At the time, the MoTrPAC consortium was one of the largest national initiatives where funding and serious attention was being allocated to studying exercise and human performance (maybe since the Harvard Fatigue Lab era, or at least to my awareness).

Then, in 2021 The Wu Tsai Human Performance Alliance was founded through a $220 million philanthropic investment from the Joe and Clara Tsai Foundation, with the following mission: “The Wu Tsai Human Performance Alliance is discovering biological principles to optimize human performance and catalyze innovations in human health for all.” The idea being that measurement of human performance gets us towards a better approximation of health for all people.

Now fast forward to the emerging massive umbrella that we call Longevity. Beyond just traditional metrics of disease detection like A1C for diabetes diagnosis, metrics of functional capacity are entering the conversation as biomarkers for healthspan that are upstream of chronic disease.

VO2 Max, Grip Strength, executive function (measured perhaps by cognitive task used in fighter pilots), and more are the new clinical benchmarks for preventive medicine and longevity. Tools that have long been used by exercise physiologists and elite performance coaches are now being repurposed. And it makes sense.

Subscribe now

Health is a spectrum, but our current diagnostic toolkit enforces an artificial binary. Normal or diseased. Benchmarks of fitness and performance can put fitness on one end and chronic disease on the other, with very clear implications for immediate and long term capabilities. While a “normal” result is constrained by disease-oriented biomarkers, human performance allows us to identify a ceiling (or reserve) much higher than population average (normal).

I’ve previously written about this trend in how biomarkers have evolved to adapt the needs of society and the technology availability at the time (see below), but I bring this up today as the race to claim the naming system for this class of biomarkers has emerged.

Intrinsic Capacity Makes Headlines

NY Times just called Intrinsic Capacity “The Most Important Longevity Term You’ve Never Heard Of.” Its based off of the World Health Organization’s method of classifying human function across five domains:

• Cognition

• Locomotion

• Vitality

• Sensory

• Psychological

The Lancet called vitality capacity the “underlying physiological determinant of intrinsic capacity” in their expert working group, which appears to be the quantitative subclassifications of five domains.

The idea being that if we can agree on how to quantify this novel classification system we can begin to understand how to get ahead of disease by measuring human function (and its decline) before disease arises or gets severe. This makes perfect sense, and given the vast majority of diseases afflicting the US and increasingly the world are largely preventable, the timing for a new toolkit is now.

But is this really a new paradigm?

Isn’t This Just Human Performance?

The idea behind moving from disease detection to improving functional capacity as the goal of modern healthcare is so that we can align interventions to ensure that people (patients, consumers, etc.) can enhance their capabilities to reach goals, overcome life’s daily challenges, and live a full and functional life for as long as possible.

Compared to the 17,698 results in PubMed for Intrinsic Capacity, there are 242,351 results for Human Performance. Wouldn’t we want to draw on the much larger evidence base, and apply these learnings (and frameworks, protocols, etc.) to proactive disease management? Maybe toss in new technology to make human performance more quantitive and scalable via wearables and minimally invasive sampling techniques.

Subscribe now

What we Call “Fitness” Today Will Simply be Health Tomorrow

The Intrinsic Capacity vs. Human Performance nomenclature example highlights a divide that I believe will close in the very near future. As healthcare shifts from reactive to proactive, we will see a continued convergence of fitness and medicine.

Performance appears to have an association with "elite athlete," which may be why the field of Intrinsic Capacity was developed by WHO rather than framed around performance. But I'll argue that the everyday athlete program will be the preventive medicine prescription of the future.

Your prescription will be sign up for a Hyrox. Compete in a 5k. Race your first triathlon.

What other prescriptions have driven SMART goal-meeting “behavior change” so effectively than joining a community that has committed to improving their health, and showing up a specific date to see how the work you have put in has allowed you to accomplish something?
I'd imagine many longevity physicians' goals are to improve their patient's cardiovascular and metabolic function, target loneliness, and improve self-efficacy.

The simplest prescription may just be sign up for a fitness competition; the same instinct that pointed me toward MoTrPAC back in 2017 is now showing up under a new name.

Usually it came right after the second most common question: what’s your specialty. Right before the third: why didn’t you do a residency after all that schooling?

I’ve never had a crisp one-line answer, so I figured I’d write out the long one. It also doubles as a decent update on what I’ve been doing for the last year and a half.

Subscribe now

Start with why I did the MD/PhD in the first place. I went back and reread my medical school applications recently, and nowhere in them did I say I wanted to be a practicing physician. I wanted to be a scientist who could run clinical trials that improved the health of a population, not just the patient in front of me. So skipping a 3-7 year residency, never mind a fellowship on top of it, was never really a departure from the plan. It was the plan.

That view took shape over three years at the NIH before medical school, in the Human Energy and Body Weight Regulation Core at the Clinical Center. It’s about as disease-agnostic a training ground as exists in medicine. As a core facility, we ran physiology studies for roughly 20 different labs at any given time, testing healthy volunteers alongside people with diabetes, obesity, rare metabolic diseases like lipodystrophy and MMA, and endocrine cancers. My job was to measure things: energy expenditure at rest, during sleep, at maximal exercise, after a meal, in the cold, on experimental drugs. Body composition by DXA, MRI, PET/CT, BIA, air displacement plethysmography. Insulin sensitivity by hyperinsulinemic-euglycemic clamp. If a sensor could pull a physiological signal off a human body, we were probably testing it.

What that job actually teaches you is harder to get anywhere else: what human function looks like when you stop sorting people by diagnosis. I remember pulling data on a group of subjects who all qualified, on paper, as the same thing — “normal, healthy volunteer.” No active diagnoses, BMI between 18.5 and 25. And within that single box, body fat percentage in the men ranged from 9% to 30%. VO2max ranged from the high 20s to the mid 50s. Insulin sensitivity varied just as widely. These people were clinically identical and physiologically nothing alike, and their future disease risk diverged accordingly — despite the chart saying “normal” for all of them. Function lived on a spectrum. The diagnosis was a single, late point on it.

Medical school at Yale complicated that picture, in a useful way. Medicine, I learned, runs almost entirely on disease classification. The diagnostics that drive clinical decisions are alarms that a pathological process has already manifested — for most chronic disease, one that’s been building for years or decades. The dominant loop is diagnose, then treat. At the VA, I saw the other side of that same coin: focus far enough upstream, on function rather than diagnosis, and you can prevent a lot of downstream disease from ever declaring itself. And during my PhD, I watched omics technology mature fast enough to make that upstream focus tractable — building algorithms across thousands of molecules at once, capturing the complexity of human function instead of reducing it to one marker.

By the end of my PhD I had a clear enough thesis. I didn’t yet have a clear plan for the funding to test it.

I’d applied for an F30 more than once, finally scored in the top 10% of applicants, and then watched the funding decision slip past the point where it could still support my remaining time in the program. That wasn’t a personal failure — it was a structural one. Timelines built for a different era, funding decisions made by people optimizing for a narrower set of priorities than the one I was working from. Tenure track offered a path to independence, eventually, on the order of a decade. Entrepreneurship offered something else: the ability to go raise money from people who cared about the same problem, on a timeline that matched how fast the science could actually move.

That’s the part people miss when they ask why an MD/PhD would leave academia for a company. It was never a rejection of the science. It was a rejection of the instrument for funding it. A company lets you interface directly with the people you’re trying to help, move at the speed the problem deserves, and pull in expertise from wherever it lives instead of wherever your institution’s org chart says it should live.

Eighteen months in, that bet is paying for itself.

We closed our pre-seed round, stood up a CLIA-certified lab from scratch in a matter of months, and have now run close to 300 VO2max tests anchored to proteomic data, with 7 academic labs partnered in to keep validating what we’re finding against the real world. We shipped our first blood test as part of our beta launch this year. None of that happens on a five-year grant cycle. It happens because a handful of people decided the upstream-function thesis was worth betting a company on, and then went and built the thing.

That’s the actual answer to “why would an MD/PhD start a company.” Not because I gave up on the science. Because starting one was the fastest way to find out if it was right.

To learn more about what we’re actually building at VO Health check us out here:

https://vohealth.co/ and reach out to hello@vohealth.co

It was imagined like this: The lab begets the discovery. The discovery begets the trial. The trial begets the treatment. The patient, waiting patiently in a hospital bed somewhere, is healed.

It’s a beautiful story. It’s also increasingly fiction.

I believe there will always continue to be a classic “bench to bedside” physician-scientist archetype— a key innovation that still falls squarely in this model is in vivo CAR T cell therapy. But for most of medicine, the “bedside” has moved. And if we don’t update our mental model of what translational science means, we’ll miss the moment entirely.

Subscribe now

The Bedside Is Now Everywhere

Technology is increasingly changing how we define healthcare, and COVID made this extremely obvious. COVID didn’t just normalize telehealth — it proved that care could be delivered without walls (I mean, as long as your license allows it).

Psychiatrists, cardiologists, neurologists: practices once tethered to physical examination rooms are now conducted over video. Meanwhile, consumer wearables have quietly crossed a threshold: they’re now FDA-cleared diagnostic tools. The Apple Watch detects AFib. The Oura ring tracks HRV and sleep architecture with clinical-grade fidelity. WHOOP’s blood pressure insights debacle demonstrates the bounds that health tech companies are expanding.

Point-of-care hormone sensors, and at-home, minimally invasive blood collections are arriving next. The line between consumer health tech and clinical diagnostics is not blurring, but really it has dissolved.

Simultaneously, the majority of chronic disease (the conditions that will determine whether your patients live well or die early) is now managed at home, in communities, in gyms, in conversations between patients and their phones. The hospital is no longer the center of the healthcare universe.

The Data Is Now the Bench

At the same time, the “bench” has transformed. Omics-based diagnostics (genomics, proteomics, metabolomics) are moving from niche research tools to real-world clinical applications faster than most academic timelines can accommodate. The UK Biobank, the All of Us cohort, and dozens of other population-scale datasets have become the new petri dishes. The majority of clinical/translational science doesn’t need a bench. You need a compute cluster and someone who knows how to use it.

This is where I’ve noticed something striking in my own recent conversations with some of the top journal editors and academic leaders: the best science can no longer stop at analysis. Mining a biomarker panel from 500,000 people is impressive. But if it can’t be translated into a scalable, operationally viable, commercially accessible diagnostic — something a primary care physician can actually order, that a patient can actually afford — then what have we accomplished? We’ve published; but is anyone getting healthier? (those of you that know me have heard my soapbox about this)

The translation is happening. But increasingly, it’s happening in startups, not in academic medical centers. It’s being funded by venture capital, not NIH R01s. And the people doing it are physician-scientists who decided to get uncomfortable — to learn cap tables alongside clinical trials, to think about reimbursement pathways alongside biological mechanisms. Shout out to Neal Amin, whom I think does a fantastic job describing his transition from academic MD/PhD to Founder/CEO on LinkedIn.

A New Category of Physician-Scientist

I think a lot about what it means to practice longevity medicine in this moment. Functional capacity (e.g. VO2 max, grip strength, gait speed, cognitive reserve) is emerging as one of the most powerful predictors of long-term health we have. The data on preventing neurodegeneration, in particular, is reaching a tipping point where even the most skeptical clinicians are paying attention.

But acting on that data requires tools that didn’t exist five years ago, ecosystems that are still being built, and a kind of fluency in data science, in AI, in product development that medical training does not teach. Making scalable diagnostics to inform the future of preventive medicine is leading this shift from reactive to proactive healthcare. And we need physician-scientists to play a key role here.

An emerging type of physician-scientist of the next decade won’t just move discoveries from bench to bedside. They will sit at the intersection of population-scale data, consumer technology, academic research, and real-world implementation. They will understand that the bottleneck is rarely the biology — it’s the operational, commercial, and regulatory infrastructure required to make a discovery matter.

Call it data server to real-world patient care. (Please, someone else suggest a better “bench to bedside” alternative)?

Subscribe now

Does the current model of training equip physician-scientists with the right skillsets?

If the physician-scientist’s role is fundamentally changing… if data literacy, industry collaboration, and startup fluency are becoming essential, does the current training model develop the necessary skillsets of the future? AI is increasingly guiding clinical decision-making, removing a lot of the clinical problem solving that physician-scientists have done for decades.
As we see more and more university incubators and accelerators, its clear that many institutions are coming to the same realization, but it is still young. Couple this with a current lull in government-funded science, and we are seeing the dire need to find alternate pathways to fund and translate innovative health science. So how do we pack even more to one of longest academic training programs that exists— and what do we remove?

I don’t think I have the answer yet. But I suspect the physicians who figure it out first will define what medicine looks like in 2040.

And almost nobody changed.

Semmelweis was ridiculed, dismissed, and eventually driven to a psychiatric institution. He died (ironically) from an infection, just two weeks after being admitted. It wasn’t until decades later, when Pasteur and Lister developed germ theory: a coherent molecular explanation for why pathogens caused disease, that physicians bought in and handwashing was finally adopted at scale.

The lesson isn’t just a medical history curiosity. It’s a blueprint for understanding why good evidence, absent a satisfying mechanistic framework, rarely changes behavior— even (especially) among experts.

I’ve been thinking about this a lot in the context of the role of fitness in medicine.

Subscribe now

The Evidence Is Already There

The data on exercise is, at this point, overwhelming. Cardiorespiratory fitness, most precisely captured by VO₂max, is among the strongest predictors of all-cause mortality we have. A landmark 2018 study in JAMA Network Open found that low cardiorespiratory fitness was associated with a greater risk of death than smoking, hypertension, or diabetes. Study after study links regular physical activity with reduced risk of cardiovascular disease, metabolic syndrome, cognitive decline, and several cancers. I’ve written a lot about this previously.

We’ve known this for decades. And yet physical fitness remains largely absent from mainstream clinical medicine. It isn’t routinely measured. It isn’t formally prescribed with the precision we apply to pharmacotherapy. It isn’t reimbursed with anywhere near the enthusiasm of a new drug approval.

Semmelweis would recognize this pattern immediately.

What’s Been Missing: A Molecular Framework

Here’s the part that I think explains a lot of the disconnect: nearly all of our current molecular diagnostics, our biomarkers, our omics panels, our liquid biopsies, were developed in service of reductionist pharmaceutical interventions. Elevated LDL leads to a statin prescription. An HbA1c above threshold leads to metformin. Circulating tumor DNA leads to targeted therapy. The molecular layer was built to support drug pipelines, not lifestyle pipelines.

That means when a cardiologist or endocrinologist looks at a patient’s lab panel, the implicit question embedded in the data is: what do I prescribe? There is no equivalent molecular infrastructure that asks: what does this person’s biology tell us about their functional capacity, and what training or nutrition intervention would move it?

This isn’t a failure of the science. It’s a failure of instrumentation.

Omics as the Germ Theory of Fitness

What multi-omic approaches (proteomics, metabolomics, transcriptomics, and their integration) are beginning to offer is something genuinely new: a molecular window into functional capacity. Not just disease risk. Not just static biomarkers of pathology. But dynamic readouts of how a body adapts, recovers, and performs.

Proteins that reflect mitochondrial capacity. Metabolites that track substrate utilization under physiological stress. Gene expression signatures that distinguish a trained from a sedentary phenotype at the cellular level. These aren’t theoretical. The data are accumulating rapidly.

What they begin to provide is a mechanistic explanation for something we already know behaviorally: that fitness works. And importantly, they provide individualized benchmarks: a molecular portrait of where someone sits on the functional capacity spectrum, and what moving that needle actually looks like at the biological level.

This is the germ theory moment for fitness. Not because the behavior was in doubt, but because the molecular substrate that explains and guides the behavior has been missing from clinical medicine.

Share

How Omics of Functional Capacity Will Change Practice

Once you have molecular benchmarks tied to functional capacity, several things follow naturally. Fitness interventions become prescribable with precision — tailored to an individual’s metabolic phenotype rather than applied generically. Nutrition protocols can be evaluated against biological readouts, not just self-reported outcomes. The field gains the rigor that pharmacological development has long demanded of itself.

More importantly, it changes the frame for the clinician. The question shifts from “should I recommend exercise?” which most physicians already answer yes in theory, to “what does this specific patient’s biology tell us about where to intervene, and how do we measure whether it’s working?”

That is the difference between Semmelweis handing out pamphlets and Pasteur publishing the germ theory.

Preventive and performance medicine has an opportunity right now to be part of the next wave of molecular medicine, rather than watching it arrive from the sidelines. The behavioral insight — that fitness is medicine — has been in hand for decades.

The framework that will make it stick is finally within reach.

Subscribe now

Nearly half involved using large language models (LLMs) to make health data more understandable to individuals.

Everyone innovation-adjacent sees the opportunity—including the major AI companies—which makes the past two weeks particularly intriguing.

The Launch of ChatGPT Health, Claude for Healthcare, and Amazon’s AI Health Assistant Marks a New Era

Within the last two weeks, OpenAI, Anthropic, and Amazon all launched an LLM that helps you interpret one’s health and medical data. There is absolutely no doubt that LLMs will guide medical and health decision-making in the future. And the rationale is simple: Healthcare has spent the last 150 years classifying diseases, defining diagnostic criteria, and developing evidence-based guidelines to standardize treatment. This is a perfectly tractable problem for AI.

Today, the prevailing workflow looks like this:

Upload Healthcare Data → Align to Medical Guidelines → Return Optimal Treatment Plan → Prescribe Rx (we also recently saw a pilot in Utah where AI can autonomously refill prescription meds without a doctor in the loop)

Yet very few people feel they truly understand how healthy they are.

There are many reasons for this—some deeply structural—but LLMs are among the first ubiquitous tools capable of translating complex health information into something people can actually understand.

Even with the largely unknown security risks of uploading one’s entire medical and health history into an LLM, the value proposition of feeling empowered to understand one’s own health is so compelling that hundreds of millions of people will choose to do it anyway.

What I’ve been thinking about most is how these LLMs reason about health—and, critically, what data and health frameworks fuel their recommendations.

Health Is Not the Absence of Disease—and LLMs Need New Benchmarks

Health, defined not as the absence of disease but as functional capacity, lacks clear, widely adopted benchmarks.

As a parallel health system is built alongside the traditional sick-care system, solving this problem becomes existential.

LLMs trained primarily on disease, decline, and pathology will naturally optimize for those outcomes. If we want models that guide health promotion, performance, and longevity, we need fundamentally different datasets.

Share

To Train the Models That Guide Health Promotion, Deep Expertise Must Shape the Data

I believe it’s a good thing that nearly everyone now has access to a model capable of helping them understand complex health information that was previously inaccessible.

But if LLMs will guide medical treatment, health promotion, and performance enhancement, the real question becomes: how do health-tech companies and academic institutions arm these models with the right datasets?

At minimum, there are three dataset components that must be optimized to guide the health plans of the future:

1. Benchmarks of Health

2. Trajectories of Health Improvement (because decline is the default dataset)

3. Personalization

Health and fitness companies that measure meaningful benchmarks of healthspan and performance—especially VO₂max, body composition, and insulin sensitivity—integrated with next-generation biomarkers (e.g., omics, biobanks) will have a sustained advantage.

Because generative AI models are designed to predict the next state in a sequence, longitudinal datasets showing how people achieve high benchmarks—or depart from healthy states—are essential. Today, those datasets are rare.

Why Personalization Has Stalled—and How Omics Changes the Equation

Personalization has yet to be fully realized. I believe omics has the potential to unlock it in ways health and longevity have not yet achieved.

Consider two people with identical VO₂max values (45 mL/kg/min) and HbA1c levels (5.4%). Even if they differ by age and sex, most AI systems today would generate nearly identical protocols.

That’s the ceiling of personalization when we rely solely on traditional clinical and fitness metrics.

Now imagine having access to the transcriptome or metabolome underlying fitness or insulin sensitivity. From decades of academic literature, we know we’d observe variation across molecular pathways related to inflammation, growth factors, carbohydrate metabolism, and fat metabolism—independently targetable biological systems responsive to diet, exercise prescription, and even specific drugs or supplements (e.g., creatine).

Simply measuring thousands of molecules isn’t enough. The real challenge—and opportunity—is identifying the meaningful few that drive outcomes.

Share

The Scientist-Led Advantage in the Healthspan Era

My thesis is that researcher-led companies—those combining integrative physiology, medicine, laboratory science, bioinformatics, and AI—are uniquely positioned to curate these datasets responsibly and effectively. (By the way, we are seeing this distinction between Anthropic’s CEO Dario Amodei and Deepmind’s Demis Hassabis— a Nobel Prize Winning scientist, vs. OpenAI’s Sam Altman— listen more to Dario’s thoughts at Davos)

They are best equipped to consider the full biological context required to usher in a new era of personalized healthspan and performance.

If you want to understand how we’re integrating omics into a practical guide to improve health and performance, reach out to my team and me at hello@vohealth.co.

Subscribe now

How does fitness, and regular exercise, impact cancer outcomes?

We’ve known for years—from epidemiology, clinical trials, and animal models—that regular aerobic exercise and higher cardiorespiratory fitness are linked to better cancer outcomes. People who are more active tend to have lower cancer incidence, slower progression, and better survival across multiple tumor types. What’s been harder to pin down is why. The molecular “middle steps” between exercise and tumor behavior remain fuzzy, especially when we think about metabolism.

Tumors are famously hungry. Since Warburg’s early observations, it’s been clear that many cancers rely heavily on glucose to fuel growth. At the same time, skeletal muscle is the body’s largest site of glucose disposal and oxidative metabolism—especially in response to exercise. That raised a key possibility: maybe exercise doesn’t just help the host broadly—it may actively compete with tumors for fuel.

We designed this study to test that idea using voluntary wheel running (a low-stress, aerobically relevant model) and a suite of metabolic readouts: glucose uptake, glucose oxidation into the TCA cycle, targeted metabolomics, stable isotope tracing, and transcriptomics. We also incorporated VO₂peak, a gold-standard measure of aerobic capacity, so we could look beyond a simple “exercise vs no exercise” binary and ask how fitness level itself shapes tumor metabolism.

Five Key Findings from the Research

1) Voluntary aerobic exercise dramatically slowed tumor growth—up to 60% in obese mice

In an obese breast cancer model, mice exposed to voluntary wheel running showed ~60% smaller tumors compared with sedentary obese controls. Exercise also improved body composition and normalized glucose/insulin levels relative to sedentary obese mice.

This is important because obesity is a strong driver of breast cancer progression, and the magnitude of tumor slowing here suggests exercise can counteract that risk environment.

2) Exercise “repartitioned” glucose away from tumors and toward muscle/heart

When we measured tissue glucose uptake using ²-deoxyglucose, exercise increased glucose uptake in skeletal and cardiac muscle, while decreasing uptake in tumors—both after an acute bout and at baseline after training.

In other words, exercise changed who gets the fuel. The host tissues became stronger glucose competitors, and tumors got less.

3) Exercise shifted glucose oxidation away from tumors and toward muscle/heart

Using [U-¹³C₆]glucose infusions, we quantified how much glucose actually fed oxidative metabolism (entry of glucose carbons into the tricarboxylic acid (TCA) cycle, AKA Kreb’s cycle). Exercise increased glucose’s contribution to the TCA cycle in muscle and heart, but reduced glucose oxidation in tumors.

So the repartitioning wasn’t superficial—exercise changed both glucose entry into tissues and its downstream fate.

4) The benefit generalized across tumor types and scaled with exercise dose/timing

We replicated the core effect in a melanoma model, a tumor less classically tied to obesity. Exercise again reduced tumor size and tumor glucose metabolism. Even more interesting: mice that exercised before tumor implantation (“prehab”) had greater tumor suppression than those starting at implantation (“rehab”).

This supports an exercise dose-response concept and suggests that being fit before cancer develops can shape the metabolic terrain tumors encounter.

5) Fitness level predicted opposing metabolic signatures in muscle vs tumor

Looking beyond exercise exposure alone, VO₂peak correlated with metabolite patterns in opposite directions in muscle and tumor. For example, higher fitness was linked to lower tumor aspartate content and glutamine-to-aspartate synthesis, while muscle showed the reverse pattern.

This finding matters because it implies a continuum: the fitter the host, the less favorable the metabolic environment for tumor energetics. It also gives us a translational handle—fitness is measurable and modifiable in humans.

Share

The Metabolic Model of Exercise’s Anti-Cancer Effects

Aerobic exercise and higher fitness enhance host tissue glucose demand and oxidation, creating a metabolic competition that constrains tumor fuel access and slows progression.

A few aspects feel especially exciting going forward:

• Metabolic competition as therapy. Exercise isn’t just “good for you” in a generic way—it may be an active metabolic intervention that reshapes nutrient flows in the tumor–host ecosystem.

• Fitness as a biomarker. VO₂peak helped explain heterogeneity in tumor metabolism far better than a binary exercise label. That suggests real clinical utility for tailoring or monitoring exercise oncology interventions.

• Glucose appears to be the main lever. Transcriptomic changes suggested shifts in amino-acid pathways, but metabolite changes there were modest. The glucose story, by contrast, was robust across models and methods.

• Timing matters. The stronger effect in prehab models mirrors what we see clinically: higher pre-diagnosis or pre-treatment fitness predicts better outcomes, and building fitness early may pay dividends later.

Of course, there are limitations—rodent exercise dose can vary, and we didn’t have human tumor metabolomics to pair with the clinical transcriptomic comparisons. But the convergence of fluxomics, metabolomics, and tumor growth phenotypes gives us confidence that this metabolic repartitioning mechanism is real and worth testing directly in people.

Subscribe now

It’s pretty well understood that start-ups demand a lot, and with venture backing, speed is non-negotiable. It’s also well understood that pacing, effort management, and consistent performance are what keep momentum high. Straddling nonstop grind and sprint-intervals, I keep finding myself wondering:

Fortunately, the concept of maximal sustainable workload is well characterized in the field of exercise physiology. The principles of peak capacity and functional thresholds can easily be applied to business and life.

Your Maximal Sustainable Workload is Set by Your Ceiling and Your Threshold Percentage

• Maximal Capacity: Your ceiling, e.g. VO₂max, when referring to aerobic exercise capacity

• Functional Threshold: Percentage of your maximal capacity at which you can sustain productive effort “indefinitely”

Subscribe now

The graph above shows that over a period of training (medical residency, the early phase of a start-up, a half-marathon plan), both maximal capacity and functional threshold can improve. In exercise physiology, functional threshold is often referred to as functional threshold power (FTP) or critical power. Increases in critical power (or maximal metabolic steady state) explain performance gains that VO₂max improvements alone don’t capture.

Functional thresholds can be measured numerous ways, via lactate threshold tests, FTP “field tests” (e.g. max effort in 20 minutes or 1 hours), which many existing calculators out there (hit up for a deeper dive on all things performance physiology).

So how do you find your functional threshold for work?

1. How do you know you’ve exceeded it?

2. More importantly, how do you know that you’ve not yet reached it?

1a. What exceeding your functional threshold looks like

I recently completed the VO₂max training study that has produced the largest increase in VO₂max ever reported in the literature. See more in this post for background.

Now I understand why no one has reproduced those gains — it’s nearly (probably actually) impossible.

The protocol called for 5-minute bike intervals at the wattage I’d hit at 100% of VO₂max in the lab earlier that week, followed by 2 minutes of “rest” at 50% wattage. Practically: hold a resistance you could sustain in the lab for <5 minutes, for a total of 30 minutes per session. It was mentally brutal (dreading something you knew you likely couldn’t finish) and physically painful — my quads couldn’t keep up with what my heart and lungs could.

In the heart-rate graph from one especially tough session, you can see that about two minutes into the third interval I couldn’t keep driving the bike hard enough to keep my heart rate rising. In the fifth interval I tried to push through the pain to drive it higher, but I couldn’t hold cadence.

So what might this look like in the business world? Ever make an intimidatingly long to-do list, crush 2/3 of it (a real win), then try the same thing for four weeks straight and find yourself sliding closer to 1/5?

In medical training and other high-performance careers, burnout is well understood. Overtraining and fatigue are too; and they can have enduring negative consequences.

1b. How do you know you may be able to accomplish more than you are?

This one is simple: you never know if you don’t try. In exercise physiology and healthcare, that starts with benchmarks- functional assessments, biomarkers, etc. - to see where you stand relative to yourself and others. In other words: KPIs.

Sign yourself in front of intimidating deadlines and do the work. I first learned to apply for abstract presentations and scientific conferences in undergrad, well before we had completed the full set of analyses, which served as a forcing function to put together a publishable piece of work. The audit your last several months after achieving that deadline— could you sign up for another presentation right away? Do you need to catch up on too many other things that had been neglected?

In other words, was this a sprint to raise your ceiling, or do you have a new maximal sustainable threshold?

Share

Benchmarking, training, and testing

In life and business, pushing both your ceiling and your maximal sustainable effort— with focused training, maybe coaching, and regular assessments— is how you unlock peak performance. At my start-up, we’re early in establishing, pushing, and measuring performance. As we build a high-performance organization, we look to people further along the path (drop your tips in the comments!).

And in the meantime, I like to use principles of physiology to understand and achieve peak sustainable performance.

When I encountered EHRs, treatment algorithms, “care pathways,” and insurance I realized my idealistic concept of measurement -> individualized care had several more regulatory steps that had little to do with physiology or health.

Turns out clinical practice is enforcedly evidence-based and guideline directed, supported by randomized controlled trials and population level outcomes data. Government and payer incentives have a substantial influence on who gets what treatment and when—these organizations require specific standards that ultimately convert individual patient-level problem-solving to measurement-driven actions standardized to a the population average.

Standardization of medicine is determined by the toolkit that takes decades of evidence to generate.

These tools include:

• Biomarkers: what measures can be associated with health, disease, and mortality outcomes at the population level?

• Clinical Trial Design: E.g. RCTs are the dominant mechanism of gold standard evidence, but treat individuals as a representative “average person” within a population

• Population-Level Databases: E.g. Framingham Heart Study, longitudinal profiling in the NIH All-of-Us, UK Biobank, and even the Apple Health Study

And, put simply, medicine evolves only as fast as its measurement tools do.

Significant technological developments are rapidly changing the toolkit of medicine, including omics-based biomarker discovery and development, N-of-1 clinical trial designs gaining momentum, and the integration of wearables, clinical data, and novel biomarkers to determine health outcomes at scale. TBH, this is the type of medicine I’d be excited to practice.

However, the practical application of these tools still must be integrated into the existing incentive structure by the systems that dictate healthcare resource allocation in the US (e.g. government and payers). (Alt view: consumers decide what data they find valuable, and pay themselves—lets save this exciting topic for another discussion)

The Influence of Government Incentives

“U.S. health care spending grew 7.5 percent in 2023, reaching $4.9 trillion or $14,570 per person. As a share of the nation’s Gross Domestic Product, health spending accounted for 17.6 percent.” – CMS National Health Statistics Group

For government agencies, the primary incentive is—and always has been—cost containment. The Centers for Medicare & Medicaid Services (CMS) and other public payers operate under the paradox of delivering comprehensive healthcare while maintaining fiscal sustainability. Every policy lever, from reimbursement structure to quality measure design, is aimed at bending the cost curve without compromising population health outcomes.

But “keeping costs low” ends up presenting as an engineering problem. By shaping incentives through mechanisms like Star Ratings, value-based payment models, and population-level quality metrics, the government effectively dictates the flow of healthcare capital. Plans and providers that prevent disease (at least as defined by current biomarker thresholds), reduce hospitalizations, and manage chronic conditions efficiently are rewarded; those that don’t are penalized through lower payments or lost contracts.

At a deeper level, these incentives push the healthcare system toward data-driven accountability. When a biomarker like HbA1c or blood pressure becomes a proxy for reimbursement, it aligns the government’s cost-saving imperative with the individual’s health optimization. The more precisely we can measure and predict physiological change, the more efficiently public dollars can be allocated. In that sense, biomarkers don’t just measure health—they define the economic architecture of modern medicine.

Keep in mind these biomarkers are assumed to be meaningful proxies of health status.

Payer Incentives

For payers (e.g. Humana, Aetna, United), the North Star—literally—is the CMS Star Ratings system, which determines not just how plans are perceived but how they get paid. Medicare Advantage plans that score four stars or higher receive substantial quality bonus payments, allowing them to offer richer benefits and attract more members. Those that fall short see declining reimbursement and enrollment—an existential problem in an increasingly quality-linked ecosystem.

The 2026 update to the Star Ratings program marks a shift from member satisfaction to measurable health outcomes. CMS is rewarding results rooted in biology: blood pressure control, glycemic stability, kidney health, and medication adherence now carry the heaviest weights in determining plan performance. These are, in essence, biomarker-driven incentives—where a population’s collective clinical data directly translates into financial outcomes for the payer.

In this system, payers are motivated to close care gaps not just for optics, but for economics. Capturing, tracking, and improving biomarkers like HbA1c, LDL-C, and systolic blood pressure isn’t just good care—it’s good business. And as these measures expand to include functional (physical activity monitoring made it on the list) and behavioral health metrics, the next generation of Star Ratings seem to be inching closer to aligning payer incentives with genuine physiological improvement, but often lag behind cutting edge innovation.

Share

So How to Use Biomarker Innovation to Improve Healthcare at Scale?

As mentioned above, biomarkers like hemoglobin A1C, LDL, fasting glucose, creatinine, blood pressure determine what drugs are recommended, what lifestyle interventions are covered by insurance, who should see a specialist, etc. This toolkit has existed in the healthcare system for decades yet has clearly not sufficiently mitigated the downstream consequences of chronic disease morbidity, mortality, AND economic strain.

Current clinical practice treats each of these biomarkers as distinct, targetable, and treatable disease entities. Modern clinical bodies are beginning to realize that they are lagging indicators of a more comprehensive physiological process. This has led to the naming of disease entities such as “cardio-renal-metabolic syndrome” (Ndumele et al 2023, Koufakis et al 2025).

Recommended treatments include GLP1-agonists, which do not target an LDL-Creatinine-A1C axis via multi-target pharmaceutical agonism, but rather through significant reductions in upstream physiological parameters—feeding behavior, body weight and fat mass. Still, clinical trials are run to demonstrate the effects of GLP1 agonists to improve long term cardiovascular health, metabolic health, liver health, addiction, etc. which treat each disease as a specific entity.

If single interventions can have multi-systems-level health benefits, why shouldn’t the next generation of resource-allocation (e.g. covered interventions, therapeutic development, etc.) be determined off multi-systems-level biomarkers?

By guiding interventions towards upstream structural and functional biomarkers (e.g. body composition, cardiorespiratory fitness, & insulin sensitivity) clinicians can directly target functional decline that underlies numerous chronic diseases.

So, how do we measure these?

Share

A New Paradigm in Biomarker-Driven Resource Allocation will Target Upstream Functional Measures

By targeting the upstream biology, the traditional reductionist (single “disease” focused) biomarkers will also improve

Gold standard measures of body composition exist today with whole-body imaging including DXA and MRI (to assess fat mass, visceral fat content, muscle mass, and bone density). Cardiorespiratory fitness can be measured via a maximal exercise test (e.g. VO2max, or cardiopulmonary exercise testing (CPET)). Insulin sensitivity could be directly assessed via a hyperinsulinemic-euglycemic clamp or approximated by oral glucose tolerance tests.

If a patient with obesity, high cholesterol, high blood pressure, and diabetes improves body composition (loses body fat, increases muscle mass index), improves insulin sensitivity, and improves their VO2max, I’d be shocked to learn that their A1C, fasting glucose, LDL, blood pressure, etc. didn’t also improve.

While these measures would be fantastic to have as a part of every primary care exam, they are practically infeasible at population scale (at least today). If the insights from these tests could be captured in blood tests that seamlessly integrate into clinical practice, a scalable biomarker toolkit could exist to drive alignment of incentives between patients, clinicians, governments, and payers.

With rapid advances in functional omics based technologies (e.g. transcriptomics, proteomics, and metabolomics), coupled with omics-based measurements in population level databases (e.g. UK Biobank, China-Kadoorie, etc.) I envision this new clinical toolkit coming very soon.

I also believe that improvements in AI and machine learning (driven largely by training on high quality, clinically-meaningful, datasets) may allow for personalized care pathways that take into account an individual’s physiology and health status, to make precision medicine a reality for chronic disease.

Not only will biomarkers measure health (rather than solely absence of disease) but will allow for cost-effective allocation of resources that preempt & intercept disease early.

A simple question with many possible answers. Some define it as the absence of metabolic disease, using criteria like BMI < 25 kg/m^2, fasting glucose < 100 mg/dL, and hemoglobin A1C < 5.7%. Others may also consider waist circumference, cholesterol levels, or inflammatory markers, often providing “optimal” ranges.

You might also encounter definitions like “metabolic health is how well your body processes metabolites,” which can feel vague—hard to measure and harder to act upon.

For a definition of health beyond simply avoiding disease, the most useful framework I’ve seen comes from the WHO’s concept of functional ability. If we apply that to metabolism—defined as “the chemical processes that occur within a living organism to maintain life”—we get:

Subscribe now

I hope you enjoy this overly simplistic graph showing a perfect correlation between high metabolic health and high functional capacities.

To illustrate differences in metabolic health as determined by legacy biomarker classifications and laboratory-assessed measures of functional capacity, let me take you through a 2004 paper in the NEJM that examined what features could explain differences in metabolic health between insulin resistant and insulin sensitive children of parents with Type 2 Diabetes (from the lab of the chair of my PhD committee).

But first, what is Insulin Resistance?

Insulin is a hormone with many actions on different target organs (muscles, fat, liver, brain, etc.). A key role of insulin is to drive glucose uptake into target organs, after binding to the insulin receptor. Insulin resistance refers to the concept that there is reduced activity of insulin action in response to a signal (like a rise in blood glucose). Thus, one manifestation of insulin resistance is the inability to lower blood glucose (or clear glucose) in response to elevated blood glucose (e.g. from food intake).

Insulin resistance is the best predictor of Type 2 Diabetes, and signs can be detected well before a person achieves a diagnosis. A landmark 1990 study in the Annals of Internal Medicine found that “One to two decades before type II diabetes is diagnosed, reduced glucose clearance is already present.” Insulin resistance is typically measured in the lab with functional tests, that I’ll describe below.

So now, back to the story.

Young, Normal A1C, Normal BMI— whats the catch?

“Healthy” young people we’re recruited to understand the mechanisms underlying insulin resistance. Here’s what the authors selected for to recruit their participants:

“All subjects…were prescreened to confirm that they were in excellent health, lean, nonsmoking, and taking no medications. A birth weight above 2.3 kg (5 lb) and a sedentary lifestyle were also required. Qualifying subjects underwent a three-hour oral glucose-tolerance test (with a 75-g oral glucose load), after which two subgroups of subjects were consecutively selected to identify extreme phenotypes for insulin resistance and increased insulin sensitivity.”

I am glad that activity was controlled for in this study, as one of the most potent modulators of insulin resistance/sensitivity is exercise. But otherwise, what you can see in these 20-30 year old (probably grad student) subjects with normal BMIs and excellent A1Cs shows nothing remarkable from traditional metrics, or even inflammatory markers. However things change when you take a deeper look.

The first functional assessment: a 3 hour glucose tolerance test

What you see in the graph is the plasma glucose in the blood of the subjects in response to 75g glucose load (~around a 20 oz bottle of Coca-Cola). The insulin-resistant subjects had a much higher peak glucose, and it remained elevated for the entire 3 hours. This type of study can be conducted at home with a continuous glucose monitor, for example, yet they aren’t quite as accurate as measuring blood glucose from an IV.

What this shows us is the 1990 definition of insulin resistance: reduced glucose clearance.

Now, since up to 90% of glucose uptake occurs by skeletal muscle, lets take a look at the measures of functional capacity in the muscles of these subjects as measured by a three-hour hyperinsulinemic–euglycemic clamp, and by a 31P magnetic resonance spectroscopy (MRS).

Reduced insulin stimulated glucose uptake as measured by hyperinsulinemic–euglycemic clamp

A hyperinsulinemic–euglycemic clamp aims to determine exactly how well the body (the majority driven by skeletal muscle) is able to use insulin to take glucose into the organs. To do this, the subject lays down in a bed with two IVs— one infusing insulin, the other infusing glucose. In response to a high infused insulin load (ie. “hyperinsulinemia”), the body begins to drive glucose into the subject’s organs.

With the other IV, the examiner aims to “clamp” blood glucose at a normal level (ie. “euglycemia”) by infusing glucose at a rate that replaces glucose as fast as it gets taken up.

The maximal rate of insulin-stimulated glucose uptake gives you the degree of metabolic functional capacity known as insulin sensitivity. The higher your insulin sensitivity, the better your metabolic health. Low insulin sensitivity is also known as insulin resistance.

In the figure above, you can see that the insulin-resistance subjects had >200% lower insulin-stimulated glucose uptake— very low functional capacity!

Subscribe now

Now, what do the muscles do with that glucose?

In addition to storing the glucose as glycogen, the mitochondria (and glycolytic machinery in the cytoplasm) make the universal energy currency, ATP. The rate at which you are able to produce ATP is an excellent measure of functional capacity. If you are developing this measure from a wearable on every organ… please talk to me!

Once again, you see impaired capacity for muscle ATP synthesis in the insulin-resistant subjects. Despite having completely normal lab values, when you seek measures that tell you the functions of the body, the picture of impaired metabolic health becomes quite clear.

The Future of Metabolic Health Requires Better Measures and a Better Definition

Poor metabolic health is perhaps the largest health crisis of our lifetime. It underlies heart disease, diabetes, dementia, stroke, depression, and more. It is completely modifiable, and the metrics exist today. However, for improvements in metabolic health at population scale, we need to embrace systems level change, new technology, and a different different definition of health.

• Current biomarkers of A1C, glucose, and BMI make prevention impossible— yet there are several well characterized measures of functional capacity that DO allow for early intervention

• While HOMA-IR (a composite insulin resistance score of fasting glucose and insulin) is better than glucose or insulin alone, it isn’t able to give a good measure of functional capacity— Continuous glucose monitors, especially when controlled challenges are conducted, could yield significant insights

• Omics (e.g. proteomics or metabolomics) have the ability to measure many molecules 10-50+ simultaneously that can provide an integrated measure that predicts glucose clearance, and can be much more powerful than 2-3 analyte composite score.

• Metabolic health as spectrum of functional capacity (rather than binary category of normal vs. abnormal) is a necessary viewpoint shift to align diagnoses and interventions with metric that matter for human health.

How do Screen for and Treat for Chronic Disease?

Chronic disease screening includes testing biomarkers for infectious diseases like HIV and Hepatitis C, blood pressure screening, cancer screening (e.g. imaging based or now cell-free DNA), depression screening, and type 2 diabetes. The frequency at which certain biomarkers are tested, and for what population, is recommended by different specialty organizations (like the American Heart Association, American College of Lifestyle Medicine, etc.) or most definitively by the US Preventive Services Task Force (USPSTF).

Regular screenings AND additional services, like “Healthy Diet and Physical Activity for Cardiovascular Disease Prevention in Adults With Cardiovascular Risk Factors” are listed in the USPSTF recommendations.

Importantly, all recommendations that are A&B rating of evidence from the USPSTF require, under the Affordable Care Act and federal code, from “group health plans and health insurance insurers offering non-grandfathered group or individual health insurance coverage provide coverage of certain specified preventive services without cost sharing.”

This is pretty cool. In theory, if the scientific evidence reviewed by a non-biased committee demonstrates a population level benefit, there is a universal code that guides the practice of medicine AND reimbursement strategy. However, the fundamental flaw is that these recommendations for disease “prevention” occur because the metrics we have available to us quantify disease presence, rather than signals prior to disease onset.

Subscribe now

USPSTF Recommendation for Intervention Require Disease Biomarkers to Be Elevated

Let’s take behavioral interventions to prevent obesity-related morbidity and mortality in adults. In order to a diet and exercise intervention to be recommended, the USPSTF requires that the individual have a BMI >30 (which is treatable obesity), or a BMI >25 with hypertension, dyslipidemia, abnormal glucose levels, or diabetes.

If an intervention is suggested when the individual already has a classifiable disease, what are we preventing?

One could argue that we are preventing the long term complications (e.g. organ failure or death) indicated by the disease biomarkers.

To dive into this argument, the NIH funded diabetes prevention program (DPP) serves as an excellent case study, as this cohort was followed for >15 years following a randomization to an intensive lifestyle intervention, metformin, or placebo in patients with pre-diabetes.

Early Lifestyle or Metformin Interventions for Pre-Diabetes Were Associated with Lower Rates of Diabetes

The NIH funded DPP endeavored to decrease the rate of diabetes incidence by prescribing metformin or an intensive lifestyle change program (with a coach) to individuals with pre-diabetes (but not yet diabetes) using an A1C cut-off of 5.7%.

After 10 years they found that the cumulative incidence of diabetes was:

• Lifestyle: 5 per 100 person years (42% cumulative incidence)

• Metformin: 7.9 per 100 person years (47% cumulative incidence)

• Placebo: 10.8 per 100 person years (52% cumulative incidence)

This demonstrated a success! Lifestyle (to a greater degree) and metformin when prescribed earlier were able to lower diabetes rates. Using A1C as the early indicator of diabetes risk, the goal of preventing diabetes diagnosis (compared to placebo) was achieved.

However, Early Metformin Did Not Prevent Long Term Organ Failure, While Lifestyle Did

It is nice to not have a diagnosis of diabetes. But it is more important to have functioning organs as a result of not having this diagnosis. Some interesting findings we’re tucked away in the supplemental tables of the 15 year follow up from the diabetes prevention program.

“In a post-hoc analysis among those whose latest HbA1c was ≥6.5%, representing ~25% of the cohort, lifestyle showed significant reductions compared with placebo in the aggregate microvascular outcome (relative risk 0.59), retinopathy (relative risk 0.51), and neuropathy (relative risk 0.39), with no significant differences between the metformin and placebo groups.”

Even if identified “early,” what was first line therapy had no benefit on prevention of long term complications in the patients with the most elevated A1C.

While this doesn’t demonstrate an issue with disease screening, it demonstrates that our interventions are often tailored towards the wrong metrics (e.g. A1C). While yes it is good to not “have a diagnosis,” what is better is not having the long term complications. This is especially true for drugs that target a single molecular pathway in order to modify a single blood-based biomarker.

Share

Early Prevention: What about the person with BMI 29, A1C 5.6%, Cholesterol 199, BP of 130/81?

Take a 44 year old man with the following metrics above. Their Framingham Risk 10 year is <5% and their PREVENT score is low risk. They do not meet current disease threshold criteria, and likely no intervention would be covered to prevent them from progressing to a quantifiable disease state.

If you were to examine their cardiovascular fitness, body composition (e.g. muscle mass index, body fat %, or visceral fat content), or insulin sensitivity, you would have a much fuller picture of their health status.

Let’s estimate his VO2max using a rough regression adjustment (Jackson et al. 1990) uses age and BMI yielding:

VO₂max ~ 60 - 0.23 * age - 0.79 * BMI)

Plugging in 44 yrs and BMI 29:

60 - (0.23 * 44) - (0.79 * 29)~ 60 - 10.1 - 22.9 = 27.0 mL/kg/min

This person’s estimated VO2max is in the bottom 10th percentile for men 40-49 from the FRIEND registry and below the 5th percentile from the Cooper Clinic (Mayo Clinic Proceedings 2015).

Rather than waiting for this individual to have a disease biomarker present, wouldn’t it make sense to focus on the low exercise capacity?

An intervention tailored not towards simply reversing a single disease biomarker, but towards improving fitness, body composition, and insulin sensitivity would improve their physical function AND prevent multi-morbidity disease & mortality risk.

Focus on Function, Rather than Presence of Disease

What if the biomarker panel measured the functions of the body, rather than signals that arise when the body can not longer compensate for nutrient overload or chronic stress?

The current paradigm focuses on disease detection through population-based screening recommendations.

Direct to consumer health platforms (e.g. Function, Superpower, Mito, Everlywell, InsideTracker, etc.) provide disease detection biomarkers on demand, outside (or inside) of USPSTF recommendations. This may allow for early disease detection.

However, if we truly want preventive and proactive healthcare, we need to focus on the earliest declines in functional capacity. We would be able to capture, and act on, early functional decline.

Wouldn’t you want to understand that muscle quality is decreasing before frailty, or significant muscle loss, kicks in? Or that your cardiovascular system is on its way down, before you need to be started on a band-aid medication?

Focusing on biomarkers of cardiovascular function (like exercise capacity), body composition (visceral fat and muscle mass), and insulin sensitivity are key functional measures that can truly indicate health status, predict multi-morbidity, and do not require that our organ systems have been unable to compensate, resulting in disease biomarker presence.

This is Part 2 of the VO₂max series.

Subscribe now

What is the Gold Standard Measure of VO₂max?

The gold standard method of VO₂max (maximum rate of oxygen consumption) determination requires maximal exercise testing using indirect calorimetry (ie. measurement of oxygen consumption via a facemask) on at least two different modalities (e.g. running and cycling) on separate days. Some people may be able to achieve higher oxygen consumption rates on a bike, while most can achieve a higher oxygen consumption rate on a treadmill test.

While there are numerous protocols, essentially all tests gradually increase exercise intensity (whether it is wattage/resistance on the bike or incline/speed on the treadmill) until the participant achieves a plateau in oxygen consumption despite an increase in exercise duration. The rate of oxygen consumption is captured every 20-60 seconds, and the peak oxygen consumption reported gives you the VO₂peak score. For most applications outside of the ex phys lab, VO₂peak and VO₂max are used interchangeably.

When I was first certified by the American College of Sports Medicine (ACSM), I learned of the at least two criteria that should be met to confirm maximal effort:

• VO₂ Plateau (≤150 mL·min⁻¹ increase)

• Heart Rate (HR) within ±10 bpm of Age-Predicted Maximum

• Respiratory Exchange Ratio (RER) ≥ 1.10

• Rating of Perceived Exertion (RPE) ≥ 17 (Borg 6–20 Scale)

• Blood Lactate ≥ 8 mmol/L (when measured)

• Volitional Fatigue (Inability to Continue)

This combination of objective and subjective metrics just suggests that the test subject just needs to be pushing as hard as possible to be a “true max.”

VO₂max can be calculated in absolute (L/min.) or relative units (mL/kg/min.) of oxygen. VO₂max relative to body weight is more important, and takes into consideration body composition. Since body weight is in the denominator, the lower the body weight, the higher one’s VO₂max.

Shown below is a typical graph of what your heart rate and oxygen consumption looks like during a maximal exercise test. The highest circle is the key metric.

This also shows why during-exercise heart rate serves as a great, but imperfect, data point that can feed into models that estimate your VO₂max.

Subscribe now

Review of Methods for Measuring/Estimating VO₂max

While the laboratory is the gold standard method for determining VO₂max, there are many simpler and more scalable estimates.

The laboratory requires specialized equipment, frequent calibration and maintenance (if you care about accuracy), trained staff, and a willing participant to carve out the time in their day to exercise completely to exhaustion.

One of the next best alternatives, that requires no equipment, is the Cooper 12 minute all-out running test. Developed over 50 years ago, this simple test can easily calculate your VO2max with relatively high accuracy. But still requires you to exercise to maximal exhaustion.

Additionally, the Whoop, Apple Watch, and many other wearables can predict your VO₂max with some degree of certainty, typically based on your resting and/or activity heart rate. These predictions get better with more data (on yourself and at the population level).

Finally, there is a last tier which can simply estimate your VO₂max from your sex, weight, age, etc. I bucket this estimate the same way that I do as BMI as a metric for body composition— at the population level it is decent, but the individual level variability (or meaninglessness) can be enormous.

How do you improve your VO₂max?

Before getting into the specifics, lets think from first principles. VO₂max represents your body’s maximal capacity to use oxygen (integrating the functions of your lung, heart, blood, muscles, and mitochondria). To improve the maximal capacity of anything, you need to challenge the system. In response to the challenge, the system adapts.

Now we can get more specific. When using VO₂max calculated relative to body weight (mL/kg/min.), the first way to increase VO₂max is to lower the denominator— that is, lose weight. Since the oxygen consumption during maximal exercise is largely driven by muscle and heart mitochondria, the more direct way is to lose fat.

Now, beyond losing weight, there are numerous exercise training programs that have been developed to increase VO₂max. When looking systematically at 37 studies there is a lot of variability in the protocols. You can see this wide distribution below:

There are two points to make from this analysis:

• The highest increases in VO₂max are from protocols that include high intensity interval training (HIIT)

• That outlier all the way to the right, is a BRUTAL protocol (but only 40 min a day)

  • Weekly VO₂max testing to determine your high intensity interval

  • 6 days a week of exercising as hard as you can for 40 minutes

    • 3 days run as far as you can

    • 3 days of intervals on the bike- six 5 minute intervals at the intensity at which you achieved your VO₂max, separated by 2 minutes at 50%

  • This protocol led to an average increase in VO₂max of 38% in just 10 weeks

However, for most people, gradual weight loss accompanied by gradual improvements in fitness can lead to large changes in this potent biomarker.

The Future of VO₂max Trainability

As interest in and understanding of VO₂max continues to improve (see Google Trends), we will begin to have large real-world datasets tied to people’s VO₂max scores. This will serve as a large-scale n-of-1 database to truly understand how we can personalize people’s protocols to maximize the benefits of exercise, nutrition, and other interventions.

How do you measure your VO₂max?

1. Start with a razor-sharp wedge and Validate the problem per domain

Sonal: “Deep-tech teams often pitch, ‘We’re going to own the entire drug-discovery automation stack.’ No. You need to identify the specific step you’re coming after—that’s your wedge.”

Your first market beachhead is a single, acute pain point you can knock out of the park. In biotech that might be liver-toxicity prediction; in space tech, orbital-debris tracking. Nail one thing, then—and only then—plot the adjacency moves.

Brooks: “A platform that claims to serve neuroscience, oncology, and autoimmunity must prove a real pain in each of those domains—because the competitive tech set is totally different.”

Domain agnosticism sounds sexy to investors, but users smell hand-waving. Before you broaden, show concrete traction—data, endorsements, or LoIs—inside the first vertical. The rest can stay on the slide deck roadmap.

2. Land-and-expand ≠ deep tech

Sonal: “Land-and-expand works in traditional SaaS, but not in drug discovery. Therapeutics demand a different playbook.”

In regulated, capital-intensive fields, each adjacent move usually targets a new budget owner (and a new scientific proof stack). Treat every expansion as if you’re launching a fresh product line: new validation data, new champions, new pricing logic.

3. Map the exact buyer inside a siloed org

Sonal: “I’ve worked in four pharma departments—each a self-contained universe. If you sell ‘to pharma’ you’ll fail. You must know which specific team, therapeutic area, and budget line you’re chasing.”

As per the above, Fortune-100 R&D, commercial, and manufacturing groups operate like federated city-states. Build stakeholder maps: who signs, who wields influence, who blocks. Your pitch, pilot scope, and evidence package should speak only to that tribe of <50 people.

4. Let customers pull you to the next layer

Sonal: “At Genentech we started with direct-to-HCP sales insights. We then asked the vendors to build personas, segmentation, and eventually large-scale brand-planning tools.”

When users beg for the sequel, you’ve earned the right to build it—and a reference base to pre-sell the upgrade. If your wedge is successful, the customer will point you in the right next direction- this can be something hinted towards but should not be shoved down the throat.

5. Ship 70 %—just an idea is too early for valuable feedback

Brooks: “Once a segment loves the concept, give them 70 % of something so they can critique it. At 25 % it’s a wasted call.”

Prototype depth matters. Enough fidelity for customers to place it in their workflow; lean enough to remain malleable. Think storyboard, clickable mock, or small-n pilot—whichever unlocks the richest feedback loop fastest.

TL;DR cheatsheet

1. Carve a hair-thin wedge—own one pain point.

2. Assume silos; pinpoint the budget holder.

3. Prove value in one domain before platform talk.

4. Let customers yank you to the next feature, not vice-versa.

5. Prototype at 70 %. Enough to spark real feedback.
   
   

The reactive system keeps getting more granular

On the reactive side, care will become far more precise—tracking disease at the molecular level and deploying novel “biological manipulators” for permanent or temporary interventions. Think targeted gene edits and neuromodulatory devices that correct dysfunction at its (hypothesized) source. Drugs will also become multi-targeted by design— or perhaps even tunable to an individual’s biology.

A parallel, proactive system shows up earlier and everywhere

Meanwhile, proactive medicine will shift further upstream, emphasizing early detection, disease-agnostic markers of healthy longevity, and truly personalized lifestyle prescriptions—tailored exercise regimens, nutrition plans, social-support networks, and sleep optimization (Hinge Health, Omada, Nourish, Oura, Whoop, etc.). Care won’t just happen in hospitals; it will flow into homes and community hubs.

Long‑term arc: human “optimization,” disease or not

Over the long term, these parallel trends—prevention, molecular biology, and neurotechnology—will converge into a healthcare ecosystem built not around illness, but around human “optimization.” In fact, with muscle-enhancing compounds already advancing through clinical trials, we’re witnessing the first steps of this evolution today.

Subscribe now

1. Whole‑genome newborn (maybe in‑utero) screening becomes standard, catching actionable variants early (National Academies are already debating the details). While ethically challenging, difficult to imagine that pressures don’t move towards whole-genome sequencing becoming the default.

2. Basic biomedical research (not immediately translational) shifts to non‑profits or FRO‑style collaboratives, while industry still funds most clinical work (NIH funded < 5 % of phase‑III spend from 2010‑2019).

3. Annual health exams focus on functional capacity will become the new “vitals”—body composition, cardiorespiratory fitness, insulin sensitivity, cognition, and mental health (which protect from ~90% of chronic morbidity and mortality)—offered at:

   • community centers/gyms

   • schools and workplaces

   • in your own home

   • concierge clinics (for a select few)

   • how many of you have a primary care doctor today, and why wouldn’t a shift towards having primary care wherever you are become the norm?

4. Early‑disease screening relies on multi‑cancer blood tests, with whole‑body MRI as the next stop. Traditional biomarkers of “disease” will become less relevant, as multi-omic biomarkers capture early declines in biological function before a “disease” is present.

5. AI‑interpretation of biomarkers and home‑sensor alerts (fall detection, silent arrhythmias via watch or wearable) nudge people into the traditional system when needed.

6. At‑home lab testing is ubiquitous, but in‑person visits remain vital for cutting‑edge tests and, honestly, high‑quality human connection. The COVID-19 “natural experiment” revealed the enormous mental health crisis and demonstrated the need to continue to foster in person connection with friends and care providers.

7. Fewer physicians, different skillsets. Driven by AI-interpreted ECGs, images, laboratory results, and implementation of treatment algorithms, there will be less of a physician shortage, and a different need for the human-centric skill sets. Care provider excellence prioritizes EQ over IQ—the MD role morphs into a blend of health coach and therapist, rather than standardized test expert. These skillsets are typically found in palliative care docs, lifestyle medicine physicians, and psychiatrists today and will be amplified.

8. Environmental toxin cleanup drops early‑onset cancer and autoimmune rates (molecular mimicry will be revealed as a mechanism underlying numerous specific etiologies).

9. Medication moves from lifetime to intermittent. Gene‑edited LDL suppression (see Lilly’s $1.3 B Verve acquisition, for example) means there will continue to be a move away from chronic dosing to single-time, or intermittent treatment periods. Treatments are increasingly designed to induce remission rather than life long management. Health plans look more seasonal—protocol blocks matched to life stage.

10. First line treatment will be diet and exercise in practice, rather than solely in theory. Driven by policy changes, the exploding longevity industry, and nutrition-therapeutic hybrid models (see Faeth therapeutics), what we know as “fitness” today will become “health” in the future.

11. Optimization therapeutics—more muscle, less visceral fat, better mitochondria, sharper memory—roll out to everyone, amplified by precision exercise‑nutrition stacks. Community health hubs and new “mini‑towns” make implementation more accessible over time and in one’s own context.

Subscribe now

1. How do we decide what roles can and should be replaced by AI? (e.g. why would we ever prefer a human cardiologist to read our EKG (80% accuracy of fatal arrhythmia detection) instead of a highly trained AI (99% accuracy))?

2. Who will be responsible for biological data ownership from newborn screenings and real world clinical trials? Clever models may include data ownership on the blockchain (e.g. AminoChain)

3. What regulatory model keeps pace with gene‑editing therapies that straddle treatment and enhancement?

4. If physician EQ trumps IQ, how should med‑school curricula and board exams change? The Step 1 USMLE already changed to Pass-Fail, and medical schools are eliminating courses on disease-centric pathology and pathobiology (in NEJM this week)

What will need to happen in order for the changes to occur in an accessible way?

During my MD and PhD training, I saw firsthand that most physicians knew very little about fitness—and that academic funding for exercise science was almost nonexistent. Less than 1% of NIH funding goes toward physical activity research. Even when I received a 14th percentile score on my F30 grant—at an institute with a 28th percentile funding cutoff—to study how exercise slows cancer growth, the grant was never awarded.

In clinical practice, it was rare to hear a physician talk about exercise with patients. Despite clear guidelines recommending lifestyle changes—like physical activity and reduced calorie intake—as first-line treatment for many chronic diseases, the focus almost always defaulted to medications. Which I don’t blame them because we didn’t learn how to do that in medical school patient interview simulations.

However, several powerful trends now make it overwhelmingly clear: the future of healthcare is fitness. And it’s no longer just the domain of gym bros.

Fitness as an Endpoint of Pharmaceutical Development

One of the clearest signs that fitness isn’t going anywhere is the pharmaceutical industry’s growing focus on fitness-related endpoints. It’s no longer “therapeutics instead of exercise”—it’s therapeutics inspired by the mechanistic benefits of exercise. (Turns out my PhD was just a few years ahead of its time.)

Just today, Eli Lilly committed up to $650 million to Juvena Therapeutics for a drug aimed at promoting “muscle health.”

““We are thrilled … [about the] transformative potential that promote healthspan in individuals facing chronic conditions like obesity and frailty,” said Dr. Hanadie Yousef, co-founder and CEO of Juvena Therapeutics. “Particularly as a strong body of evidence demonstrates the link between muscle health and enhanced metabolism, mobility, and even long-term disease prevention and overall health”

Outside of vaccines, it’s been exceedingly rare for drugs to be positioned in the context of prevention and health promotion—until now.

We’ve moved beyond testosterone and Olympic weightlifters. Today, myostatin inhibitors and other muscle-building pathways are getting serious attention, catalyzed by the formal classification of obesity as a treatable condition.

It won’t be long before we see drugs developed specifically to boost VO₂ max—even in the absence of disease.

Subscribe now

Longevity Movement Pushing the Boundaries of Healthspan

Whether it’s a VO₂ max test, methylene blue, or sober “Coffee & Chill” cold plunges in metropolitan cities, the booming longevity movement is getting people to talk about—and invest in—healthspan. People don’t just want to live longer; they want to feel what it means to be healthy. While the science behind some of these frontier interventions is still catching up, one thing is clear: the public discourse around health is shifting.

Exercise as Medicine Goes Mainstream

A recent randomized controlled trial published in The New England Journal of Medicine showed that a structured exercise program improves survival in cancer survivors—a landmark validation of movement as medicine.

At the same time, the popularity of figures like Andrew Huberman, Peter Attia, and Eric Topol reflects a cultural shift: science-backed lifestyle medicine is going mainstream. These voices aren’t just promoting fitness, nutrition, and sleep—they’re translating cutting-edge research into actionable habits for the public. And people are listening.

Subscribe now

Necessary Infrastructure for the Emerging Fitness-Oriented Healthcare Industry

• Biomarkers that define health and fitness: Whether used to quantify the effects of lifestyle interventions or pharmaceutical ones, a new class of diagnostics will be required. To promote healthspan—and secure reimbursement—we need to define what health looks like in the absence of disease. These biomarkers will form the foundation of both prevention-focused care and personalized therapeutic development.

• Integrated medical fitness centers: Primary care will increasingly resemble a hybrid of community center, gym, and wellness clinic—with a phlebotomist down the hall. While the traditional sick-care hospital system will persist, the default for prevention and health promotion will be a new kind of clinic: one rooted in movement, measurement, and behavior change. The Medical Fitness Association, founded in 1991, has laid the groundwork for this model—but what comes next will be more integrated, more tech-enabled, and far more scalable.

• Reimbursement Pathways for Structured Exercise Protocols: Insurance already covers nutrition counseling and rehabilitation for patients who are sick or recovering from surgery. But true prevention requires coverage before disease manifests. Structured exercise and nutritional interventions must be reimbursable as first-line therapies—not just reactive afterthoughts.

The Science Behind VO₂max

Let’s start 2 billion years ago.

For the first time, a cell and a mitochondrial symbiont assimilated, which endowed early organisms with oxygen-dependent energy production. As mitochondria hold the necessary machinery to transfer electrons onto oxygen in the process of ATP production, this made cellular functions significantly more energy efficient. Aerobic (oxygen-dependent) energy production can yield significantly more ATP per gram of glucose, for example, than anaerobic energy production (oxygen-INdependent).

I say this to orient you to human being’s historical physiological origins. Throughout evolution, our ability to utilize oxygen has endured, and represents a fundamental function of the capacity for survival.

Over the last 2 billion years, however, our machinery has become significantly more complex. Human have organs (!), and those organs function together in an integrated and coordinated fashion. How we get oxygen from the atmosphere to our mitochondria follows the oxygen cascade.

• The cascade begins as we breathe air (~20.93% of the air at sea level on Earth is oxygen) into our lungs.

• Red blood cells with low oxygen content accept that oxygen in the tiny vessels in our lungs, and store that oxygen in the protein hemoglobin which lives inside the red blood cells.

• Newly oxygenated red blood cells then make it back to the heart, which pumps oxygen-rich blood throughout the body.

• Oxygen is delivered to all the organs and is extracted by organs in which there is energy demand to fuel ATP production in the mitochondria.

• In the mitochondria, oxygen serves as the final electron acceptor in the electron transport chain, which are derived from the chemical reactions that break down fats, carbohydrates, and amino acids for fuel.

Even while resting, sleeping, etc. the body requires oxygen to generate and expend energy (that is, resting energy expenditure (REE)). However, during maximal exercise, your muscles can demand up to 80-90% of the body’s blood flow to get the oxygen it needs to support the energy demands imposed.

You can see (figure below) that oxygen-rich blood is shifted away from your other organs (e.g. kidneys, liver, GI) to maintain blood flow to the brain, and increase blood flow to your pumping heart and exercised muscles. Note also, the ability to increase blood flow to your heart and muscles can be trained, as athletes are able to get multiples of the maximal rate of blood flow to exercising organs during maximal exercise.

The body has numerous mechanisms in order to coordinate the provision of oxygenated blood to the organs with the most demand. Thus the single metric at maximal exercise, VO₂max, can serve as an integration of our system’s ability to meet the demands of maximal physiological stress.

When broken down into components (ie. lungs, heart, blood, vasculature, muscle, nervous system), we can also of course study those in isolation. There are a number of tests that lead to biomarkers that reflect the function or status of each of these systems independently.

• Pulmonary function tests can be performed to understand lung mechanics, volumes, and gas exchange in relative isolation from the rest of the body.

• One can study the heart in isolation with ECGs and echocardiograms (ultrasound)

• Standard labs can tell you the amount of red blood cells, and the fraction of your blood comprised of red blood cells (hematocrit), and the amount of hemoglobin.

The isolation of each of these components serves a role in describing disease in modern medicine. While the coordination of all of these components into a single metric tells you how the entire system is functioning.

Subscribe now

VO₂ Max is one of the strongest independent predictors of all-cause mortality

When studied in hundreds of thousands of people across all races, ethnicities, ages, etc., one’s VO₂max is an extremely strong predictor of risk of death from any cause. If you’re 30–49 and in the highest fitness category (top 2%), your risk of dying from any cause is only 16% of the risk faced by someone in the lowest 20% of fitness.

Even moving into the next VO₂max category (21-40^th percentile) can lower your risk of death by 20-40%, depending on your age.

There any exceedingly few biomarkers that exist that show such a clear linear relationship for any person at any age that predicts your risk of death from any cause. Further, higher fitness is associated with lower risks of having diagnoses of chronic diseases or having drug prescriptions for chronic disease management.

The higher the fitness level, the lower rates of chronic diseases, biomarker abnormalities, and use of prescription medications. This graph is showing the rates of disease and medication use at an average age of 61.3 years old. The trend with smoking and increased fitness level suggests also that healthy habits (regular physical activity and lack of smoking) go hand in hand at the population level.

The trends are clear: higher fitness is associated with lower risk of death, less disease, and fewer prescriptions.

The clinical importance of measuring VO₂max annually (or other proxies of cardiorespiratory fitness) has been recommended for nearly a decade by physicians and scientists at the American Heart Association.

They in fact recommend that all adults should have their fitness measured every year as a part of their annual exam. While I never saw this while training in medical school, physicians, coaches, and trainers are increasingly offering fitness testing as many begin to truly recognize its importance. Perhaps in the coming years we will see VO₂max as a vital sign…

Stay tuned for the major methods to measure VO₂max, and how to improve it…

You likely have an intuitive sense of what health feels like—perhaps a combination of your day-to-day well-being compared to when you're sick, the public health messages you've absorbed, and interactions you've had over the years with the healthcare system.

Your personal definition of health is probably shaped by family beliefs, education, societal norms, and advice from healthcare providers. But defining health in a universally meaningful way remains challenging precisely because it is inherently personal and context-dependent.

Society’s collective definition of health has evolved significantly over time, directly influencing how resources are allocated toward maintaining population-wide wellbeing. Historically, modern medicine's successes have shaped a health system largely centered on the absence of disease. Though we've largely succeeded in eliminating infectious disease and have strong mechanisms of containing emerging ones (though still imperfect, e.g. COVID 19), we are in the midst of a growing chronic disease crisis and have the largest healthspan lifespan gap in the world.

This article explores how we arrived at this point: How has the U.S. historically defined health? Why has our definition led us to our current challenges? And importantly, what shifts are necessary to build a future where health means more than merely the absence of disease?

The Biomedical Model: Health as the Absence of Disease

Scientific breakthroughs in the late 1800s and early 1900s dramatically transformed medicine and public health. Discovering specific causes of diseases—such as bacteria, viruses, and nutritional deficiencies—enabled precise, targeted interventions like vaccines and antibiotics. Consequently, medicine's primary goal became the elimination of these disease agents, with health defined as the default state achieved once diseases were removed.

The period saw numerous Nobel Prizes awarded rapidly for breakthroughs in infectious diseases: 1901 for diphtheria, 1902 and 1907 for malaria, 1905 for tuberculosis, and 1907 for trypanosomiasis (“disease causing protozoa”).

Soon after, surgical innovations gained recognition, notably in 1909 for thyroid gland surgery and in 1912 for advancements in blood vessel suturing and organ transplantation techniques.

Thus, modern medicine established a clear template: identify a disease, then develop acute treatments like antibiotics or surgical interventions. Defining health as "the absence of disease or infirmity" logically followed this successful approach.

This biomedical model achieved remarkable results, exemplified by a dramatic rise in life expectancy from 47 years in 1900 to 68 years by 1950.

Treating Infectious Disease Became the Model for the Modern Healthcare System

Infectious disease eradication significantly influenced biomedical innovation and the standardization of healthcare practices. A landmark event was the publication of the first-ever randomized controlled trial (RCT) in 1948, which evaluated the effectiveness of the antibiotic streptomycin in treating tuberculosis. Previous research had demonstrated streptomycin’s efficacy in a dish (in vitro) and in guinea pigs. Preliminary non-controlled clinical trials were “encouraging but nonconclusive.”

The trial was able to demonstrate a statistically significant effect of streptomycin at preventing death and leading to significant improvements in young patients with acute progressive tuberculosis, clearly linking the presence of this drug to improvements in symptoms.

The RCT set the new standard for how to approach health and disease. This revolutionized healthcare and helped standardized the provision of safe and effective therapies to the public.

Building upon this successful paradigm, the medical field focused their efforts on identifying and classifying disease and developing a treatment to eradicate this cause. However, as treating infectious diseases and surgical techniques improved, chronic diseases became the major killers.

Mid 20th Century: Chronic Disease and the Rise of "Risk Factors"

Already by 1950 chronic diseases including heart disease, cancer, and stroke caused more than half of all deaths in the US.

This prompted multiple national efforts to identify factors related to chronic diseases.

In 1948, the Framingham Heart Study launched as one of the first major endeavors to “identify the common factors or characteristics that contribute to cardiovascular disease.” In the initial 6 year follow up study published in 1961, the concept of “risk factors” for heart disease was introduced, highlighting modifiable behaviors and objective metrics such as smoking, high blood pressure, obesity, high cholesterol, and physical inactivity.

In 1956, the National Health Survey Act authorized ongoing statistical surveys to measure illness and disability across the United States. This initiative further solidified the disease-centric approach, embedding metrics like blood pressure, lipid profiles, hemoglobin A1C, glucose levels, body weight, and body mass index as standards for determining health status. The goal was to normalize these metrics with lifestyle and therapeutic interventions.

We began to realize that using the highly successful model of identifying and treating infectious diseases had shortcomings when it came to mitigating chronic disease—obesity rates rose significantly, and the proportion of US adults living with chronic diseases continued to rise.

Towards the end of the 1900s and early 2000s, there were key signs that demonstrate there was a need to shift the view of health from one that simply avoids disease to one that promotes promotes health status (real-time function and quality of life) and longevity.

Measuring Population Health: From Disease Indicators to Functional Capacity

Next-generation population-wide initiatives in the United States began shifting focus toward promoting overall health and longevity. The US Surgeon General’s “Healthy People 1990” released in 1980, established measurable objectives for improving health and well-being nationwide. Initially focused on mortality reduction, subsequent iterations emphasized longevity and quality of life.

Healthy People 2000 explicitly aimed to "increase the span of healthy life," while Healthy People 2010 emphasized improving overall quality of life. The latest edition, Healthy People 2030, further emphasizes thriving, health maximization, and prevention of disease and premature death. Its primary goal is to "attain healthy, thriving lives and well-being free of preventable disease, disability, injury, and premature death."

In theory, this is fantastic, given the rising healthspan-lifespan gap. In practice, how will this happen?

To truly drive a society geared toward health, we must measure health not merely by the absence of disease, but by an individual's functional capacity. Healthspan refers to the number of quality years a person lives in full functional health relative to their age and life stage—able to perform essential physical, cognitive, and daily life activities with ease, and maintain resilience to life’s stressors. Achieving this requires scaling research-grade assessment methods to the broader population.

Early indications of this shift appear in the latest updates of the Framingham Heart Study, which now includes comprehensive evaluations like cardiopulmonary exercise testing (CPET) to measure VO2max, grip strength, high-resolution bone density tests, omics-based blood metrics, and functional MRI scans of the brain.

Interestingly, CPET was added only in 2015. In 2016, the American Heart Association declared that fitness should be a clinical vital sign. Cardiovascular fitness is one of the most powerful prognostic markers for mortality.

There is a very clear, dose-response reduction in relative risk of death associated with cardiorespiratory fitness, and no upper bound of the protective effects of fitness (Kokkinos et al 2022). This key metric displays health as a spectrum where lower is worse and higher is better.

Grip strength also has a very clear association where lower is worse and higher is better, at least for all-cause dementia incidence of dementia associated mortality.

Compare this to LDL and A1C, where lower/higher is not always better. (deeper dive into some of the world’s hottest biomarkers coming in future posts)

From Kip et al 2024 (LDL) and Selvin et al 2010 (A1C) demonstrate J-shaped curves related to blood levels and hazard ratio of all-cause mortality. Interestingly, having an A1C <5.0% was associated with a statistically significant increased hazard ratio for all-cause mortality (HR = 1.43 (1.17-1.74)) compared to 5.0-5.5%.

Future health metrics must therefore focus on assessing individuals' functional capacities—such as intrinsic or vitality capacity—to provide clearer, more actionable measures of health, quality of life, and longevity.

Rethinking Health: Critiques and Emerging Frameworks

While population-level lab values have led to significant advances in mortality, there appears to be a relative plateau in healthspan.

Additionally, when health is defined on a spectrum, recommendations are clear and actionable at any point—one can always improve their health status and decrease mortality risk, rather than wait in the “normal zone” until signs of disease are present.

Now say these new metrics of health are adopted: how might we use this information practically?

1. Top down recommendations from healthcare providers (clinicians, dieticians, health coaches)

2. Peers or digital tools to encourage healthy behaviors

Changing the Healthcare Provider Relationship

Changing healthcare recommendations takes a long time. Clinicians in training have already learned a gargantuan set of rules by the time they are practicing in the real world. But if novel metrics of health become commonplace, clinicians may be able to focus on the 3-5 metrics that truly matter, making personalized interventions much more straightforward, rather than complex interpretation of many lab values that show up in the “normal” range.

This simplified view focusing on the few proven metrics that actually matter (e.g. VO₂max, body composition, grip strength) that determine both health status and are associated with lower rates of all-cause mortality, will serve as the baseline for health exams. While disease screening labs should be ordered in the presence of symptoms or at specified intervals of age, the relationship between the healthcare provider and the client may largely be focused on lifestyle and behaviors to optimize these key measures of health.

Achieving Health Outside of the Traditional Medical System

With the rise of wearable health monitoring, direct to consumer lab testing, the growing model of gym/wellness centers as clinic, and recent commercial traction of digital health companies (e.g. Hinge Health’s IPO, and Omada Health’s S1 filing), how people access their health data and healthcare is changing rapidly. As “behavior change” is often hailed as the necessary target to improve chronic disease, technology offers a unique ability to modify healthy behavior.

By the way, this topic deserves a post of its own.

Recently when in LA, a friend (Zach Poll) and I discussed how to techify health behaviors. He referred me to a study that showed that basketball teams 1 point behind at half time were more likely to win than teams 1 point ahead at half time. This study points to the “comeback effect,” where there may be a psychological advantage or improved motivation related to being slightly behind a competitor.

What if you were matched with an individual just ahead (or tied) with you on a certain health metric, and you agreed to participate in a protocol to improve your health scores?

A tool like stadium science provides a compelling platform for competitive sleeping. Leaderboards in gyms through apps like Wodify allow one to compare not only to other gym-goers, but to oneself over time.

In a proactive health-promotion focused system, the traditional clinic will likely not be the space in which the majority receives primary care.

Conclusions

The evolution of health has been shaped by each era’s specific needs, initially driven by the remarkable success of the biomedical model in combating infectious diseases through targeted identification and eradication. We’ve achieved enormous strides over the last century and this is worth celebrating. But now we need to pivot, and quickly. Chronic diseases are complex and multifactorial, highlighting the limitations of viewing health merely as the absence of disease. Despite changing attitudes, the prevailing disease-centric biomarkers still primarily detect illness after it manifests.

To achieve genuine improvements in population health, we must shift our perception and adopt scalable measures focused on functional health. Disease screening will always play a role, but future health assessments must prioritize evaluating functional capacities and their scalable proxies.

Whether optimal health is achieved within clinics, gyms, or through digital technologies in everyday environments, focusing on the right metrics is essential for advancing individual and public health.

Adaptability is arguably one of the most essential qualities influencing the health and longevity of living organisms. It represents the ability to respond effectively—and ideally, improve—in the face of diverse challenges, minor or major.

As society grows increasingly sedentary and chronic diseases become ever more prevalent, we face a pressing question: how can we retain and enhance our adaptability?

In this article, I’ll first introduce key concepts to illustrate what adaptability entails. I’ll then outline general principles related to measuring adaptability. Ultimately, adaptability should be recognized as a foundational element when identifying meaningful biomarkers of health.

Subscribe now

Principles

SAID Principle

In one of my kinesiology 101 courses back in the day, I first encountered the SAID principle: Specific Adaptations to Imposed Demands. The question posed was how effectively long-term endurance cycling training translates to improvements in a running-based VO₂max test (technically VO₂peak, but I’ll use them interchangeably). The short answer: your running VO₂max will indeed improve, but it would improve even more if tested pre- and post-training with a running-specific protocol. Simply put, your physiology adapts most specifically to the demands you're directly targeting.

Later, in medical school, I observed a significant shift in perspective. The nuanced distinctions of exercise type, modality, duration, intensity, and other factors were often simplified and lumped together as merely “exercise.” While any exercise is undoubtedly better than none, this generalization neglects specificity and leaves clinicians with few precise metrics to evaluate efficacy. Consequently, it is uncommon for physicians to prescribe detailed exercise regimens, and even less common for them to help patients understand if their efforts yield measurable results.

This gap highlights a key disconnect between the medical and fitness communities, contributing to a growing divergence between medicine and health.

So what do you do with these anecdotal stories about the SAID principle?

If you are about to embark on a new lifestyle or therapeutic intervention, remember that your body responds very specifically to what you impose on it. A drug targeting a single protein will induce changes related specifically to that protein’s function (assuming no off-target effects). A ketogenic diet will lead to physiological shifts directly tied to reduced carbohydrate stores and elevated blood ketones, causing water loss and changing the fuel sources used by your heart, brain, and muscles.

Another practical consideration is having a clear "north star" to evaluate efficacy. For endurance training, metrics like sports performance times and VO₂max tests clearly indicate whether a training protocol is effective at improving fitness and health. In medicine, such clear indicators for diet or exercise effectiveness are rare, leading practitioners to default to metrics that merely demonstrate the absence of disease.

This could perhaps be a whole post if you find it interesting.

Share

Early Principle

Have you ever tried to learn another language? Were you 4 or 40? Which was easier? How about riding a bike?

It's commonly understood that learning a new skill is easier and more enduring the earlier in life you learn it. This advantage partly stems from factors like reduced anxiety during childhood and implicit learning, as opposed to consciously applying memorized rules like many adults do. However, another widely accepted explanation is the heightened neuroplasticity present in childhood and early adulthood. You could just as easily call this phenomenon neuroadaptability, highlighting that your brain adapts more readily when it remains highly malleable.

The practical takeaway here is simple: if you're contemplating taking on a new challenge to improve yourself, start as soon as possible. The earlier you begin, the greater the impact of your efforts. Additionally, starting early likely influences the maximum threshold of skill or health status you can maintain throughout your later years (see next section).

Ceiling Effects

You’re only as capable as your maximum capacity—your "ceiling." Therefore, it makes sense to raise this ceiling as high as possible in any domain, with health perhaps being the most vital. Although you can still become fitter and healthier even later in life, it becomes progressively more difficult due to the SAID and Early principles discussed above.

The traditional lifespan-healthspan curve demonstrates this ceiling effect clearly—the goal of existing metrics typically centers on remaining disease-free for as long as possible.

However, it is possible that your ceiling can be raised, further protecting you from health decline and even enhancing your overall health. For example, lifelong exercise is known to be a critical countermeasure against age-related fitness decline (Valenzuela et al 2020). The earlier you raise the ceiling, the stronger your defense against decline—“healthspan.”

Additionally, cognitive adaptability—such as learning a new language late in life, engaging in activities like chess, or even regular socialization—can both defend against disease and actively improve health.

While a ceiling determines your maximum capacity, actively fostering adaptability can help raise this ceiling even further.

Subscribe now

Rest & Work

Doesn’t more work equal more productivity? Not necessarily—there are limits.

For beneficial adaptations to occur, you need both a potent stimulus and adequate rest periods for repair, renewal, and recovery. I resonate with Peter Attia’s observation that most people are undertrained, considering fewer than a quarter meet the recommended USPTF physical activity goals for both cardio and resistance training. Nonetheless, gradually ramping up intensity and ensuring sufficient rest are essential for achieving optimal adaptive benefits.

An Interesting Application to Immunotherapy

I remember reading a fascinating paper (during my PhD, peak science twitter era) demonstrating that immunotherapy could be made more effective at treating cancer by allowing cells periodic rest. A major issue in CAR-T cell immunotherapy is T-cell exhaustion. Evan Weber, from the Crystal Mackall lab at Stanford, characterized the transcriptome, epigenome, and functional status of CAR-T cells. They demonstrated that strategically providing transient rest periods could reprogram the epigenome of these cells, ultimately prolonging their anti-tumor effects in living animals.

From cellular to organismal levels, rest can indeed make you more effective.

Applied Adaptability

Whether applied to life, business, fitness, medicine, or elsewhere, this phrase rings true: "The best predictor of your ability to do something is having previously done it." Although you can't have done everything before, there are strategies to assess your capacity. Measuring adaptability—improvements in capacity—can be especially valuable when aiming to improve in any endeavor.

Most biological tests today rely on static measures, usually blood draws taken at rest from large populations. However, certain tests aim to reveal our body’s true capacities through stress tests, assessing responses to (patho)physiological challenges.

Of course, VO₂max testing exemplifies this concept. By pushing your body to maximal exercise, you determine how effectively it utilizes oxygen. As your fitness improves through regular exercise, the stress test becomes relatively easier, reflecting an increase in your capacity.

Another interesting example is how the body responds to repeated cold pressor tasks. One study aimed to determine if dunking your bare feet in ice became less stressful with repetition. Over a series of five cold pressor tasks, subjects gradually experienced reduced pain, a lower heart rate response, diminished pupillary dilation, and notable changes in alpha and beta frequency bands as measured by EEG. Generally, the same stressor became less stressful and less painful over time.

If you want difficult things to become easier, embrace repeated stressors. Adaptability is both essential for and indicative of health. Continuous improvement and growth serve as powerful defenses against illness.

Simply measuring EEG, heart rate, oxygen consumption, or any single metric alone provides limited insight—these are static biomarkers. Future health assessments will increasingly focus on identifying responsive, dynamic markers indicative of physiological adaptability.

“Man is the measure of all things” - Protagoras (ca. 485-410 BC)

Humans have long been interested in measuring things. It’s no surprise that one of humankind’s favorite things to measure is humans themselves.

Biomarkers are measurable indicators of different biological states, ranging from simple vital signs to more complex physiological or molecular signatures. In the context of human health, they serve roles in diagnosing disease, predicting mortality risk, directing therapies, and determining one’s exercise capacity. Throughout the last century biomarker discovery, development, and deployment have evolved in response to both technological developments and the needs of society at the time.

In the early 1900s, acute diseases like infection, trauma, and nutrient deficiencies necessitated methods of infectious disease and vitamin deficiency detection. As chronic diseases dominated in the mid to late 1900s, emphasis shifted towards quantifying risk of heart disease, diabetes, cancer, etc. With the advent of computational biology, molecular-level population analyses (like the human genome project) and multi-omics methodologies precision medicine became possible with much more sophisticated biomarkers.

Moving beyond the “absence of disease” definition of health

While the last 100+ years has gotten the US much more capable of diagnosing disease, the payoff in healthspan has diminished.

This is in part because in all of the advances in the diagnosis of disease vs. “normal,” health has never been truly defined.

In this article I will briefly describe trends in biomarker development and deployment in the US Healthcare system, and suggest that for healthspan to improve, there needs to be a partitioning of biomarker development:

1. The natural evolution of precision medicine biomarkers is to continue to molecularly define disease subsets.

2. There is a glaring unmet need for biomarkers of healthspan to be defined. Clearly, defining health as the “absence of disease” has diminishing returns— there is no North star for achieving health and quality of life.

I have previously written on what makes a good biomarker and potential functional biomarkers that could redefine “health”. I would love to hear your thoughts to help refine a collective vision of healthspan.

Subscribe now

Early 20th Century: The Era of Acute Diseases

Much of the modern US healthcare infrastructure began in the early 1900s, when the dominant health challenges were related to preventing death from acute diseases. Infectious diseases, trauma from war (and the need for safe blood transfusions), and nutritional deficiencies (e.g. scurvy) were leading causes of illness and death. This necessitated innovation in detecting microbial insults, screening blood for incompatibilities, and early biochemical assays of disease status. Several key innovations occurred:

1. Direct measurement of pathogens for diagnosis and prevention

   1. Malaria parasite blood smears in the 1890s

   2. First serologic blood test for syphilis in 1906 – the Wassermann reaction – had been developed, enabling diagnosis of this “social disease” via antibody response

2. The identification of ABO human blood groups in 1901 (Karl Landsteiner)

   1. These became one of the first widely implemented “biomarkers” in hospitals and eventually in blood banks

   2. This was driven by an acute need to support lifesaving surgeries and injuries in the wars of the early 1900s

3. Innovations in biochemical assays

   1. The quantitation of blood glucose in the 1910s marked the beginning of a transition from physicians relying on tasting sweet urine to diagnose diabetes, to using quantitative assays for glycemic control.

   2. Nutritional and vitamin deficiencies could be defined and measured

   3. Urea, creatinine, and uric acid for kidney function were also defined, to diagnose kidney disease prior to severe side effects

With the ability to mitigate acute illnesses, and a massive jump in life expectancy, the needs of healthcare changed significantly.

Mid 20th Century: Chronic Diseases Emerge

As Americans began to live longer, the predominant health concerns became chronic diseases like heart disease, cancer, stroke, and diabetes. The solution at the time was to be able to diagnose and define each of these diseases objectively, and ideally as early as possible. The concept of a biomarker as an indicator of future disease risk emerged, as many of these diseases are asymptomatic until very late.

Heart disease for the first time topped the list as the major killer in the US. This prompted the launch of the Framingham Heart Study in 1948, allowing for longitudinal monitoring of heart disease risk factors (in a proactive fashion). Key biomarkers that served as “risk factors” for heart disease included elevated cholesterol, high blood pressure, and ECG abnormalities.

This led to the classification of hypertension and hyperlipidemia as independent, treatable “diseases.” As the concept of asymptomatic population screening emerged, technology adapted. Multi-test “panels” were developed, which allowed physicians to have a standardized set of biomarkers of disease risk. This still dominates the practice of medicine today and include

• Electrolyte panels

• Lipid panels (HDL, LDL, triglycerides)

• Kidney function panels

• Liver enzymes

Endocrinology also saw a big increase in biomarker development, with markers like hemoglobin A1C (an integrated measure of long term glycemic control), thyroid function tests (T4, T3, and TSH), and reproductive hormones, cortisol, and insulin (though the latter are still measured mostly in specialty settings today).

In general, there began the movement of biomarkers that enabled early diagnosis and early intervention (regardless of symptoms). While diseases were better classified, some chronic diseases saw fewer early deaths, it has been more difficult to capture the effects of early diagnosis on healthspan.

Subscribe now

Late 20th Century: Technological Evolution of Disease Classification

The late 20th century and the early 2000s marked a period of intense technological innovation, beginning the the human genome project, innovations in mass spectrometry, and systems biology. There was an increasingly large aging population enabled by the biomarkers and interventions developed earlier in the century. Diseases related to obesity, metabolic syndrome, dementia, and mental health then dominated society. Innovation was largely driven by the new technologies available (genome sequencing, proteomics), and less because it directly addressed the needs patients living was chronic disease.

The largest beneficiary of the Personalized Medicine movement enabled by the sequencing of the human genome has largely been the field of cancer diagnosis and treatment. The concept of companion diagnostics emerged, where treatments could be targeted towards a very specific molecular biomarker. A prime example was the approval of trastuzumab (Herceptin) for HER2-positive breast cancer in 1998, which coincided with the first FDA-approved companion diagnostic test (the HercepTest immunohistochemistry assay for HER2 overexpression). In fact, the vast majority of the 188 FDA approved companion diagnostics are for cancer indications.

Interestingly, we are increasingly seeing objective metrics of complex diseases, like Parkinson’s and Alzheimers, such as the recently published MTBR-Tau243 blood test that correlates with AD disease severity in a dose response fashion. However, early trials aimed at targeting asymptomatic Alzheimer’s disease with anti-amyloid antibodies have demonstrated significant side effects with very little benefit so far.

While the US adopted many more sophisticated biomarkers, the return on healthspan has been minimal. In 2022, life expectancy at birth was 77.5 years for the total U.S. population. Though an imperfect measure of health, we’re not much further ahead than 72 years ago.

To both take advantage of novel technologies that can better diagnose and classify disease AND increase population health, there needs to be a shift in the evolution of biomarker development.

The Next Era of Health & Disease Biomarkers

I believe that natural evolution of biomarker development will diverge into two key classifications: biomarkers of disease (traditional approach, novel techniques) and biomarkers of healthspan (novel approach, new frontier)

Biomarkers of Disease will continue to be dominated by genetic underpinnings. Because drug development strongly emphasizes genetic support (there is evidence that genetic validation of a drug target increases clinical trial success), this is unlikely to change soon. However, the further classification of disease into molecular subsets will not rely on genetics alone. Proteomic biomarkers, especially those measurable through minimally invasive methods, are already emerging as secondary endpoints in Parkinson’s and Alzheimer’s diseases—and might eventually serve as surrogate endpoints, similar to Amyloid PET scans. Necessarily, within the disease-focused healthcare model, biomarkers will remain strongly linked to the diagnosis, prognosis, and mortality associated with a given disease.

Biomarkers of Healthspan are much less likely to be genomic. Although genetic associations with athletic performance have been identified, genes themselves are not responsive to interventions (except gene editing—a topic for another time). For guiding healthspan programs, functional genomics (e.g. RNA, proteins, metabolomics, epigenetics) are much more likely to play a key role given their responsiveness to environmental, lifestyle, and therapeutic interventions. Furthermore, these biomarkers will be disease-agnostic, shifting the definition of health away from merely the absence of disease toward functional capacity and quality of life. Measures of functional capacity (like VO₂max or insulin sensitivity) predict not only quality of life and resilience but also mortality risk.

Excitingly, we will increasingly see biomarkers measured at home, or on a person’s own time, enabled by wearable technology, increasing consumer health options, and eventual AI-enabled efficiency of health systems.

The Future is Healthspan

Without clearly defining health, quality of life will always be secondary to an intervention’s ability to produce measurable changes in biomarkers.

I envision a future where healthspan biomarkers, supported by sufficient real-world evidence, become the primary metrics for measuring population health. As individuals begin to show early signs of disease—identified through traditional metrics measured infrequently or in response to symptoms—they will transition into precision medicine approaches for disease management. The ultimate goal will not necessarily be to keep individuals within a “normal” range and outside of a “disease” range; rather, it will be to align lifestyle and medical interventions with objective measures of healthspan.

Subscribe now

Key One-liners for Consumer Wearables Start-ups:

1. Start with a niche!

2. New, continuous measures of health may be on the horizon

3. Form factor matters

4. Consider privacy concerns

5. Success requires both deep technical expertise and consumer marketing savvy

6. Marketing & brand are crucial

Conversation Below

Sonal:

“It looks like we’re both really interested in this space of consumer wearables. There’s so much happening—new companies, new technologies, and evolving markets. I thought we could kick off a discussion about what we’re seeing out there, how we think the market will shape up, and what a ‘company of the future’ might bring to the table. Maybe I’ll start by asking you: in your view, what kind of technologies exist right now in consumer wearables?”

Brooks:

“There are so many ways to slice it. One framework I like is to look at what sort of sensing the device is doing. First, there’s physiologic sensing—heart rate, temperature, respiration, oxygen saturation. That’s the most common and widely distributed. You see it in smart watches, rings, and bands using existing technology packaged into a wearable.

Next is molecular sensing, where you’re looking at proteins or small molecules like metabolites in different biofluids—blood, interstitial fluid, sweat, even potentially urine, though that’s less convenient for a ‘wearable.’ Within that category, you might have hormone sensing, which can measure things like cortisol or other biomarkers indicative of stress or reproductive health. All of these land in a few broad buckets: physiology, metabolism, and hormones.

Sonal:

“It’s super interesting that you bring up the possibility of integrating all those elements. My ideal vision for wearables is having one sleek device—maybe a ring or a band—that tracks my physiological changes, my hormones, and delivers predictive analytics about stress or mood. That kind of all-in-one wearable is the dream.

But each technology has trade-offs. We’ve been doing a lot of diligence on aptamer-based sensors, for example. As you said, they can measure physiological and hormonal biomarkers. The question is: how many different markers can you measure before you’re forced to expand into a larger form factor or develop multiple layers of sensing? That’s often the bottleneck—fitting many complex assays into one tiny ring or band.”

Brooks:

“Aptamers—or even antibody-based approaches—run into the challenge of binding affinity and turnover. If you engineer an aptamer or antibody to be super specific for, say, beta-hydroxybutyrate (a ketone), it might bind that molecule so tightly that once it’s bound, it doesn’t ‘un-bind’ easily. You end up saturating your sensor. That’s not ideal for continuous monitoring.

On the other hand, if the binding is too weak, you might get cross-reactivity with other molecules. It’s a tricky balance between specificity and the practical reality of continuous sensing. I’ve seen some companies—like Persperion—try to do sweat-based sensing. But sweat itself can be unreliable unless you force sweat production through something like iontophoresis or exercise. Their method requires pressing a fingertip on a sensor for thirty seconds, saturating that sensor, and then re-testing. So, it’s not strictly continuous.

One big question is whether these devices need to be fully continuous 24/7. Maybe you only need repeated tests throughout the day to get actionable insights. Where do you think the market stands on the need for truly continuous versus intermittent monitoring?”

Subscribe now

Sonal:

“You bring up a really good point, and that’s where market segmentation comes into play. Let’s take two different examples. If you have a chronic condition like PCOS, it could be super valuable to see perpetual hormone readouts—how your levels fluctuate throughout the day or month, and how diet or medication might help. But do I need a minute-by-minute update? Probably not. Checking daily or hourly might be sufficient. On the other hand, for a heart failure patient, minute-by-minute readings might be critical to predict cardiac events.

This suggests that maybe you start with conditions or markets where extremely high-frequency data isn’t necessary—where having data every few hours or once a day is still a game-changer. Then, as your technology matures, you move into more demanding indications. That approach also dovetails nicely with how to choose a beachhead market: start somewhere you can deliver real value with the technology you currently have, then expand outward.

What are your thoughts on consumer wearables that claim they can do it all—target athletes, patients, and general wellness? Do you think it’s better to start narrow or broad?”

Brooks:

“That’s such a key question: depth versus breadth. In healthcare, you could go the route of tackling a very specific disease state—diabetes, for example, where you focus on continuous glucose monitoring. Or heart failure. Or a particular hormone imbalance. You do one thing really well, and that’s enough to drive adoption in that population.

By contrast, the purely consumer side can focus more on lifestyle. For instance, Oura staked out ‘sleep optimization’ as its core. Whoop targeted the elite athlete. But functionally, Oura and Whoop measure extremely similar things: PPG-based heart rate, SpO2, HRV, etc. The difference is brand and marketing. Oura says, ‘We own sleep;’ Whoop says, ‘We’re for athletes trying to optimize performance.’ It’s about how you position yourself to the consumer.

You asked about starting narrow or broad. I lean toward starting narrow—like you said, picking a very precise group such as people training for their first half-marathon. Offer them a device with a tailored approach: a specific set of biomarkers, real-time or near real-time feedback, and advice that’s unique to that journey. You earn trust, collect meaningful data, and then expand. I think the same logic applies in a clinical context: pick a single chronic condition (e.g., heart failure or PCOS), prove your value, and then add new indications over time.”

Sonal:

“Yes! I completely agree. A company we were recently looking at decided to start with heart failure. I love that approach. They’ll develop the technology and hone in on that patient group first, then expand outwards as their technology gets more sophisticated. You also highlighted something important: form factor. I have an Oura ring; most of my female friends do as well. Meanwhile, most of my male friends use Whoop. There’s definitely a gender or style dynamic to this. If you’re targeting PCOS, you probably want a sleek, subtle ring that doesn’t scream ‘medical device.’ If you’re targeting hardcore athletes, maybe a big wristband is acceptable—it can even look cool, like a badge of honor. So yes, there’s a big brand and design aspect that can’t be overlooked.”

Brooks:

“Exactly. And I like how you framed that half-marathon example. It’s so precise that the group feels special and recognized. That’s a proven startup playbook: serve a niche deeply, then expand.

Another thing I’m thinking about is how we’re still measuring mostly existing clinical biomarkers. There’s an exciting possibility, though, to develop new, more holistic biomarkers—ones that aren’t just about diagnosing disease but measure how close you are to your optimal health state. Right now, so many biomarkers are binary: normal or abnormal. But for those hovering on the borderline, we don’t have a lot of nuanced guidance beyond ‘exercise more, eat fewer sweets, sleep better, drink less.’

In an ideal world, we could track continuous markers that reflect a range—like, instead of just labeling you ‘healthy’ vs. ‘pre-diabetic,’ we’d say, ‘You’re at a 4.7 out of 10 on your metabolic resilience scale. Here are the specific interventions to push that to a 6.0 or beyond.’ That prescriptive element could be huge. But developing novel biomarkers means educating the market, and that’s a big lift.”

Sonal:

“Exactly. We’re in the era of first measuring the basics. But as consumer datasets grow, we can start building that deeper, more personalized analysis. Some groups are already feeding Apple’s internal health data—heart rate, motion data from AirPods, even iPads or iPhones—into large-scale studies. The long-term goal is discovering novel biomarkers for mental health conditions, or for overall wellness. That’s the best-case scenario, where large-scale consumer data leads to real breakthroughs.

But there’s an ethical dimension. Apple (and other tech giants) can gather all this data, like they’re doing with their new health study, but how do we trust them to use it only for health innovation versus, say, ad targeting or differential pricing? It’s a sticky situation. Part of me loves that consumer data could drive new diagnostic markers. Another part worries we’re giving giant companies the keys to deeply personal health data.”

Brooks:

“That’s the core tension. In traditional pharma, there’s a clear incentive to keep patients on medication—chronic conditions are a revenue driver. On the consumer side, big tech’s business models often revolve around advertising or data monetization. Neither scenario is purely altruistic.

But, like you said, we can still drive positive change. If we can encourage biomarkers of health—rather than disease—companies might compete on getting people to a better state of well-being instead of just labeling them as ‘sick.’ The question is how we reorient the system away from normal vs. abnormal, and toward a continuum where each person can see progress and strive for improvement.

In truth, we have a mismatch between the concept of ‘normal’ (usually population averages) and what’s actually healthy or optimal for an individual. We might not want to be ‘average’ if the average population is in poor metabolic shape. So maybe these new wearables and biomarkers can define a better standard than the standard reference ranges we see in clinics. Still, that’s a big challenge—like you said, it requires large datasets and careful framing.”

Subscribe now

Sonal:

“And I wonder if that’s where we come full circle: going after specific market segments—like half-marathon runners in training—can help. They become the reference group for each other, not the general population. Same goes for a PCOS community, or for an older adult population focusing on cardiovascular health. Once you build enough data, you can define new ‘healthy ranges’ for those more precise segments instead of defaulting to broad references tied to age and BMI.”

The technology side is crucial, but so is market focus—the right customer, the right set of biomarkers, and eventually the creation of novel, aspirational health metrics. All of that leads me to a final question: Who are the right teams to build these companies? I’ve seen amazing technologies that never get traction because they’re missing the marketing and consumer-savvy piece. Oura and Whoop are as much lifestyle brands as they are health-tracking devices. The ‘coolness’ factor is part of why people buy them, not just the sensor specs.

So you really need a team that’s deeply technical and deeply consumer-aware. That’s not easy to find. Sometimes the founder is a PhD who’s never launched a brand. Other times it’s a marketing genius who struggles with sensor engineering. Striking that balance is going to be vital.”

Brooks:

“I agree 100%. If you look at how some consumer wearables have succeeded, they often have a brilliant marketing approach right from the start. Consider companies that tout their waitlists of tens of thousands of eager users—often, it’s not because those users understand the ins and outs of the tech, but because the company told a compelling story about what this device could do for them. Then you look at Oura’s pitch at Slush from 2017—where they literally had a magician on stage! That’s the level of showmanship it takes to stand out as a consumer brand.

For purely pharma-focused solutions, you can have a more clinical, data-driven pitch. But for general consumer wearables, you often have to be showy, emotional, or aspirational to spark interest. Technical brilliance alone isn’t enough.”

Sonal:

“Completely agree. That’s such an important insight. Marketing, distribution, brand positioning—those can make or break a consumer health startup. Anyway, this has been a wonderful conversation, Brooks. Hopefully, we’ll find something interesting to invest in or even build together.”

Brooks:

“Absolutely. Thanks so much, Sonal. Looking forward to seeing how all of this develops.”

(End of Conversation)

Key Takeaways

1. Segments & Niche Use-Cases

• Start with a tightly defined user base (e.g., half-marathoners, PCOS patients, or heart failure patients). Prove out the device’s value and data quality in that market, then scale outward.

2. Biomarkers Beyond Disease

• Moving from traditional clinical ‘normal vs. abnormal’ lab values toward new, continuous measures of health could transform how we track and improve wellness.

3. Form Factor Matters

• Rings, wristbands, patches—each appeal to different demographics. Design must align with both comfort and consumer identity (fashion, lifestyle, medical discretion, etc.).

4. Ethical & Data Considerations

• Large-scale consumer data can unlock novel biomarkers and major health insights, but also raises privacy concerns when held by big tech.

5. Team Composition

• Success requires both deep technical expertise and consumer marketing savvy. A purely technical founding team may struggle with brand-building; a purely marketing-focused team may not execute the R&D effectively.

6. Marketing & Brand Are Crucial

• Wearables are lifestyle products as much as health monitors. Companies like Oura and Whoop differentiate themselves less on the science and more on how they position their product in consumers’ lives.

Thanks for reading! If you enjoyed this discussion, feel free to share your thoughts in the comments or reach out with any questions. We’re excited to see where consumer wearables go next—from enabling healthier lifestyles to potentially revolutionizing chronic condition management.

—Brooks & Sonal

As of 2023, 44% of Americans own wearable health tracking devices. The global wearable technology market is expected to reach $152 billion by 2029 with a CAGR of 16.8%. People purchase wearables for a variety of reasons—from tracking fitness and monitoring sleep to understanding how your body’s glucose levels respond to bagels. Now, even watches are emerging as a more convenient way to take phone calls than your cell phone.

As a millennial, I remember FitBits as the original wearable to measure step count. Turns out the concept of a pedometer dates back to the 1500s (potentially originally designed by Leonardo da Vinci). As technology has progressed, the quantified health movement has evolved dramatically.

In this brief article, I want to highlight four key areas where we’re seeing high levels of innovation. The main idea is how wearables will evolve over the next several years. Previously, I pointed out a major issue with current wearables: the insight-to-action gap. In other words, while you might see your sleep quality, glucose curve, or hydration status (insight), it’s still up to you to identify and implement the necessary intervention (action).

We’re beginning to see haptic feedback, automated alerts from wearable devices, direct brain stimulation, and more—all triggered by signals detected from the body. Since we humans love to categorize, I’ve grouped these wearables into several types. Although this categorization is neither exhaustive nor overly complex, its how my brain works. The groups include clinical applications, brain health, physiology, and metabolic health monitoring.

Additionally, I’ll briefly cover go-to-market strategies, especially for companies straddling the line between clinical utility and consumer appeal.

Subscribe now

Clinical Wearables

We’re increasingly seeing clinical-grade wearables featured in studies or implemented in clinical trials. This is fantastic—we can now gather real-time, objective biomarkers from patients in real-world settings. I see two primary ways this is being done.

1. Turning existing clinical grade devices into wearable devices (that is, EEG, ECG, SpO2, Temperature, etc.).

At least one study demonstrated that the Apple Watch can achieve 100% accuracy on measure heart rate both at rest and during supraventricular tachycardia. AI can now detect potential fatal arrhythmias from ambulatory ECGs with up to 98.6% accuracy (compared to 80.3% by human technicians).

How long before we close the loop with healthcare system alerts based off heart rhythms from a watch?

While there are FDA cleared EEG devices like from Zeto, we will eventually move to clinical grade EEG with even less invasive methodologies (see Brain Health applications).

Diagnosing epilepsy typically requires observing two seizures on separate occasions—a process that can take a long time and involve numerous hospital visits, especially for children. Imagine if an at-home wearable device could automatically detect seizures, allowing for an objective diagnosis and treatment while saving patients and parents precious time and reducing uncertainty.

2. Generating novel biomarkers from wearable devices correlated with disease status

Koios Care monitors Parkinson’s status by detecting bradykinesia (a hallmark symptom of Parkinson’s) from the wrist while patients are eating, etc. These real time biomarkers are being implemented in clinical trials with pharma to assess treatment efficacy, and ultimately to inform long term dosing.

With wearable datasets growing every day, the enhanced statistical power and advanced algorithms are bound to uncover associations between novel wearable features and clinical status. This breakthrough could pave the way for closed-loop medication management and delivery, helping patients remain symptom-free.

Subscribe now

Brain Health

Smart headphones can tell you when you’re losing focus. Neurable’s algorithms can convert low and messy signal to interpretable EEG signals. This allows the user to understand their level of attention and focus and take breaks in real time and prevent burnout. Neurode is creating a closed-loop version of attention monitoring with a headband that can determine brain activity via fNIRS, AND deliver a mild non-invasive brain stimulation via a headband.

As technology evolves, not only will we be able to get clinical grade EEGs in every day devices (like the Apple Airpods patent for EEG detection), but much more. I envision the capacity to get quantitative fMRI, MEG, and perhaps even MRS in a discreet wearable device.

This would enable us to capture real-time, quantitative metrics of the brain’s physiological, electrical, and metabolic activity, offering high-resolution, interpretable insights into brain health over a lifetime. While these various measures might ultimately point to the same fundamental health interventions—exercise regularly, eat vegetables, and get good sleep—the early detection of disease could lead to just-in-time interventions, potentially enhanced by closed-loop systems to optimize our health.

Physiology

Much of wearable physiology tech has focused on leveraging the ECG. For instance, Whoop offers a strain score based largely on heart rate during exercise and a recovery score primarily derived from overnight heart rate variability (HRV). While HRV is widely regarded as one of the most valuable health metrics, much about it remains to be understood—I still don’t find it particularly explainable (see features of a great biomarker). Moreover, the industry hasn’t agreed on the best way to calculate HRV; each company seems to rely on different ECG features. (e.g. “Apple watch uses SDNN, Whoop and most other consumer devices (including Garmin & Oura) use RMSSD.” – Karl Etzel).

See Samir Akre’s self-experiment demonstrating the lack of correlation between HRV as determined by Apple and Whoop.

Sleep is perhaps one of the most insightful metrics. I hear many people finding relief when their device confirms they slept like crap. Perhaps others have seen how much alcohol or late caffeine intake affects their REM sleep.

While the gold standard in sleep testing is a polysomnography test in a lab, devices like Oura can combine plethysmography (tracking blood flow and heart rate), accelerometry (monitoring motion), and temperature to determine sleep phases.

When researchers conducted the head to head comparison between the Oura Ring, Fitbit and Apple Watch were compared to polysomnography, the Oura Ring Gen 3 appeared to have the highest level of agreement and ability to distinguish between sleep cycles (REM, light, and deep sleep). An interesting problem—how to close the loop with sleep? Automated bed temperature adjustments? Automated ambient lighting?

Closed loop hydration: Epicore Biosystems recently published the ability to administer real time haptic feedback (vibrations) when the wearable device detects too much sweat and electrolyte loss. Rather than looking at your phone after an exercise bout, you can understand in the moment the need for hydration intervention. Interestingly, Epicore is also collaborating with the US anti-doping agency to detect doping in athletes.

Subscribe now

Metabolic Health

Continuous glucose monitoring is the most widely adopted tool for tracking metabolic health. Companies like Levels (consumer-first), Dexcom, and Abbott (clinical-first) offer devices that measure glucose in the interstitial fluid using a tiny needle inserted under the skin. This technology allows consumers to see how their blood sugar responds to meals and enables patients with diabetes to assess their insulin response. However, in highly controlled conditions like those at the National Institutes of Health, research still demonstrates the high variability of glucose responses. The Dexcom device yields measurements varying as much as 30+ mg/dL between their device and venous blood. The study’s authors (at my old workplace) emphasize that for personalized nutrition recommendations to be truly effective, these measurements need to be even more accurate.

Beyond Glucose: Biolinq has developed wearables that are capable of real time glucose, lactate, and blood alcohol monitoring (according to it’s SBIR grants). Its website also demonstrates the capacity to measure ketones. These metrics could guide precision nutrition interventions and be applied in clinical settings where ketosis serves as a key biomarker.

Emteq Labs is developing smart glasses that detect facial musculature to determine chewing rate real time—this is a great metric for a closed loop solution for more mindful eating. Slowing chewing rate is a modifiable health behavior that has been demonstrated to be associated with lower calorie intake. I mean, I feel like I do chew salads a lot more slowly than mac and cheese.

Wearable Go to Market

A major question for new wearable start-ups is whether to pursue a consumer-first or clinical-first go-to-market strategy. My take is this: if you’re making a clinical-grade device more user friendly, you have an immediate clinical application—offering the benefits of less invasive technology. Take Dexcom, for example. They developed a high-quality glucose sensor that began with patients (eliminating the need for constant finger pricks) and later became available to consumers.

On the other hand, if you’re developing novel biomarkers, you often need large datasets to train your algorithms. You might either build this dataset through long-term clinical trials in academia or consider acquiring a large real-world dataset by offering value to a specific consumer market. The key question if pursuing a consumer-first strategy: how do you provide value to your first 1,000 users, especially when the biomarker isn’t widely understood?

If you decide never to make clinical claims, that simplifies your path. However, consumer companies face a lower threshold for market entry but must contend with more competition over time.

Finally, consider your company’s personality. Are you an energetic, social-media-first marketing company? Or are you an engineer-turned-entrepreneur who has never, and likely never will, download Instagram? All routes have their challenges, so choosing the one that best fits your company’s identity is the ultimate winning strategy.

Subscribe now

Beyond Health

Wearable devices are being used for aspects much beyond health. Brain computer interfaces (BCIs) is an entirely other topic, that is gradually blurring the lines between invasive technology and every day wearables. I first remember seeing the Meta Ray-Ban glasses, for example.

Even if not everyone embraces smart glasses, Meta recently released their vision for real world brain computer interface via a wrist watch that measure muscle activity from the wrist via surface electromyography (sEMG). With simple wrist motions, you could control specific features on your personal computer. While their visuals show users wearing both smart glasses and the wrist watch, I can easily envision a wrist-worn EMG device being useful on its own—whether for taking calls or triggering other pre-programmed actions.

A final ethical thought:

Eventually we will be faced with how much autonomy we decide to give technology. What if we let all our physiological and metabolic signals determine all inputs and outputs to our health? What will we do with all the time that we otherwise spend deciding where to eat or commuting to a care facility? What decisions will we fret over if not when is the best bed time for me, or whats the best alcohol alternative?