses both insulin and glucose. It increases GLUT4 transporters in muscle cells, which is literally the definition of improved insulin sensitivity. And it’s the primary driver of non-insulin mediated glucose uptake (NIMGU)—your muscles pulling glucose directly from the bloodstream without insulin.

The glucose chaos T1Ds experience during exercise can be managed be more intelligent dosing protocols, and those protocols cannot be programmed into a pump. You have to do it manually, precisely because each exercise event is going to affect glucose levels differently. But don’t let that idea overwhelm you—it’s far easier than it seems.

In fact, managing insulin/carb dosing for exercise becomes easier as metabolic fitness improves. The spikes get smaller. The crashes become avoidable. Your insulin needs decline. The very thing that makes exercise “hard” at first is exactly what makes it essential—your body is adapting, and those adaptations compound over time.

For more, see T1D and Exercise: Insulin, Carbs and Glucose Control.

Summary: T1D is Not All About A1c

If there’s a throughline connecting these ten myths, it’s this: T1D management has become dangerously fixated on A1c as the singular measure of success. That’s why the first article in this top-10 list is so essential to read.

The irony is that many T1Ds with “good” A1c and TIR percentages are metabolically unhealthy, while others with higher A1c’s live long, complication-free lives because they got the other factors right.

T1Ds can achieve these outcomes through these principles:

• Cardiovascular health: ASCVD is the leading cause of death, and LDL management is more predictive of longevity than A1c.

• Insulin dosing strategy: Overbasalization and blind reliance on IOB/COB calculators cause weight gain, hypoglycemia, and daily chaos—none of which show up in your A1c.

• Metabolic fitness: Exercise is the single most powerful intervention for insulin sensitivity and longevity, yet it’s undertreated because glucose volatility scares people off.

Obviously, there are technologies today that can help. Ok, one: the CGM. In my article, The Four Habits of Healthy T1Ds, the first thing I explain is that, before you can become healthy, it requires a frame-of-mind: engage with your diabetes, don’t put it in the background. You don’t need goals or targets. Targets are distracting at best and dispiriting at worst, and either way they pull attention away from the things that work.

Ask any T1D what their A1c is, and most will probably know. Ask them what their LDL levels are, and most won’t know—or if they do, they won’t understand what it means. Or what to do about it. And that’s a big deal, as it can affect your lifespan a lot.

Let’s start with some startling facts that few T1Ds are aware of:

• According to CDC Heart Disease Facts, 2024, cardiovascular disease causes about one in three deaths in the US. That is, a Major Adverse Cardiovascular Event (MACE). This is when plaque in your arteries (created by high LDL levels) break off and block blood flow, causing a heart attack or stroke.

• For T1Ds, MACE is the leading cause of death, striking ~11 years earlier than in the non-diabetic population. More concerning is that some Americans die up to 15-17 sooner if they don’t have access to minimal healthcare.

• Young adults with T1D under age 40 face a 20- to 40-fold higher risk of dying from heart disease compared to their peers.

This data puts to rest one of the most pervasive myths about T1D death: that you can die from hypoglycemia. The so-called “dead in bed syndrome”, as it was called, was thought the brain would die due to insufficient glucose. No autopsies have ever found that to be the case, except for severe and intentional overdoses of insulin by suicide, and even then, most attempts are unsuccessful.

For the vast majority of T1Ds, what severe hypoglycemia does is trigger acute and intense counterregulatory hormones to drive the liver to produce glucose. These raise blood pressure and cause other systemic responses that cause vulnerable plaques within the arteries to break off, triggering the heart attack or stroke.

It’s not the hypo that kills you, it’s the plaque in your arteries.

A note on younger T1Ds: Some may wonder how MACE explains “dead in bed” deaths in T1Ds under 40, who presumably haven’t developed significant atherosclerosis yet. Recent research clarifies this. A 2025 study in Frontiers in Endocrinology examined cardiac electrical changes during hypoglycemia in healthy T1Ds without cardiovascular disease. The authors explicitly state that pre-existing cardiac vulnerability, such as undiagnosed structural abnormalities or autonomic neuropathy in younger people, can result in the same outcome.

The saddest part is that the interventions that can avert heart disease are easier, cheaper, safer and in all other ways, simpler than struggling with the daily management of glucose levels. Better still, they can be done independently of glucose management.

That doesn’t mean glucose management isn’t important—it is. But it’s not the only thing T1Ds should focus on. And that’s what brings us back to those LDL levels that few T1Ds talk about or properly understand. Reduce the plaque, and you reduce your risk of death, potentially adding years to your life.

The aim of this article is to unpack all this to live a longer, healthier life. For that, let’s start at the end and work backwards.

Subscribe now

The basics of cardiovascular disease

Cardiovascular disease (CVD) leads to a fatal heart attack or stroke—what doctors call Major Adverse Cardiovascular Events, or MACE. And it’s triggered by plaque that breaks off from artery walls, travels along until the path narrows too small to pass through, forms a blood clot, and, well, that’s it.

This is where LDLs come in. LDL stands for low density lipoproteins, which is colloquially called “bad cholesterol”. Once LDLs penetrate arterial walls, they trigger an immune response, causing macrophages to accumulate and form that plaque mentioned above. Over time, plaque builds up and becomes increasingly less stable. All you need is one big “event” that unlodges some plaque, and that’s it.

Let’s say your arteries are like the garden hose you connect to your water spigot. Plaque is like mud that’s stuck on the interior of the hose. If you turn the water spigot on, the amount of pressure might be strong enough that the shearing force can rupture the mud, where it then travels through until it gets to a narrow spot—like one of the hose kinks you often see—and then the mud gets stuck. Boom. Heart attack.

Short story: Eliminate the mud and there’s no risk of a heart attack. Put in terms of LDL levels: Clear all that out, and you reduce the plaque levels, thereby reducing your risk of death. By a LOT.

Glucose has nothing to do with this.

This is why I mentioned earlier that the intervention is relatively simple, cheap, and safe: a variety of medications are perfectly suited.

So, are we done here?

Not quite. It’s only 50% of the story.

It’s not feasible to reduce LDL levels to zero, so risk isn’t zero. But we can address the next thing in the cascade: Reduce the ability for those LDL particles to attach to artery walls in the first place. To do that, we need to bring glucose back into the conversation. Indirectly, at first.

Subscribe now

Endothelium: The protective armor

The reason LDL can do this damage is that it penetrates a protective armor that lines the interior of your arteries: endothelium. It’s more like thin latex gloves protecting your hands. If endothelium is healthy, the LDL particles don’t break through. But if it’s damaged or destroyed, not only do LDLs get in, but the amount of damage correlates to the rate at which LDLs get in.

So, how do we keep the endothelium from getting destroyed?

That brings us to insulin. Yes, insulin.

You can spend a lifetime achieving non-diabetic A1c numbers and still die prematurely of MACE. A1c doesn’t capture your LDL, your blood pressure, your insulin resistance, or the epigenetic residue of every excursion you had before you got good at this.

That sentence should be uncomfortable, because the entire apparatus of diabetes care — the quarterly lab draw, the clinic conversation, the pat on the back or the gentle scolding — is built around a single number that doesn’t measure the thing most likely to kill you.

But there’s a second problem underneath the first, and it’s stranger. It isn’t just that A1c misses things. It’s that the way we get A1c down can create some of the things it misses.

Let’s clear away a myth first

Starting in the 1980s, large studies of non-diabetic men found that those with high insulin levels in their blood were more likely to have heart attacks years later. The Quebec Cardiovascular Study found that men who went on to develop heart disease had fasting insulin levels 18% higher than matched men who didn’t. The Helsinki Policemen Study followed 970 men for 22 years and found the same pattern.

Insulin high, hearts bad. Case closed?

Not quite. They had high insulin because they were insulin resistant — their tissues had stopped responding well, and their pancreases were working overtime to compensate. The high insulin wasn’t the disease. It was the pancreas succeeding at a job that was getting harder.

The researchers who ran the Quebec study figured this out themselves. Two years after that famous paper, Després’ own group reframed their finding as the “atherogenic metabolic triad” — high insulin plus high apolipoprotein B plus small, dense LDL particles. By 2000 they’d built a screening tool for it that didn’t measure insulin at all. Just a tape measure around the waist and a triglyceride level.

And when someone finally put all three variables into the same analysis — 65 studies, 516,325 people — the answer was unambiguous. Per standard deviation increase, the relative risk of coronary heart disease was 1.21 for glucose, 1.46 for insulin resistance.

Insulin isn’t the villain. Insulin resistance is. And it’s a stronger villain than glucose.

Let that sink in for a moment.

The strategy of chasing glucose down with insulin as your only lever creates the conditions for insulin resistance, and that’s what kills you.

The loop works like this:

Your A1c is high, so you take more insulin. Reasonable. It’s the tool you have, and it works.

Insulin is a storage hormone. Its job description isn’t only “move glucose out of the blood.” It’s also “store energy, stop burning fat, build tissue.” More insulin, at the same food intake, means more storage.

More insulin means more lows. And lows mean eating — juice, glucose tabs, whatever’s within reach at 3 a.m. Nobody counts those calories. Then comes defensive eating: the snack before the drive, the extra carbs before the workout, the bedtime buffer. It adds up quietly.

Weight goes on, and it goes on around the middle. Visceral fat, which surrounds organs—not the visual fat you see around your waist—is the tissue most strongly associated with insulin resistance.

Now you’re insulin resistant, so your insulin requirement rises. Which means more insulin, more lows, more defensive eating, more weight.

And you’re back at the top of the loop, one turn worse. Your A1c may look identical the whole way around. That’s the part that should get your attention.

This isn’t theoretical. It’s in the DCCT.

The Diabetes Control and Complications Trial is the study that proved intensive glucose control prevents complications. It’s the foundation of modern T1D care, and its findings hold.

It also has a subplot nobody quotes at the clinic.

The intensively treated group is the one that achieved better glucose levels, and as a result, reduced microvascular complications, such as retinopathy and kidney disease.

But they also gained substantially more weight than the conventional group.

Purnell and colleagues sorted participants by how much they gained and looked at what happened to the ones in the top quartile. Their conclusion, published in JAMA in 1998: the changes in lipids and blood pressure in that group were similar to those seen in the insulin resistance syndrome, and might raise coronary risk over time.

They were right. In the long-term follow-up, the excessive-weight-gain group’s cardiovascular event rate began rising at year 14, and by year 20 it had converged with the conventionally treated group — hazard ratio 1.99 compared to intensively treated people who hadn’t gained the weight. When the analysis adjusted for the weight-related risk factors, the difference disappeared, which is the tell: it was the metabolic profile doing the damage, not the treatment intensity.

Two decades of excellent glucose control, and the cardiovascular benefit eroded because of how it was achieved.

Type 1 diabetes plus acquired insulin resistance is now called “double diabetes”. It’s more common than the label suggests, partly because nobody screens for it. There’s no lab test on the standard panel. Your A1c can be 6.2% and your insulin sensitivity can be in the basement, and nothing in your quarterly visit will surface the discrepancy.

What to measure instead

There’s a calculation called eGDR — estimated glucose disposal rate. It takes three things you already have or can get in five minutes: waist circumference, whether you have hypertension, and your A1c.

eGDR = 21.158 − (0.09 × waist in cm) − (3.407 × hypertension) − (0.551 × A1c)

Lower numbers mean more insulin resistance. And unlike A1c, it predicts hard outcomes — each 1 mg/kg/min increase is associated with meaningfully lower cardiovascular risk and lower all-cause mortality in T1D.

Notice what’s in that equation. Waist circumference carries the most weight. Hypertension is the single largest term. A1c is in there, but it’s the smallest contributor of the three.

Notice also what’s not in it: your total daily insulin dose. That’s not an oversight. It’s the finding. In one study of young women with T1D, eGDR correlated with carotid artery thickness and plaque. A1c did not. Neither did diabetes duration, BMI, or cholesterol.

Let’s be clear here: Just because BMI or cholesterol wasn’t correlated, it doesn’t mean those aren’t harmful. It means that eGDR is a risk indicator because insulin resistance predicts devastating health outcomes later. This is not unlike finding that your immune system is killing your beta cells years before you’ve been diagnosed with T1D. If you can detect it early, and you had the tools to stop it, you do it. Unfortunately, we can’t yet stop the immune system from killing beta cells, but you can keep insulin resistance from happening in the first place.

What this is not an argument for

Let me be direct, because this material gets misused.

This is not an argument for running your glucose high. Intensive control in the DCCT reduced complications across the board, and in the imaging follow-up, prior intensive treatment was associated with less coronary calcification, largely through lower A1c. Glucose is genuinely dangerous. Nothing here changes that.

This is not an argument for taking less insulin than you need. Insulin restriction is one of the most dangerous behaviors in T1D, and it kills people.

The argument is narrower and more specific: insulin should not be your only lever.

Exercise moves glucose into muscle through a pathway that doesn’t require insulin at all. Simply walking after meals for 15 minutes has been shown to reduce insulin requirements significantly. It improves insulin sensitivity, and mitochondrial health.

And then there’s basal insulin profiles. I’ve got an extensive, four-part series on basal insulin myths and facts, starting with Ye Olde Basal Rate: Its Effect on T1D Management and Long Term Complications, which begins by citing section 9.27 of the ADA’s 2025 guidelines, which now recognize overbasalization as a major health concern for T1Ds. Symptoms include big differences in glucose levels between bedtime and morning—or, before and after meals; occurrences of hypoglycemia (aware or unaware); and high glycemic variability.

Yeah, you’d think every T1D suffers from those symptoms, right?

Exactly. And nearly all T1Ds take too much basal insulin (injected or as a basal rate in their pumps). The four-part series walks through the physiology of how, when, and why “background insulin” plays a role in daily life. And one of the primary benefits of basal insulin reduction is combating those unexpected lows. The hypos that force you to eat when you don’t want or need to. Taking care of that also keeps weight off.

This brings us all the way back to LDLs: Treating those like it’s a separate disease is exactly right, because functionally it is.

None of that shows up on your A1c. All of it shows up in your arteries.

The uncomfortable part of the A1c value

If you’ve been doing this a long time, you’ve probably built your identity around a number—the dreaded HbA1c. Most of us have. There’s real accomplishment in achieving low numbers, and enormous distress when it’s higher.

But the number is not a proxy for overall health; it does not reflect your true risk profile; it is a single number among several that matter. And not only is the the absolute hardest number to optimize, the very acts people to do in order to lower A1c values almost always work against you.

Two people can walk into the same clinic with the same 6.1%. One of them got there with exercise, stable dosing, and a waistline that hasn’t moved in a decade. The other got there by climbing the loop — more insulin, more lows, more eating, more insulin.

They get the same congratulations.

They do not have the same arteries.

Who knew that insulin—the hormone that saves our life and keeps us from dying every day—is also responsible for a shortened lifespan.

Glucose: The Lethal Bullet

Up to this point, we’ve focused primarily on LDL and insulin levels. Optimize those, and your risk of death and serious complications is reduced. So now, let’s look at the role glucose plays.

The primary reason to reduce glucose levels is because of glycation. Glucose binds irreversibly to proteins and lipids, they then oxidize, and the end products resulting from that are called Advanced Glycation End-products (AGEs). The higher one’s glucose levels, the greater the number of AGEs in your body. Over time, this makes tissue less flexible and more brittle, affecting virtually every part of you. Frozen shoulder is one of the most common ailments that T1Ds get from certain kinds of exercise that involves the rotator cuff, and yet, no one talks about the true cause: soft tissue that’s been glycated by AGEs.

There is no tissue that’s spared.

When LDL proteins get glycated, the harms they’re already doing—forming plaque—are amplified. A lot.

Why our LDL is not the same LDL

Here is the part that makes this specific to us rather than general cardiology advice.

A paper in Science Direct explains that glycation concentrates on small, dense LDL particles — the subfraction already most associated with heart disease. In people without diabetes, roughly 2 percent of the apoB in large buoyant particles is glycated versus 17 percent in the small dense ones, and about two-thirds of all glycated LDL sits in that small dense fraction. Diabetes roughly doubles the glycated share even when control is reasonably good, according to another NIH paper.

Modified particles then behave differently. Rather than being cleared normally, they are swallowed by scavenger receptors on immune cells in the artery wall — an uptake route with no off switch — producing the foam cells that start a plaque, and they do this without needing to be oxidized first, as explained in this NIH paper.

High glucose also increases the number of those scavenger receptors on vascular cells, which has no shutoff, so they gorge on it and become the foam cells that make up plaque.

This is why an LDL of 70 in a person with type 1 diabetes is far worse off than a non-diabetic with an LDL of 70. That’s why our targets should be lower rather than merely equal. There’s no way to formally quantify this, but we can see it in long-term outcomes: it’s not just that T1Ds die of heart attacks and strokes, the amplification explains why T1Ds die 11-15 sooner than the general population.

(For the extensive detail, see the article, How hyperglycemia promotes atherosclerosis: molecular mechanisms and expanded in The Diabetes Mellitus–Atherosclerosis Connection: The Role of Lipid and Glucose Metabolism.)

So, now we see all three come together: LDLs are the killers, insulin resistance removes the protecting armor, and excess AGEs assure that each bullet is more lethal than it would otherwise be.

The easiest solution is to start by reducing the number of killers (reduce LDLs), because that requires safe, inexpensive, and easily-managed lipid-lowering drugs, like statins. Next, keep the armor intact (reduce insulin resistance), by keeping weight lower and setting basal rates much lower. Then comes the last step, making sure the bullets aren’t as lethal by reducing AGEs.

Yes, lower A1c levels can also lower AGEs, but there’s not a direct line to that. Glucose itself isn’t the problem — it’s oxidative stress. That’s what converts glycated proteins into AGEs. Lowering glucose levels may not do that if it’s achieved by higher insulin levels because insulin itself can lead to oxidative stress if managed inappropriately.

This is all explained in my article, A1c and T1D Complications: What It Reveals, What It Hides, and How the Damage Actually Happens. A healthier way to lower A1c levels is to keep insulin levels within physiologic norms, and to use exercise to generate the antioxidant enzymes that absorb the oxygen species that create AGEs.

Again, a quick reminder: Heart attacks of exactly this nature are the leading cause of death for everyone in the US and most developed nations, including non-diabetics who have perfectly normal A1c levels. Lipid mismanagement, poor dietary patterns, and lack of exercise leads to poor health outcomes, are the primary causes of early death. And are far easier to manage than A1c levels.

Some might be thinking, low-carb diets! You can get pristine A1c levels, and low insulin levels, and low body weight.

Why Low-Carb Diets Don’t Work

Ironically, low-carb diets not only don’t help, they can actually make the problem worse, but from a different physiological angle: When carbohydrate intake drops too low, the body must mobilize massive amounts of stored fat to use as energy. To convert fat to glucose, the liver must package these free fatty acids into VLDL particles for transport, which eventually become the LDL that drives cardiovascular disease.

The amount of LDLs generated rises along with the amount of fat that needs to be converted. The lower the carb intake, the more fat is used. And the more energy you need—especially for exercise—the amount of fat utilization increases.

According to a meta-analysis in PLOS ONE, people with the lowest carbohydrate intake—such as the Bernstein diet—had the highest risk of overall mortality (32% increase), cardiovascular disease mortality (50% increase), and cerebrovascular mortality (51% increase).

But what about the low insulin levels? Shouldn’t that keep the endothelium intact?

Insulin isn’t only a glucose hormone — it also signals blood vessels to stay healthy and relaxed, and does this through two branches: one that produces nitric oxide and relaxes the vessel, one that produces endothelin-1 and constricts it.

The plain English version: At physiologic levels they balance, and insulin’s net effect on a healthy vessel is dilation. If you lower insulin levels, you reduce the benefits of nitric oxide and the vessel no longer relaxes. The lower the insulin levels, the less protection you get. Now add even more LDLs into the bloodstream because of the low-carb diet and you have a perfect storm: more LDL does more damage to a compromised vessel wall, amplifying the risk of ASCVD.

The key is avoiding extremes. Moderate carbohydrate intake (150-250g daily) and moderate insulin intake is the sweet spot. For a complete analysis of low-carb diets, see my article, The Paradox of Low-Carb Diets: A1c vs. Metabolic Health.

The sweet spot for ideal insulin levels—not too high and not too low—is 35-40 units of insulin per day, according to Dr. Ralph DeFronzo during this interview with Dr. Peter Attia. Dr. DeFronzo is one of the foremost experts in diabetes research, especially in the area of glucose and insulin research.

One’s total daily dose is one thing, but balancing between basal and food-based boluses is another challenge completely. For more, see my article, Basal Insulin Reduction: A How-To Guide using “Active Learning”.

Some may find it challenging to maintain healthy glucose levels with that amount of insulin per day, so let’s put glucose levels back in the spotlight.

Share

Can’t I just lower my LDL with diet?

If not low-carb, then what about just reducing the kinds of fat we eat?

Yes, technically, that can lower LDL. The mechanisms are not in dispute: saturated fat suppresses the liver’s LDL receptors so that less LDL gets cleared from the blood. True, and therefore, replacing animal protein with plant protein lowers LDL. Eating more beans, oats, and vegetables and less processed meat is a good idea for reasons that have nothing to do with this article.

The question is not whether diet lowers LDL. It’s whether diet lowers LDL enough, for us, given what our LDL is doing. Remember, our LDL isn’t like everyone else’s. So, let’s look at how much LDL reduction we get from switching from meat to plant-based diets.

The number is smaller than the advocacy suggests

The most-cited source for the plant-based cholesterol claim is a 2017 meta-analysis by Yokoyama, Levin, and Barnard covering 30 observational studies and 19 clinical trials. It’s usually quoted as showing a 15 to 30 percent LDL reduction. That range comes from the observational half of the paper, where vegetarians had LDL levels 22.9 mg/dL lower than omnivores.

But people who choose vegetarian diets also differ in lifestyle. Smoking and generalized caloric intake typically far exceeds those who follow vegetarian diets, who also happen to exercise more. So, it’s not just the food type that’s favoring a lower LDL level. The clinical trials from that paper, where researchers actually assigned people to the diet, found a more modest 12.2 mg/dL reduction. Off a typical starting point, that’s roughly 10 percent. Not bad, but it doesn’t move the needle on risk profiles.

Two other results in that same paper rarely make it into the summaries: HDL also dropped by about 3.5 mg/dL, and triglycerides did not fall at all. In the intervention trials they drifted slightly upward.

Twelve points is a real effect. It is not a large one.

What we are comparing it against

A high-intensity statin lowers LDL by about half. Across 26 randomized trials, every 1.0 mmol/L of LDL reduction (about 39 mg/dL) cut major vascular events by 21 percent, with no threshold — the benefit kept accruing as LDL went lower. In the diabetes-specific analysis of 18,686 people, the proportional benefit was identical to that in people without diabetes, and because our baseline risk is higher, the absolute benefit was larger: 42 fewer major vascular events per 1,000 people over five years.

These are not two comparable tools. They differ by a factor of four or five on the thing being measured.

“Reactive” is exactly backwards

A common framing holds that statins should be held in reserve — a reactive measure, deployed once something has gone wrong. This gets the biology inverted, and the plant-based literature itself explains why.

The conclusion is not either/or

The amplification argument cuts toward more aggressive lipid lowering, not less. There is no way to un-glycate a particle that is already circulating. The only lever that reaches the damaged fraction is reducing how many particles there are in the first place — which is what a 50 percent reduction does and a 10 percent reduction largely doesn’t.

Meanwhile, glycemic control operates on a different term entirely: it reduces the rate at which particles get modified. Particle count and modification rate multiply. They don’t trade off.

So the answer to “diet or statins” is that the question is malformed. Eat the beans and the oats; the fiber is doing real work and the rest of your metabolic health benefits. Then recognize that in a disease that chemically alters your lipoproteins from the inside, a 12 mg/dL improvement is a floor to build on, not a ceiling to stop at.

Balancing Insulin, Glucose and LDL

To see how to balance insulin, glucose, and LDL, let’s look at those who live a long time with T1D. What are their profiles like?

Fortunately, we have lots of data to look at, beginning with the Golden Years Cohort, a group of 400 T1Ds in the UK who’ve had the disease for over 50 years. Their mean HbA1c was 7.6% (± 1.4), with some levels as high as 8.5%-9%. None had an HbA1c below 7%.

The Joslin Medalists, American T1Ds who’ve also had the disease for over 50 years, had an average current A1c of 7.3%, and 7.7% among the 73-person longitudinal subgroup followed for 15 years. Across the full cohort, individual A1c values ranged from 5.0% to 14.0% — and current/longitudinal glycemic control was not related to complications.

What they shared was metabolic health: normal weight, low insulin requirements, and crucially, low lipid levels. Neither group achieved this through perfect glucose control. It’s not low-carb diets, insulin pumps, or any other drugs (other than statins for lipid management).

Ok wait — If glucose is so important, how is it that their A1c’s are so high and they lived so long?

They lived long because they had low LDL levels, but they also suffered a lot of microvascular complications due to their elevated glucose levels. In an analysis published in Diabetes Care, among the Joslin 50-Year Medalists with mean A1c of 7.7%:

• 57.4% had proliferative diabetic retinopathy (PDR)

• 13.1% had nephropathy (kidney disease), and

• 60.6% suffered from neuropathy.

Importantly, only 21.2% were free of all three microvascular complications. So, A1c levels do matter.

Let’s put these concepts together:

• T1Ds who lived to 70+ were able to do so because they had low lipid levels.

• However, decades of A1c levels > 7% resulted in eye, kidney and nerve diseases.

• THEREFORE:

  • Lower lipids = live longer, and

  • Lower glucose levels = live better.

Most of my other articles deal with glucose management—exercise, diet, CGM use, insulin absorption variability, food absorption, and many other topics. I won’t cover those here. But what none of them cover is how to track and manage lipid levels, so now that we’ve covered how cardiovascular disease happens, let’s talk about lipid management.

Tracking and Evaluating Lipids

Virtually everyone who has any kind of diabetes should monitor lipid levels just as often as they get their A1c tests done. No exceptions, and no age limit. Yes, children too.

Lipids should be included in all blood work orders, and should include a ‘lipid panel’ that includes total cholesterol, LDL cholesterol, HDL cholesterol, and triglycerides. These are the minimum, but you should also include apoB and Lp(a), if those aren’t included in that default panel.

The particles that erode your endothelium are LDL, VLDL, IDL, and Lp(a), but they all share one structural feature: they each contain exactly one apolipoprotein B (apoB), and it’s this protein marker that indicates atherosclerotic risk. This is why apoB is the most important one to test for.

Lp(a) is entirely genetic, so you only need to test for it once. If it’s elevated, you need to be hyper-aggressive in treatment (next section).

If you don’t get apoB, knowing your LDL levels is the next best thing, but note that people with insulin resistance or elevated triglycerides may appear to have acceptable LDL levels, but their particle count (apoB) could be dangerously high. This discordance explains why some people with “normal” LDL still experience cardiovascular events.

When you get results, labs report the values along a line that also indicates the “normal” range. Here’s an illustration, with the first item being “total cholesterol”.

In this lab report, everything is in the green “normal” zone, but this should not be confused with “safe”. In this case, the LDL level is 38, which is unusually low, but this person is taking medications to reduce their LDL levels because they are aware that risk rises with higher levels.

Without this medication, levels would be considerably higher—even as high as 80—at which point, doctors will say, “we’ll keep an eye on it.”

Unfortunately, few clinicians are as up to date on cardiovascular risk as they are on the latest insulin pumps, so when they see “normal ranges” for lipids, they’re not thinking about T1Ds vs. non-diabetics. So, they typically don’t act. And this is one of the leading reasons why T1Ds die from MACE so early in life: Because elevated insulin levels can erode endothelium, even “normal” levels of LDL can be harmful because the protection is weaker, and doctors aren’t prescribing medications to lower LDL levels.

And remember, there are also AGEs from elevated glucose levels, which make those same LDLs even more lethal, far more than that of the general population. By the time you’re in your mid-50s, you may have had “normal” LDL levels, but your compromised endothelium and weaponized LDLs dramatically elevates your risk. An acute hypoglycemic event could be the big one that dislodges vulnerable plaque.

So, your response to the doctor is, “No, you’re not going to keep an eye on it. You’re going to lower it.”

That leads to the next step: interventions.

Subscribe now

Lipid Management

There is no healthy or safe level of LDL, apoB or Lp(a). The fewer harmful particles there are, the less able they are to create the “vulnerable plaque” that leads to MACE. You want to go as low as medically possible.

An analysis in PMC emphasized the dose-response relationship: the lower the LDL, the lower the risk, with no apparent threshold. This isn’t just correlation—it’s mechanistic confirmation. Fewer atherogenic particles mean less substrate for plaque formation.

So, how should you reduce LDL levels?

Doctors and media say that lipid levels can be lowered through diet, exercise and pharmacology. This is technically true…for the general population, where risk profiles are much lower. But, T1Ds’ risk of MACE is so substantial, that the goal is not to compare yourself against the general population. It’s impossible to eat low fat food and exercise a lot and hope to reduce LDL and apoB. Yes, those are good habits to have, and there are other benefits, particularly for diabetics. But they are rarely going to lower harmful lipid levels on their own. You will need pharmacological interventions (drugs).

Fortunately, the drugs to do this are old, well-established, well-tolerated, inexpensive, and generally available through most medical insurance.

Statins are the foundation of lipid therapy, blocking cholesterol synthesis in the liver and typically reducing LDL by 30-50%. They’re well-tolerated, inexpensive ($4-20/month generic), and have decades of safety data. Common options include atorvastatin and rosuvastatin. About 5% of the population can experience muscle soreness to varying degrees with statins.

A detail that comes up is that statins can induce insulin resistance, but it manifests more dramatically in non-diabetics because it can degrade beta cell function in a small set of individuals.

As for T1Ds, the beta cell function is not an issue of course, but then we get into T1D obesity, which is now “double diabetes” territory, and we have the same situation as noted above: Weight gain is the real problem.

Ezetimibe (Zetia) works differently—it blocks cholesterol absorption in the intestine, adding another 15-20% LDL reduction when combined with statins. It has minimal side effects and costs roughly $10-30/month generic. Note that many studies show Ezetimibe has minimal effect when used alone—you need to take it with either statins or PCSK9 inhibitors.

PCSK9 inhibitors (evolocumab/Repatha, alirocumab/Praluent) are the most powerful option, reducing LDL by an additional 50-60% on top of statins by increasing the liver’s ability to clear LDL particles from the bloodstream. Because they don’t have the confounding effects of statins, they’re becoming very popular in this domain. They are currently taken by injection every other week, but an oral version is in trials and expected to be released in coming years.

While expensive ($500-600/month without insurance), many insurance plans now cover PCSK9 inhibitors for high-risk patients—yes, T1Ds—and patient assistance programs exist.

The results speak for themselves: The FOURIER trial, published in the New England Journal of Medicine, followed 27,564 patients with established cardiovascular disease. Those treated with evolocumab (a PCSK9 inhibitor) achieved median LDL levels of 30 mg/dL—far below what was thought safe or beneficial. The result: a 15% reduction in major cardiovascular events and, critically, continued benefit down to LDL levels as low as 10 mg/dL with no safety signal.

The recently published VESALIUS-CV trial in NEJM went further, studying patients without prior heart attacks or strokes (primary prevention). Among 13,970 participants, evolocumab produced a 25% reduction in major cardiovascular events and a stunning 36% reduction in myocardial infarction. The benefit was consistent across all subgroups, including those with diabetes.

A 2024 review in Drugs on long-term PCSK9 therapy documented both excellent tolerability (minimal side effects even after years of treatment) and evidence of plaque modification on imaging. The substrate itself becomes less dangerous.

Children and Lipid Medications

All the medications listed above are FDA approved and appropriate for children, typically starting around age 8-10. Statins have 20 years of safety data showing no effects on growth, development, or long-term health. PCSK9 inhibitors and ezetimibe also have strong pediatric safety profiles. The key is working with a pediatric lipidologist who understands appropriate dosing and monitoring.

A 20-year statin study showed that starting treatment in childhood prevented early cardiovascular death compared to starting in adulthood. Zero cardiovascular deaths in the children who started statins early. At age 39: 100% survival.

The cumulative LDL burden matters - every year of elevated LDL damages arteries. Starting early = more years protected.

Summary

The conventional wisdom in T1D care has been almost entirely on glucose control—chasing perfect A1c numbers. And yet, cardiovascular disease has silently claimed lives decades too early.

The big takeaway is that you can win at both: Medications to lower apoB levels lead to a longer life. Managing glucose levels leads to a healthier one.

The Joslin Medalists survived to 70+ without perfect glucose control—but they also paid the price with eye, kidney, and nerve damage. You don’t have to accept that trade-off. With modern CGM technology and the knowledge in this article, you can optimize both.

Your next steps are simple:

• At your next appointment, request a complete lipid panel including apoB and Lp(a).

• If elevated, advocate for treatment - the medications are safe, proven, and inexpensive.

• Continue managing your glucose, but now with the confidence that you’re addressing the real killer.

Building intelligent glucose-management habits is 20% technique and 80% psychology. For more, see my article, The Four Habits of Healthy T1Ds.

As for me, you can read how I managed through all this in my article, Why I Haven’t Died Yet: My Fifty Years with Diabetes.

Share

Subscribe now

At a recent T1D conference, a clinician spoke about a child whose A1c was over 10%, and time-in-range (TIR) under 20%. Her parents were gripped with anxiety. So, their endocrinologist put her on an automated insulin delivery (AID) system, and immediately, her TIR went up to 70% within a week. By the end of three months, her A1c had dropped to 7.5%. Nothing was required of her or her parents.

The crowd gasped and applauded.

In another session on DIY (do it yourself) automated pump therapy, speakers explained how to tweak the algorithms in these open-source looping systems to set basal rates to zero, allowing for far tighter glycemic control, achieving A1c levels <7% and TIR rates as high as 85%.

The crowd gasped and applauded.

What attendees took away from these sessions is that if those users are benefitting in such headline-grabbing and extraordinary ways, everyone could benefit. And, indeed, that’s exactly the message echoed by advocacy groups, diabetes organizations, health care providers (HCPs) and diabetes educators.

And, of course, pump manufacturers.

The predominating message is that automation is not just better, it’s easier. “Diabetes can finally be put in the background.”

What is not discussed is the large middle group—people who are capable of self-management: They’re mature adults, have had T1D for years, but for whatever reason—and there are many—they choose automation over manual insulin administration. They may or may not have been necessarily fully adept at carb counting, dosing ratios, and other technical aspects, but may have otherwise just managed their disease adequately.

A decade’s worth of medical research, clinical trials and epidemiological studies published in peer-reviewed medical journals are revealing something unexpected and counterintuitive: The paradox of the Barbell Distribution of Clinical Benefit, where the true beneficiaries of AID systems are the far outliers in T1D demographic profiles—those who are either entirely unable to manage themselves, or are hyper-engaged in self-management—leaving those in the middle as either not realizing any material benefit from automation, or may be potentially harmed by it.

These harms, as the literature shows, are more subtle, and take time to evolve, including:

• A rise in obesity and deterioration of metabolic health (cardiovascular disease, aerobic fitness, insulin resistance, and other markers of type 2 diabetes)

• Loss of agency and stress resilience

• Increased costs

• Cognitive disengagement, including “de-skilling” of self-management

As a result, ethicists have been publishing about the moral hazard of promoting AID systems to those who are not able to benefit from them. Studies show that people can—and usually do—yield improved health outcomes when they engage with their self-management, especially for those factors that are entirely independent of A1c levels, such as managing lipids, exercise, and weight.

It’s not that people can’t do this with automated systems, they just don’t, or gradually drift away. The source of the problem, the literature shows, is the “attractive nuisance” of convenience: Why do the hard work if automation is thought to do better? And if it does all that, people tend to pay less attention to their overall health.

This article aim to unpack all these topics citing peer-reviewed articles published in high quality medical journals. Here’s a roadmap that we’ll follow:

• The Barbell Distribution of Clinical Benefit: The three primary groups of users who may or may not benefit from AID systems.

• Tracking AID system performance: We’ll review articles from 2018 to present day, and understand how the technology has progressed—or not.

• The MDI Comparison: How users on multiple daily injections compare to those who use automation.

• Physiological Barriers for AID algorithm improvement: AID systems may have hit an impenetrable performance ceiling.

• Hypoglycemia, Quality of Life, Financial Burden, Mental Relief: Measuring these other potential benefits of automation.

• The Creeping Rise of Obesity: The causal link between automation and obesity is no longer deniable.

• AID Systems and Exercise: To achieve healthy levels of exercise, automation is one too many cooks in the kitchen. Just self-manage.

• The Moral Hazard of Automation: Health care providers and advocacy groups need to change their messaging on T1D from “put it in the background” to “take agency for yourself.”

A Word about A1c Levels

Much of this article discusses A1c levels, because that’s the metric every study uses and every clinician tracks. But A1c is not the thing that determines your health, and it’s worth being clear about that up front.

In my article, A1c and T1D Complications: What It Reveals, What It Hides, and How the Damage Actually Happens, I lay out the three-stage process. Glucose binds to proteins — that's glycation. The bond stabilizes. And then, under oxidative stress, it hardens into irreversible cross-linked Advanced Glycation End-products, or AGEs.

Two people with identical A1c can end up in completely different places. Your A1c could measure X number of glycated proteins, but your lifestyle choices and other interventions might keep many of them from becoming oxidized AGEs. Or, worse, you could engage in things that make more of those oxidized AGEs.

If you’re looking at AID systems because you’re aiming to lower A1c levels, you’re not seeing the bigger picture. AID systems are a tool, and the risk to using them is what defines the paradox that’s the title of this article.

If you reach 7.5% using an AID system while you’re also not overweight, eat well, and get good exercise, you’re in good shape. If you reach it through more insulin, weight gain, defensive snacking, and giving up exercise because the algorithm can’t handle it — you’ve optimized the number and degraded everything underneath it.

It’s vital to keep all this in mind, because the sole aim of AID systems is to promote one of two ideas: First, that its algorithm can achieve better A1c levels than you can, which we’ll see is not true for most capable adults; and second, that automation can relieve the daily burden of self-management. Again, this is more illusory than it first appears.

Now, if this is surprising to you, let’s first review what the American Diabetes Association actually says about AID systems.

ADA Recommendations: Standard of Care?

Many people believe that AID systems are the “standard of care” for T1D. Medtronic’s website claims “the American Diabetes Association calls automated insulin delivery systems the standard of care for people with type 1 diabetes”—which mischaracterizes the actual ADA language.

This language can also be found in clinical trials and other published journal articles, but each of them were conducted and paid for by pump companies. T1D advocacy groups—including BreakthroughT1D and the Helmsley Charitable Trust—also make unqualified “standard of care” claims. For example, Helmsley’s website states: “Standard of care for T1D includes utilization of devices such as Continuous Glucose Monitors (CGMs) and Automated Insulin Delivery (AID) Systems, which are proven to improve health outcomes.”

This language actually exceeds what the ADA says. These organizations are not standards-setting bodies, yet their claims are stronger than the most authoritative professional medical association’s position.

These statements are all mischaracterizations of the American Diabetes Association’s actual 2026 guidelines, which say they “Recommend early initiation, including at diagnosis, of CGM, Continuous Subcutaneous Insulin Infusion (CSII), and Automated Insulin Delivery (AID) depending on a person’s or caregiver’s needs and preferences.”

To be clear, “preferred” is decidedly NOT the same as “standard of care”, as there are specific legal and clinical implications. For example, CGMs are considered “standard of care”, as per the current standards, which explicitly state: “Its use is now considered standard of care for most people with type 1 diabetes”.

And THAT is where the nuance comes in.

This article is not arguing that AID systems are inherently unsafe or will result in poor health outcomes in exactly the same way that the ADA is not saying that AID systems are “standard of care”.

In a way, this article mirrors ADA’s recommendations and guidelines, but gets into greater detail and nuance using the ADA’s own peer-reviewed articles. I’m just delving further into that research to help the reader understand with greater clarity who benefits, under what conditions, and when are flags raised that clinicians should consider. And those “who” and “what” exceptions are not only important, they are precisely why AID systems are not “standard of care”.

This is why the “specific legal and clinical implications” are important. If a patient suffers an acute event that leads to a legal claim, the defense’s position differs on whether the treatment was “standard of care” versus “their judgement, based on the facts and circumstances of the individual.” If a doctor prescribes an AID system believing that it is the standard of care, they have recourse if something goes wrong. If it’s not the standard, then they bear more responsibility.

What emerges is an inflation cascade: The ADA (the most rigorous body) says “recommend... depending on individual needs.” Medtronic’s marketing claims “the ADA calls it standard of care.” BreakthroughT1D and Helmsley declare it “standard of care” without qualification. Each actor claims more than the source supports, and the inflation is then falsely attributed back to the ADA. Patients and physicians hear “standard of care” repeatedly and assume it reflects consensus—without realizing the most authoritative body explicitly preserves physician discretion and individualized assessment.

This should be sufficient evidence alone that AID systems can and do have variability that, as the ADA guidelines suggest, should be taken into consideration by the healthcare provider. They know—as their own medical literature shows—that not everyone benefits the same way, and that there are limitations to these benefits, particularly when those limitations can lead to poor health outcomes.

Lastly, the European and American Diabetes Associations put out a joint position statement on AID system adoption titled, Automated insulin delivery: benefits, challenges, and recommendations. A Consensus Report of the Joint Diabetes Technology Working Group of the European Association for the Study of Diabetes and the American Diabetes Association. Much of the concerns expressed in that statement are similarly expressed here, though this article contextualizes much of it for the benefit of non-technical readers, and cites further journal papers that further demonstrate support.

Barbell Distribution of Clinical Benefit

For context, a “clinical benefit” is defined as the level of glycemic control (A1c and time-in-range, or TIR) that a person achieves with the AID system versus how well that same person might perform without it. And again, we’re only looking at A1c levels, which is not the full picture. But it’s still useful to understand how we go from “benefit” to harm.

We can identify the three groups in this barbell distribution as such:

• The first group are those who are more likely to achieve better A1c levels with the AID system than they can do on their own. These include children, cognitively impaired, those with several mental health conditions, hospitalized, or those who have not had T1D for long enough to have learned the techniques and empirical experiences that could make them better able to manage themselves.

• On the polar opposite end are those who are hyper-engaged. They not only understand their disease, but use this knowledge to fully engage with AID algorithms: meal announcements, carb-counting, pre-bolusing for meals, engaging in (and announcing) exercise, and other acts. Many are employing modified AID system algorithms that either meet their personalized individual profiles, or are hyper-aggressive in insulin dosing that may not be appropriate for the general T1D population. Because of their hyper-engagement, they could do as well with MDI if they transferred the same knowledge and focused drive, but the psychological enjoyment of using the AID technology is likely the primary motivator. Hey, if you’re healthy, you do you, babe.

• Those in the middle are everyone else—people who are capable of engaging with self-management, are mature, fully informed on self-management, and have had T1D for at least five years. But (for any number of reasons) have chosen to use AID systems. As the studies we’ll be citing later show, they tend to perform worse than their counterparts of the same characteristics, but who happen to use MDI. By “perform”, we’re not just talking about A1c levels — we’re talking about being engaged with your overall health. Those who manually dose tend to be more attentive to those other factors. (And you can manually dose with pumps too.)

Share

Tracking AID system performance

To track AID system performance, we are looking at two separate datasets. The first is the performance of AID systems themselves, and second, how they perform for different demographics.

To set a reference baseline, we’ll begin with data starting from 2018, when a research study by the T1D exchange that covered all age groups, and included AID system users and MDI users alike. Note that this data precedes AID systems, so we have a baseline for comparison.

The researchers found that “there isn’t a substantial difference between A1c outcomes for those who manually take insulin injections compared to those who use pumps,” (among those wearing CGMs). Below is a figure from that study. (The age group on the far right are those >26.)

Note that the majority of users above the age of 26 had A1c levels at roughly 7.5% (+/- .5%) for AID and MDI users who wear CGMs. And although we can’t differentiate the degree of automation in the pump group, it’s the baseline A1c levels that we’re interested in. This will become clear soon.

Now, let’s compare this data to more recent trials. At the 2024 ADA Scientific Sessions, researchers presented results from a Medtronic 780G study in 80 high-risk youth with T1D with an average A1c of 10.5%. After adoption of the automated system, they achieved an impressive 2.5 percentage point reduction, bringing their average A1c down to 8.0%.

Note that 8% is higher than the ~7.5% in the 2018 T1D Exchange study.

Similar results emerged from a 12-month prospective study by Michaels, Boucsein, and colleagues, published in Diabetes Technology & Therapeutics in 2024. Participants entered the study with an average A1c of 10.5% and saw the same 2.5 percentage point improvement to 8.0% and time-in-range nearly tripled, from 27.6% to 62.5%, representing a dramatic improvement in day-to-day glucose stability.

A published in Lancet Diabetes Endocrinology, 2023, titled, Advanced hybrid closed loop therapy versus conventional treatment in adults with type 1 diabetes (ADAPT): a randomised controlled study, enrolled patients with A1c > 8%, and within six months, A1c decreased from 9% to 7.32%.

Ok, we get it: AID systems can improve your A1c levels if you’re starting at higher levels than 8%.

What is never promoted, however, are studies that include participants whose A1c levels are lower from the outset. An example is a 2021 study titled, Predictors of the effectiveness of insulin pumps in patients with type 1 diabetes mellitus, where the authors conclude that the subjects whose A1c levels were >8% going into the study saw an improvement of 0.9 ± 1.2% when started on insulin pumps, whereas those whose A1c levels were <8% saw no benefit to pumps at all, irrespective of age, race, and income levels.

These are not unique trials—a literature review yields hundreds of trials with similar results.

Subscribe now

The MDI Comparison

Now let’s compare this to MDI users: In a 2024 paper titled, Shining the Spotlight on Multiple Daily Insulin Therapy: Real-World Evidence of the InPen Smart Insulin Pen (Diabetes Technol Therapeutics), the study included 3,793 T1D users using smart insulin pens with CGM. Researchers found that users who dosed at least three times daily achieved 70.3% time-in-range and A1c equivalent of 7.0%, matching or exceeding typical AID system performance.

That will probably tilt some heads: Could it really be that simple? Matching or exceeding AID outcomes with only 3x daily dosing?

It is true, but it’s also not the full story. Those who dosed >6x/day had worse outcomes: TIR 61% (10% lower) and A1c of 7.3% (.3% higher).

How could that be?

Because it’s not about quantity, it’s about quality. In this particular study, those who dosed more were also less efficient—they didn’t pair boluses with meals as tightly, and missed more meals as well.

As T1Ds know, once you miss a meal bolus—or even misdose it—you’re now on the glucose roller coaster and you’re bolusing more frequently to “catch up” to the sugar, which also leads to hypos from overcorrecting. In the literature review article, “Optimal prandial timing of bolus insulin in diabetes management: a review,” T1Ds who pay attention to glucose patterns are far more likely to remember to bolus for meals than those who do not watch their CGMs. Forgetting to bolus for only two meals per week results in A1c levels 0.5% higher than those who remember to bolus.

Think about the huge benefits of simply being engaged: A1c levels drop from 8% to 7.5% by doing nothing more than bolusing for meals that you might have otherwise missed. This applies whether you use a pump or not.

The limiting factor isn’t the technology, it’s the engagement. And MDI, by its very nature, is more likely for users to be engaged.

To be clear, MDI does not guarantee engagement. There are certainly plenty of MDI users whose A1c levels are far north of 9% as well. But it’s the long game that we’re after: Engagement comes from experience and a modicum of training. Basic foundational facts like insulin-carb ratios, correction ratios, best practices on carb counting, dosing for exercise are like the training wheels on a bicycle that children use to keep from falling. But, like training wheels, there’s a point where there are too many exceptions to the rule, and you have to diverge from the regular pattern and “read the road”, as it were. This is hard, but once engaged, it’s a quick learning curve.

In cognitive psychology, there is something known as “desirable difficulties.” These are obstacles that make learning more challenging, slower, and sometimes more frustrating in the short-term, but deeper in the long-term. One example of a desirable difficulty involves what is known as the “generation effect.” Simply: forcing yourself to come up with an answer to something before having one given to you primes your brain for subsequent learning.

The very nature of T1D forces this upon a person, and those who engage over the long term are those who struggled through the learning phase. And yet, the institutions are not only failing to encourage this, they’re encouraging the opposite—to adopt automation. Those on pumps are not challenged to learn their own disease, dooming them to a life of slowly eroding health.

Some may argue that automation does not necessarily remove all engagement - people are told that meal announcements help the algorithms perform better. But these announcements are more performative than genuinely thoughtful. Users enter a mindset of selective disengagement where users actively participate in prescribed tasks (bolusing for announced meals) but passively accept algorithmic decisions for everything else (basal rates, corrections, insulin-on-board calculations).

We can compare the outcomes of these groups using worldwide data.

In the United States, where approximately 66% of adults with type 1 diabetes use insulin pumps (2021), mean A1c levels range from 8.0-8.4% with only 21% achieving the recommended target of A1c <7%.

In stark contrast, European pump adoption varies dramatically by country—from less than 5% in many nations to 37% in Germany and Austria—yet glycemic outcomes range in the 7.6% range across eight high-income countries, with several registries demonstrating improving trends. The MDI users are performing so much better that it’s bringing down the national averages.

The temporal trends further underscore this paradox: mean A1c levels increased in the United States while Sweden saw decreases across all age groups and Germany and Austria reported declines among children, despite Sweden and Germany having substantially lower pump adoption rates than the United States.

The fact remains, the rise of automation has disincentivized user engagement, leading to higher A1c levels. And as more T1Ds in America migrate to automation, we’re finding a deviation towards the mean: Those with high A1c levels are improving towards the middle, whereas those with lower/better A1c levels are drifting upwards to the mean.

And that “mean” is an A1c of 7.5% and TIR of 70%.

One last argument proponents make—especially those who themselves perform well with AID systems—is that there is a population that does well that isn’t in that group of those who cannot manage their own T1D.

Interestingly, what these people are doing without realizing it is conforming their behavior to meet what the algorithm expects.

A 2022 review of AID’s psychological implications — Nefs, Frontiers in Clinical Diabetes and Healthcare 2022 — documented exactly these dynamics. Users weren’t just passively accepting the algorithm’s output. They were actively restructuring their lives to produce the numbers the algorithm rewarded: eating predictably, avoiding unannounced snacks, timing activities to coincide with low insulin-on-board, and — most consequentially — reducing or eliminating exercise that would introduce the kind of glucose variability the algorithm couldn’t handle. As Nefs noted, frustrations with and distrust in the system led to workarounds like entering fake carbohydrates, while some users actively limited physical activity or carbohydrate intake to increase time in range.

All this raises the next obvious question: What about the future? Surely, the algorithms will improve, right?

Why the Algorithm Can’t Improve — And Who Fills the Gap

All of this data converges on the same number: an A1c of ~7.5% and a TIR of ~70%. Across systems, across manufacturers, across years. The natural question is whether better algorithms can puncture through that ceiling. AID system advocates and data scientists — especially in the DIY community — believe they can.

They can’t. And the reason goes deeper than any individual engineering constraint.

In a companion article, Why AID Algorithms Don’t Improve: The Category Error, I lay out the full case: glucose metabolism is not a complicated engineering problem — like cruise control holding a speed on a highway — but a complex adaptive system, governed by the same mathematical principles as weather.

The short answer for why AID systems cannot possibly improve is chaos theory. There are simply too many parts that interact with one another to drive glucose levels in different directions, and this feedback loop (of which insulin is only one of many, many parts) is the quintessential archetype of a chaotic system.

The physiological details bear this out: CGM readings is the one and only data available to the algorithm, but it is blind to insulin absorption rates, hepatic glucose output, counter-regulatory hormone activity, and dozens of other state variables that are feeding back into the system itself. That is, those hormones and activities also affect glucose readings—it’s not just insulin alone.

This is called a category error because “control theory” only works if you can measure all the variables, and if those variables are reliably consistent. None of that applies to human physiology.

So what do the algorithms actually do? Far less than advertised. Tandem’s Basal-IQ engages a process called Predictive Low-Glucose Suspend (PLGS). According to Tandem’s own study, users on Basal-IQ achieved a TIR of approximately ~64%. No dosing algorithms, no correction boluses, no meal detection. That single safety mechanism — stop delivering a dangerous drug when the patient is heading toward a dangerous low — captured the majority of the achievable benefit. Everything above that 64% floor is the territory where an AID algorithm operates.

So now the algorithm has to figure out how to deal with hyperglycemia. Fine, but when you factor in safety controls and other factors, there’s not a whole lot the algorithm can do. In the study One Year Real-World Use of the Control-IQ Advanced Hybrid Closed-Loop Technology, (Diabetes Technology & Therapeutics, 2022) approximately 9,000 users went from a baseline TIR of ~63% (on Basal-IQ) to about 74% on Control-IQ’s full system.

So, there you have it. You can have no algorithm and just use the low-end PLGS and achieve 64% TIR, or you can apply an algorithm that gets you to 74%. That’s only a ten percentage point difference. And it’s a physiological constraint, not a matter of better algorithms.

When you strip away the control-theory language, AID systems operate as conservative safety heuristics: threshold-based rules, cautious correction doses, and wide risk margins designed to avoid catastrophic lows. They work within a narrow operational band — not because the engineers lacked ambition, but because the mathematics left them no room to be ambitious. This is why every commercial system — Medtronic 780G, Omnipod 5, Tandem Control-IQ, and the open-source alternatives — converges on the same outcomes. The algorithm isn’t the differentiator. It was never going to be.

Which means the only variable with room to move the needle is the one the algorithm can’t touch: the user’s behavior.

And here is where the paradox sharpens into something darker. If the algorithm is constrained to a narrow band, then the users who report the best numbers on AID systems are not the ones whose algorithms performed best — they’re the ones who adapted their lives to make the algorithm’s job easier. They eat on schedule. They avoid foods with unpredictable absorption. They reduce or eliminate vigorous exercise, because exercise introduces the very volatility the algorithm can’t handle. They simplify their routines. They become, in effect, compliant with what the algorithm demands: sitting quietly after a controlled meal.

And compliance has a cost.

The lifestyle modifications that optimize TIR metrics — predictable meals, minimal exercise, defensive snacking to prevent lows — are precisely the behaviors that erode metabolic health. They promote weight gain, insulin resistance, cardiovascular risk, and the slow accumulation of the very comorbidities that ultimately kill people with T1D. The numbers on the CGM look fine. The body underneath them is deteriorating. This is why the clinical trials that run for 3-6 months will never catch it — behavioral conformity is a slow process, and its consequences unfold over years, not weeks.

The irony is that an engaged user — someone actively managing their own insulin — has access to the one form of predictive power that actually works in a chaotic system: intentional knowledge. You know when you’re going to eat, sleep, and exercise. You know what kind of exercise. You know whether you slept badly or are coming down with something. None of this information is available to the algorithm, and none of it can be, because it exists in your mind before it manifests in your glucose. When you take iterative actions — dosing in advance of meals, reducing insulin before a run, adjusting for a stressful day — you’re not fighting the chaos. You’re navigating it with contextual knowledge the algorithm structurally cannot possess.

The user’s behavior is, and always has been, the disproportionately largest variable in determining both glycemic control and health outcomes. The question is whether that behavior is shaped by engagement with one’s own physiology — which improves health — or by compliance with an algorithm’s limitations — which quietly degrades it.

Hypoglycemia

Even with these limitations, advocates point out other advantages to automation, the first of which is that the algorithm can suspend insulin delivery when glucose levels drop to avoid hypoglycemia. This is especially problematic at night.

In theory, that’d be great, but in practice, it doesn’t pan out as well as hoped.

In a study published in the June 2024 issue of Diabetes Care, titled, Severe Hypoglycemia and Impaired Awareness of Hypoglycemia Persist in People With Type 1 Diabetes Despite Use of Diabetes Technology: Results From a Cross-sectional Survey, the authors collected data from 2,074 T1Ds, half of whom used AID systems, whereas the rest wore non-AID pump systems. The authors wrote, “the number of severe hypoglycemic events from automated systems reached 16% for those using AID systems, versus 19% of non-AID pump users.”

Sure, there’s a 3% improvement from non-AID users, but 16% is still a high incident rate for a system that purports to effectively eliminate hypos.

The most likely explanation is insulin absorption variability described earlier. That is, there’s likely a pool of insulin suspended in tissue that the pump doesn’t know hasn’t yet been absorbed. The pump may have stopped infusion, but the insulin pool from earlier infusion just keeps seeping through the tissue. When glucose levels drop, it’s already too late.

When faced with these outcomes, advocates then pivot to claiming other benefits of AID systems: reduced cognitive burden, better quality of life, decreased mental stress.

In theory, yes, but is this assertion seen in medical literature?

Cost and Quality of Life Considerations

As of 2025, a paper by Kommareddi & Wherry (AJMC April 2025) found that the Medtronic 780G treatment runs $10,258/year vs. $4,604-$4,688/year for MDI, plus another $5,600/year premium for the full AID system. It’s notable that the authors also work for Medtronic.

Multiple cost-effectiveness analyses, one from Sweden and one from Finland—two countries with the highest rates of T1D—demonstrate that insulin pumps are cost-effective in patients with A1c levels above 8%, but the benefit flatlines at—you guessed it—7.5%. At this level, insulin pens are not only cheaper for the healthcare system, but for the individuals themselves.

In the U.S., medicare’s requirement for A1c >7% for pump coverage isn’t arbitrary bureaucracy; it reflects actuarial analysis. And, of course, private insurers often follow suit. The coverage restrictions essentially codify the barbell distribution: pumps are approved for those who need help achieving 7.5% A1c, but denied to those already there.

The economic calculation appears validated: European healthcare systems with restricted pump reimbursement policies achieve equal or superior outcomes at lower cost, supporting the hypothesis that widespread pump adoption is not necessary for optimal glycemic control in capable patients.

Quality of Life

In the DIAMOND trial cited earlier, researchers unexpectedly found that pump users’ quality-of-life factors decreased by 0.71 compared to MDI, and life expectancy dropped by 0.48 years (based on higher risk of DKA, among other risk factors). Minimal differences from MDI to be sure, but these numbers certainly didn’t improve, as pump advocates would like to believe.

A more detailed understanding of this reduction on quality of life analysis can be found in the paper, Diabetes Device Use in Adults With Type 1 Diabetes: Barriers to Uptake and Potential Intervention Targets, published in Diabetes Care, 2017, which found that 47% of all participants endorsed the idea that “hassle of wearing devices” was a barrier to using them (whether they wore them or not), and 35% also endorsed the idea that “not liking devices on one’s body” was a barrier. These were ideas that the study participants agreed with, most of whom continued using the pumps. Interestingly, the authors wrote: “Individuals endorsing [these] barriers to device use also report more diabetes distress.”

Unfortunately, the paper does not ask or answer the questions: Why continue using the devices, despite the problems? Do these users believe benefits outweigh burdens? Are they being pressured by HCPs? Do they feel like everyone is moving in this direction, so they should too?

The most likely explanation is a commonly known phenomenon called, Social influence in device adoption. In a paper published in Social Science Medicine regarding adoption of healthcare technology, researchers found that while mean peer beliefs didn’t predict individual device use, “sharedness of belief” within peer networks did. That is, individuals’ own positive beliefs only translated into actual system use when those beliefs were shared with their peers.

In Type 1 diabetes specifically, the article, Barriers and facilitators to diabetes device adoption for people with type 1 diabetes, the authors noted that peer connections within the T1D community are noted to have great impact in decision-making and utilization of diabetes devices. And when device companies and T1D advocacy groups are the sole messengers of information, the results speak for themselves. This is particularly effective when self-management via MDI is downplayed by care providers and the messaging from the T1D community.

The Creeping Rise of Obesity

As the adoption of automated systems rises, medical literature has been revealing the association to obesity.

Once again, we’ll start in 2018 to see trends.

This 2021 article in Lancet titled, Obesity in people living with type 1 diabetes, shows that the rate of obesity reached 37% in 2018 compared to only 3.4% in 1986, where it refers to this phenomenon of overinsulinization. In the excerpt below, “insulin profiles” are part of the algorithms that determine insulin dosing for food, rest, exercise, and other conditions. The article states that:

… insulin profiles that do not match basal and mealtime insulin needs, defensive snacking to avoid hypoglycaemia, or a combination of these, are believed to affect body composition and drive excessive accumulation of body fat in people with type 1 diabetes.

By 2023, the rate of AID system adoption rapidly accelerated, and along with it, obesity rates. In a large U.S. study titled, Prevalence and Management of Obesity in U.S. Adults With Type 1 Diabetes (Johns Hopkins/CDC NHANES data, March, 2023), the authors found that 62% of U.S. adults with T1D were overweight or obese. This rate is catching up to the 66% of those using automated insulin pumps from the study cited earlier.

The Hopkins study found T1Ds were the LEAST likely of all groups to increase physical activity or reduce caloric intake, due to the complexity of adjusting insulin to prevent hypoglycemia.

The rise in obesity is not limited to T1Ds, of course—it’s an epidemic everywhere, and the T1D rate strongly parallels the general population. Hence, it’s tempting to think that T1Ds are just eating more along with everyone else. The difference is that T1Ds take insulin independently of food in the form of basal rates (for pumps) and basal insulin (for MDI).

In other words, non-diabetics will produce their own insulin to match the food they eat, whereas T1Ds have a steady drip of insulin that requires the person to eat in order to avoid hypoglycemia. This additional level of basal insulin goes beyond one’s physiological needs, and has been termed, overbasalization.

The situation has gotten so bad, the ADA’s latest (2025) guidelines identify overbasalization as being a major health concern for T1Ds. I get into extensive analysis of this phenomenon in a four-part series, the first of which is titled, Ye Olde Basal Rate: Its Effect on T1D Management and Long Term Complications.

While basal rates can and should be reduced in AID systems, that comes with tradeoffs that are hard to generalize to the population at large, which is covered in my four-part series. But tied very closely to basal rates—and AID systems in general—is the ability to engage in the healthiest thing anyone can do: Exercise.

AID systems and Exercise

As is commonly known, one great way to lose weight and improve overall health is exercise. And here’s where automation can truly be problematic. Once you engage with progressive levels of exercise, your insulin needs change, your body burns more fat, and glucose sensitivity changes.

Most importantly, your muscles absorb glucose during exercise without insulin mediation, a process called “non-insulin mediated glucose uptake”, or NIMGU. This efficiency increases as you become more physically fit. But the progression is irregular, so it can’t be predicted or programmed. Glucose movements during exercise are far too fast and volatile—as they should be in order to support the energy requirements—so, dosing decisions must be done “in the moment”, not delayed after glucose levels have already moved. See my article, T1D and Exercise: Insulin, Carbs and Glucose Control.

A meta-analysis of dozens of high quality, random controlled trials among T1Ds who exercise can be found in the paper, Time in Range for Closed-Loop Systems versus Standard of Care during Physical Exercise in People with Type 1 Diabetes: A Systematic Review and Meta-Analysis. Here, the authors concluded that pump users were not able to achieve a meaningful statistical improvement over control groups who didn’t use an insulin pump, irrespective of socio-economic background.

Weight control and exercise are far more important for long-term health, so A1c levels should not be the sole aim for ideal health.

The Moral Hazard of Automation

Decades of medical professionals have been calling out the ethical quandary of automation, according to the article, A critical review and analysis of ethical issues associated with the artificial pancreas, published in the journal PMC, 2018:

…the adoption of these systems involves complex questions about “patient agency“ and reliance on automation. If the technology makes the sedentary life “easy,” the path of metabolic adaptation becomes only “harder” by comparison.

In another paper, A Qualitative Study of the Views of Individuals With Type 1 Diabetes on the Ethical Considerations Raised by the Artificial Pancreas, individual T1Ds were interviewed about how automated systems make them feel. Among the expressed concerns include a loss of “personal identity and agency”. This is interesting because it reflects an important part of the T1D narrative, where many feel that T1D is dehumanizing—that they feel a loss of agency. This study suggests that it’s the pump/automation that does that, or at least, makes it worse.

Patient agency is essential for health, not just psychologically, but in the actual operations of the systems themselves.

And this sentiment was echoed again in 2022, in a systematic review of closed-loop systems in the Diabetes Journal article, Diabetes Technology: Standards of Medical Care in Diabetes—2022:

The most important component in all of these systems is the patient. [...] Simply having a device or application does not change outcomes unless the human being engages with it to create positive health benefits. [...] Expectations must be tempered by reality—we do not yet have technology that completely eliminates the self-care tasks necessary for treating diabetes.

Yes, it all comes down to engagement. The more a person is engaged with their self-management, the healthier they are. There are no studies at all that examine T1Ds with A1c levels < 6.5%, but the group does fall into demographic surveys, and this group has only one thing in common: engagement. It doesn’t matter whether one uses a pump or MDI.

But one thing that is known, as shown by the studies above and the commentary by researchers: Automation disincentivizes engagement.

What should change?

Perhaps most importantly, T1D advocacy groups should be more mindful about their messaging that T1D is “hard” and “you can put it in the background”. They should also be more circumspect in their universal advocacy for pumps and automation. This is the gateway message that leads people to opt for “easy” over “healthy”.

Automation should not be promoted as a one-size-fits-all solution. Yes, automation is good for children and those who are unable to self-manage. But people should be advised that automation is not a replacement for self-management and can lead to poor health outcomes.

T1D Literacy: There should be more and better materials that effectively teach basic mechanics for T1D management stratified for different ages and stages of T1D onset. Newly diagnosed T1Ds are given nearly nothing, and most of it is vacuous. Moreover, as automation and pump adoption increases, there’s less incentive for patients to learn self-management—and, more worrisome—healthcare providers are less likely to teach it. This lack of education may actually result in fewer people being “able to self-manage”.

Weaning Protocols: Every AID user who’s had T1D for at least five years and has the maturity to learn manual self-management, should take steps to do so proactively. In my article, Basal Insulin Reduction: A How-To Guide using “Active Learning”, I outline a two-week boot camp, where you glue yourself to your CGM, log everything, and learn how to read and react to glucose patterns. Observe the cause and effect from insulin, food and exercise to see how your body works, and manually administer insulin using an insulin pen. If you truly immerse yourself for two weeks, you can return to automation and see the difference in how the algorithm manages you. Nothing teaches like hands-on experience, making you better prepared to override settings when you know yourself better than it does.

The automation paradox isn’t an argument against technology—it’s an argument for honesty about whom that technology truly serves. For children, the cognitively impaired, and those otherwise unable to self-manage, AID systems are lifesaving. For the hyper-engaged willing to hack and tweak, they’re powerful tools.

But for the vast middle—those capable of effective self-management if encouraged to stay engaged—automation may be solving a problem that doesn’t exist while creating problems we’re only beginning to measure. Until we acknowledge this openly, we’re optimizing metrics that don’t improve outcomes, while ignoring the harm hiding in plain sight.

Subscribe now

On Halloween day, 2025, a very scary thing happened to Dexcom’s stock price: It dropped nearly 15% following management’s lowered revenue expectations. Worse, this came on the heels of an 11% decline on September 19, when a report by Hunterbrook Capital titled, Dexcom’s Fatal Flaws, revealed shaky business and tech shenanigans that the market was unaware of.

As bad as that may appear, the market may be missing something far more dire that’s hiding in plain sight. And it’s a lesson in “common knowledge” — something that you know, and others know, but also the fact that you know that the others know it. This will become clear very soon.

The background is that Dexcom, maker of continuous glucose monitors (CGMs) introduced their G7 sensor in 2023, as a follow-on to their G6 sensor that was wildly successful. The G7 was promoted as being far more “accurate” than the G6, surpassing all other CGMs on the market.

Soon after the sensor was released, however, users were complaining about its “erratic” readings. In November of 2023, I published my own deep dive technical analysis of it, predicting that it’s performance was so bad, it couldn’t possibly last for long, possibly causing potential harm. And my prediction, sadly, has come true: Many people have been harmed by it.

Hunterbrook conducted its own independent research for its report, and found that users were abandoning the G7, and doctors were migrating patients off as well. A Facebook group of 57,000 (and growing) of disaffected G7 users talk about horrible experiences. All the while, the number of serious “adverse events” had been growing, where 56 people had died. All of it is due to the G7’s “poor accuracy”.

The snowballing of problems continues, as class action suits are increasing. The first was filed in California’s Central District by a parent whose child uses the G7 to manage Type 1 diabetes. They claim the G7 yields “dangerously inaccurate glucose readings“ that put users at risk.

Though this lawsuit will have a nearly immaterial financial effect on Dexcom, it does reveal the consequences of chasing a marketing term—”accuracy”—to an unsophisticated general public. The reality of medical devices will always emerge.

The MARD Problem

What could have possibly caused everything to unravel so badly?

The part hiding in plain sight is the word, accuracy. The problem with this term is a technicality that, at first blush, seems counterintuitive, but once understood, seems obvious. And this was the inside secret that all CGM makers have known about, but didn’t think it would really snowball into a catastrophe.

Let’s rewind a bit: Back in the 1980s, the first finger-prick blood glucose monitor (BGM) was made available for home use. It was the only way for a diabetic to know their glucose levels. They would use it to make self-management decisions, such as whether to dose insulin, and by how much. This had become the “standard of care”, because it was the best technology available. It was far from perfect, but better than nothing.

When CGMs were developed, the only way to determine their “accuracy” was to compare readings from the CGM to those BGM readings. And that’s what brings us to the term, MARD, which stands for Mean Absolute Relative Difference. The “relative difference” refers to how closely CGM readings match blood glucose analyzer readings from a finger-prick blood sample (capillary blood). The lower the percentage difference, the more similar the CGM reading is to the blood sample, the more “accurate” the device.

The G7’s MARD is 8.2%, compared to the G6’s 9.1%, which means that the G7 is a more “accurate” sensor—and at least at the time it was introduced, was the most accurate sensor on the market, and the company has always marketed itself vigorously with this claim.

I get into far more technical detail with more charts and data in my article, Continuous Glucose Monitors: Does Better Accuracy Mean Better Glycemic Control?

To be clear, the FDA does NOT mandate MARD—it is merely a measurement that the FDA uses to assess whether the device technically “works” — that it detects glucose. The problem is that CGM manufacturers are building hardware and algorithms to try to optimize for MARD as a marketing term to compete against one another, and to make end-users feel confident.

That was all well and good, but what was discovered in the process is that frequent blood glucose readings didn’t really help diabetics as much as assumed. Better, yes, but there were real problems with the readings: they were all over the map.

That’s when scientists investigated more closely and found that blood glucose is highly volatile, making it hard to measure with precision. When each fluidic sample can vary from one another, the ability to infer what a person’s “systemic" glucose levels are” yields too much error. And that makes actual dosing decisions really hard. And, as it turns out, potentially dangerous.

Because this science was well-established, Dexcom’s G6 sensor addressed this by doing some sophisticated internal data analysis within the sensor to “smooth out” those readings in ways that more closely resembled “systemic glucose levels”. Indeed, self-management was highly improved with the G6’s data.

The problem was that competitors started to come out with newer sensors that advertised themselves as “more accurate” than the G6, which put Dexcom in a bind. They were not, in fact, “better”, but they just happened to calibrate more closely with blood samples, which allowed them to claim “higher accuracy”. This was the technicality that they were leveraging to win the marketing wars.

To preserve their market leadership, they produced the G7, which reverted back to the old BGM-like data, which was technically “accurate”, but highly volatile.

This sleight of hand around this term “accuracy” became something that deep science researchers knew, and they know that it’s “common knowledge” among other industry players. But as long as the customer didn’t know, that would be fine. At least everyone could compete with one another and keep the arms race going.

Bad Marketing Gone Awry

Now, why not tell the customer? Largely because it would reveal there’s an upper limit on a CGM’s potential. CGMs would become a commodity, and there would be no way to differentiate them from one another. This is exactly what happened to blood glucose monitors, which are now dirt cheap from your corner pharmacy. If this happened to CGMs, it would move a multi-billion dollar industry to China, who would sell G6-quality CGMs for about $10/each. It would be a race to the bottom.

By at least giving the appearance of competition, the CGM manufacturers could continue getting top dollar from insurance reimbursements.

CGMs are a huge industry.

This dynamic—where everyone in an industry knows a truth but continues acting as if they don’t—is what cognitive psychologist Steven Pinker calls “common knowledge” in his 2025 book, When Everyone Knows That Everyone Knows...Common Knowledge and the Mysteries of Money, Power, and Everyday Life (New York: Scribner, 2025). It’s not just that everyone knows something; it’s that everyone knows that everyone else knows it, creating a recursive loop of awareness. The business continues so long as this knowledge remains confined to insiders.

This leads to the question of why the FDA doesn’t change its standard for approving CGMs. Therein lies the rub, as the saying goes. What else can a CGM be compared against if not blood glucose? No one’s come up with an answer.

But that class-action lawsuit — the one that the market is ignoring (for now) — that’s the child in the crowd saying that the emperor isn’t wearing any clothes. Financially, it’s nothing more than a spanking. But it bodes very differently for the future of CGMs because now everyone knows that everyone else knows.

Subscribe now

The MARD Critiques: A Decade of Warnings

None of this is new — it’s the worst kept secret in the CGM world, and scientists and researchers have known it for decades. The article, “Limits to the Evaluation of the Accuracy of Continuous Glucose Monitoring Systems by Clinical Trials”, describes the erratic and random patterns of glucose fluctuations, and calls into question the appropriateness of how clinical trials for CGMs are conducted in the first place. Popular metrics (like MARD) can be misleading without considering trends and consistency.

But it gets worse—MARD itself can be fudged by device design and how the clinical trials are done.

In 2024, Medtronic’s Chief Medical Officer Robert Vigersky published a review titled “The Myth of MARD” in Diabetes Technology & Therapeutics. The title alone tells you where this is going.

Key findings:

• “There is limited evidence that MARD has meaningful clinical relevance in the current era of sensor technology”

• MARD is “substantially impacted by the design of the data collection in an accuracy study”—meaning study design alone can change MARD values

• Kovatchev et al. recognized that “the use of metrics such as mean absolute relative difference (MARD)... as a sole source for accuracy assessment undervalues substantially the information carried by CGM about the glucose fluctuation process”

In a 2024 presentation, Vigersky’s team demonstrated that for a sensor with an overall MARD of 9.2%, simply adding 15% more glucose readings in the hypoglycemic range changes MARD to 10.3%, while adding them in the hyperglycemic range changes it to 8.8%. Study design—not actual sensor performance—drives the number.

More damning: “Pivotal trials and real-world data of AID systems using sensors with MARDs of 9% (Dexcom’s G6) or 10% (Medtronic’s GS3) demonstrate that AID algorithms compensated for any of these minor differences with most system users achieving consensus recommended TIR or A1C targets”.

Translation: MARD differences don’t predict clinical outcomes.

German diabetes technology researchers published similar findings in 2017 and 2020, noting that “MARD should not be used as the sole parameter to characterize CGM systems, especially when it comes to nonadjunctive usage” and that “MARD does not take into account the key feature of CGM sensors, the frequency of the measurements”.

A 2015 study bluntly concluded: “MARD Values Are Not Always a Reliable Indicator of CGM System Accuracy”.

MARD became the marketing standard without being a regulatory requirement, and without evidence it predicts what matters.

Subscribe now

What’s Missing: Effectiveness Trials

In my 2023 article cited earlier, I noted there are no independent, third-party effectiveness trials comparing real-world outcomes between different CGM sensors.

None.

We have efficacy trials (does it measure glucose? Give it a MARD value) but not effectiveness trials (does it improve diabetes management?). We have manufacturer-sponsored studies, but not independent comparisons. We have MARD numbers, but not head-to-head time-in-range (TIR) data.

Think about that. Every drug is eventually put to the test by independent clinical trials that show whether it improves patient outcomes. But CGMs—which inform life-or-death insulin dosing decisions—are not.

This cannot be a regulatory requirement, but it is something that independent researchers could do. But who? This is an exceedingly expensive task, and no one has the financial incentive to spend that kind of money, except for the manufacturers themselves. And they don’t because they know there’s no “there” there. It’s in no one’s best interests to expose a limitation that they all share and that everyone knows. If there’s an upper limit on what the tech can possibly do, what’s there to show?

To confirm it myself, I designed a cross-over study that I self-administered (which is included in my article), where I wore both the G6 and G7 at the same time, but swapped between which one I used to manage my day-to-day dosing and feeding choices. The results were clear—my time-in-range (TIR) was a stunning 95% when I only looked at data from the G6, and it dropped to ~70% when I only used the G7. Here’s the data from that study, which, again, you can see in my longer article on the topic.

It’s an N=1 study. Not an RTC. But those could be done, especially with automated pumps.

And that’s what brings us to the lawsuit—it’s a harbinger of more harm to come because “better” MARD values actually translates to adverse events. And so long as the CGM business strategy relies on ever-improved MARD, the CGM business segment faces some unrelenting headwinds.

Dexcom’s lowered earning forecasts may continue downward much further into the future so long as their technology is geared towards “accuracy” claims. Not even glucagon can bring these levels back up.

The Real Solution: Focus on the Future

The ceiling on CGM technology has been hit, so the CGM market is technically “mature”. There’s no new ground to conquer. Now that everyone knows what everyone knows, CGMs should devolve into a commodity product, where new devices are no longer about being technically better, but by having an improved user experience.

You know, like phones. You don’t see Apple and Samsung competing on their ability to make a phone call.

The G7 (when it actually works) offers far and away a better user experience than the G6, with its smaller size, simpler insertion process, shorter warmup window, and more. It’s the data that’s bad. Like a phone that has noise and drops the call. We are way past that, and yet, by trying to market a “better” CGM, they’re only making it worse.

It also calls into question: why bother to compete in the T1D market anyway? No, do not abandon the market, just remain as the market leader and invest in the future.

And to Dexcom’s credit, they have. There are approximately 4 million T1Ds in the US (representing 55.6% of current CGM revenue), but Type 2 diabetes accounts for over 90% of diabetes cases worldwide (Future Market Insights), and it’s the faster-growing market.

For T2Ds, CGMs are just entering the market as behavior modification tools, exemplified by Dexcom’s new Stelo OTC device. And Dexcom’s latest Q3 2025 earnings report validates this strategy: Stelo surpassed $100 million in revenue within its first 12 months on the market, while overall revenue hit $1.21 billion (up 21.6% year-over-year) with full-year 2025 guidance raised to $4.63-4.65 billion, according to Investing.com.

Dexcom could easily streamline their entire manufacturing process by producing a new single CGM that has the improved user experience of the G7, but all the precision benefits of the G6. It’s the “uni-sensor” that works for everyone.

This is insanely simple and easy because they’ve already sunk their R&D costs into everything: The G6 tech has already been built and approved—it just needs to be repackaged into the G7 applicator and other “experiences”.

Manufacturing a uni-sensor allows for shifting focus to the T2D market while also entering into less-expensive/higher-margin revenue streams like apps, data services, and so on.

From a marketing perspective, they could finance “effectiveness trials” that demonstrate this new uni-sensor yields healthier outcomes among diabetics of all types.

And now that everyone knows that everyone knows, it’s a race to see who the “first movers” in this direction. It’s an economically efficient, high-margin business model. When things become this commoditized, expect to see very good, useful, and inexpensive knock-offs from Asia, where manufacturing is dead simple.

In fact, this is already happening: Abbott has been involved in patent infringement lawsuits against two Chinese companies—Sinocare and SiBio Technology—over CGM technology, primarily in Europe. The litigation is ongoing, but Abbott is not seeing slam-dunk support from the courts. They haven’t lost yet, but it’s more likely than not that the Chinese firms will just find another way around the patents. This tech is mature enough that it isn’t that hard.

The point is, there’s a logical endpoint to the profitability of CGMs as a novel technology, much the same way glucose test strips can now be purchased for pennies (when they used to cost $1.25 each!). Dexcom’s G6 precision is as good as it can get, and it’s not rocket science on how to replicate that. Now that everyone knows that MARD was never really useful in the first place, the future of CGMs is likely going to be a commodity.

So, are we done here?

Not quite. The company’s mistake in focusing on MARD was a lesson in the business risk of pretending that no one knows what everyone knows. And that leads to the next risk on the horizon for Dexcom—ketone sensing—where the same phenomenon is happening.

Share

Ketone Testing

Dexcom’s next CGM (the G8) is slated to also detect ketone levels. This is driven by the same rationale that drove MARD: A perceived market opportunity by tapping into T1Ds’ fears of diabetic ketoacidosis (DKA). At face value, this makes sense: Up to 40% of T1Ds were diagnosed with the disease because they went to the ER with DKA. It’s horrible. So, the fear is real. Dexcom is looking to tap into that by offering ketone detection, where alerts will give people the chance to “correct” for it—namely, take insulin.

The problem is, ketones are very healthy aspects of normal human physiology. You produce them when you lose weight and exercise. In fact, statistics show that once someone is diagnosed with T1D, the incident rates are vanishingly low, with the predominant cases are those with severe eating disorders who intentionally withhold insulin to lose weight.

This video explainer may be very useful.

Once again, researchers and scientists know this. Everyone knows this. It’s “common knowledge” within these scientific communities, just as MARD was common knowledge that it’s a useless marketing term.

So, Dexcom’s chasing this market risks making the same mistake. What the company underestimated about MARD was the risk of producing a CGM whose better accuracy might translate into users being harmed. That seemed counter-intuitive, so it caught them by surprise. But lesson learned. The same can be said of marketing a device that detects ketones, even though there’s no real, tangible, clinical benefit, and carries health risks.

What are those risks? I answer that in my article, Ketones: The Unjustly Demonized Villain in T1D Management (and is also explained in the above video). If users see ketones, and they are told to take insulin to treat them, there can be some pretty serious “adverse events” if actual glucose levels do not warrant that, and those incident levels will rise if Dexcom’s CGM is also using the same erroneous MARD assumption.

Even before people are harmed, the problem may take care of itself through insurers: Once they see that the additional expense associated with ketone testing doesn’t actually translate into lower DKA incidents (which are already very low), they may not cover these sensors, even before adverse events take place. This could be another big hit on Dexcom and others.

There is one potential business opportunity to ketone sensing in CGMs: the potential uptake of GLP-1 and SGLT2 drugs in T1D management. This applies only to T1Ds, because DKA risk is not a concern for the T2D market, which, as noted above, is where the real market opportunity lies.

The reason that ketone testing would be useful here is because these drugs lower glucose levels in ways that do not require insulin. For those looking to lose weight, the GLP-1 drugs will suppress appetite, which reduced glucose levels, while also requiring lower insulin levels.

This presents two very risky circumstances: taking any insulin when your glucose levels are low risks severe hypoglycemia, whereas, taking no insulin is exactly how DKA sets in. This is a no-win scenario on both sides.

Of course, dosing the drugs to prevent these outcomes is the real challenge, and it’s not yet been established that this is achievable, at least, while assuring minimal risk in two critical areas. Indeed, the researchers need a CGM that detects ketones. As one diabetes expert put it in ADA Meeting News: “I don’t think we will get additional SGLT2i trials in type 1 diabetes until we get continuous ketone monitoring to better assess DKA risk”.

That may well be, but should Dexcom take the business risk? I would argue that there’s a trifecta of risks for pursing ketone testing: The market is not that big to begin with, it’s not that profitable, and it’s still unproven that these drugs actually provide the potential benefit to T1Ds that they appear to benefit the T2D community.

Even in the best case scenario, the “massive market opportunity” is there because drugs are insanely expensive and profitable. By comparison, CGMs are not—at least, not at the low volumes this potential market presents, especially compared to the massive T2D market for the uni-sensor.

It makes sense for Abbott to make a CGM that detects ketones because they have a drug business to support. This is the textbook case of a “companion diagnostic” to support a profit center (drugs). But Dexcom?

Well, they could chase that market because they hear the screams from unsophisticated and frightened end users, but they’d be putting a lot of resources into a very small percentage of T1Ds that don’t actually need it, and who could inadvertently suffer harm.

Conclusion: When Lawsuits Can’t Fix What Markets Must

To summarize:

1. MARD is a flawed metric that doesn’t predict clinical outcomes, and people are harmed by “greater accuracy” CGMs.

2. Current CGM technology is at its ceiling due to fundamental physiological constraints.

3. The regulatory framework incentivizes the wrong optimization targets.

4. The marketing and product strategy is wrong for a commodity technology with limited differentiation potential.

Sure, the company can recover by improving customer service and other public-facing operations, and that will go a long way. That’s what a market spanking will do. But the long-term lesson is clear: Don’t be fooled into thinking you can continue pretending that no one knows what everyone knows. That’s what happened to the naked emperor.

Share

Welcome to the next sequel in the T1D Cinematic Universe franchise, featuring the Islet League, a group of superheroes that aim to make life better for T1Ds. At least, that’s their goal.

In the first film, we introduced GlucoGirl, the well-intentioned, but dysregulated master controller of your Alpha Cells. They produce glucagon that signals your liver to produce glucose.

Buuuuut, she’s not always that good at it. And yet, she is super smart about the science behind liver’s glucose production, as she helped three T1Ds learn how to manage exercise and T1D.

In this sequel, GlucoGirl takes on a new invention that aims to fix her own dysregulation while promising to provide T1Ds with effortless glycemic control: The Iron Pump Duo: The Insulin+Glucagon Delivery System. (It also comes as an injectable “basal” formula.) Note that this is a preview because these technologies are currently under development. None are available commercially, as they are going through trials to get FDA approval.

The premise of this robotic duo is that, by incorporating insulin and glucagon into a single automated pump, the T1D has a safety net on both hyper- and hypoglycemic excursions. Insulin clips off the highs, and glucagon treats the lows by stimulating the liver to produce glucose when GlucoGirl falls asleep on the job.

Let’s preview the opening scene:

Scene One: The Setup

[ Dramatic, haunting, suspenseful music. It’s a dark alley. A T1D just left a restaurant on his way to his car. He notices his CGM shows 190. He sighs. A shadowy figure in a trench coat and fedora approaches him and speaks in a low, deep voice. ]

“I have just the thing for your high blood sugar.” He holds out what looks like a briefcase, opens it slowly, and a golden glow emerges. “This will throw your blood pressure through the roof, your heart rate will skyrocket, your blood glucose levels will shoot higher, and then plummet all the way down again. Sweat will pour out of you like a fire hose, and you’ll be panting hard to catch your breath. But it makes you feel really good. It’s perfect for your blood sugar, really. It’s really euphoric. Best of all, it’s highly addictive. Once you adapt to it, you’ll be a dedicated user, probably for life.”

The T1D initially looks interested because of the golden glow, but shies away, worried he may be in the wrong movie genre.

Instantly, a bright light emerges from behind them: The Iron Pump Duo swoops in, glowing brightly, landing between the guy and the shady character. The duo, each dressed something like Iron Man, but different enough to not trigger a copyright violation (but it wouldn’t, because this is satire). One has an “I” on his chest, and the other a “G” – and with blond hair coming out from the back of her helmet. They speak in unison in monotonic, male+female robotic voices:

“Ignore that man. We have a new insulin pump that delivers both insulin and glucagon. It would mean you would rarely ever get hypoglycemia again, your TIR could go over 85%, and your daily management would be far simpler. Almost hands-free. For life.”

The T1D looks at the duo in amazement. He excitedly replies, “Of course I would use that! Who wouldn’t?!”

The Iron Pump Duo continues in their flat, robotic voice: “Preliminary studies show remarkable results.”

One article, Dual hormone fully closed loop in type 1 diabetes: a randomised trial in the Netherlands – study protocol in the journal BMJ Open, 2023, the authors note that these systems: “remove any diabetes management burden related to meals or exercise“ and cite preliminary data showing median Time-in-Range (TIR) of 86.6% with lower time in hypoglycemia compared to traditional insulin pump treatment.

“Wow!”, says the guy. “Removes ANY diabetes management burden! – that’s awesome!”

The Duo continues:

Similarly, the concept of a smart insulin is where an insulin-glucagon fusion protein is formulated like a basal insulin, where you take one injection and it sits in your body, waiting for glucose movements. When levels rise, one protein dissolves proportionally to release the right amount of insulin; when glucose levels fall, another protein dissolves to release a glucagon-like component to raise glucose levels.

According to the article, A novel mixture of insulin and glucagon, published in the journal Type 1 Diabetes Grand Challenge, 2024, the aim of the research is to ‘protect against hypos by activating glucagon if blood glucose levels fall too low,’ without the need for constant monitoring.

The guy is more amazed. “Wow – yet another amazing claim! Sign me up!”

But before he takes the device, a huge burst of light appears from the sky—GlucoGirl swoops in. As the mist clears, she’s seen standing between the T1D and the Duo, arms outstretched to keep them away, “No! It’s not that simple!”

The guy is shocked. “Why not? The dual-hormone technologies will keep glucose levels really stable and low! With no effort on my part! Isn’t that the biggest challenge in T1D management?”

GlucoGirl gives a side-eye glance at the Duo with a stern look of disapproval. She then swiftly turns her head to look directly at the guy, tossing her hair dramatically. She shifts to a gleeful smile; a cinematic twinkle in one eye catches his attention.

Subscribe now

“Here’s the thing, my little sugar-laden friend. Low A1c levels are great, but it’s only one key to a long and healthy life. You need all seven keys in the T1D universe. The rest include cardiovascular fitness, mitochondrial function, lipid management, blood pressure control, appropriate insulin dosing, and weight management.

Low A1c levels are good, yes! In fact, high A1c levels can make things worse. But if your A1c’s are like non-diabetics, you can still suffer from debilitating diseases. Low glucose does not, by itself, confer metabolic health.”

The T1D guy thinks to himself. Sugar-laden? Really? He looks at his CGM again… uh… ok. Yeah, I’m sugar-laden.

GlucoGirl continues: “As it happens, that crazy substance that mysterious figure was offering you was vigorous exercise, and it’s the healthiest thing you can do for a long, healthy, happy life. It makes your muscles more insulin and glucose sensitive, improves blood pressure, lowers cholesterol levels, stabilizes hunger, reduces stress, improves gastric emptying, speeds recovery, and above all, improves your cardiovascular and metabolic health. And exercise is the only way to obtain all these benefits.”

“But what about all those scary things he described, like elevated blood pressure, heart rate, and all that?”, asks the guy. “Those are all the things that you said are bad!”

“Ah yes, but those are just short-term effects that happen during exercise. He said all that for dramatic effect. To build tension and drama. This is the cinematic universe, my sugar-puff daddy.”

Sugar-puff daddy!? Really?

“T1Ds—or anyone, for that matter—live longer and healthier with exercise, but they also say it’s highly euphoric, and yes, addictive. Just as he said. It’s part of their lifestyles, as it involves the great outdoors, social activities, or just fun playing sports. Come walk with me as I explain how that dual-hormone pump may not be what you really think it is.”

She and the guy start walking. The Duo, feeling left out, follow along. GlucoGirl then explains…

Share

Scene Two: The Faustian Bargain

“You see, my little sugar-plum, here in the T1D universe, metabolic health is more directly associated with long-term health than A1c levels. One study called, the Golden Years Cohort, examined those who’ve had T1D >50 years. Surprisingly, their A1c levels averaged 7.6% (some as high as 8.5-9%). None had A1c levels in the normal range, yet they avoided severe complications. They all had very healthy metabolic biomarkers—low cholesterol levels, cardiovascular health, and other metrics.”

Before GlucoGirl could go on, the guy chimes in, “Fine! You convinced me. I’ll exercise! But why can’t I also use the dual-hormone pump? Wouldn’t that be the best of both worlds? Exercise AND low A1c levels?”

GlucoGirl looks at him with one raised eyebrow and then breaks the fourth wall to wink at the camera.

She turns back to the guy and continues: “Yes, you can get both low A1c levels and exercise. But the problem with the dual-hormone system, my little sugar-cube: by clamping your glucose levels (and other hormones, especially glucagon) within a tight range, it would prevent you from the level of exercise needed to improve metabolic health.”

Sugar-cube! WTF, girl! But the guy holds it inside.

She continues: “That’s right, sitting around is mostly what you’d be forced to do using these dual-hormone technologies, because vigorous exercise requires both insulin and glucagon levels to go much further out of range than what these systems will tolerate, even though it’s for a very brief period.

And since exercise would also change your physiology into a much healthier state—progressively over time—it would mean those systems could never keep up.”

The guy pipes up: “What if I turn it off during exercise!”

GlucoGirl continues, “Yes, even if you temporarily turned them off during exercise, that’s not the problem. Exercise improves—thereby changing—insulin and glucose sensitivity, which changes your insulin-carb ratios and many other aspects of glycemic control. Automated systems can’t track this—they can’t even estimate it. Your muscles’ glucose uptake accelerates—they have no idea. Your glycogen store replenishment at night—they have no idea. Nothing about your complex metabolic activity is visible to them. They don’t expect change; they expect—and impose—stasis. If you change, they can’t help.

The decision between easy and sedentary versus vigorous exercise is the classic Faustian Bargain: Choosing something that looks really attractive comes at a tragic cost that you don’t see till it’s way too late: Poor metabolic health. You’d never obtain all seven keys in the T1D universe.”

At this point, the Iron Pump Duo speeds up to walk alongside the two in order to participate in the conversation. They say sheepishly, but in a really off-putting, monotone, dual-gender voice:

“GlucoGirl has a point, but we’re not that bad. We’re also a tool that can be used for good. In fact, there are many T1Ds that are, sadly, unable to exercise: bedridden, hospitalized, or immobile for any number of reasons. There are small children, those with mental health disorders, or frankly, even really busy adults who either don’t have the time or inclination for exercise. Look, in the real world, they all can’t be superheroes. Those people don’t go through metabolic changes. They are, as you said, “static”.

GlucoGirl looks at them with a sympathetic look. “Aww! That’s so sweet! You’re right! There are those who may need you! But this guy!? My little sugardrop? You offered it to him without even knowing whether he was a good candidate for it! Shame on you!”

The Duo stop and look at each other in their signature herky-jerky robotic movements. Realizing what GlucoGirl said, they hang their heads in shame.

GlucoGirl continues, looking back at the guy. “Besides, these systems are not even through the FDA process, so we don’t even know if they deliver on their promise, even for their target audience. But we can deconstruct the underlying science, and also look at independent studies on both the systems themselves, and their theoretical underpinnings. Let’s keep walking; pick up the pace, kids…”

Subscribe now

Plot Twist: We Already Have GlucoGirl. She’s just dysregulated.

See, the Iron Pump Duo delivers both glucagon and insulin. At one level, this makes sense: We understand why T1Ds take insulin: because there are no beta cells to produce it at all. That’s why my little sugar-dumpling over here has high glucose levels right now. [ GlucoGirl reaches over and pinches his cheeks. The guy rolls his eyes. ]

But that’s not the case with alpha cells. Those are still in his body! That’s me, by the way. You know that right? I’m GlucoGirl, the metaphor for Alpha Cells. [ The Duo nods in sync. ]

I used to be awesome at controlling alpha cells. And that’s when my twin brother, Beta Bro, controlled the beta cells that make insulin. He was my partner in ultimate glycemic control. He would signal me to produce just the right amount of glucagon to keep glucose levels balanced with his insulin production.

But when that rogue assassin from the autoimmune division came and offed him, I was alone! I can’t regulate glucagon production properly without his making insulin!

That does not mean that I don’t produce glucagon at all. I am alive and continuing to create glucagon. Being “dysregulated” means that I make glucagon at unexpected times in unexpected amounts. I’m highly dysregulated.”

GlucoGirl whimpers; The guy reaches over and pinches her cheek. GlucoGirl feigns a girlish shyness. Then continues…

“But this doesn’t keep me from being a T1D superhero; I genuinely offer more benefit than harm. The T1D who learns and understands my dysregulation learns to manage around those erratic behaviors, and often uses them to their own advantage.

But they can’t do that with a dual-hormone system: The Duo’s basic design assumes I’m dead, so little Miss GlucoBot over there just injects glucagon without considering whether she’s adding more to what I may already be making.

Insulin I understand, but glucagon? No girl. Bad idea. You and I are going to have a major fight scene. I hear the special effects team is currently working on some explosive ideas to depict this: BAM! POW! KAPOOEY!

And by the way, who does your hair, girl! And why!? You’re a BOT! You don’t need hair!

Anyway, even under the conditions where this dual-hormone system is intended to cover—namely, hypoglycemia—Miss GlucoBot’s exogenous glucagon still has another major hurdle: HER glucagon is not delivered through the portal vein the way I do. Remember, I may be dysregulated, but I have a direct, instant pathway to where glucagon is supposed to go: to the Liver. You, GlucoBot, do not.

Instead, the Pump Duo delivers both glucagon and insulin through subcutaneous tissue, where it travels along a tortuous and inefficient pathway into the bloodstream where it eventually lands in the liver. By then, some unknown and inconsistent portion of it will have arrived. This could take 15, 20, 30, or even 60 minutes. Everyone’s physiology is different.

It’s like an actor trying to act their part over email using an old dial-up AOL connection. It’s just going to be a bad performance.”

GlucoGirl looks at the guy, who looks confused.

“AOL, dude. Sugar-booger! Forget it. Ask your parents.”

Finally, the guy pipes up: “Okay, Miss GlucoGirl with the flaming hair and a cape that doesn’t make any sense at all. Given these pitfalls with the Iron Pump Duo’s delivery of glucagon, how does it square with the initial published articles on these dual-hormone pumps, showing TIR values of 86% and users not having to interact with the system at all?”

GlucoGirl continues: “Ah, my sugar-fairy, as is always the case with reports from the developers themselves, they conduct small, highly controlled lab conditions in order to establish a best-case scenario. Like a movie shot entirely on a soundstage with perfect lighting and no weather, these ‘proof of concept’ experiments never face the chaos of real life—the unpredictable nature of living.

C’mon, walk faster! Keep up!

You see, under real world conditions, dual hormone pumps don’t actually work so well, according to many independent studies:

A 2023 Frontiers study (Lindkvist et al.), used dual hormone systems that involved exercise, where the researchers noted “the risk of hypoglycemia is higher due to increased insulin sensitivity, insulin absorption, glucose uptake in combination with an impaired glucagon secretion“.

A more comprehensive review of many independent studies was published in a 2022 meta-analysis (Zeng et al., Diabetes Obes Metab). Here the authors state that “Postprandial glucose excursions and exercise-induced hypoglycemia represent major hurdles” with dual-hormone systems, and that there were “gastrointestinal symptoms” as side effects. (Excess glucagon levels can have other disruptive effects that are only now being understood.)

From a 2024 Lancet real-world study, the authors state, “Even after 1 year, management of insulin administration for intake of carbohydrates and physical activity can be burdensome.”

Even without exercise, researchers have found that simple day-to-day use does not perform nearly as well as the controlled studies. In the article titled, Glucose-responsive insulin and glucagon delivery (dual-hormone artificial pancreas) in adults with type 1 diabetes: a randomized crossover controlled trial in the journal CMAJ, 2013, researchers demonstrated that while the dual-hormone delivery system resulted in a lower percentage of time in hypoglycemia, it did not result in a significant improvement in overall TIR compared to a conventional insulin pump.

That paper was all the way back to 2013! That’s over 12 years ago, and the studies it reviewed were even older! Heck, teenagers weren’t even addicted to cell phones yet, or forming relationships with AI chatbots! And the dual-hormone system still isn’t released yet?”

The Iron Pump Duo look at each other and shrug in an I-Don’t-Know sort of way.

“Even if they were to eventually work, the ideal users are those who, as suggested earlier, are sedentary. For them, this could be a quiet, unassuming semi-hero that keeps their glucose levels in a tighter range than what they can do on their own.

For everyone else, they’d be stuck using a technology that forces them to be sedentary. And if that’s the case, guess what happens next…”

When Heroes Collide: Exercise vs. Automation

A sedentary lifestyle—even with ideal A1c levels—has been long known to be the most unhealthy, and we see this in over 150 million Americans. Yes, these are people whose bodies create an infinite supply of insulin that just adds increasingly more weight. And those people eventually become type 2 diabetics.

And yet, that exact same thing is happening with T1Ds. In a large U.S. study titled, Prevalence and Management of Obesity in U.S. Adults With Type 1 Diabetes (Johns Hopkins/CDC NHANES data, March, 2023), the authors found that 62% of U.S. adults with T1D were overweight or obese, where the majority were those using automated systems: “The very tools used for optimal glycemic control clash with weight management goals, leading to a high prevalence of overweight and obesity.”

This is why there is now an epidemic of “double diabetes” among the T1D population.

In fact, the ADA has identified “overbasalization” as being a major health risk in their 2025 guidelines. Yes, it’s that big of a deal. For more, see my four-part series on overbasalization.

And these dual-hormone technologies, whenever they hit the market, can exacerbate that very problem.

That’s the sad reality and the moral hazard of all automated systems. They perform best when they’re not in full automation mode, where users engage with them by announcing meals, exercise, and so on. The T1Ds who do that do not “put diabetes in the background”. In fact, the more engaged you are, the less you need automation. If anything, automation creates more work than if you just managed T1D yourself.

Decades of medical professionals have been calling out the ethical quandary of automation, according to the article, A critical review and analysis of ethical issues associated with the artificial pancreas, published in the journal PMC, 2018:

…the adoption of these systems involves complex questions about “patient agency“ and reliance on automation. If the technology makes the sedentary life “easy,” the path of metabolic adaptation becomes only “harder” by comparison.

[ Director’s Cut: For a long list of medical literature that validates the harms of automation, emphasizing the need for user engagement, see my article, Medical Literature Analysis: The Performance Paradox of AID Systems. ]

GlucoGirl looks at the Iron Pump Duo, who are both hanging their heads low, trying to avoid eye contact. Glucogirl continues:

“Look guys, I know you mean well. And you may in fact, eventually, achieve the level of glycemic control that you hope to offer your real target audience. But there’s a slippery slope here. In fact, before this movie is even released—which it hasn’t—the pharmaceutical industry is already writing the next sequel, “GLP-1: The Hunger Games”, which raises the premise that requires an even greater suspension of disbelief, what if we combined GLP-1 agonists with the Iron Pump Duo?

Kidding. Not kidding. Keep walking…

Is that where we’re going, guys? Inventing another technology to fix the problem created by the prior technology? In fact, you guys—the Iron Pump Duo—are only here because your predecessor, RoboPump, led to this obesity epidemic in the first place. 37% of T1Ds! Obese! Not good!”

The group keeps walking into the cityscape as the camera rises above. The credits start to roll.

Post-Credits Scene: The Real Hero

As per the latest SAG contract, movie credits are now required to roll at a pace that seems to last forever, allowing friends and relatives of people actually involved in making the movie a chance to see their names displayed.

Meanwhile, the audience is really just focused on the group of four meandering through city streets, alleys, and eventually back where they started.

The final credit scroll finishes, leaving all of them in the scene standing in a small circle.

“I hope you learned something from this, my friend,” GlucoGirl says in a 1960’s style health documentary style dictation you might have seen in high school (even as late as the 1990s).

“Oh!? I’m your friend now? No sugar-this or sugar-that? What happened? Run out of witticisms?”

“No, my friend. Your blood sugar dropped. Take a look.”

The guy looks at his CGM. A clean 99. His trend is flat as a calm lake. He looks like he realized he was pranked, but in a good way.

“That’s what walking after meals will do for you, my friend. Just keep that up, and maybe do a little jog or bike ride now and then. But even if you don’t, look at what walking just did. It’s really amazing. That’s the healthy way to keep glucose levels lower.”

“Not to plug my own movie, but I will: To learn how to manage T1D and exercise, see the aptly titled prequel, GlucoGirl Explains T1D and Exercise: Insulin, Carbs and Glucose Control. If you liked this movie, you’ll love that one!”

The Iron Pump Duo looks at the camera, like they’re being left out. GlucoGirl hands GlucoBot a piece of paper and nods to her that she should read it to the camera. GlucoBot faces the camera and awkwardly starts reading in a robotic voice, stumbling through the words like someone just learning English:

“The biggest villains in the T1D cinematic universe aren’t any of these technologies, because some subset of T1Ds who are unable or unwilling to care for themselves can benefit from them. But that’s not everyone. It’s when automation controls you, that’s when things go awry. And disengagement is a truly slippery slope.”

GlucoGirl takes the paper from GlucoBot, looks at the camera and says, “The real hero in the T1D cinematic universe is the person who takes agency in themselves, engages with their disease, and develops physiological literacy.”

Fade to Black.

[ On your way out, consider sneaking into theater 2 to watch that prequel GlucoGirl mentioned. It’s still playing! Tickets are free! And don’t forget to sign up for the mailing list to be notified of new releases! ]

Share