Jake Kelly, MD

Three Minutes

Jake Kelly MD — Mon, 31 Aug 2026 14:42:45 GMT

Fire Ops 101, Feel the Burn, Anchorage, August 29, 2026.

On Saturday, I put on roughly 40 pounds of turnout gear during Fire Ops 101, to experience what the Anchorage firefighters feel responding to 80% of all 911 calls in this city.

Fire Ops 101 is an IAFF (International Association of Fire Fighters) program that has been developed to show community leaders the diverse skillset required to perform the duties of a firefighter every day. I breathed bottled air, cut a car apart and crawled (attempted to carry) a hose line toward an actual fire. I performed chest compressions on a mannequin in 2 minute intervals, alternatively with my teammates while my shoulders and core burned just like our first responders do daily (except on real people). Anchorage runs 350 codes a year. My day was exhausting. Firefighters do this for 24 hour intervals repeatedly for an entire career.

I also learned about the expansive and expensive tools, equipment and infrastructure needed to keep a city safe. Before Saturday, I thought a “fire truck” was a red vehicle with a siren. Inside the fire service, there are engines and trucks. The engine carries the pump and hoses; their job is carrying and distributing water. The truck carries ground ladders, saws and forcible entry tools; their job is everything and it can be anything: prying (or removing) doors open, searching for people, cutting ventilation holes and pulling ceilings apart to find hidden fire.

Three Minutes

The 40 pounds of turnout gear is heavy: the firefighters’ clothes protect them against a completely different chemical (fire) enemy than it had to do fifty years ago.

In 1975, victims had roughly 17 minutes to escape a house fire. Today, you have three minutes.

Underwriters Laboratories, a global safety science organization that tests, inspects and certifies electronics, appliances, building materials and industrial equipment for safety hazards, ran paired room burns to understand how house furnishing and fire burn rates have changed over the years using two rooms with the same dimensions and amount of furniture. One room was furnished like in 1950 with cotton, wool and solid wood. The second room was furnished the ways homes and furniture are built now with polyurethane foam and polyester.

Every modern room reached flashover in under five minutes. In three of the four experiments, the legacy room never flashed over before the test was stopped at 30 minutes.

Sofas, mattresses, carpet and the foam in our dining room chairs are all petroleum products. Even though “legacy” rooms hold more fuel by weight, “modern” rooms with their synthetic makeup release their energy exponentially faster and hotter than 50 years ago.

Flashover

A room fire moves through stages, but flashover is the terminal event. It happens when the upper gas layer hits about 1,100 degrees Fahrenheit. Every exposed surface in the room then ignites at once.

The room is lethal well before that. A room approaching flashover is roughly 100 degrees Fahrenheit at the floor and over 600 degrees at eye level. Human skin is destroyed instantly at 162 degrees. Fire is not the main reason humans die in house fires. In residential fire deaths, thermal burns account for 6 percent, smoke inhalation accounts for 84 percent. Thirty-one percent of victims are asleep. Thirty-seven percent of victims die trying to get out.

Get Low

Most of us were taught the fire safety phrase as children: Stop, drop, and roll. Unless your clothes are actively on fire, it is effectively useless.

The real phrase is two words: Get low.

Heat, smoke, carbon monoxide and hydrogen cyanide all rise. The temperature gap is enormous between heights: 100 degrees at the floor versus 600 degrees at eye level, which is the difference between air you can breathe and air you cannot. Get low and get out. Do not stop to get dressed or to find the cat. Teach your children that no one goes back in. Do not hide from a firefighter. Get low and get out.

Eighty Seconds

Under the 2026 NFPA (National Fire Protection Association) standard, a career department has 80 seconds of turnout time for a fire call and 60 seconds for EMS. Imagine being asleep in a bunk to immediately jumping out of bed and rolling out the door in under 80 seconds, at 3 a.m., repeatedly throughout a night.

The clock does not start with the firefighters. It starts with a dispatcher. NFPA gives them 15 seconds to answer 95 percent of incoming calls and 64 seconds to process and dispatch 90 percent of them. Only then does the 80 seconds of turnout time clock start. In Anchorage the dispatchers belong to the same local as the firefighters and work 12 hour rotations responding to and triaging over 50 to 60,000 calls a year.

As a cardiologist, all of these statistics and numbers feel different to me. Heart disease causes 45 percent of on-duty firefighter deaths. Responding to alarms carries up to 14 times the odds of cardiac death compared with routine duty. It is not normal (or healthy) to go from dead sleep to a full catecholamine dump in seconds…bradycardia to tachycardia, normal blood pressure to higher blood pressures and unfathomable shifts in cellular and blood particle changes over and over across a 25-30 year career.

Anchorage firefighters work a Kelly schedule: 24 on, 24 off, 24 on, 24 off, 24 on, then four days off. The AFD (Anchorage Fire Department) crossed 50,000 calls in 2025, roughly double from a decade ago on with similar staffing (I bet your city has seen similar upticks in calls). Every potential life saving technique I performed poorly on Saturday, firefighters do flawlessly at 3 a.m.

Shut Your Door

There is one piece of firefighter vocabulary I would like you to remember: VEIS (Vent, Enter, Isolate, Search).

When a firefighter enters a bedroom window from a ladder, the first thing they do is close the bedroom door. Isolate. Closing the one door buys enough time to find a person and get them out alive. Firefighters entering a burning house with air on their backs close the door first.

You have the same door in your home.

Close your bedroom door tonight. Behind a closed, hollow-core door, UL measured under 100 degrees and carbon monoxide around 100 parts per million. On the open side? Over 1,000 degrees and 10,000 ppm. Your door buys you survivable air.

While you’re at it, take the “Tot Finder” sticker off the window. Kids grow up and move out and the sticker stays, sending a crew to search an empty room and telling a child to wait rather than leave. Firefighters search every room regardless.

Somebody you know has a smoke alarm older than they think and sleeps with the bedroom door open. Send them this article to remind them. It only takes ten seconds and it could save someone’s life.

My Ask

From September 6 to 10, I will be riding 343 miles from Boston to New York with 500 other people: one mile for each firefighter lost on 9/11, on the 25th year anniversary. The First Responder Ride for Resilience supports charities that exist because firefighters have an increased risk of cancer (and cardiovascular disease).

I am personally matching donations up to $4,000. Every dollar you give between now and the ride, I will match one for one. Donate to my ride here.

If my ride is not your thing, please consider giving to your own community. In Anchorage, the Anchorage Firefighters Charitable Foundation is a 501(c)(3) run by the members of Local 1264. They do Fill the Boot for MDA (Muscular Dystrophy Association), coats for kids and support more than 30 local charities. Donate here.

Close your bedroom door tonight, check the date on your smoke alarms this weekend and give something to the people who will be awake at 3 a.m. whether you are or not.

Jake

Jake Kelly, MD, MHS, FACC is a board-certified performance and preventive cardiologist, Wasserman-certified in cardiopulmonary exercise testing, ACC Governor for Alaska, at Alaska Heart & Vascular Institute and founder of ROOL Health, a cardiometabolic longevity practice in Anchorage, AK and virtually in CO, FL, NC and TX. He practices where circulation, metabolism and cardiorespiratory fitness are treated as a single system. To learn more, work together, or get in touch, start at jakekellymd.com.

References

Kerber S. Analysis of changing residential fire dynamics and its implications on firefighter operational timeframes. Fire Technol. 2012;48:865-891.
UL Fire Safety Research Institute. Comparison of Natural and Synthetic Home Furnishings (2020); Close Before You Doze (fsri.org/programs/close-before-you-doze).
Bukowski RW, Peacock RD, Averill JD, et al. Performance of Home Smoke Alarms. NIST Technical Note 1455-1. 2008.
U.S. Fire Administration. Civilian Fire Fatalities in Residential Buildings, 2017-2019 (2021); Home Fire Sprinklers; Orman JM, Stop, Drop and Roll: Is the Message Still Relevant? (Executive Fire Officer Program, 2004).
NFPA statistical reports: Hall S, Fire Loss in the United States During 2024 (2025); Smoke Alarms in U.S. Home Fires (2024); Home Structure Fires (2021).
NFPA 1750, Standard for the Organization and Deployment of Fire Suppression Operations, EMS, Special Operations and Fire Prevention Activities. 2026 edition.
Kales SN, Soteriades ES, Christophi CA, Christiani DC. Emergency duties and deaths from heart disease among firefighters in the United States. N Engl J Med. 2007;356(12):1207-1215.

Medical disclaimer

The information provided here, including articles, newsletters, social-media posts, videos, and downloadable resources, is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment. Your use of this content does not establish a physician-patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement. Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.

The Readiness Score Isn’t Ready

Jake Kelly MD — Fri, 28 Aug 2026 14:21:06 GMT

For readers new to the series:

In Part I, I introduced the Wearable Experiment in What’s Worth Measuring.
In Part II, I wrote about What Your Sleep Score Doesn’t Tell.
In Part III, I shared A Cardiologist’s guide for wearing a CGM in patients without diabetes.

Today, we’re mostly writing (Jake and Katie) about what Katie has learned about herself and her wearable data over the last 7 weeks of data overload.

JAKE
What is it that people actually want from their wearable data? Why crave new metrics? Some want validation that they’re safe. Others are competitive and want proof they’re outperforming the person next to them. Some are tracking a specific symptom looking for peace of mind; some have genuine FOMO; and, others are struggling, grasping at any tool or experiment that might help them get through the day.

What I wanted to learn from Katie’s experiment is how she actually feels and how that subjective sense lines up with the numbers. More importantly, I wanted to see what real data looks like on a real person living a normal, busy life across three different devices. She wore a Garmin, a Whoop and an Oura ring simultaneously for seven weeks.

Below we continue to write back and forth on her personal take home data interpretations. Let’s start with the number she ignores.

The readiness score doesn’t help me get ready

KATIE

As great as it is that these devices analyze my data and give a readiness score: I train early in the morning because if I don’t, it won’t happen. Even if my readiness score is low, the workout happens. We run long once on the weekend and ride the trainer for an hour-plus the second day and no matter what the app says, the training continues.

I might not feel 100% during the session and if I’m sore from a heavy lift, I’ll keep it easy. The weekly routine is largely fixed. I listen to my body, but I also have to work within the window life gives me.

JAKE

The data agrees with Katie’s memory. Across 47 wearable tracked days, there were zero days on which she did no exercise. The correlation between her morning readiness score and what she actually did that day was +0.05 for strain and -0.05 for workout minutes.

Statistically, the readiness score had zero impact on her behavior.

_{Zero days off out of 47. If anything, she did slightly more on red mornings.}

KATIE

Is it smart to push through? Sometimes exercise is the best cure when I don’t feel one hundred percent. I trained throughout pregnancy when I was nauseous and I continue now because the end result is almost always better focus and a feeling of well-being. Jake and I enter different races and adventures that might require longer rides or more swimming, but in general, my training is for life and consistency is my strength.

JAKE

Katie’s data shows zero physiological cost to her decision.

Looking strictly at alcohol-free nights so alcohol isn’t skewing the baseline: after exercising for 40+ minutes on a low-recovery morning, her next-morning HRV was 79.1 ms. After training that same amount on a normal recovery day, her next-morning HRV was 79.3 ms. Virtually identical HRV scores.

Her overall recovery score the next morning was slightly lower: 63 versus 70, but I suspect this primarily reflects where she started the previous day, not the effect of her workout itself.

Seven low-recovery training days is a small sample, so this isn’t proof that her exercising based on schedule rather than readiness is harmless. But minimal HRV deviations across seven weeks suggests that training through a “red” morning did not affect her HRV the following day. Her wearable data suggest what Katie already knew from her internal feelings: she felt fine, she was ready, life goes on and wearables don’t have access to these conclusions.

What I thought I already knew

KATIE

When we started monitoring everything, I assumed the different devices would surface different insights. But broadly speaking, I already knew wine disrupted my sleep the most. I also knew I benefited from going to bed early; I’m a natural early riser, so the first third of the night is where I bank my deep sleep.

In general, I eat clean. We eat out about once a week, but I cook the rest of the time. I average around 120 to 140 grams of protein a day, get some Zone 2 cardio in daily and lift twice a week. I had a solid health baseline to start with (even if I do love dessert after dinner).

JAKE

Over the seven weeks, the three devices logged 84 workouts across 49 days and Katie walked roughly 14,500 steps per day. Her body weight held steady between 131 and 134 pounds.

This is not a study of someone starting from zero.

Alcohol and Sleep

JAKE

Across 13 drinking nights and 21 alcohol-free nights, every device told the exact same story.

p = approx 0.001 across all three platforms

Every single one of her six worst recovery mornings of the summer followed an evening with alcohol. On 21 confirmed “dry” nights, her HRV never once dipped below 62 ms.

Yet, total sleep duration barely budged and Katie slept just as long on drinking nights as dry nights. Although the sleep was far less restorative. This is the finding that tends to surprise people most: a full eight hours in bed feels like it should count for something but the autonomic nervous system apparently knows the difference between restorative and fragmented sleep.

(Figure 1: Seven weeks of Whoop Recovery. Every red marker represents a morning following alcohol consumption.)

But there are nuances

KATIE

I knew I needed to avoid alcohol and I knew I needed to get to bed early.

JAKE

Katie thought bedtime and alcohol were two separate health rules. Katie’s data revealed that bedtime and alcohol compound on top of each other.

She had three consecutive nights including alcohol while camping in Fairbanks: all with normal, early bedtimes and these night’s yielded recovery scores of 52, 66 and 50. No large crash in recovery. However, her three worst recovery nights of the entire summer (July 24, August 4 and August 15) had bedtimes of 10:48 PM, 10:41 PM and 10:51 PM.

The actionable takeaway isn’t necessarily “never have an alcoholic drink.” An alcoholic beverage paired with a normal bedtime reduce Katie by 15 recovery points. The same drink paired with a late bedtime cost her 35 recovery points. Alcohol and the subsequent decisions it leads to, such as later bedtime (or late night ice ream for me) compound for the worst. The lever she (we) actually control is the clock.

But a late bedtime on its own does almost nothing

JAKE

Here is a result we didn’t anticipate: looking strictly at Katie’s alcohol-free nights, deviating from her usual bedtime barely disturbed her recovery or HRV on any device.

A gentle slope, three nights after and no statistical penalty. On its own, staying up a bit later cost Katie almost nothing from recovery and HRV perspective (though many wouldn’t consider 10:30 PM “late”). The bedtime penalty primarily occurs in combination with alcohol: going to bed late doesn’t directly harm her autonomic recovery, it multiplies the physiological hit from alcohol.

What I actually want from this

KATIE

At the end of the day, wearables are a useful diagnostic tool. If I still had a regular menstrual cycle, I’d like to track my symptoms closely and I think many women could benefit from the insight Whoop and Oura provide from self-reported symptoms and cycle-related HRV changes.

However, my routine and consistency will always override what a device tells me in the morning. The device trackers uncovered a real and actionable insight about the interaction between drinking wine and my bedtime. But the devices I wore offered zero evidence that I should skip a workout because that question was never theirs to answer.

(And as an aside: while we didn’t wear all three devices in Europe this summer, I kept my Garmin on. My sleep scores were solid and my HRV stayed in the high upper 60s to low 70s despite alcohol use and jet lag. An endless number of people say they tolerate food and wine better in Europe than in the States. That’s too large a topic to unpack here, but I will continue to enjoy traveling!)

What about the science?

JAKE

We deliberately avoided framing this post around other people’s published studies; these numbers are Katie’s and single-subject (N=1) data is as compelling and as limited as this implies. But two areas of literature related to our discussions are worth exploring more deeply:

On sleep regularity and consistency

On alcohol, heart rate variability and sleep

The population-level literature on sleep regularity is robust, yet Katie’s “dry” nights showed minimal sensitivity to bedtime shifts. This isn’t a flaw in the research; this is what an individual dataset can look like compared to an epidemiological average of thousands.

It is also the single best reason to track and trend your own data if you have questions: to find where the averages fit you and where they don’t.

Medical disclaimer

This is general education, not medical advice. If you have questions about your glucose, metabolic health, or wearable data, talk with your own clinician about what’s right for you.

A Cardiologist’s Guide to Wearing a CGM in Patients Without Diabetes

Jake Kelly MD — Thu, 20 Aug 2026 12:03:39 GMT

My hemoglobin A1c had risen to 5.8 and I was concerned. That was about 18 months ago. I am a cardiologist who tells patients to measure what matters and my own metabolic marker had moved in the wrong direction. So I put on a continuous glucose monitor.

I do not routinely prescribe CGMs to my patients. For most people with a normal HbA1c of 5.4 or less, a normal HOMA-IR and a healthy triglyceride to HDL ratio, they are likely to be unhelpful and not teach you a whole lot. But for people whose numbers are trending in the wrong direction, particularly hemoglobin A1cs in the high fives to low sixes, I think a targeted two-week approach can show you things that quarterly (or annual) labs cannot.

Over the past 18 months, Katie and I have worn CGMs about every six months for a few weeks. My average glucose this round was 91 mg/dL with a time in range of 99%. This corresponds to an estimated A1c around 5.0. Eighteen months ago I was at 5.8. The CGM did not fix that. But it showed me what was likely driving it and I changed those things.

Katie decided to wear a CGM because she is curious and enjoys learning about technology and wearing devices. She happily goes along with some of my experiments and if anything I’ve learned some of my curiosity from her. She has always enjoyed trying things like a Paleo-type diet, Whole 30 or a sugar detox to really understand and feel her body better. I learned from my wife, as many great spouses do.

What a CGM is

A continuous glucose monitor is a small subcutaneous filament that samples blood sugar continuously in the tissue beneath the skin. The sensor measures interstitial fluid as an estimate of blood glucose. There is a physiological lag of about 5-10 minutes between blood glucose and interstitial fluid in steady conditions. When glucose is rising or falling quickly, such as after a meal, that lag can stretch to 15-20 minutes. For practical purposes, what you see on the app is close to real time but not exactly real time.

CGMs are relatively accurate, but some sensors tend to read a little low or a little high compared to what may come out with your actual hemoglobin A1c or 90-day blood glucose average.

CGM for a diabetic versus a non-diabetic is a different story. One group of patients has significant insulin resistance. The other may have some mild insulin resistance or no insulin resistance. These devices were originally made to help people determine how to dose insulin for patients with type I diabetes mellitus and then translated over to people with type II diabetes mellitus. The question now is whether they’re useful for people who have neither.

How to wear it

The best places to wear them are on the back of your arm, but you can also wear it on your abdomen or off label on the upper outer part of your thigh. I find that using it on my non-dominant arm makes me less likely to knock it off, although this week I knocked my CGM sensor out when I was wrapping up an electric cord connected to my electric dog grooming tool this weekend. It ripped it right off easily and painlessly. Katie’s came off on day 8 and she was without it for 12-16 hours before applying a new sensor.

When the CGM is deployed, sometimes I feel a little bit of a slight prick sensation and often times I feel nothing. The first time is a little anxiety-provoking because you want the skin to be clean and place it in just the right spot. But the second or third time you deploy the sensor, or if you have your spouse do it, it’s easy.

Sweating, hot weather, hot baths, hot tubs, swimming or sauna can make the sensor more likely to fall off. There’s a nice patch you can place on to help adhere it better in these environments.

Sensors last about 14 days. There’s no charging at all. The most likely time that you feel it is when you’re bathing and using a washrag with soap on your arms. I try to be careful not to knock it off. I’ve also noticed that tight fitting clothing can catch it.

The first week: just watch

I think the best strategy in week 1 is just to observe and not change your behaviors whatsoever. The first time you’re wearing a CGM you really want to eat your normal diet, have your normal treats, eat at your normal times and really see how your body (the CGM) are responding.

If anything, say there’s a behavior you do once a month such as go out to dinner or eat dessert late or have a couple late night beverages or long exercise habits. I’d like to include these activities in the first week so that you can see what your body is doing and how your glucose is responding.

Ideally we’d know what insulin is doing as well, but we know that insulin follows glucose so we can get some idea of what insulin may be doing. The apps typically let you log what you ate, when you ate it, and what else was happening. Then you can start seeing patterns.

What you’ll likely see:

If you’re metabolically healthy, your trace will probably be boring. Katie’s seven-day Lingo average was 86 mg/dL with a time in range of 94% and variability of 14%. Mine was 91 mg/dL average, 99% in healthy range, 10% variability. A flat, boring trace is good news. It confirms what your labs showed. Not every device needs to reveal a problem to be useful. Sometimes the value is confirmation.

If you’re borderline, you may see things your labs missed. Research suggests that about one-quarter of people with normal HbA1c levels actually reach prediabetic glucose ranges up to 15% of the time on CGM. Standard labs average over 90 days. The CGM shows you the days that raised the average.

Things that surprised me the first time I wore one: Potato chips after skipping breakfast spiked me to 180. Late-night eating, particularly ice cream or alcohol, raised my blood sugars the following day by about 10 from baseline and it could take up to 48 hours to return to normal with a normal diet pattern afterwards. Those findings changed my behavior permanently.

Something most people don’t expect: Stress raises glucose without eating. Cortisol drives fasting glucose up more than almost any food. You may see your glucose climb during a stressful day at work, an argument, or after a terrible night of sleep, without having eaten anything bad for you. For high-performing professionals who already eat well but live hard, this is often the most actionable finding from a CGM.

Just by wearing a CGM, my glucose time in healthy range gets better and my average glucose goes down.

What the numbers mean

Time in range: Greater than 90% is normal for a non-diabetic. This tells you that your body is adequately responding to glucose across your full macronutrient intake.

Average glucose: Ideally less than 5.8 A1c (about 120 mg/dL) and potentially closer to 5.0 for excellent (average glucose around 97).

Post-meal spikes: Normal, but getting outside 140 mg/dL suggests excess or unopposed carbohydrates. Your body should clear the spike within one to two hours. A banana that takes you to 130 (or 160) and back down in 45 minutes is a healthy response. Do not develop food fears over these changes. A bowl of sweetened cereal that takes you to 175 and stays elevated for 2 hours is a different story.

Fasting glucose vs CGM glucose: These should be fairly similar, but due to the differences in measurement, one from interstitial fluid and one from capillary blood, there may be a slight lag and small differences are expected.

The dawn phenomenon: In the early morning hours, typically between 3 AM and 8 AM, the body releases cortisol, growth hormone and glucagon as it prepares to wake up. These hormones signal the liver to release stored glucose. In a healthy person, the pancreas releases enough insulin to keep this rise in check. You may see your CGM glucose rise 10-20 mg/dL before your alarm goes off. For non-diabetics, this is a normal physiological process and not a reason for concern.

Week two: experiment

Once you have a week of observation, that’s when I think you need to have some insight. This is where the CGM earns its value.

What happens on nights that I stop eating three hours before bedtime, as I recommend? How does my fasting blood sugar look the following morning and on the day afterwards?

If you typically eat a bowl of Cheerios or sweetened cereal in the morning and then you change that to yogurt with berries, you may see that your glucose spike is not as steep or more flattened. When I started adding peanut butter or nuts to an afternoon banana I noticed that my blood sugars were not spiking nearly as high. Actually it would be very rounded and reasonable.

A patient of mine had always eaten healthy cereal, but it was still fairly carbohydrate-dense in the mornings with a small amount of milk. When she wore her CGM, she noticed that most days she ate cereal, her blood sugars would be higher, and throughout the day her blood sugars would be higher. She noted that when she had an omelette on the weekends, her blood sugars would be steadier by about 8 to 12 throughout the day. This observation led her to start having more fiber and protein in her diet throughout the day and starting her day with an omelette or adding yogurt to oatmeal, blueberries and raspberries to increase the fiber content in her breakfast.

Walking after meals. Pre-meal walks are not particularly helpful compared to post-meal walks, but on direct experience they help a little. A 2025 study found that as little as 2-5 minutes of light walking after a meal reduced glucose levels for up to 2 hours. A 10-minute walk immediately after eating lowered peak glucose by about 10%. The key is timing: immediately after eating is more effective than waiting 30-60 minutes. If you work out too hard post-meal, you can actually drive glucose too low, so casual walks are the sweet spot.

What CGM does not tell you

This is where I think the wellness CGM market overstates their case.

A flat glucose tracing could still occur in a patient who’s developing significant insulin resistance. CGM doesn’t tell you exactly what’s going on with insulin. It doesn’t replace labs such as hemoglobin A1c, fasting insulin, calculated HOMA-IR (fasting glucose in mg/dL multiplied by fasting insulin in uIU/mL, divided by 405), or triglyceride to HDL ratio. It doesn’t tell you directly about cardiovascular risk.

CGMs can give you food fears, particularly with healthy fruits such as bananas or cantaloupe, because you might spike your blood sugar. In reality your blood sugar is going up appropriately and coming down fine. There is a term for this: “glucorexia,” an obsession with glucose data that doesn’t reflect meaningful metabolic dysfunction. If you find yourself avoiding fruit because of a spike on your CGM, you have gone too far (I did too, albeit briefly). A cardiologist is telling you: eat the banana.

CGM data can mislead or cause unnecessary anxiety because you’re checking all the time and seeing these spikes. Appropriately eating nutrition based on a broad base of macronutrients and total calories is more important than specific intakes or macronutrient limitations.

Who should actually wear one

Honestly, for people that have a very normal hemoglobin A1c, HOMA-IR, and triglyceride to HDL ratio, a CGM is probably not going to be very helpful.

But if people have borderline values bordering on prediabetes and they go in trying to learn, I think they can really see that certain dietary patterns and lifestyle patterns make a large difference. If and when you exercise, if and when you walk after meals, if and when you have a sedentary day at work. Incorporating a little more fiber and healthy fats with your carbohydrates, protein at each meal and even some short activities throughout the day can really make a large difference in rounding out your peaks and probably reducing your average blood sugars and potentially improving your energy levels.

The A1c that you get every three months is a little late. It doesn’t give you that direct feedback. The CGM is a feedback loop. You can see your blood sugars going up after you’re eating.

I do think CGMs should be used with a physician to help you understand what’s going on and interpret anything that’s really challenging. I do discuss indications and data from CGMs with my patients, but I do not routinely prescribe the use of CGMs.

The wellness CGM market probably believes you need to wear these much more frequently. For most patients that are curious or have a borderline hemoglobin A1c, doing a targeted two-week approach to learn something with a follow-up in two to six months to see that you’re maintaining these habits is a healthy strategy. Ideally, the use of a CGM would be monitored under the guidance of a physician, nutritionist, dietitian or endocrinologist who understands how such an instrument can help you.

The two-week framework

One week passive. One week intentional. Then take it off.

The devices

We use the Lingo because we’ve used it before and already had the app. The cost is typically about $45 to $50 per two-week sensor for these over-the-counter devices.

The OTC devices are not quite as high fidelity as their prescription counterparts, but they’re pretty similar.

Next week: Katie's data. Different physiology, different patterns, and likely a few surprises.

If you have been told your A1c is borderline, or if you have been curious about what your body does with the food you eat, a two-week CGM may teach you more than another round of labs. If your trace is boring, that is good news. Send this to someone who has been thinking about trying one. It might change how they approach it.

References

Danne T, et al. International Consensus on Use of Continuous Glucose Monitoring. Diabetes Care. 2017;40(12):1631-1640.

Shah VN, et al. Continuous Glucose Monitoring Profiles in Healthy Nondiabetic Participants: A Multicenter Prospective Study. J Clin Endocrinol Metab. 2019;104(10):4356-4364.

Buffey AJ, et al. The Acute Effects of Interrupting Prolonged Sitting Time in Adults with Standing and Light-Intensity Walking on Biomarkers of Cardiometabolic Health. Sports Med. 2022;52:1765-1787.

Freckmann G, et al. Continuous Glucose Profiles in Healthy Subjects under Everyday Life Conditions and after Different Meals. J Diabetes Sci Technol. 2007;1(5):695-703.

Medical disclaimer

This is general education, not medical advice. If you have questions about your glucose, metabolic health, or wearable data, talk with your own clinician about what’s right for you.

What Your Sleep Score Doesn’t Tell You

Jake Kelly MD — Thu, 13 Aug 2026 14:16:46 GMT

We flew to Juneau on a Saturday to race a triathlon. We drove to Fairbanks the following week to watch our daughter play soccer. In between we worked full clinic days, packed a 34-foot trailer, fed three kids and two grandparents and tried to remember to keep four wearable devices synced and charged.

This was supposed to be the first structured week of our experiment. It was not structured. It was chaos. The data that came back taught us more about sleep than a routine week.

The week that destroyed our recovery

Here is what happened, day by day, mapped against what the devices recorded.

Sunday, August 2. Olympic triathlon in Juneau. Cool weather, good race. That evening we had dinner with friends, two glasses of red wine and went to bed later than usual. My Garmin sleep score the next morning: 84, Good. Katie’s Whoop recovery the next morning: 29%, red. Her resting heart rate spiked to 59 bpm, thirteen beats above her recent average. Her HRV dropped to 46 ms.

Tuesday, August 4. We put on the CGM. I saw clinic patients all day. That evening we had an amazing dinner out with Katie’s parents. I had a beer and a glass of wine, Katie had 2 glasses of wine and we had dessert. We went to bed with full stomachs finishing dinner at 9 pm. My Garmin sleep score said 84, Good. Katie’s Whoop recovery: 19%. Her HRV: 39 ms. Her resting heart rate: 62 bpm. This was her worst recovery of the entire month and it arrived two full days after the triathlon. The combination of the triathlon and wine had a huge impact on her recovery.

Wednesday, August 5. I woke up exhausted. My HRV had dropped to 58 ms, the lowest reading of the month. My Body Battery was 44, also a low. This was the morning after the late dinner with alcohol and dessert. My Garmin sleep score still said Fair, 77. Katie’s Whoop: 41% recovery. We were both in a recovery hole and we both had a full day of work ahead.

Thursday, August 6. We ran 30 minutes in the morning, I finished my introductory Wearable experiment article and we drove 7.5 hours towing a 34-foot trailer to Fairbanks with 2 giant schnauzers. Katie’s parents managed our three kids. Halfway into the drive, Katie shifted to our second car with her mother and I drove the remaining distance with my father-in-law. Katie and Gammy arrived to our daughter’s game in time for warm-ups. Papa and I parked and set-up the trailer just in time to see the beginning of game one. There was no time for recovery. We were on to the next thing.

Thursday through Saturday. Soccer games every day. One beer Thursday, one Friday, none Saturday. Sleeping in the trailer. My sleep scores bounced between Good and Excellent. Katie’s recovery stayed yellow: 66%, 50%, 58%. Her body never got back to green.

Sunday, August 9. We drove 7.5 hours home. Lunch was plantain chips, sliced ham, and cookies in the car. Dinner was freezer compartment scavenging. We found Alaska Pasta Company, frozen meatballs and some leftover spinach and broccoli. We were running on fumes.

Monday and Tuesday. Clinic both days. Katie’s parents flew home. We were both depleted in a way the sleep scores did not accurately reflect. My Garmin said Good and Excellent. Katie’s Whoop said yellow. We felt grey.

The recovery deficit compounds

The most important pattern in this data in my interpretation is that there was not any specific bad night. It is the adding up and compounding (for the worse).

Katie’s lowest recovery, 19%, arrived two days after the triathlon. Not the morning after the race. Two mornings later. The physiology behind this is straightforward: exercise-induced muscle damage triggers an inflammatory response that peaks 24-48 hours after the event, not immediately. Her autonomic nervous system was still processing the race while she was already back at work, eating late dinners and drinking wine.

Then the Fairbanks trip stacked on top of our incomplete recovery. Each new stressor, the long drive, the disrupted sleep environment, the schedule compression kept raining down on our central nervous systems that had not yet recovered and returned to baseline. Katie’s HRV went from 90 ms (Aug 3, the morning she felt good) to 39 ms (Aug 4) to 57 ms (Aug 5) and never climbed back above 77 ms for the rest of the period.

My pattern was similar but less volatile. My HRV dropped from 88 ms to 58 ms after the late dinner night, partially recovered to the 70s during camping and was still bouncing between 67 and 85 ms a week later.

The lesson: one day of recovery is not enough to erase one day of stress. The ratio is closer to 2:1, sometimes more. If you stack three hard days in a row (race, travel, late night), you may need most of a week to get back to your baseline (or longer). Our sleep scores never reflected this. Our bodies did.

What your sleep score actually measures (and what it misses)

My Garmin gave me a sleep score of 84 on the night I went to bed with a full stomach, a beer and a glass of wine. The next morning I had the lowest HRV and lowest Body Battery of the month. The score said Good. My body thought otherwise.

Sleep scores are composites. They blend duration, time in each sleep stage, movement and sometimes HRV into a single number. The problem I see is that the weighting is proprietary to the company, the inputs are estimated and the score does not know what happened during the day before you fell asleep. The formula that calculates the score does not know you ate at 9:30 PM. Or that you drove for seven hours pulling an RV with your right foot constantly on the gas pedal or break. It does not process how you felt about how well your daughter played or how much game time she saw. The score captures what happened between the sheets from a device peering in from your right (or left) wrist or finger. The devices don’t capture what you brought into bed.

Why do all of these other inputs matter? Because the score is what most people check. It is the number on the home screen. If the score says 84, you move on. You may not look at the HRV trend underneath, deep sleep duration, REM time or the Body Battery recharge. The score often becomes the story and the story is frequently incomplete.

Two devices, two answers about deep sleep

This is where it gets uncomfortable for device manufacturers.

Katie wore an Oura Ring and a Whoop MG every night for 30 days. Both measure sleep staging. Both report deep sleep duration. Here is what they said:

Oura Ring 4: Average deep sleep per night: 57 minutes.

Whoop MG: Average deep sleep per night: 128 minutes.

Same person. Same nights. One device says she gets about an hour of deep sleep. The other device says she gets over two hours.

Both devices use accelerometry and photoplethysmography (optical heart rate) to infer sleep stages. Neither is polysomnography. Neither is measuring brain waves directly. They are estimating sleep stages from wrist or finger movement and heart rate patterns and their algorithms appear to disagree substantially where to draw the line between deep sleep, REM and light sleep.

Which one is right? Neither, probably. The absolute number is unreliable. A 2022 validation study found that consumer wearables agree with polysomnography on total sleep time reasonably well but diverge significantly on sleep staging, particularly deep sleep. Oura tends to undercount it. Whoop tends to overcount it. The truth is probably somewhere between 57 and 128 minutes and you cannot really know unless you validate it in a sleep lab.

So what do I think the the number may be good for? The trend. If your deep sleep drops from 90 minutes to 50 minutes over two weeks and your energy levels confirm this change, this is a signal. If you are comparing your deep sleep minutes to a friend’s, or to a target you read online, you are probably comparing noise.

What alcohol actually did to our data

Katie said something in her journal this week that I think captures it better than any chart: “Wine is evil.”

She elaborated: beer seems less disruptive to her sleep. One or two beers during camping did not tank her scores. But wine, even two glasses, correlated with lower HRV, worse motivation the following day and a subjective heaviness that lingered.

My data showed a similar but less volatile pattern. The night I had a beer and a glass of wine with a late dinner (August 4), my HRV dropped to 58 ms the following morning, which was the lowest of the month. My Body Battery recharged to only 44. But a single beer during the camping trip (August 8) preceded my best sleep score of the month: 96, Excellent.

Is this proof that wine is worse than beer for sleep? No. These are two people over ten days. But the pattern is consistent with what the literature shows: higher ethanol doses suppress REM sleep more aggressively and wine tends to be consumed in larger absolute alcohol volumes than beer. A standard glass of wine delivers roughly 50% more alcohol than a standard light beer. The dose matters more than the type, but the type often lead to a higher dose.

The broader point is that we both already knew alcohol was not helping our sleep. I wrote about breaking my daily drinking habit through wearable feedback in the introductory article. Katie tracks it through Whoop’s journal. The data is not telling us something new. But staring at the data repeatedly makes it harder to ignore.

What the CGM showed (and what it didn’t)

My glucose was remarkably stable through the entire week. Average: 91 mg/dL. Time in range: 99.2%. Even on the night of the late dinner with wine and dessert, my glucose barely rose to 145 before dropping back to the 80s.

For a non-diabetic with good insulin sensitivity, which is what my baseline labs confirmed, CGM during a chaotic travel week was the least informative device. The glucose trace was flat while my HRV was crashing, my Body Battery was bottoming out and my subjective energy was low. The metabolic system held steady even when my autonomic nervous system was under strain.

This is worth mentioning because CGM is often positioned as a window into how your body is handling stress. For some people it may be but for a metabolically healthy person eating a varied but not extreme diet, the sleep and recovery data told a far richer story this week than the glucose data did. I will write more about how to use a CGM to learn individual patterns around life, specific meals and exergcise training. This week our CGMs mostly confirmed what the labs already revealed.

Katie’s CGM fell off on day 5 and she was without it for 12-16 hours before applying a new sensor. This is a reminder that the devices are physical objects: they unstick, they need charging, they rotate on your finger and the band gets wet in the shower. The logistics of wearing four devices every day are annoying, problematic and can easily lead to long absences of wearing them.

What the journal revealed

Katie has been logging into the Whoop journal more consistently than I have. Her data over 31 days paints a picture that no single device metric captures: dark chocolate consumed 94% of days, added sugar 97% of days, calories tracked only 26% of days. She described a pattern that resonates: consistent at breakfast and lunch, flexible at dinner. She does not want to live so strictly that she gives up the things that make food enjoyable. But Katie wonders whether certain habits are serving her health or just comfortable or routine habits.

For the next two weeks, Katie plans to cut out gum and be more deliberate about evening sweets. She is not on a detox. Rather an experiment within the experiment. We will dig deeper into nutrition, metabolism and the hormonal context that shapes Katie’s data in a future article.

Device impressions at the two-week mark

Katie’s assessment:

Garmin (Forerunner 265): Still her first choice. GPS for runs, good sleep and HRV trends, connects to the scale. If she could only keep one device, this is it.

Whoop MG: She likes the app and the journal feature. The daily questionnaire forces reflection in a way the other devices do not. She does not love the band, particularly showering in it and she finds the bicep heart rate monitor more comfortable than a chest strap (chafing under the breast line with any chest-based strap). She likes that it charges without removing it.

Oura Ring 4: Small, unobtrusive. She switches it between two fingers to avoid irritation. The tags feature is less intuitive than Whoop’s journal, more tedious to fill in each day. She does not find the data superior to Garmin for sleep trends.

My assessment:

Garmin (Forerunner 965): My go to device. I track my GPS, it automatically updates to Strava. Light, comfortable. I do feel that I have to replace them every 2 years.

Whoop MG: It took me 2 weeks to get comfortable wearing the device. I love the app, the journal and the data. In particular, it gives me the best sleep scores, which I like.

Oura Ring 4: Easy to wear. Annoying to shave with it. I need to try using the app more. It is a beautiful app but I really haven’t dug in. Undecided.

Lingo CGM: Easy to put on. I’ve had issues in the past. Tolerated sauna. Harmless.

The common thread: none of these devices interprets the data for you. They measure. They score. They do not explain. The gap between measurement and meaning is what this experiment is trying to close.

What we are changing for the next two weeks

Based on what this week taught us:

Data routine. We are committing to a Sunday night export: CSVs from Whoop, Garmin, and Oura, plus a voice memo summary of the week. The biggest challenge of this experiment is not wearing the devices. It is aggregating the data. Five apps, three ecosystems, no shared dashboard. The friction is strong and it is the main reason most people stop after the first week.

Katie’s sugar experiment. Cutting gum, being intentional about evening sweets. Not elimination, just awareness.

Alcohol awareness. We are not going dry. But we are labeling alcohol nights more deliberately and looking at the +1 and +2 day effects, not just the next morning.

Recovery windows. After this week, we both understand that stacking travel, training, late dinners and full work days without a genuine rest day creates a deficit that does not resolve in 24 hours. We need to build in at least one true low-strain day after any high-strain sequence.

If you checked your sleep score this morning and felt fine about it, look at the trend underneath. One number does not tell you whether you are recovering or just sleeping. Send this to someone who lives by their sleep score. It might change what they look at tomorrow morning.

References

Haghayegh S, et al. Accuracy of Wristband Fitbit Models in Assessing Sleep: Systematic Review and Meta-Analysis. J Med Internet Res. 2019;21(11):e16273.

de Zambotti M, et al. Wearable Sleep Technology in Clinical and Research Settings. Med Sci Sports Exerc. 2019;51(7):1538-1557.

Ebrahim IO, et al. Alcohol and Sleep I: Effects on Normal Sleep. Alcohol Clin Exp Res. 2013;37(4):539-549.

Peake JM, et al. Recovery of the Immune System After Exercise. J Appl Physiol. 2017;122(5):1077-1087.

Medical disclaimer

This is general education, not medical advice. If you have questions about your sleep, recovery, or wearable data, talk with your own clinician about what’s right for you.

What’s Worth Measuring?

Jake Kelly MD — Thu, 06 Aug 2026 14:33:55 GMT

The Wall Street Journal asked which wearable device is the best. STAT News asked if any wearable device is ready for the clinic. My fellow cardiologists have criticized the lack of precision and that the wearable numbers don’t lead to changes in clinical decisions. A viral LinkedIn post shared that they as a health coach had “graduated from wearables.”

All the authors and physicians share valid concerns, criticisms and conclusions. But I think they may have asked the wrong questions.

I’m writing a series to explore and describe my personal journey using wearables in a more structured way to show all of you how to review wearable data in an objective way while I hopefully incorporate a subjective perspective. My hope is to identify some of my misaligned behaviors, teach some physiology along the way and just maybe show all of you what they’re all missing.

The experiment

I’ve been tracking my resting and exercise heart rates and heart rate variability (HRV) since 1999 and 2022, respectively. I started wearing Polar watches at the beginning of my triathlon career. I became interested in HRV during my Cardiology fellowship (2012-2015) forgot about it and rekindled my personal interest and began monitoring daily in 2022. I’ve prescribed and performed lactate testing, DEXA scans, run and cycle heart-rate and watt ramp studies along VO₂ max testing in athletes and patients throughout this same time period. But I’ve never monitored all of these data points in a structured, documented way for myself until now.

My wife Katie and I have been wearing a Garmin watch, an Oura Ring 4 and a Whoop MG every day since early July. We drew baseline labs. We underwent a DEXA scan, resting metabolic rate and submaximal lactate and VO₂ max testing. We tracked meals loosely, exercised normally, lived normally (which is not that normal in Alaska with all of our sunlight).

For the past 3-4 weeks, we just let the data sit there like most consumers do.

Our sleep scores were interesting. Our HRV trends were mildly informative. We checked the apps most mornings, nodded at the numbers, and went about our day. Nothing structurally changed. The devices were measuring us. We were not using the measurements.

Today we start the real experiment.

Day 1. 4 wearable devices, two guinea pigs, one scorecard.

For the next 30 days, we’re adding two things. First, a continuous glucose monitor, the Abbott Lingo, for the first two weeks. Second, and most importantly, a structured daily journal. This is the same weekly scorecard I use with my patients. Every day we’ll log sleep quality, energy, stress, mood, soreness, motivation to train, macronutrient intake, training type and duration, weight, blood pressure, sauna and any alcohol. At the end of every week we’ll score ourselves across eight categories that I have tried to map to cardiovascular physiology.

I believe that the difference between wearing devices and changing behavior is using a framework.

Why use wearable devices to change behavior?

Up to 80 percent of cardiovascular disease is considered preventable. The variables that drive prevention are measurable, modifiable and mostly behavioral: what you eat, how you move, exercise and train, how you sleep, how you manage body composition and sometimes measuring, monitoring and attempting to change biomarkers that are associated with cardiovascular events.

Consumer wearables touch some of these variables: heart rate, heart rate variability, sleep duration, training load but miss many of these biomarkers that associate more strongly with cardiovascular disease: ApoB, Lp(a), blood pressure trajectory, fasting insulin, visceral fat.

Your wearable doesn’t know any of these biomakers.

I think what can be useful, insightful and eye-opening is under-appreciated; the value of a wearable was never supposed to be diagnostic. A ring is not a blood test. A watch is not a stress echocardiogram or a coronary CT. The question isn’t whether the devices are precise enough to diagnose disease. The question is whether a wearable device can help the user identify misaligned behaviors and change the behavior.

I’ll tell you when I had my aha! moment. Aches, pains and subtle brain fog set in sometime in my early 40s. A fascinating article in Nature Aging identified 2 time periods of more rapid aging (44 and 60) that provided some relative comfort to what I was experiencing. At the time, I was wearing my Garmin as I had for the past 2 decades but not in a structured or disciplined way. I decided I wanted to feel better. So I used wearable feedback, my sleep scores, HRV trends and an honest subjective assessment of how I felt each morning to first identify and then to break an almost daily drinking habit. I went from having a beer or 2 most days of the week to drinking seldom. My data didn’t make the decision. But the pattern and response to my behavior revealed worse sleep, lower HRV and brain fog in the morning on the days after I drank. This feedback made me recognize that my behaviors were causing me to feel how I did. This is not a clinical trial. It’s one person paying attention to his own trajectory. But I changed a behavior that years of knowing better had not changed.

Recent evidence suggests that werables might be able to change behavior, but only with structure. A 2026 review in the behavioral medicine literature found that hybrid interventions, a wearable paired with a structured framework, outperform wearable-only or app-only approaches for sustained physical activity. The devices provide awareness. The framework that is used can convert awareness into action. Without the framework, you get what researchers have called the “guidance infrastructure gap”: data without interpretation, trends without context, numbers without a plan.

The gap between casually observing wearable data and individually, personally associating behaviors with wearable data is what I’m trying to assess, teach and show in this series.

The population-level data on wearables is less encouraging. Broad deployment of wearable devices does not seem to meaningfully change outcomes across large groups. But if you focus in really sharply on the individual level, at the person who actually pairs their data with a specific framework and a willingness to change, the outcomes can be different. I’m writing this series to help that person.

The tools

Four wearable devices. Each one maps to a distinct physiologic domain. If the devices measure the same variable and several of them do, the overlap might be the lesson. Same trends, potentially different absolute number. The signal is in the slope.

Garmin watch (Forerunner 965 for me, 265 for Katie). Heart rate, overnight HRV, sleep staging, VO₂ max estimate, training load, Body Battery. This is the primary training and cardiorespiratory fitness instrument. I chose Garmin over Apple Watch because Garmin is purpose-built for structured training: native Zone 2 and Zone 4 tracking, training load metrics and an ecosystem that connects directly to the scale. Also, full disclosure, I’ve used Garmin dating to the late 2000s.

Oura Ring 4. Finger-based HR and HRV (RMSSD), sleep staging, readiness score, temperature deviation. This is the primary sleep and autonomic recovery instrument. Is there overlap with the Garmin? Absolutely. This is part of the experiment.

Whoop MG. Strain, recovery, respiratory rate, HRV, ECG and an emerging blood pressure estimation feature that is currently under FDA regulatory review. Katie and I wear it on our wrists but it can also be worn on the bicep during exercise. Optical blood pressure estimation from the wrist is a promising technology but it is a poor substitute for a validated BP cuff. We use an Omron Evolv for blood pressure measurement, same arm, same time, every morning.

Abbott Lingo CGM. Continuous interstitial glucose for 14 days. Applied yesterday. We’ll wear it for the first two weeks of the experiment, then remove it. CGM teaches fast and plateaus by day 14 for non-diabetics. I’ll explain why in the series.

Garmin Index S2 scale. Daily weight. Same dashboard and App ecosystem as the watch. Weigh every morning, post-void, pre-food.

Katie tracks daily macronutrients with MacroTracker. I’m logging nutrition through Gemini and direct measurements.

A brief note on what we didn’t include: the Hume Band is an interesting emerging alternative at a lower price point, but it has a smaller validation track record and a proprietary “Metabolic Capacity” metric that was difficult for me to interpret clinically. I may revisit it and/or other of the myriad wearables at a future date.

The ground rules

All devices are self-purchased. No sponsorships. No affiliate links. This is stated plainly and will not change.

This is an n=2 experiment. Two adults, different physiology, same protocol, same instruments. Katie and I will attempt to identify and teach principles, not prove absolute physiology. Every article in this series will attempt to be grounded in published evidence. Our data makes the evidence tangible. It does not replace previous experiences and more clinically rigorous data.

Another interesting fact to share: Katie is 41, trains hard, and uses a Mirena IUD, which means she doesn’t have a predictable menstrual cycle to overlay on her wearable data. What Katie may have are the early, uneven hormonal shifts of perimenopause: disrupted sleep, increased soreness or changes in recovery that don’t follow a pattern the wearable algorithms expect. Most wearable data and content about female physiology focuses on cycle tracking. Katie’s reality is different here and it’s one that millions of women in their late 30s and 40s share. Her data will be interpreted in this context throughout the series. If this series does one useful thing, I’d like it to be this: showing that context matters more than any number or score.

The scorecard

This is the behavioral bridge between data and action.

Every week, for both Katie and me, we’ll fill in the same card that I give my patients:

Strength: 30+ minute sessions, compound lifts. Goal: 3 per week.
Zone 2: Total Zone 2 minutes. Goal: build to 90 min/week.
Zone 4: High-intensity minutes. Goal: 0 until Zone 2 floor is met.
Nutrition: Days meeting macro goals. Goal: 5–7 days.
Sleep: Days with 7+ hours. Goal: 7/7.
Mental clarity: Days you felt good. Goal: 5–7.
Exercise motivation: Days you felt like exercising. Goal: 5-7.
Stress: Days stress was manageable. Goal: 5–7.
Steps: Daily average. Goal: 8,000+.

The scorecard doesn’t require a single device. It requires honesty with yourself and the data. The devices can sharpen the inputs: Zone 2 minutes need a heart rate monitor, sleep hours are easier to track with a ring or a watch. But the scorecard is the tool that turns passive measurement into weekly accountability.

Week 1. The card that turns data into action.

The question driving 30 days

Can consumer wearables, interpreted through clinical physiology and paired with a behavioral scorecard, change what we actually do?

Not what we know, what we track or which device is most accurate. What we do.

The critique is fair: the advice is the same regardless of the numbers. Eat well, sleep enough, exercise and manage your stress. True. The question is whether the numbers help you actually follow the advice. This is what we’re testing.

The goal of this series is not to determine which device is best. It’s not to convince you to wear four things on your wrist. My goal is to show you how these tools, used with structure, ideally with the guidance of a coach, trainer, or clinician, can help you monitor, track and improve the specific variables that matter for your health and life: cardiorespiratory fitness, body composition, sleep, quality of life.

These tools can be empowering. They can also be overpowering. I’m about to find out which, in real time, with my own data overload. I’ll try to carefully balance the subjective and the objective so that what I share with you makes your life a little simpler and potentially a little better.

Let’s go on a journey together.

Follow along

If you want to run a simplified version alongside us: pick one device, whatever you already own. Download the scorecard, log your macros and track for 30 days. I’ll review reader data in a follow-up post.

Download the Weekly Scorecard PDF

This series documents a personal experiment, not medical advice. Discuss any changes with your physician.

If this reframed how you think about your wearable, not as a gadget to obsess over but as a tool worth using right, send it to one person who’s been staring at their sleep score wondering what to do with it. Let’s replace the confusion with a framework.

Medical disclaimer

This is general education, not medical advice. If you have questions about your wearable data, training, or cardiovascular risk, talk with your own clinician about what’s right for you.

Caffeine and Your Heart: It’s Not the Cup, It’s the Kind, the Count and the Cream

Jake Kelly MD — Wed, 22 Jul 2026 14:14:26 GMT

When I was going through my training, the teaching was simple: coffee is a stimulant, stimulants provoke the heart and if you had atrial fibrillation, you better put the mug down. I repeated it. If I could feel my heart after a strong cup, then others could too and that jolt could be trouble waiting to happen.

Then I spent the better part of two decades caring for patients (while sipping coffee day & night) and the association of coffee drinking and atrial fibrillation I’d been taught never transpired. The coffee drinkers weren’t the ones filling my clinic with new atrial fibrillation. I also watched my electrophysiologist colleagues, the rhythm specialists, the people you’d expect to be strictest about this, change their tune. This week the American Heart Association released a new scientific statement in Circulation that lines up with what my patients had already been showing me for years.

When I saw these guidelines released, the coffee lover and practicing cardiologist in me stayed up late reading. Here is my breakdown.

The answer depends on three things

Here’s the framing I use with my patients, because “is caffeine bad for my heart” is not a yes-or-no answer. It depends on how much caffeine is in your coffee, what kind and what you put in it. Get these three choices right and the evidence is genuinely reassuring. If you get these choices wrong, you can absolutely hurt yourself. Here’s my summary in my triangle framework.

My biggest take home message in the scientific statement is that the dose isn’t linear. More caffeine is not worse in stepwise fashion and it’s not steadily better either. For most cardiovascular outcomes, the risk drops to a low around three to four cups a day and then trickles back up. Blood pressure actually runs the opposite direction, a little higher risk at one to three cups, then risk lowering with more. This is why “just cut back” is an incomplete answer.

The metabolic surprise

Lets start with the good news: habitual coffee is tied, over and over, to a lower risk of type 2 diabetes, ~ 20 to 30 percent lower in people drinking three and a half to five cups a day. There is an interesting twist, it isn’t a caffeine buzz doing the work: decaf coffee shows the same benefit. The credit probably goes to other compounds in the bean: chlorogenic acid, magnesium and chromium. Coffee is not just a caffeine delivery system.

The rhythm story

What about the old teaching that coffee should be avoided in patients with or prone to atrial fibrillation?

In atrial fibrillation, light has been shined on the available data. Observational studies show coffee drinkers are at no higher, and often lower, risk of atrial fibrillation. It is not just epidemiological studies anymore. In a randomized trial of patients with atrial fibrillation planning for cardioversion (a procedure where we shock you back into rhythm in a controlled way), drinking caffeinated coffee was associated with 39 percent less recurrence. The updated ACC/AHA/HRS guidance now specifically tells clinicians not to advise patients to stop caffeine to prevent atrial fibrillation. If you were told years ago to quit coffee for your fibrillation, this advice has expired.

However, individual sensitivity to caffeine is real. I have patients who can tell you exactly which cup of coffee set their heart racing and I’ve had a few patients swear off caffeine entirely and watch their atrial fibrillation disappear completely (including from chocolate). The population data and the person in front of you are two different clinical situations. Genetics drive how each of us metabolizes caffeine and some people are simply wired to respond to caffeine more. If caffeine reliably triggers your atrial fibrillation, you should continue to avoid it.

While caffeine looks protective against atrial fibrillation, the guideline notes it can actually increase premature ventricular contractions, or PVCs, which are extra skipped or thumping beats. Same molecule, two different rhythms, opposite directions.

What actually leads people to my care

Can you drink too much coffee? When caffeine does cause a cardiac catastrophe, it almost shows up in overcaffeinated retirees or Cardiologists with stained teeth. What we see in emergency rooms are the stories of overindulgence and this rarely includes an overused French press. It’s energy drinks, often in high volume, frequently potentiated by alcohol and sometimes stacked on top of illicit stimulants. High doses and aggressive additives in hard-working individuals trying to extend their days, staying awake, working all-night long. The AHA statement is clear that energy drink data are limited but point toward harm, with case reports of dangerous ventricular arrhythmias and even sudden death at extreme doses. These are different exposures from your morning coffee.

Two things about the cup itself

The filter matters more than you would think. Unfiltered coffee, meaning French press, Turkish, moka pot, and yes, espresso, carries a compound called cafestol that nudges your LDL cholesterol up. A paper filter traps it. Instant coffee does too. Same beans, different cholesterol effect. If your numbers are borderline, switching to a paper filtered drip is one of the easiest wins in this whole conversation.

Black coffee is what the evidence supports. The moment you turn it into a meal or dessert drink, with syrups, whipped cream and a few hundred sugar calories, you’re drinking a different product and the AHA says so plainly. The benefit doesn’t include the sugar.

The prescription

So, what is the prescription worth putting on your fridge? Keep it moderate, up to three to five cups a day. This means real eight ounce cups, not the trough sized ones. Drink it black. Use a paper filter if your cholesterol needs the help. Skip the energy drinks. If you’re one of the genuinely caffeine sensitive folks, the coffee lover in me is sorry for you. The guidelines describe the average response and you don’t get a hall pass as the exception.

Your morning coffee, it turns out, was never the bad guy. It’s what we’ve been doing to coffee and what we’ve been reaching for instead of it, that earns the worry.

If this reframed how you think about your coffee, not as a vice to ration but as a habit worth getting right, send it to one person who’s been told to give up their morning cup. Let’s replace the fear with what the evidence actually says.

References

Marcus GM, Hu FB, van Dam RM, et al. Caffeine and Cardiovascular Disease: A Scientific Statement From the American Heart Association. Circulation. 2026. https://www.ahajournals.org/doi/10.1161/CIR.0000000000001454

Medical disclaimer

This is general education, not medical advice. If you have atrial fibrillation or a heart condition, talk with your own clinician about what’s right for you, especially if you’re one of the people whose rhythm reacts to caffeine.

A Diet for People Who Don’t Do Diets

Jake Kelly MD — Wed, 15 Jul 2026 12:03:56 GMT

A 1990s cardiology trial I recently reread rearranged how I think about food in the exam room.

In the Lyon Diet Heart Study, researchers enrolled people who had already survived a heart attack, who are the highest-risk patients we have, and randomized them to either their usual post-MI care or a Mediterranean-style way of eating. The diet group went on to have somewhere between half and two-thirds fewer cardiac events over the next four years. For the hardest endpoint, cardiac death and/or another heart attack, the reduction was around 70%: 14 events in the diet group versus 44 in the usual-care group. These are numbers we usually associate with a very good drug.

In secondary prevention studies, statins lower major vascular events by roughly a quarter to a third, whereas the Lyon’s diet arm posted something closer to two-thirds reduction in event rates. This is not a head-to-head comparison and I am not telling anyone to trade a statin for a salad, but it should reset how you think about eating for health.

The Mediterranean style of eating (treatment group) group’s LDL cholesterol looked essentially the same as the control group’s. The number I as a Cardiologist counsel on, screen for, and medicate toward barely moved. And yet the patients kept dying less.

So if the benefit didn’t come from the number we obsess over, where did it come from?

The answer to that question is the core of today’s article. It’s the reason I stopped handing patients a list of nutrients and starting writing them a food prescription instead.

We keep losing the war on single nutrients

Modern nutrition advice has spent decades hunting for villains. First it was dietary cholesterol, so we feared eggs. Then it was fat of any kind, so the 1990s gave us the fat-free cookies, which were, of course, packed with sugar. Then saturated fat took the throne. Then sugar. Each villian was oversold as the cause and each one was silently walked back once the evidence (lack of) caught up.

Notice what happened every time: we isolated a nutrient, demonized or sanctified it and the health of the population remain unchanged (debatably sicker).

The clinical trials that actually moved outcomes did something different. They changed the whole pattern of eating. Lyon didn’t lower a nutrient; it swapped a way of eating and cut events without touching lipids. PREDIMED (the largest Mediterranean-diet trial we have, with roughly 7,400 high-risk adults) found about a 30% reduction in major cardiovascular events. And critically, that protective diet was relatively high in fat, around 40% of calories, mostly from olive oil and nuts. This is the opposite of the low-fat gospel many of us were trained on.

Skeptics will note that PREDIMED was retracted and re-analyzed in 2018 after irregularities in how some sites randomized patients. It’s worth acknowledging. When the investigators corrected for these irregularities and reran the numbers, the conclusion remain unchanged: the Mediterranean diet groups still had fewer cardiovascular events. Nutrition trials are hard, messy and rarely as clean as drug studies. But when independent trials, decades apart, on different continents, keep pointing in the same direction, I start to agree with the compass.

The guidelines have finally caught up to the trials. The American Heart Association’s 2026 dietary guidance opens by reinforcing “the importance of focusing on heart-healthy dietary patterns, rather than on single foods or nutrients.” Patterns. Not nutrients.

The patient-facing translation is simpler than any of this: stop grading individual foods as good or bad. We’re building a plate, not auditing a molecule. It’s the logic behind the cardiovascular triangle, too: pull a single nutrient and you tug one corner; change the whole pattern and you move all three legs at once.

Okay, but what about macros?

Whenever I say “patterns, not nutrients,” a sharp patient (or a sharper colleague) pushes back: So macronutrients don’t matter?

They matter. Just not the way the internet thinks. Macros matter as quality, far more than quantity.

Take fat. Lyon didn’t cut fat; it upgraded it. The intervention was richer in alpha-linolenic acid, an omega-3 fat, and it rebalanced the ratio of omega-6 to omega-3 in a way that’s plausibly anti-inflammatory. The lesson isn’t “eat less fat.” It’s eat better fat. It’s the kind of fat (or macronutrient), not the gram count.

Carbohydrate is the same story wearing different clothes. The fault line isn’t “carbs are bad”; it’s refined versus intact. An apple and a glass of apple juice are the same sugar on paper and completely different events in your bloodstream, because fiber and food structure change how fast and how far your metabolism swings. Whole, structured carbohydrate behaves; stripped, refined carbohydrate doesn’t.

Protein? Source beats grams. Lentils, fish, and nuts pull in a different direction than processed and red meat, even at identical protein counts.

So here’s how I explain it to my patients: macros are a lens, not a target. Nobody sits down to eat 30% of their calories from fat. You eat olive oil, fish and a handful of walnuts, and the macros quietly take care of themselves.

How I actually write the prescription

This is where clinic diverges from the pamphlet. The AHA gives us nine sensible features: more vegetables and fruit, whole grains over refined, unsaturated over saturated fat, minimally processed over ultra-processed, less added sugar and sodium. All true. But you cannot hand a real person, on a real budget, with a family and twenty minutes to cook, a nine-point checklist and call it care.

So I build the prescription like this.

Start from what they already eat and love. Culture and habit are assets, not obstacles to bulldoze. If someone’s grandmother’s cooking anchors their week, we build from it, not against it.

A few high-leverage swaps beat thirty rules. Change the cooking fat to olive oil. Move red or processed meat to fish or beans a couple of nights (I’m not going to give up my hamburger or steak…I just eat them less frequently). Trade the refined grain for its intact version. Make the default snack nuts and fruit. Four durable swaps outperform a diet plan nobody remembers by Thursday.

Add before you subtract. The fastest way to make eating well feel like punishment is to lead with a list of bans. So I lead with abundance: more vegetables, more legumes, more fruit, letting the good food crowd out the rest. Generosity is more sustainable than restriction and it happens to work.

If it helps, I’ll literally write it down like a script: Olive oil as your kitchen default. Fish twice a week. Beans on the plate most days. Fruit or nuts when you snack. A vegetable at every dinner. Five lines. That’s a prescription a person can actually fill.

Take a patient I’ll call Rob. He told me flat out that he didn’t do diets. He’d tried, they never stuck, and he wasn’t about to start weighing his food. Fair enough. I don’t really do diets either. So we didn’t diet. We just changed the kind of short list above: olive oil in the pan, beans a few nights a week, almonds and an apple instead of the afternoon chips, a vegetable at every dinner. No macros, no app, nothing to memorize. Six months later his triglycerides had come down and, in his words, he “didn’t feel like he was on anything.” That’s the whole trick: the plan he’ll still be eating next year beats the perfect one he quits by Friday.

The part that makes it stick: culture

Here’s what took me too long to appreciate. The Mediterranean diet isn’t protective because it’s Italian. The same backbone shows up in the Okinawan diet, in the Nordic diet, in traditional plant-forward eating patterns all over the world: mostly plants, whole and minimally processed foods, healthy fats and very little of the ultra-processed stuff. Different cuisines, same skeleton, same outcomes.

The AHA’s 2026 guidance says this out loud: eat this way “regardless of where food is procured, prepared or consumed.” In plain English, this pattern is portable. It adapts to any kitchen.

Which means the message to patients is not you have to eat like a Sicilian. It’s make your own food culture heart-healthy. Tacos, curry, jollof rice, congee: every one of them can carry the pattern. Beans, vegetables, good oils, whole grains, fish and herbs are not the property of one coastline.

And this is where the fun comes back in, because food was never only fuel. It’s pleasure, it’s the table, it’s the people around it. Adherence, the thing every diet study lives and dies on, comes from joy, not deprivation. The best diet is not the theoretically optimal one. It’s the healthful one you’ll still be eating next year because you actually like it.

Start With One Thing

The diet worked even though the cholesterol number didn’t move, because we were never really treating a number. We were treating a way of eating, a pattern, a life lived around a certain kind of plate.

So what do you actually do tomorrow morning? Not a cleanse, not a 30-day reset, not a fridge purge you’ll resent by the weekend. Pick one swap and let it go automatic before you reach for the next. If you cook, make olive oil your default fat this week and stop thinking about it. If you snack, put fruit and nuts where the chips used to live. If dinner looks bare, add a single vegetable, any vegetable, cooked any way you’ll actually eat it. One change, held long enough that it stops feeling like a change. Next month, add the second, and build each new swap out of food you already love rather than food a stranger online told you to fear.

That is the whole method. It is unglamorous, it is slow and it is the only version that still works a year from now.

And if you’re the one writing the prescription rather than filling it, the same rule holds: one durable swap a patient actually keeps beats a perfect plan they abandon by the next visit.

So tell me what you’re starting with. Reply with the one swap you’re making this week, I read every response, and take this with you, because I’ll come back to it on Friday: prescribe the pattern, honor the culture, make it delicious, and the biochemistry follows.

If this reframed how you think about food, not as a strict spreadsheet of molecules, but as a pattern of living, send it to one person who’s tired of the constant cycle of restrictive diets. Let’s make eating well feel like abundance again.

References

de Lorgeril M, Salen P, Martin JL, et al. Mediterranean Diet, Traditional Risk Factors, and the Rate of Cardiovascular Complications After Myocardial Infarction: Final Report of the Lyon Diet Heart Study. Circulation. 1999;99(6):779–785.
de Lorgeril M, Renaud S, Mamelle N, et al. Mediterranean alpha-linolenic acid-rich diet in secondary prevention of coronary heart disease. Lancet. 1994;343(8911):1454–1459.
American College of Cardiology. Lyon Heart Study (trial summary). https://www.acc.org/Latest-in-Cardiology/Clinical-Trials/2010/02/23/19/10/Lyon-Heart-Study
Estruch R, Ros E, Salas-Salvadó J, et al. Primary Prevention of Cardiovascular Disease with a Mediterranean Diet Supplemented with Extra-Virgin Olive Oil or Nuts (PREDIMED). New England Journal of Medicine. 2018;378:e34.
Harvard T.H. Chan School of Public Health, The Nutrition Source. PREDIMED Study Retraction and Republication. 2018. https://nutritionsource.hsph.harvard.edu/2018/06/22/predimed-retraction-republication/
Lichtenstein AH, Khera A, Anderson CAM, Appel LJ, et al. 2026 Dietary Guidance to Improve Cardiovascular Health: A Scientific Statement From the American Heart Association. Circulation. 2026.
Lichtenstein AH, Appel LJ, Vadiveloo M, et al. 2021 Dietary Guidance to Improve Cardiovascular Health: A Scientific Statement From the American Heart Association. Circulation. 2021;144:e472–e487.
Cholesterol Treatment Trialists’ (CTT) Collaboration. Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. Lancet. 2010;376(9753):1670–1681. (Statin scale for the comparison above: ~22% reduction in major vascular events per 1 mmol/L LDL lowering.)

Medical disclaimer

What a Year of Writing Taught Me About Health

Jake Kelly MD — Wed, 08 Jul 2026 14:38:49 GMT

Today I turn 46.

It’s not that I doubted I’d get here. It’s that it’s hard to believe how quickly here arrived. Not long ago I was walking across a stage… high school, then college, then medical school, residency, fellowship. Then my first year in practice. Then 40. And now, somehow, 46.

A birthday is a threshold. You cross it in a single day.

But health doesn’t work that way. Health is a trajectory. A slope you’ve been walking for years, mostly without noticing. The number on the cake tells you almost nothing. The direction you’re heading tells you everything.

This is the lens I’ve spent this year writing through.

One year ago, on my birthday, I published my first newsletter. No followers, no readers. I reached out to friends and family and shared it across my feeds (I didn’t even know you called them socials then). The writing itself started a year earlier. My mother (you all know her now) always encouraged me to journal and I’d found a handful of half-finished ones over the years. Slowly, I read about how to journal. Write. Think. Reflect. I wanted my thoughts on paper. I needed to. I released the fears and the negative energy, and positive, creative thoughts emerged. I wrote to understand what I actually believed and to become a clearer thinker. Maybe a better human.

What I didn’t expect was that it would give me something with legs.

Along the way I think I became a better version of most of the things I care about being. And continuing to become. A better dad, a better husband, a better doctor. And, to my surprise, a better writer.

Here’s what a year of writing taught me about health: the work is in the trajectory, not the milestone. In medicine, we wait for thresholds: a diagnosis, a number that finally crosses a line, a birthday that feels significant. But the heart doesn’t announce itself at thresholds. It changes gradually, for decades before the first symptom. The people who do well aren’t the ones who react at the line. They’re the ones who measured the slope early and changed it while it was still easy to change course.

If you take one thing from my writing into your next year, take this: measure what matters to you and your health and measure it before you have to.

Now, the numbers, because you’re a physiology audience and you like evidence.

This year the newsletter grew to 2,571 subscribers, across 68 articles, read in 49 states and 78 countries. When I started (and even now) I would have found any of those numbers hard to imagine.

But I have a confession. I once had readers in all 50 states. Somewhere this year, I lost West Virginia.

So, if you know anyone in West Virginia, send them my way. Beautiful state. I’d like readers from all 50 states back.

More than the numbers, though, are the relationships. Patients. Doctors. Mentors. Thinkers. Writers. New friends. This year taught me that an audience isn’t a metric; it’s a set of people you’re genuinely learning from.

So what’s next? What am I going to do this upcoming year?

I’m going to keep writing about once a week, going a little deeper. I’m building out my website jakekellymd.com, which will archive and organize (along with Substack) all my previous articles. I’m collaborating with other writers I admire, and I’m exploring a few exciting projects I look forward to sharing over the coming year.

And I’m choosing the endurance challenge of my late 40s. I’m close to figuring out what I want to do to challenge me mentally and physically and show my kids (and myself) that I can still do something daunting. My goal is simple, and it’s the whole point of everything I write about: build physiologic reserve. Raise capacity. Prove that the slope can bend the right way in your 40s, not just hold flat.

Thank you. All of you. Each one. For reading, for writing back, for commenting, for believing in this. For believing in me.

I believe in you too. And I believe that together we can raise the health and fitness of our families, our communities and our country as long as we keep each other motivated. Walk a little more. Sing a little louder. Be a little kinder.

On my birthday, I want all of you to do something for me. Take that step outside and go for a walk. Ideally, invite a family member or a friend. Even a stranger. Every step counts.

And somebody find me a reader in West Virginia.

One request before you go: drop a comment and tell me where you’re reading from. State, country, town. Let’s see how many of the 49 we can light up in the thread and let’s find that 50th.

Your Summer Is a VO₂ Max Opportunity

Jake Kelly MD — Thu, 02 Jul 2026 16:53:49 GMT

The mask seals over my nose and mouth and for the next several minutes it counts every breath I take.

I’m on a bike trainer in my basement, the Alaska light already long at six in the morning, watching a number climb while my legs and lungs argue with each other about who will be the limiting factor in continuing. The mask measures, breath by breath, how much oxygen my body is actually using. Near the end, when the effort stops being negotiable, the number tops out. This ceiling is my VO₂ max.

I do this type of workout purpose and not because I enjoy the last two minutes. No one enjoys the last two minutes. I do it because it is one of. the most useful numbers I know about my own physiology and because I spend my days asking patients to watch theirs. I had better know mine.

Here is what makes it worth the discomfort.

There is a vital sign that predicts your risk of dying more accurately than your cholesterol, your blood pressure, or your blood sugar. Almost no one has it measured or is currently tracking it.

Many of us track the others numbers obsessively: LDL to the single digit, blood pressure on my home machines, hemoglobin A1c to a tenth of a point every 3 months. But the number that, in the largest datasets we have, separates the longest-lived from the shortest-lived more powerfully than any of them, we leave blank.

This number is VO₂ max: the most oxygen your body can take in, deliver and use. It is the ceiling on your aerobic engine: a measure of how much physiologic work you can do and how much reserve you have when life, illness or age demands more than the ordinary.

I’m making this argument now, at the top of summer, on purpose. Summer is when capacity is easy to build. Winter is when our aerobic fitness slowly erodes. So this isn’t a “get more exercise this summer” essay. It’s an argument to bank physiologic reserve while it’s cheap and carry it into the dark months instead of surrendering it.

Find your number. Then begin to understand what it means, how it works and how to raise it.

The vital sign almost no one has measured

In 2018, the Cleveland Clinic published one of the largest studies of fitness and mortality ever assembled: 122,007 patients who underwent treadmill testing were followed a median of 8+ years.

The findings were striking. Comparing the least-fit to the elite, the adjusted risk of death was more than fivefold higher: a hazard ratio of 5.04. In the same analysis, coronary artery disease carried a hazard ratio of about 1.29 while smoking was 1.41.

Read that again. Low cardiorespiratory fitness was associated with a greater adjusted risk of death than coronary disease, smoking or diabetes. And there was no ceiling… the benefit kept accruing all the way up.

This isn’t one outlier study and the relationship is quantifiable. A 2009 JAMA meta-analysis put a number on it: each one-MET increment in fitness, about 3.5 mL/kg/min, the difference a few months of exercise training can produce led to roughly 13% lower all-cause mortality and 15% lower cardiovascular risk. Fitness isn’t pass or fail. It’s a dose-response curve and every step up it pays more dividends.

The pattern is old. In 1989, Steven Blair followed more than 13,000 adults and found the least-fit died at roughly three times the rate of the most-fit. In 2016, the American Heart Association made it official, calling cardiorespiratory fitness as strong a predictor of mortality as the risk factors we screen for routinely and recommended it be measured or estimated every year.

A clinical vital sign. Recommended annually. Measured almost never.

One honest caveat, since physicians are reading: this is observational data. No one has randomized people to a high VO₂ max and followed them along to accrue outcomes (CV death, stroke, heart failure and overall mortality). But the association is large, consistent across cohorts, dose-dependent and as we’re about to see exactly, what you’d expect from a high-capacity cardiovascular system. It clears a bar few risk markers ever reach.

One engine, three payoffs

What makes VO₂ max more than a longevity statistic is that it isn’t mysterious. Physiologically it reduces to one equation: the Fick principle:

VO₂ max = maximum cardiac output × maximum arteriovenous oxygen difference.

How much blood your heart can pump, times how much oxygen your muscles can pull from it. A central pump and a peripheral extraction system. Train either side and the number rises, which is why training the engine moves all three vertices of the Triangle at once.

Circulation. Aerobic training lowers blood pressure, improves endothelial function, and shifts the lipid picture: HDL up, triglycerides down. A heart that pumps more blood at lower pressure does less damage to the arterial wall with every beat.

Metabolism. Few interventions improve insulin sensitivity as reliably. Trained muscle is a glucose sink: it clears sugar with and without insulin, lowers fasting glucose and improves the triglyceride-to-HDL ratio that tracks insulin resistance.

Cardiorespiratory fitness. The number itself, the reserve, is the ceiling on what your physiology can deliver under load.

One engine, three payoffs. That’s what makes raising VO₂ max the master lever. You aren’t choosing between lower blood pressure, better glucose and more fitness.

Know your number and whether it’s any good

You can’t bank what you can’t see. So get a number, even a rough one.

Estimate it this week, no lab required. Several validated field tests will put you in the right range. Pick one that fits your fitness and your appetite for suffering:

Cooper 12-minute run: cover as much ground as you can in 12 minutes on a track; the classic estimate.
1.5-mile run: the military standard; time yourself over 1.5 miles and convert with a standard table.
Multistage shuttle, or “beep test” 20-meter shuttles that speed up until you can’t hold the pace; the level you reach maps to VO₂ max.
Bruce protocol: the staged treadmill test that gets steeper and faster every three minutes; time to exhaustion estimates VO₂ max (technically peak) and it’s the same protocol many clinics run.

Most modern watches also estimate VO₂ max from heart rate and pace. All of these are estimates with the absolute value frequently off by few points (and sometimes several), but the direction over months is what matters clinically.

Get the real thing: a CPET. The gold standard is cardiopulmonary exercise testing, which is a graded treadmill or bike test with a mask that measures the oxygen you actually use, returning a true VO₂ max and your ventilatory thresholds. In most cities, you can get them at the local university exercise physiology or human performance lab. There are various centers popping up across the US so more options are becoming available. Ask for a “VO₂ max test” and whether they report ventilatory thresholds, since those are what you’ll train against. The test is worth doing once to anchor the estimates you track all year.

Now the question everyone actually asks: is my number good? From the FRIEND registry, the largest database of directly measured fitness in the U.S.

Two anchors. At the top, there’s no meaningful ceiling. Every step above the median buys risk reduction. At the bottom, there’s a floor: independent living in later life takes roughly 18 mL/kg/min for men and 15 for women. Below it, ordinary daily tasks consume your entire aerobic capacity. Most people never think about that floor until they’re near it. The point of building reserve now is to stay far above it for as long as possible.

And the key interpretive point: trajectory beats any single reading. A median number climbing is worth more than a good number sliding down.

What’s actually happening when the number moves

It’s worth knowing why VO₂ max is trainable, because it changes how you train.

The first thing that changes isn’t your heart. It’s your blood. Within days of beginning an exercise program, plasma volume expands; over weeks, red-cell and total hemoglobin mass rise 5 to 8%. How much does that matter? In one elegant experiment, people who had raised their VO₂ peak about 9% in six weeks had just enough blood drawn off to return red-cell volume to baseline. Their entire gains vanished.

Then the heart adapts: sustained aerobic work raises stroke volume, which is the blood pumped per beat by 10 to 15%. Over months the ventricles grows more compliant. And the muscle remodels with new capillaries, more mitochondria, higher oxidative-enzyme activity extracting more oxygen from each liter of blood.

Blood, heart, muscle. That’s the whole story and it’s why the number responds to exercise training at almost any age.

It also explains why gains come fast, then wait for you. VO₂ max adapts exponentially, with a half-time of about eleven days. Push a new load and most of the response to that load arrives within three weeks, then plateaus. Over a year, roughly half the total gain lands in the first three months.

The plateau isn’t a ceiling; it’s equilibrium at a given stimulus. Raise the stimulus and a new cycle begins. Which brings us to the part everyone wants.

How to raise the roof

Here’s the reassuring truth, up front: you can raise your VO₂ max substantially without any hard or specific workouts. Easy exercise volume and a couple of strength sessions move the number for most people. The hard session is an accelerant, not a prerequisite.

Build in this order.

Foundation: move all day. The largest and most underrated lever isn’t your workout. It’s everything around it: non-exercise activity thermogenesis, or NEAT. Walking, standing, carrying, gardening, general restlessness. NEAT adds up to more energy turnover and better glucose handling than any single training block. Be a generally active person first.

Cornerstone: Zone 1/2. Most of your aerobic volume should be easy: conversational, nose-breathing pace. This is where the base is built and now you know what “built” means: capillaries, mitochondria, plasma, hemoglobin, stroke volume. It feels too easy to be working. It’s doing the most.

Strength is not separate. Resistance training raises fitness too. In adults over 60, strength programs improved VO₂ peak by about 1.9 mL/kg/min. And the old worry that lifting blunts aerobic gains doesn’t hold: across dozens of trials, adding strength to endurance training doesn’t compromise VO₂ max, while adding muscle, bone and metabolic function. Muscle is a longevity organ. We lose muscle tissue with age and but it is the this very muscle that keeps us independent. Two sessions a week protect the engine and the chassis.

One hard session to consider after you have a base. A single weekly high-intensity session accelerates gains: VO₂ intervals, or for the well-conditioned, non-maximal strides. If you’re newer, you don’t need sprints and you shouldn’t be at 100%. Build the base first; add the hard session as the accelerant, not before.

When you’re ready, the best-studied prescription is the 4×4: four four-minute efforts at 90–95% of max heart rate, three easy minutes between. In a work-matched trial, those intervals raised VO₂ max more than the same time spent easy and continuous. The mechanism is specific: it’s the accumulated time spent near VO₂ max that overloads the heart and drives stroke volume up. That’s also why all-out 30-second sprints with long rests do surprisingly little for the already-fit: the effort ends before oxygen uptake ever reaches the ceiling. Intensity with duration, not intensity alone.

Personally, I keep it short. I like 6 to 8 rounds of 30 seconds on, 30 seconds easy at about 50%. The brief recovery keeps oxygen uptake high across the whole set. Build over a few weeks to two sets and potentially three sets for the overly ambitious or well-trained. When an athlete wants more, we mix efforts of one to four minutes with varying recovery until the hard work totals 12 to 16 minutes. The specific protocol matters less than the principle: accumulate time at a high percentage of VO₂ max, then repeat it, week after week.

The program is a menu, not a mandate

Pick the tier that fits your life. Any of them raises the number.

~90 minutes/week (the floor that still works): two to three easy Zone 2 sessions of 20–30 minutes; two short strength sessions; NEAT every day.
~3 hours/week (the sweet spot for most): three to four Zone 2 sessions of 30–45 minutes; two full strength sessions; add one VO₂ interval session only once Zone 2 feels easy; NEAT as the background.
~6 hours/week (for the committed): four to five Zone 2 sessions of 45–60 minutes; two to three strength sessions; one dedicated VO₂ interval session; high daily NEAT.

Nobody is actually a non-responder

For years, researchers described “non-responders” whose VO₂ max wouldn’t budge. A 2017 study reframed it: non-response ran 69% at one session a week and 0% at four to five. Given two extra weekly sessions, every apparent non-responder responded. It wasn’t a fixed trait; it was under-dosing. (Interestingly, each person’s gain again tracked their hemoglobin mass.)

There’s a real genetic component to how much you improve: the HERITAGE Family Study put the heritability of trainability near 50%. But that governs the magnitude, not whether you respond. Some people’s aerobic fitness climbs faster than others; with adequate dose, essentially everyone increases their aerobic fitness. If your number hasn’t moved, the honest first question isn’t “am I a non-responder?” It’s “was the dose enough?”

The summer play

I write from Alaska, where the seasonal swing is extreme. But you don’t need the far north for this. Wherever you are, summer changes the shape of your day. More light, more energy, and, with kids, often more chaos. School’s out, vacations stack up and the season that should be easy for fitness is exactly when routines fall apart.

So the goal isn’t to train harder this summer. It’s to use the season to set a rhythm that outlasts it.

Protect three mornings. The single highest-yield move and summer is when it’s easiest to install. Use the early light: pick three mornings. 5 or 6 a.m., whatever fits and treat them like appointments. The rule is forgiving: do something for 30 to 60 minutes. Wake up flat? Start with a power walk; you’ll usually warm into it and lift when you get home if the legs come back. Some mornings it’s an easy run, some it’s the treadmill because you didn’t feel like lifting. The habit is the asset. Three mornings ingrained in July are still there in January, when there’s no light and no motivation to build them from scratch.

Guard the night, too. You can’t protect a 5 a.m. session if the light keeps you up to midnight. This is the half people skip. The long evenings that make morning training possible also wreck sleep. I use an eye mask to simulate darkness and it helps me get to sleep earlier. Aim to be in bed about seven and a half hours before you want to wake. The morning and the bedtime are one system.

On vacation, walk. Vacation is where fitness usually goes to die. Routine gone, gym far away. Reframe it as a walking windfall. At home I run about 25 miles a week and average just under 15,000 steps a day; on vacation, when the running drops off, my wife and I just walk. Morning, midday, on the adventure, after dinner and the count climbs toward 25,000. That’s NEAT, the most underrated lever there is. Hold a few mornings, add some bodyweight work, but mostly: walk. You’ll come home fitter than you left.

Stack the hours with your people. The honest part: three mornings is the floor, not the ceiling. To really move fitness you have to stack minutes into hours and the sustainable way isn’t solo grinding. It’s with the people around you. If you have kids, fold them in. Early it’s almost nothing. A one- or three-mile bike ride while you walk alongside. It adds up though. Do it a few times a week, year after year and capacity compounds. My kids are ten, eight, and five; we ride ten miles a couple times a week now, the youngest tags along for four or five and by summer’s end we push toward fifteen or twenty. I call it secret Zone 1 training, sometimes Zone 0, but hours that never feel like training. No kids? Same move, with friends: the evening walk, the weekend ride.

Keep one foot indoors. A caution learned the hard way: train outside all summer and never touch the gym and going back inside in the fall feels like starting over. Keep one indoor session in the mix, weekly or every other week, so when the light goes, the door’s already open.

That’s the play. More light, more energy, more time with people you love spent building something so important that your later years depend on it. Get outside, get a rhythm into the ground and carry this rhythm into the season when both daylight and motivation run short.

The reserve close

I build in summer because I know what winter does.

Up here, capacity slowly erodes when the light goes. The runs shorten, then stop; the walks move indoors, then shrink. And it’s fast. Stop for more than a few weeks and VO₂ max falls around 9%; in the highly trained, twelve weeks off can cost 20%. You don’t feel it trickle down, any more than you feel your ApoB rise. The loss is invisible until it’s already cost you something.

But here’s the hopeful half: what erodes can be rebuilt. Detrained masters athletes who lose a quarter of their capacity recover nearly all of it within weeks of structured retraining. The engine remembers. Reserve isn’t a one-time deposit; it’s an account you draw down and refill.

The ceiling with age is higher than most assume. Lifelong endurance athletes in their eighties have measured VO₂ max near 38 mL/kg/min: the aerobic capacity of untrained men forty years younger, with muscle oxidative enzymes 40–50% higher than sedentary peers. They aren’t genetic freaks. They never stopped making deposits. Maintaining training roughly halves the rate of decline: about 5% a decade instead of 10%.

That’s the real argument. Not exercise more. Build reserve above the demands of daily life, so when winter, stress, age or illness asks more of your body than usual, there’s something in the tank.

Trajectory over threshold. Capacity above daily demand. Build the reserve before you need it.

The body you build in July is the one that carries you through January.

Three moves this week

Estimate your VO₂ max. Pick one of the field tests above, a Cooper 12-minute or 1.5-mile run is simplest, or read the estimate off your watch. Compare it to the median for your age and sex, and mark the number to retest in a few months.
Anchor the week on the base. Zone 1/2 volume plus daily NEAT plus two strength sessions. Add a single hard session only once the easy work feels easy.
Block three mornings and guard the nights that make them possible. Put the sessions on the calendar now, while the light makes it easy; set a bedtime about seven and a half hours before the alarm; protect both into the dark.

Build the reserve while it’s cheap.

If this reframed how you think about fitness, as a vital sign you can measure and bank, not just a habit you should keep, send it to one person who’s been meaning to get back in shape this summer. The number is easiest to raise right now. Most people don’t know that. Tell them.

Work with me. I practice performance and preventive cardiology at ROOL Health in Anchorage, where circulation, metabolism, and cardiorespiratory fitness are treated as a single system. To learn more, work together, or get in touch, start at jakekellymd.com.

Jake Kelly, MD, MHS, FACC is a board-certified performance and preventive cardiologist, Wasserman-certified in cardiopulmonary exercise testing, ACC Governor for Alaska and founder of ROOL Health, a cardiometabolic longevity practice in Anchorage, AK and virtually in CO, FL, NC, TX.

References

Mandsager K, et al. Association of Cardiorespiratory Fitness With Long-term Mortality Among Adults Undergoing Exercise Treadmill Testing. JAMA Network Open. 2018;1(6):e183605.
Kodama S, et al. Cardiorespiratory Fitness as a Quantitative Predictor of All-Cause Mortality and Cardiovascular Events. JAMA. 2009;301(19):2024–2035.
Blair SN, et al. Physical Fitness and All-Cause Mortality. JAMA. 1989;262(17):2395–2401.
Ross R, et al. Cardiorespiratory Fitness as a Clinical Vital Sign — A Scientific Statement From the American Heart Association. Circulation. 2016;134(24):e653–e699.
Kaminsky LA, Arena R, Myers J. Reference Standards for Cardiorespiratory Fitness (FRIEND Registry). Mayo Clinic Proceedings. 2015;90(11):1515–1523.
Helgerud J, et al. Aerobic High-Intensity Intervals Improve VO₂max More Than Moderate Training. Med Sci Sports Exerc. 2007;39(4):665–671.
Smart TFF, et al. Resistance Exercise Training and Cardiorespiratory Fitness in Healthy Older Adults: A Meta-Analysis. Age and Ageing. 2022;51(6):afac143.
Khalafi M, et al. Concurrent vs. Aerobic or Resistance Training on Cardiorespiratory Fitness and Strength: A Meta-Analysis. Physiology & Behavior. 2022;254:113888.
Montero D, Lundby C. ‘Non-Responders’ Do Respond to Higher Dose of Training. J Physiol. 2017;595(11):3377–3387. (See also Montero et al., J Physiol. 2015 — the hematological basis of the gain.)
Bouchard C, et al. Familial Aggregation of VO₂max Response (HERITAGE Family Study). J Appl Physiol. 1999;87(3):1003–1008.
Scharhag-Rosenberger F, et al. Time Course of Changes in Endurance Capacity: A 1-Yr Training Study. Med Sci Sports Exerc. 2009;41(5):1130–1137.
Trappe S, et al. New Records in Aerobic Power Among Octogenarian Lifelong Endurance Athletes. J Appl Physiol. 2013;114(1):3–10.
Rogers MA, et al. Decline in VO₂max With Aging in Master Athletes and Sedentary Men. J Appl Physiol. 1990;68(5):2195–2199. (~5.5% per decade in athletes vs. ~12% in sedentary men.)
Physical Activity Guidelines Advisory Committee Scientific Report; and ACSM/AHA consensus statements on the effects of aerobic and resistance exercise on blood pressure, insulin sensitivity, glycemic control, and lipids.

Medical disclaimer

The Heart Attack You Can Now Prevent, Before You Ever Have One

Jake Kelly MD — Fri, 26 Jun 2026 11:03:58 GMT

Two patients. I’ll tell their clinical stories to demonstrate how the practice of Cardiology is changing. I’ll call them Alan and Alana.

Alan is 63. I met him the way cardiologists have met most of their patients for a generation: after. For two or three weeks a pressure had been settling into his chest when he climbed stairs. One morning the pressure started while walking up a single flight of stairs to his office and wouldn’t ease up; a coworker took one look and told him to call 911. By the time I saw him in the ER, his troponin was in the thousands and his ECG showed the deep precordial T-wave inversions we call a Wellens’ warning, the pattern that suggests the proximal left anterior descending artery is about to close. The next morning his coronary angiogram showed a 99% narrowing in the proximal LAD. We opened it, post-dilated the stent to 4.5 millimeters and got him home.

Alan had not ignored his health. For years a doctor had told him his blood pressure ran high and his cholesterol was “up there, but not dangerous.” He’d buckle down, walk his BP and cholesterol back toward normal for a few months, then he would lose focus and his levels would drift upwards. The numbers drawn the week before his heart attack told a more accurate story: total cholesterol 244, HDL 36, LDL 170, triglycerides 190. No one had ever checked his Lp(a). The old model didn’t require it. The old model required an event. And he finally had one.

Alana is 57, and she came to me before anything happened, because she wanted to prevent an event. She’d had kids in her late thirties and was determined to be there when they had their own children. Her cholesterol panel was the kind every doctor has waved through: total cholesterol 205, HDL 51, LDL 132. “Borderline, work on your diet.” No symptoms. A stress test would have been clean. On paper, a healthy woman doing fine.

Except I measured the numbers nobody had ordered. Her apoB, the actual count of the particles that drive plaque, was 153. Her Lp(a), the inherited risk almost no one checks, came back at 210. And the family history she’d been told didn’t matter turned out to rhyme: an aunt who needed a stent at 57, the age Alana is now; a grandmother bypassed in her seventies. We started, as I always do, with the things that don’t need a prescription. And they worked. More fiber, psyllium husk, learn about macronutrients, increase physical activity, and her LDL fell into the low 120s, her apoB toward the high 130s. Better. Not optimal. So she decided to obtain a coronary calcium score after a shared decision discussion. It came back 74, with plaque in two arteries. No symptoms. The disease had already started.

One patient is how we’ve always treated heart disease. The other is how we can finally get ahead of it. The difference between them is how we can chose to practice Cardiology today.

A few months ago, I wrote about exactly this woman, in the article Before the Heart Attack: What Actually Shows Up First. The argument was simple and, for a lot of readers, unsettling: the disease shows up long before the symptom. The plaque, the particles, the silent drift. They arrive first. The chest pain arrives last.

That piece named the problem. It left a question open, and a lot of you wrote to ask it directly.

If the disease shows up first, what do we actually do about it before the event?

This is the answer. And for the first time in my career, the honest answer has changed.

The old contract

For a generation, cardiology organized itself around the event.

The heart attack was the entry point. It was the moment a patient became, officially, a heart patient, the moment the system mobilized. Statin, stent, secondary prevention. We became extraordinarily good at the rescue. We built cath labs that can open a closed artery in under ninety minutes. We saved lives that would have been lost a few decades earlier.

But look closely at what we were rewarding.

We were waiting. We were treating the heart attack as the diagnosis instead of the failure. “Your numbers are normal” so often meant “we don’t have enough evidence of damage yet to act,” which is a different sentence than it sounds like. It means: come back when it’s worse.

Alan is the model working exactly as designed. He survived, the machinery did its job, and he is thriving today. But for the person whose first heart attack is fatal, and for too many people the first one is the last one, that contract never pays out. You can’t run secondary prevention on someone who didn’t survive the primary event.

The old model had a blind spot, and the blind spot was Alana. An apoB of 153. An Lp(a) of 210. A calcium score of 74. Real disease, already underway, years before any symptom. And under the old rules, no permission to treat it.

What changed this year

This year, a large clinical trial closed that gap.

VESALIUS-CV enrolled 12,257 high-risk patients, people with atherosclerosis or high-risk diabetes, who had never had a heart attack or a stroke. Their LDL was still elevated on a statin. They were randomized to add a PCSK9 inhibitor, evolocumab, or a placebo, and followed for a median of about 4.6 years.

These were not rescue patients. This was the upstream population, the people the old model told to wait. People, in other words, a great deal like Alana.

The results, in plain endpoints:

A 25% reduction in the risk of a first major cardiovascular event. A 36% reduction in the risk of a first heart attack. Revascularizations down. The composite of cardiovascular death, heart attack, and stroke down by 27%. LDL driven down roughly 55%, to a median around 45 mg/dL.

Then the detail that reframes everything.

In the patients without established significant atherosclerosis, the furthest upstream, the ones with the least visible disease, the relative risk reduction was larger, around 31%. The benefit didn’t shrink as you moved earlier in the disease. It grew.

There was also a mortality signal: fewer cardiovascular deaths, fewer deaths overall. Precision matters here. These findings were exploratory: nominal, secondary, not what the trial was powered to prove. It’s the first primary-prevention mortality signal we’ve seen for this class. Notable, and not the same as a confirmed mortality benefit. Both are true. Hold both thoughts together for now.

What the trial proved cleanly is this: in people whose first event hasn’t happened yet, lowering the particle burden further, earlier, prevents events.

The rescue drug works best before there’s anything to rescue.

The reframe: trajectory, not threshold

If you’ve read my writings, you’ve heard me say it before. Risk is not a number you cross. It’s a slope you ride.

Heart attacks are not caused by your LDL this morning. They are caused by cumulative apoB-particle exposure over decades: the total area under the curve, every particle, every year, grinding against the artery wall. A “normal” number held for forty years still deposits a lifetime of cargo. Threshold thinking asks, are you over the line today? Trajectory thinking asks, where is this line headed, and how long have you been on it?

PCSK9 is not a new idea. It is a more powerful lever on the same physiology I’ve been writing about all along.

A statin lowers the slope. A PCSK9 inhibitor lowers it further, and lower, and the VESALIUS-CV signal says that bending the curve sooner changes the destination, not just the readout at the next visit. Lower for longer beats lower for now. That isn’t a marketing line. It’s the whole mechanism of atherosclerosis read backwards.

Alan rode a high slope for decades (high pressure, high particles, an Lp(a) no one ever saw), and we only met him at the moment it crossed the line. Alana’s slope was bending the wrong way too. The difference is that we found hers with the curve still low enough to change.

This is trajectory over threshold, written in pharmacology.

The pill on the horizon

There’s a second reason this conversation arrived all at once.

For the entire history of this drug class, “PCSK9 inhibitor” meant an injection. Effective, but a barrier: for cost, for access, for the simple human friction of a needle. A lot of the physiology never reached the patient because the delivery didn’t.

This barrier is starting to come down. An oral PCSK9 inhibitor, enlicitide, has now posted three positive Phase 3 trials, lowering LDL roughly 55 to 60% as a daily pill. In December 2025 the FDA granted it a priority review voucher. If it clears, it would be the first PCSK9 inhibitor you swallow instead of inject.

A pill changes the arithmetic of prevention. It changes who can realistically stay on therapy for the decades that trajectory actually requires.

But the caveat is important to note for now: what enlicitide has shown so far is LDL lowering, not fewer heart attacks. The outcomes trials aren’t in yet. We have every mechanistic reason to expect events to follow the particles (that link is among the most established in cardiovascular medicine), but expectation is not evidence. For now: a powerful new way to lower LDL/ApoB, with the final proof on the effect still pending.

I’d rather tell you this now than walk it back later.

And it isn’t the only one on the way. Obicetrapib, an oral drug from a different family, a CETP inhibitor, is in late Phase 3, lowering LDL and apoB by roughly a third, raising HDL and dropping Lp(a) along with them. Its cardiovascular outcomes trial reads out in late 2026 or early 2027. The same caveat applies, in capital letters: we have the particle numbers, not yet the event numbers.

The story isn’t enlicitide, or obicetrapib, or evolocumab. It’s that the Circulation leg of the Triangle is gaining more levers on the same target, the apoB-carrying particles, in more forms, by more mechanisms, taken more easily. You don’t have to bet on a brand. The strategy is older and simpler than any molecule: know your number, lower it with whatever tool fits you, and hold your levels down for as long as you can. The drugs are just how we keep getting better at the same job.

The ladder, not the leap

It would be easy to read all of this as “everyone needs the strongest drug.” That’s not the message. The point is that we finally have a full set of tools, and the right move is to match the lever to the person.

Statins are the oldest tool here and the most studied lipid therapy in history: cheap, effective, and a reasonable starting point for many people. But they are not a gate everyone has to pass through. Intolerance is real, mostly at higher doses, where muscle aches and a small, dose-related rise in new-onset diabetes show up. One honest caveat: in blinded trials, much of what patients call “statin symptoms” doesn’t reproduce when they can’t tell the drug from a placebo. Real for some, overstated for many. If a statin works for you, it’s a good option. If it doesn’t, or if your numbers call for more, the rest of the ladder is right there, and nothing says you have to climb it one rung at a time.

That ladder now has real rungs:

Ezetimibe lowers LDL roughly 15 to 20% and has been around for years: cheap, oral, well-tolerated, and an easy add-on. Bempedoic acid lowers it by a similar margin, with its own cardiovascular outcomes data, and is useful precisely for the person who can’t tolerate a full statin dose. Inclisiran is a twice-a-year injection that silences PCSK9 production at the genetic level and cuts LDL by roughly half. And the PCSK9 antibodies (and soon, potentially, the oral) sit at the top, for when you want the largest single lever, whether that’s first or last.

Alan and Alana picked different starting points for the same job. Alan was certain he couldn’t tolerate high-intensity statins but wanted to keep some on board for the benefit beyond LDL, so he takes low-dose rosuvastatin three days a week and adds a PCSK9 injection, and he now sits comfortably under his secondary-prevention goals: an LDL under 55 and an apoB under 60. Alana works the same numbers from the other end: rosuvastatin 10 milligrams and ezetimibe 10, a PCSK9 held in reserve, toward the same pair of targets. Same physiology, two routes, one direction of travel.

The goal was never a particular drug. The goal is the lowest tolerable particle burden, held for as long as possible. There’s more than one way up.

Where the 2026 guideline lands

The science doesn’t change practice until the guidelines move. This year, they moved.

The 2026 dyslipidemia guideline does three things worth knowing.

It puts apoB next to LDL at the center of the targets, both tiered by risk, to be hit together. As risk climbs, the apoB goal tightens from under 90 to under 70 to under 55 mg/dL, alongside the matching LDL goal of under 100, 70, or 55. Not LDL alone. The particle count is the thing that actually drives the disease.

It recommends measuring Lp(a) once in every adult, a Class I recommendation. Lp(a) is the inherited risk almost nobody checks, the one that can drive an early heart attack in someone who “did everything right.” Once in a lifetime. One blood draw. It belongs in everyone’s chart. Alana’s 210 is exactly the kind of number that rewrites a person’s risk the moment you see it, and it had gone unseen for 57 years. Alan’s was never drawn at all. A value like his, left in the dark for six decades, is how a careful person ends up with a Wellens’ warning on an ordinary Tuesday.

And it positions a PCSK9 inhibitor as a tool for patients with elevated Lp(a) and established disease, among others.

For the framework I keep coming back to (the Triangle of Circulation, Metabolism, and Cardiorespiratory Fitness), this is the Circulation vertex getting sharper. Better targets. Better measurement. A stronger lever. Same physiology, finally matched with tools equal to it.

The honest boundaries

A paradigm shift is the moment people overshoot. So let me draw the lines clearly, because the decisions made should be with the patient, not for the patient.

A PCSK9 inhibitor is not a magic exemption, and neither is a statin. The honest boundary was never which drug comes first. It’s that no pill or injection replaces the rest of the Triangle. Which lever you start with, lifestyle measures, a statin, a PCSK9 inhibitor, ezetimibe, a combination, depends on your numbers, your tolerance, your access, and how far you have (want?) to go. If a PCSK9 inhibitor is the right first move for you, that’s a reasonable choice, not a luxury. Lever-first thinking is the trap. Target-first thinking is the real goal.

What no drug does is substitute for the rest of the Triangle, and this is the part I want to say louder, not softer. Any lipid therapy moves one input: circulating particles. It does nothing for the other two legs, and the other two legs are not fixed. They are some of the most modifiable physiology in medicine.

Insulin resistance is not a life sentence. It can be improved, and in many people functionally resolved, through training, nutrition, weight loss and where appropriate the metabolic therapies we now have. Move someone off the path toward type 2 diabetes and you’ve changed their cardiovascular trajectory at least as much as any lipid drug, because you’ve pulled the Metabolism leg back into line.

Cardiorespiratory fitness tells the same story, in the other direction. VO₂ max is among the strongest predictors of how long and how well a person lives, and it responds to physical activity & exercise training at almost any age. You can raise it. You can build strength and physiologic reserve on top of it. No pill does that for you, and nothing you can swallow comes close to what a rising VO₂ max buys.

So hold the whole Triangle in view. The drug lowers the cause in one vertex. The work that improves insulin sensitivity and raises fitness strengthens the other two, and those are still yours to earn. Which is the good news, because they’re the levers most under your control.

Cost and access are real. Even with a pill on the horizon, this is not yet a universal therapy and pretending otherwise helps no one.

Primary-prevention treatment is individualized, not automatic. VESALIUS-CV studied high-risk people, not everyone. A clinical trial result is population level data; your decision is a personal one, made with your own physician, against your own numbers and your own trajectory.

I won’t promise anyone a longer life. I can tell you the first event is now, for many people, a preventable one. That is a large enough claim to stand on without inflating it.

Would I take one myself?

Recently, I wrote about my own decision: my apoB, my trajectory, the conversation I had with myself about lowering my own numbers. Several of you have now asked the obvious follow-up.

Would you ever consider a PCSK9i for yourself?

Here’s how I actually think about it, transparently, without making it a recommendation for you.

I think in trajectory, not threshold. If my apoB is well-controlled on my current approach and my slope is flat, the answer is no. There’s nothing to add. But if my numbers told me the slope was still pointed the wrong way despite doing the work, especially with any inherited risk in the mix, I would not wait for an event to earn the right to act. I’d treat the trajectory.

That’s the entire point. The decision isn’t do I have a problem today. It’s where is this line going, and how long am I willing to ride it before I bend it.

I’d answer that question for myself the same way I’d want every reader to answer it for themselves: with the actual numbers, honestly.

What to actually do now

Three moves. None of them require a prescription to start.

Know your apoB, not just your LDL. It’s a standard, inexpensive blood test that counts the particles driving the disease instead of estimating their cargo. If you’re high-risk, you want both numbers low: an LDL under 55 and an apoB under about 60.

Get your Lp(a) measured once. The guideline now says every adult. It’s a single draw that can explain a risk you’d otherwise never see coming.

And if you’re high-risk and not at your apoB and LDL goals, ask your physician a specific question. Not “are my numbers okay,” but “am I a candidate for lowering them further now, rather than waiting, with whatever therapy fits me best.” The wording changes the discussion. One asks for permission to relax. The other asks to change the slope.

Alan and Alana tell two potential endings of the same story. Alan is the model we built: heroic, expensive and one event too late. He is thriving now, and I am glad of it every time he walks into clinic. Alana is the model we can finally choose instead: treated early, across all three legs of the Triangle, while the disease is still small enough to bend. With any luck, she never becomes Alan. This isn’t a smaller kind of medicine. It’s the better one.

We spent a generation treating the heart attack. The work now is to make sure it never has to happen.

Patient details have been changed to protect privacy; clinical accuracy has been preserved. This article is educational and not a substitute for individualized medical advice.

Share this. If someone you love thinks “normal cholesterol” means they’re safe, or has an apoB or Lp(a) sitting in a chart that nobody ever explained, send them this. The most preventable heart attack is the first one.

References

2026 ACC/AHA/Multisociety Guideline on the Management of Dyslipidemia. JACC / Circulation, 2026.
VESALIUS-CV: Effect of Evolocumab in Patients at High Cardiovascular Risk Without Prior Myocardial Infarction or Stroke. Presented AHA 2025; published New England Journal of Medicine, 2025; featured at ACC 2026 (NCT03872401).
Merck/MSD: CORALreef Lipids and CORALreef HeFH Phase 3 results, enlicitide decanoate (2025 to 2026).
FDA Commissioner’s National Priority Voucher, enlicitide decanoate (December 2025).
2026 Dyslipidemia Guideline-at-a-Glance, JACC, 2026.
ADA 2026: Lipid management and cardiovascular prevention in diabetes.
Family Heart Foundation: ongoing Lp(a) outcomes trials (HORIZON, OCEAN(a)).
TCTMD / ACC coverage of VESALIUS-CV (2025 to 2026).
NewAmsterdam Pharma: BROADWAY Phase 3 results and PREVAIL cardiovascular outcomes trial, obicetrapib (oral CETP inhibitor); outcomes anticipated late 2026 / early 2027 (NCT05202509).

Medical disclaimer

My Dad Is My Hero

Jake Kelly MD — Sun, 21 Jun 2026 14:31:14 GMT

This is a complicated topic for me to write about.

Now that I have three kids of my own and I’m in midlife, I understand something differently. I’ve been a child. A teenager. A young adult. An adult. And now I’m the head of a household that depends on me.

Life is egocentric at first. Your own needs. Your own desires. Then those compete with the needs and desires of your spouse and your children. If you choose a spouse who can stand on her own two feet, as I did, you don’t worry about her. You admire her. You love her. But the weight of providing and protecting still sits on your shoulders.

If you had asked me growing up who my hero was, I might have said a famous left-handed baseball player like Lou Gehrig. I bonded with him through a book because we shared left-handedness and a pursuit of greatness in sport. In high school, I might have said John F. Kennedy. Or Martin Luther King Jr., who walked into the line of fire with grace and courage and taught a nation something about love and strength.

In college, it would have been Lance Armstrong, Mark Allen or Dave Scott: cycling & triathlon legends. After college, maybe Dr. Herbert J. Thomas III, the surgeon who put me back together when I was broken. Or my triathlon coach, Chuckie “V” who taught me how to live simply and chase excellence on a shoestring budget.

But somewhere on the back nine of life, something changed.
My hero is my dad.

My mom was and is my foundation. She is powerful and outspoken and formative in my life. But fathers are often the unsung heroes. Quiet. Steady. Always thinking about their family.

Now I understand what it means to lie awake at night wanting your children safe. Protected. Happy. Given every opportunity.

Looking back, I had all of that. Here’s the complicated part. I remember throwing a football or baseball with my dad once or twice. But not often. I don’t think he was naturally drawn to that. I’ve never asked him why. Maybe it wasn’t modeled for him. Maybe there was pain there. Maybe it simply wasn’t his wiring.

What I do know is this:
He was a star football player in high school.
He earned a full scholarship to Rice University.
Football was a full-time job. It limited what he could explore academically and socially.
People might say, “Full scholarship. What a gift.”

It was.

But late adolescence also requires time. Time to think. Time to wander. Time that looks wasteful on paper but is essential for growing up. I’m not sure he had much of that.

He married my mom and adopted my older brothers. He became their father. He helped them build engines. Fix cars. Move across states. My brothers are thirteen years older than me, so when I came along, I mostly played on the edges of those projects.

My dad is not a natural teacher. But I was always allowed to be involved.
Sometimes required.

And from that I learned patience. Follow-through. The importance of finishing what you start. My dad can fix anything.

I remember learning to ski between my dad’s legs. The cold came up fast off the snow but I felt warm and safe. His skis were on either side of mine, the boundary that made the speed feel safe. I couldn’t see much ahead. I didn’t need to. He could. I taught my kids the same way.

He came to all of my football games. Track meets. Most of my soccer games. He was quiet. Relatively positive. Yes, he yelled when frustrated at home… I recognize that feeling now. When you work hard to keep a house stable and kids create noise and break things and kick soccer balls against the wall.

But there was always food on the table.
A roof overhead.
Camping trips.
Cross-country road trips.
Skiing twice a year.
A love of mountains.
That was love.

When Katie and I moved from Texas to North Carolina, my parents showed up and drove their airstream trailer across the country to help us along our new voyage, camped in our yard so our dogs were walked and our house was ready. When we needed a bigger deck, my dad built one.

When we moved from North Carolina to Alaska, my dad drove 4,350 miles with me in a Subaru Outback and two dogs.

He has always been there.

When I was nearly killed after college and had to relearn how to walk, my parents carried me. Fed me. Endured the stress. My dad physically carried me at times. Then he let me get back on the bike.

It really is all about the bike.

When we moved to Alaska with young children, my mom felt the pull to be near her grandchildren. I’m not sure that was my dad’s dream. But he moved. He supported his wife. He supported his son.

That’s who he is.
He is quiet.
A deep thinker.
A perfectionist with perfect handwriting even now.
An educated, thoughtful American who votes and reads and thinks about others.
He is often overpowered by the energy of my mother and, honestly, of me. But he has always been steady.
He has always shown up.

I think fathers are the unsung heroes in a boy’s life. We cannot live without a mother’s love. But a father’s quiet belief in his son, his steadiness, is a superpower.

My dad and I don’t always have deep emotional conversations. That’s not really how we’re wired.

But I know he loves me.
And now, on the back nine, I know this:

He is my hero.

I only hope that someday, my son feels the same way about me.

What a Calcium Score of Zero Really Means

Jake Kelly MD — Thu, 18 Jun 2026 13:38:51 GMT

There is a test that lets me look directly into your arteries to see whether coronary artery disease has already started.

Not predict it. See it.

It takes ten minutes. No IV. No contrast. No treadmill. You lie still, the scanner makes a few passes across your chest and a number comes back. This number is one of the more honest truths in cardiovascular medicine, because it doesn’t tell you what might happen. It tells you what already has happened.

Last week I shared Justin’s story. Forty-nine years old. He runs a sub-20 minute 5K. Resting heart rate 47. HbA1c of 5.1, fasting insulin of 4.1, blood pressure of 118/72. Two of the three legs of his cardiovascular triangle, metabolism and fitness, were as strong as I see in any patient. But his Lp(a) was 187, his father had bypass surgery at 54 and when I sent him for a coronary calcium scan, the number came back positive.

Calcium. In the coronary arteries of a man who runs a sub-20 5K.

A lot of you finished that article with the same question Justin had when he sat across from me: what does that number actually mean and should I get one?

Let me answer both.

An illustrative calcium result: a score of 49, all of it in the LAD. Any score above zero means atherosclerosis is already present.

What the scan measures, and what it misses

The test is a coronary artery calcium scan. A CAC score. It’s a non-contrast CT of the heart that measures calcified plaque in the walls of your coronary arteries and returns a single number, the Agatston score.

Here is the first thing to understand, because most people get it backwards.

Calcium in a coronary artery is not the disease. It’s the scar left behind by the disease. Atherosclerosis begins as soft, lipid-rich plaque. Over years, as that plaque is injured and stabilized, the body calcifies it, the same way it lays down scar tissue anywhere else. So a CAC score doesn’t measure how much soft, dangerous plaque you have. It measures how much old, healed plaque you have. The presence of calcium is proof that the atherosclerotic process has been running long enough to leave a mark.

This has two consequences and they move in opposite directions.

A positive score is not a heart attack waiting to happen. Calcified plaque is, in many ways, the stable kind. Most heart attacks are triggered by softer, non-calcified plaque that ruptures. So a positive CAC does not mean a blockage, it does not mean you need a stent and it does not mean an event is imminent. It means the process is underway and you now have a reason to take it seriously.

But a zero score is not a free pass either. CAC sees calcium. It does not see soft plaque that hasn’t calcified yet. A young person with aggressive, early, lipid-driven disease can have meaningful soft plaque and a CAC of zero, because the disease hasn’t been present long enough to scar. A 2025 analysis in the Journal of the American College of Cardiology made this concrete: even among people with a calcium score of zero, higher LDL cholesterol still carried higher long-term cardiovascular risk. Zero is reassuring. Zero is not immunity.

Hold onto both of those. The whole framework depends on them.

Calcium forms late. By the time a CAC score turns positive, soft plaque has been accumulating for years, which is why a zero isn’t the same as low risk.

What changed in 2026

For years, the calcium score lived in a strange place. We knew it predicted risk better than almost anything. But the guidelines treated it as a tiebreaker, something to order when you couldn’t decide whether to start a statin.

The 2026 ACC/AHA Dyslipidemia Guideline, the first major update since 2018, gave CAC a real job. It didn’t just say who to scan. It said what the number means for your treatment target.

The logic runs through what the guideline calls the CPR model. Calculate your 10-year risk with the PREVENT equations. Personalize it using factors the equations miss: family history, Lp(a), inflammatory conditions, ancestry. If the decision is still genuinely uncertain, Reclassify with a calcium score. CAC is no longer the first move. It’s the deciding move when the first two leave you on the fence.

Then the guideline did something it had never done before. It tied your LDL target to your calcium.

Any detectable calcium supports an LDL goal under 100. A score of 100 to 299 calls for at least a 50% LDL reduction and a target under 70. A score of 300 or higher pushes the target under 55, the same territory we treat people who’ve already had a heart attack. Above 1,000, under 55 is explicitly on the table.

Read that again, because it’s the part most patients and a surprising number of clinicians don’t yet know. Your calcium score doesn’t just tell you whether to treat. It tells you how aggressively. The same LDL of 110 means one thing in a man with a CAC of zero and something entirely different in a man with a CAC of 350. The number in your arteries sets the number we’re aiming for in your blood.

The 2026 guideline ties your LDL target to your calcium score: the higher the score, the lower the goal.

Who should actually get one

Not everyone needs this scan. The guideline is specific.

The clearest candidates are men 40 and older and women 45 and older sitting at borderline or intermediate 10-year risk, the people for whom the statin decision is a genuine coin flip. That’s where a calcium score earns its keep. It moves you off the fence in one direction or the other.

Some factors lower that threshold. A strong family history of premature heart disease. Elevated Lp(a). South Asian ancestry. Chronic inflammatory conditions like rheumatoid arthritis or lupus. If you carry one of these, the conversation about coronary artery calcium scanning can start earlier.

Some people don’t need it at all. If you already have established cardiovascular disease, you’re getting treated regardless and a calcium scan adds nothing. If your risk is already clearly high or clearly low, the number won’t change what we do. CAC is a tool for resolving uncertainty, not for confirming what’s already obvious.

The practical details matter, because they’re often what stops people. In most markets a calcium score is $100 to $400 cash, frequently without needing insurance to weigh in. There’s no IV and no contrast dye. The radiation is roughly that of a mammogram. You’re in and out in about ten minutes.

The calcium you didn’t go looking for

Here’s a newer wrinkle the 2026 guideline addressed directly and it’s one of the most practical things in it.

Every year, millions of people get CT scans of the chest for reasons that have nothing to do with their heart: lung cancer screening, a persistent cough, trauma, a pre-op workup. Those scans almost always capture the coronary arteries too. For years, the calcium sitting right there in the images was routinely ignored, because the scan wasn’t ordered to look for it.

The guideline now says to stop ignoring and commenting on CT scans that have already been performed.

If a chest CT shows coronary calcium, that finding should be reported, graded simply as none, mild, moderate or severe. Moderate or severe incidental calcium corresponds to a formal CAC score of 100 or more, which is itself a guideline-level reason to sit down and talk about statin or other cholesterol lowering therapy. You don’t need a second, dedicated scan to act on calcium that has already been seen and documented.

One caveat is important to note: the absence of calcium on a non-gated chest CT is not the same as a true zero on a dedicated, gated calcium scan. These scans aren’t optimized to rule calcium out. So incidental calcium that’s present is meaningful and actionable. Incidental calcium that’s absent doesn’t earn you the warranty I’m about to describe.

The power and the limits, of zero

A calcium score of zero is one of the strongest negative findings in cardiovascular imaging.

In a person at borderline or intermediate risk with no major risk enhancers, a CAC of zero is good evidence that near-term risk is low, low enough, in carefully selected patients, to justify holding off on a statin and leaning hard into lifestyle instead. Clinicians call this the warranty period. Data from MESA, the large multi-ethnic cohort that defined much of this, suggests that warranty runs somewhere between five and ten years for most people.

But the “power of zero” is also one of the most contested ideas in this field.

The warranty shortens in people with diabetes, in smokers, in older adults and in anyone whose metabolic profile is drifting the wrong way. MESA was explicit on this point: even at a calcium score of zero, smoking, diabetes and high blood pressure still independently predicted events. Zero lowers your risk. It does not delete your risk factors. As that JACC analysis showed, cumulative LDL exposure keeps compounding underneath a zero score, because soft plaque forms before it calcifies.

There’s another important point worth stating plainly. After three decades of use, no randomized trial has shown that a CAC-guided strategy reduces heart attacks, strokes, or deaths compared with simply treating risk factors directly. The score predicts events well, but predicting an event and proving that acting on the prediction changes it are not the same thing. Much of a calcium score’s real-world value may be motivational: a number that moves people to start and stay on, treatment they were already candidates for.

That’s why I treat a calcium score as one input, not a verdict and why a zero never makes the lipid conversation disappear.

So zero buys you time. It does not buy you out.

I know this one personally.

I got my own calcium score in September 2022. Not because a guideline told me to. Fear did. I’d had some reflux that I couldn’t entirely talk myself out of worrying about and I’d been seeing too many young patients land in the hospital with heart attacks that no one saw coming. I wanted to look inside my own arteries before I told one more person to look inside theirs.

It was zero.

A clean scan. The warranty. And about a year later, in 2023, I started rosuvastatin and ezetimibe anyway. In the last article I told you why: an LDL of 136 and an ApoB of 125, both higher than I want them for a 30-year horizon. The ApoB is now 53.

This sequence isn’t a contradiction. It’s the entire point of this article. My zero told me my near-term risk was low. It said nothing about the cumulative atherogenic exposure my elevated ApoB would deliver over the next three decades if I left it alone. The calcium score answers what has already happened. It does not answer where is this heading. My answer to the first was reassuring. My answer to the second is what keeps me taking the pill.

I’ll add two more tidbits, because they’re the same ones I’d want from my own cardiologist. I didn’t check my Lp(a) when I got the scan; I only thought to test it in late 2023. It came back at 13, which is low, so unlike Justin, genetics are not the lever driving my risk. Mine is cumulative LDL, full stop. A zero score never fully rules out soft plaque that simply hasn’t calcified yet. There’s also irony in the biology: when you start a statin, some of that soft plaque stabilizes and calcifies, which means a perfectly clean baseline scan can read higher years later precisely because the treatment is working. The number going up wouldn’t mean I’m getting worse. It might mean the opposite.

Trajectory, not threshold. A zero score is a snapshot. I’m treating the slope.

What Justin’s number meant

Justin’s score didn’t change his biology. The Lp(a) of 187 was still there. The ApoB was still elevated. The father with bypass surgery at 54 was still part of his story. None of that moved.

Here’s the part worth being precise about: Justin already had every reason to treat before he ever laid down in the scanner. An Lp(a) of 187, a first-degree relative with bypass at 54 and an elevated ApoB are, together, a clear indication on their own. The calcium score didn’t create his case. If his calcium score had come back zero, the case would have stood anyway, the way mine did. This is important to realize, because the danger identifying risk from a positive scan is the inverse it teaches: a zero score is a reason not to treat. It isn’t. The scan didn’t make Justin’s decision. It made his decision real to him.

What the scan changed was the conversation.

Before the scan, Justin’s risk was an argument. A projection. A set of numbers I was asking him to take on faith against a lifetime of doing everything right. He’s an athlete with pristine metabolic health and the abstraction of “elevated lifetime risk” is easy to wave off when you can run a sub-20 5K.

The calcium score took the argument away. It wasn’t a probability anymore. It was a finding. The plaque wasn’t something that might develop in his future. It was already there, behind the wall of his fitness, accumulating for years while every other number told a reassuring story.

This is what moved him. Not me. Not the guideline. The evidence of his own arteries.

We set his LDL target using the number. We started rosuvastatin. We also talked about the future: the Lp(a)-lowering therapies now in late-stage trials that may, within a few years, give him a lever for the one risk factor he currently can’t touch.

Two men. One with exceptional fitness and a positive scan. One with elevated cholesterol and a clean one. Same framework, opposite numbers and in both cases the calcium score did exactly what it’s supposed to do: not deliver a verdict, but sharpen the decision.

Two different risk pictures and the same decision. The score sharpened each conversation; it didn’t dictate the treatment.

What you can do about it

If you’re a patient: If you’re a man over 40 or a woman over 45 and your statin decision feels like a coin flip, or if you have a family history of early heart disease, an elevated Lp(a), or South Asian ancestry, ask your doctor whether a calcium score makes sense for you. If you’ve had a chest CT for any reason, ask whether it showed coronary calcium. The answer may already be in your chart.

If you’ve had a scan: Know what your number actually means. A positive score is a reason to act, not a reason to panic and under the 2026 guideline it now sets a specific LDL target. A zero is genuinely reassuring, but it’s a warranty with an expiration date, not a lifetime guarantee. Ask what your target is.

If you train seriously: Justin is the cautionary tale. A sub-20 minute 5K, a resting heart rate of 47 and calcium in his coronary arteries at 49. Fitness is one leg of the triangle. It does not protect you from the other two. If you haven’t checked your ApoB, your Lp(a), and, where it’s warranted, your calcium score, you are guessing about the one part of your cardiovascular health your training can’t fix.

Coronary calcium measures just one leg: circulation. Fitness and metabolism still decide your risk.

The scan doesn’t tell you what might happen.

It tells you what already has, so you can change what comes next.

Patient details have been changed to protect privacy. Clinical accuracy has been preserved. This article is educational and not a substitute for individualized medical advice.

If someone you know is weighing whether to get a calcium score, or has one sitting in a chart that nobody explained, send them this.

For individuals who want this integrated approach, where circulation, metabolism and cardiorespiratory fitness are treated as a single system: jakekellymd.com.

References

Blumenthal RS, Morris PB, Gaudino M, et al. 2026 ACC/AHA/AACVPR/ABC/ACPM/ADA/AGS/APHA/ASPC/NLA/PCNA Guideline on the Management of Dyslipidemia: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation. 2026;153.
Dzaye O, Blaha MJ, et al. Warranty Period of a Calcium Score of Zero: Comprehensive Analysis From MESA. JACC: Cardiovascular Imaging. 2021;14(5):990–1002.
Peng AW, Razavi AC, Wilkins JT et al. Impact of LDL-Cholesterol When the Coronary Artery Calcium Score Is 0: Long-Term Cardiovascular Events. JACC. 2025;86(9):676–680.
Sabouret Y, Giamundo A, Rosencher J, Figliozzi S. Coronary artery calcium scoring in 2026: strengths, limitations, and optimized clinical use. Frontiers in Radiology. 2026;6:1822303.
Budoff, M. J., Young, R., Burke, G. et al. Ten-year association of coronary artery calcium with atherosclerotic cardiovascular disease (ASCVD) events: the multi-ethnic study of atherosclerosis (MESA). European Heart Journal. 2018. 39(25), 2401–2408.

Legal Medical Disclaimer

The information provided on this site including articles, newsletters, social-media posts, videos, and downloadable resources is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment. Your use of this content does not establish a physician-patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement. Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.

62 Out of 66

Jake Kelly MD — Fri, 12 Jun 2026 14:59:09 GMT

I used to prescribe statins like it was a job requirement.

Early in my career, fresh out of fellowship, the conversation was unidirectional. Your LDL is abnormal. Take a statin. If you stop, you’re increasing your cardiovascular risk of a heart attack or death. I wasn’t wrong about the risk. I was wrong about the conversation.

The truth is, I was afraid. Afraid that if I didn’t prescribe the statin and something happened, I’d be liable. Afraid that deviating from the guideline-driven default would make me a bad doctor. So myconversations had a predictable structure: I presented the number, I recommended the pill and the patient either agreed or quietly never filled the prescription.

Some patients would nod, seemingly planning to take the pill on my recommendation. Some nodded and didn’t. Some started, felt something, stopped and never told me. And some patients pushed back, citing something they had read or recanting a friend’s experience. Many shared their general sense that the statin pill was doing more harm than good. I didn’t handle those conversations well. I heard resistance where I should have heard a desire to under more about the underlying lipid biology. My patients desired to understand their full cardiovascular risk profile and how to address from all of their life inputs.

It took me years and thousands of these conversations, to realize the problem wasn’t the patient. It was the framing. The question was never “statin or not.” The question was always: how do we lower my overall cardiovascular risk?

This reframe in my thinking has changed how I approach discussing cardiovascular risk.

And then Justin walked into my office.

The patient who didn’t look like he needed a cardiologist

Justin is 49. He lifts weights three times a week. He mountain bikes. He runs 20 to 25 miles a week and recently ran a 5K in 19 minutes and 55 seconds. His resting heart rate is 47. His HRV sits in the 70s.

Based on his race time, his estimated VO2 max is roughly 49 to 51 ml/kg/min. That puts him above the 85th percentile for men his age in the FRIEND registry. By every exercise physiology metric, Justin has exceptional cardiorespiratory fitness.

His metabolic markers were equally clean. HbA1c: 5.1 percent. Fasting insulin: 4.1. Blood pressure: 118/72. hsCRP: less than 0.2. Triglycerides: 74. HDL: 52. No insulin resistance. No inflammation. No metabolic syndrome.

I think about cardiovascular risk through three lenses: circulation, metabolism and cardiorespiratory fitness. Two of Justin’s three legs of the triangle were as strong as I see in any patient.

But his father had coronary artery bypass surgery at 54.

And his Lp(a) was 187 nmol/L.

Lp(a) is the strongest inherited lipid risk factor for atherosclerotic cardiovascular disease. Roughly 20 percent of people have levels above the threshold where risk accelerates and there is currently no medication that meaningfully lowers it. You can’t exercise it down. You can’t diet it away. It’s genetic. You either have elevated Lp(a) or you don’t. Justin has it.

His LDL cholesterol was 124 mg/dL. His ApoB was 95 mg/dL. Concordant. In isolation, those numbers wouldn’t alarm most physicians. But in the context of Lp(a) 187 and a father who needed bypass surgery at an age Justin was rapidly approaching, those numbers meant something different. His cumulative atherogenic exposure was higher than it appeared, and it had been accumulating silently for decades.

Justin didn’t want to take a statin. He’d read the side effect lists. He’d heard the stories. He was doing everything else right and the idea that he still needed a pill didn’t sit well.

I understood completely. Because that’s exactly the conversation I can use the data to frame overall cardiovascular risk.

What 123,940 patients actually showed

In February 2026, the Cholesterol Treatment Trialists’ Collaboration published the most comprehensive analysis of statin side effects ever conducted. They pooled individual participant data from 19 large-scale, double-blind, randomized controlled trials, 123,940 people, with a median follow-up of 4.5 years and systematically evaluated every adverse effect listed on statin product labels.

There are 66 undesirable effects listed.

62 of them were not supported by the trial evidence.

Statins did not cause memory loss. They did not cause depression. No sleep disturbance, erectile dysfunction, fatigue, headache, or peripheral neuropathy. Not at rates distinguishable from placebo. The confirmed harms were limited to a modest increase in liver enzyme elevations, minor changes in urinary composition, a small signal for edema and the one most patients worry about: muscle symptoms, which occurred at an absolute excess of roughly 0.5 percentage points over placebo, concentrated in the first year.

The well-established risk of a small increase in blood glucose, mainly in patients already trending toward diabetes, was confirmed. Serious myopathy, rhabdomyolysis, remains real but exceedingly rare, roughly 1 in 10,000 per year.

These are all of the side effects. Four confirmed signals out of 66 listed. All with absolute annual excess risks below 0.1 percent for the non-muscle findings.

The Lancet published an accompanying editorial titled “Statin safety: when warnings outlive the evidence.” The editors wrote that more than 30 years after the Scandinavian Simvastatin Survival Study showed that statins save lives, their full public health potential remains unrealized. The editorial called on regulatory authorities worldwide to review statin product labels and remove or recategorize harms unsupported by causal evidence.

Three months later, JACC’s editor-in-chief Harlan Krumholz published a pointed editorial: “Time to Move Beyond the Statin Nocebo Effect.” His argument was direct. Every unsupported adverse effect listed on a statin label functions as a suggestion planted in the minds of patients and clinicians, one that the best available evidence says is unfounded. The nocebo effect, symptoms arising from the expectation of harm rather than from the drug itself, is not the result of prescribing a medication. It is amplified by misleading labels, uncontested online narratives and cautious clinical framing. The nocebo effect is also at least partly reversible.

Krumholz cited two studies to explain the nocebo effect.

The first is ASCOT-LLA. In the blinded phase of this large randomized trial, muscle-related adverse events were reported at nearly identical rates in statin and placebo groups: 2.03 percent versus 2.0 percent. Then the trial moved to an open-label extension, where patients and their doctors knew who was taking atorvastatin. Muscle complaints in the statin group rose significantly. The drug hadn’t changed. The knowledge had.

The second is the SAMSON trial, published in JACC in 2021. Sixty patients who had previously quit statins due to side effects were given a randomized sequence of statin, placebo, and empty bottles. The result: 90 percent of the symptom burden patients attributed to statins was reproduced identically by placebo. Neither the intensity of symptoms on starting nor the relief on stopping distinguished statin from sugar pill. After seeing their own data, half the patients went back on their statins.

I want to be clear about something. If you took a statin and experienced real muscle pain, I’m not dismissing your pain. Roughly 1 in 100 patients have a genuine, pharmacologically mediated muscle response. It’s real, it’s uncomfortable and if it happened to you, I believe you and I’m sorry. The point is not that everyone’s symptoms are imaginary. The point is that 62 of the 66 listed concerns aren’t supported by the trial evidence and for the small number of patients who do experience real effects, there are now four other drug classes that work through entirely different mechanisms. You’re no longer out of luck.

I want to be honest about clinical trial data and the CTT Collaboration in particular, which has faced legitimate criticism about data transparency. The individual participant data from contributing trials was shared under agreements that prevent release to independent researchers. Some have argued this limits the ability of the broader scientific community to verify the findings. This criticism is valid and important. Transparency in clinical research is paramount and I’d welcome more open access to this data. But the study’s design, 19 double-blind RCTs with nearly 124,000 participants, represents the highest tier of evidence we have available. The findings are consistent with what every other well-designed nocebo trial has shown: most of what people fear about statins isn’t caused by statins.

This is the conversation Justin and I needed to have. Not “take a statin.” But: let me show you what the evidence actually says about these medications, because the conversation you’re afraid of isn’t the one we need to have.

The conversation I used to have

Patient sits down. Labs are back. LDL is 145.

Old me: “Your cholesterol is high. I’d like to start you on a statin.”

Patient: “I’ve heard those have a lot of side effects.”

Old me: “The benefits outweigh the risks.”

Patient: Nods. Leaves. Never fills the prescription.

Or: Patient fills the prescription. Reads the package insert. He or she then develops fatigue, muscle aches, brain fog within the first week. Patient stops the medicine. Tells three friends. The narrative spreads.

This was not a failure of the drug. It was a failure of the conversation. I was asking the wrong question. The question isn’t whether you should take a statin. The question is: what is your ApoB and what levers do we have to lower it?

Five levers, one goal

The goal isn’t “take a statin.” The goal is lower ApoB. Lower cumulative LDL exposure. Less atherosclerotic plaque building silently in your coronary arteries over 30 years.

The 2026 ACC/AHA Dyslipidemia Guideline, the first major update since 2018, made this more explicit than ever. For the first time, ApoB is formally recommended as a measurement in US guidelines, particularly in patients with diabetes, elevated triglycerides, or achieved LDL below 70 where standard lipid panels may underestimate true risk. Universal Lp(a) testing is now a Class I recommendation for all adults. The guideline also emphasizes cumulative lipid exposure over time, not just a snapshot number.

ApoB is the number I care about. It counts every atherogenic particle in your blood, LDL, VLDL, IDL, Lp(a), all of them. One ApoB, one particle, one chance to enter the arterial wall. Your LDL cholesterol can look fine while your particle count is elevated. That’s the gap ApoB closes.

Here’s how I think about lowering it. Five levers. The mix depends on the patient.

Lever 1: Nutrition. Dietary changes can lower LDL by 10 to 30 percent depending on baseline diet and commitment. Dean Ornish demonstrated in randomized trials that a program built around plant-based nutrition, 10 percent fat, reversed coronary artery disease. LDL dropped from roughly 123 to 102 mg/dL at three years. Meaningful. More moderate approaches, Mediterranean diet, reducing saturated fat, increasing soluble fiber, plant sterols, typically deliver 10 to 15 percent LDL reduction. Nutrition is a real lever. But I’m honest with patients: for most people with significant ApoB elevation, diet alone won’t narrow the gap. It’s a lever. Rarely the only one you need.

Lever 2: Exercise. A meta-analysis of 148 trials showed exercise lowers LDL by roughly 7 mg/dL. Modest. But exercise doesn’t earn its place in the framework through LDL lowering. It earns it through everything else: VO2 max improvement, blood pressure reduction, insulin sensitivity, inflammation, endothelial function, autonomic tone. Exercise is essential for cardiovascular risk reduction. I wrote an entire article about how cardiorespiratory fitness predicts death better than any lipid number. But if your ApoB is 130, exercise alone won’t get you to target. You exercise to raise your capacity, not to lower your cholesterol. These are different levers.

Lever 3: Body composition and metabolic health. Visceral fat drives insulin resistance, which drives atherogenic dyslipidemia: high triglycerides, low HDL, small dense LDL, elevated ApoB. Reducing visceral fat, whether through sustained caloric deficit, structured exercise, or GLP-1 receptor agonists, can substantially improve the entire lipid profile. This is the metabolic leg of the Triangle. It’s often the most under-appreciated lever.

Justin was already pulling levers one through three harder than most patients I see. His nutrition was dialed in. His exercise capacity was exceptional, above the 85th percentile for his age. His metabolic health was pristine. Triglycerides 74. Fasting insulin 4.1. HbA1c 5.1 percent. There was nothing left to optimize on those three levers.

But Lp(a) doesn’t respond to lifestyle. It doesn’t drop with better nutrition. It doesn’t improve with more miles. It’s set by genetics. And with his Lp(a) at 187, plus a father who needed bypass at 54, the cumulative atherogenic burden in Justin’s coronary arteries was a problem only lever four could address.

Lever 4: Medications. This is where the landscape has fundamentally changed. Ten years ago, the conversation was statins or nothing. Today there are five drug classes:

Statins lower LDL by 30 to 50 percent or more depending on intensity. They remain the most studied cardiovascular drug in history, with decades of outcomes data.

Ezetimibe blocks intestinal cholesterol absorption and adds roughly 20 percent additional LDL reduction. It costs a few dollars a month.

Bempedoic acid works upstream of statins in the same cholesterol synthesis pathway, but because it’s not active in muscle tissue, it doesn’t cause the myalgia that some patients experience with statins. The CLEAR Outcomes trial showed cardiovascular event reduction in statin-intolerant patients.

PCSK9 inhibitors, evolocumab and alirocumab, lower LDL by 50 to 60 percent. Injectable, every two to four weeks. Originally expensive, now increasingly accessible.

Inclisiran uses small interfering RNA to silence PCSK9 production. Two injections per year after loading doses. Roughly 50 percent LDL reduction.

A patient who says “I don’t want to take a statin” now has four other drug classes available. This sentence used to end the conversation. It no longer has to.

Lever 5: Genetics, and the frontier. About 2 percent of Black individuals carry natural PCSK9 loss-of-function mutations. These people have LDL levels roughly 28 percent lower than average, and an 88 percent reduction in coronary heart disease risk over their lifetimes. They’re healthy into their late eighties and nineties. Normal liver function. No adverse consequences. This is nature’s proof of concept: sustained very low LDL, from birth, is profoundly protective.

Now gene editing is mimicking what nature does.

On May 25, 2026, Eli Lilly released results from the VERVE-102 Phase 1b trial. A single intravenous infusion of a base-editing therapy that turns off the PCSK9 gene in the liver reduced PCSK9 protein by up to 88 percent and LDL cholesterol by up to 62 percent, with reductions sustained for up to 18 months. No treatment-related serious adverse events. No dose-limiting toxicities. Lilly plans to initiate Phase 2 by the end of this year.

One dose. Potentially permanent.

We are moving from a world where lowering cholesterol required daily discipline to a world where it might require a single infusion. The question was never “statin or nothing.” It was always “how do we get your ApoB where it needs to be?”

What Justin decided

I ordered a coronary artery calcium scan.

For a 49-year-old man with no symptoms, excellent fitness, and clean metabolic markers, many physicians would have relied on risk calculators alone. But risk calculators underestimate lifetime risk in younger patients and they don’t account for Lp(a). The CAC scan does something different: it shows you what has already happened. Not what might happen. What has already begun.

Justin’s calcium test or CAC score came back positive at 49.

Calcium in the coronary arteries of a 49-year-old man who runs a sub-20 5K. This finding reframed the entire conversation. The plaque wasn’t a future risk. It was already there, silent, building behind a wall of excellent fitness and perfect metabolic numbers. Two legs of the Triangle were strong. The third had been quietly accumulating damage for years.

We sat with the results together. I walked him through the Five Levers framework. He was already maximizing levers one through three. His Lp(a) was fixed by genetics. That left lever four.

I didn’t tell Justin to take a statin. I showed him the data. 123,940 patients. 62 of 66 feared side effects unsupported by trial evidence. A calcium score showing that subclinical atherosclerosis was already present. A father who had bypass surgery at the age Justin is approaching. And a drug class with more outcomes data behind it than almost anything else in cardiovascular medicine.

We discussed the options. For Justin, with his Lp(a) driving risk on top of a moderately elevated ApoB, the goal was meaningful LDL reduction. We started rosuvastatin 10 mg.

We also talked about the future. There are no approved therapies that target lower Lp(a) today, but multiple agents are in late-stage clinical trials. Pelacarsen. Olpasiran. Lepodisiran. Zerlasiran. Within the next few years, Justin may have a lever for the one risk factor he currently cannot meaninfully modify. Until then, we control what we can.

Justin didn’t take a statin because I told him to. He took it because he understood the math. The conversation wasn’t “take this pill.” The conversation was: here is your number, here is what’s already happening in your arteries, here are five levers and here is why lever four matters for you specifically.

The conversation is different for me. And it ends differently for us now.

What I decided for myself

I wrote about this briefly in my ApoB article. My LDL was 136. My ApoB was 125. Concordant. Both elevated above where I want them for a 30-year time horizon.

I changed my diet. Reduced saturated fat. Increased fiber. Added more plant-forward meals while keeping protein adequate for training. I improved my body composition and dialed in my metabolic markers.

Then I started rosuvastatin 10 mg and ezetimibe 10 mg. Two levers in one decision. A moderate-intensity statin paired with an intestinal cholesterol absorption blocker that costs a few dollars a month.

Not because I was afraid. Because I ran the numbers. Cumulative LDL exposure over the next 30 years carries more risk than any single reading. I had three levers working, nutrition, exercise, metabolic health and adding a fourth gave me the trajectory I wanted.

My ApoB went from 125 to 53. My LDL from 136 to 49.

This is one person’s decision. Yours may be different. The levers available to you may look different. But the framework is the same: know the number, understand the options, make an informed choice with your physician. The days of “take the pill or don’t” are over.

What you can do about it

If you’re a patient: Ask your doctor for your ApoB level, not just your LDL. Ask about your Lp(a), which the 2026 guidelines now recommend testing in every adult. If your numbers are elevated, ask about all five levers, not just statins. If you’ve been avoiding a statin because of side effect fears, the Lancet data is worth reading: 62 of 66 listed effects weren’t caused by the drug. If you genuinely experience side effects on a statin, know that there are four other drug classes available now. You’re not stuck.

If you’re a clinician: Lead with the number, not the drug. “Your ApoB is X” opens a different conversation than “I’d like to start you on a statin.” Present the lever framework. It changes the dynamic from confrontation to collaboration. The 2026 Lancet meta-analysis and the Krumholz JACC editorial are tools for the nocebo conversation. Use them gently. Consider how a CAC scan can move a patient from abstract risk to tangible evidence, the way it did for Justin.

If you train seriously: High cardiorespiratory fitness does not guarantee low ApoB. Justin runs a sub-20 5K with a resting heart rate of 47 and he had calcium in his coronary arteries at 49. Both capacity and circulation matter. If you haven’t checked your ApoB and Lp(a), check them.

Patient details have been changed to protect privacy. Clinical accuracy has been preserved.

If you know someone who’s been debating whether to take a statin, or someone who stopped taking one, send them this.

If this changed how you think about statin safety, share it with someone whose fear of side effects may be costing them years of protection.

For individuals who want this integrated approach, where circulation, metabolism and cardiorespiratory fitness are treated as a single system: jakekellymd.com.

References

Cholesterol Treatment Trialists’ Collaboration. Assessment of adverse effects attributed to statin therapy in product labels: a meta-analysis of double-blind randomised controlled trials. Lancet. 2026;407(10529):689-703.
Statin safety: when warnings outlive the evidence. Lancet. 2026;407(10529):649.
Krumholz HM. Time to move beyond the statin nocebo effect. J Am Coll Cardiol. 2026;87(19):2549-2551.
Gupta A, Thompson D, Whitehouse A, et al. Adverse events associated with unblinded, but not with blinded, statin therapy in the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT-LLA). Lancet. 2017;389(10088):2473-2481.
Howard JP, Wood FA, Finegold JA, et al. Side effect patterns in a crossover trial of statin, placebo and no treatment. J Am Coll Cardiol. 2021;78(12):1210-1222.
Grundy SM, Stone NJ, Bailey AL, et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol. J Am Coll Cardiol. 2019;73(24):e285-e350.
2026 ACC/AHA/AACVPR/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of dyslipidemia. Circulation. 2026.
Cohen JC, Boerwinkle E, Mosley TH, Hobbs HH. Sequence variations in PCSK9, low LDL, and protection against coronary heart disease. N Engl J Med. 2006;354(12):1264-1272.
Nissen SE, Lincoff AM, Brennan D, et al. Bempedoic acid and cardiovascular outcomes in statin-intolerant patients. N Engl J Med. 2023;388(15):1353-1364.
Sabatine MS, Giugliano RP, Keech AC, et al. Evolocumab and clinical outcomes in patients with cardiovascular disease. N Engl J Med. 2017;376(18):1713-1722.
Verve Therapeutics / Eli Lilly. VERVE-102 Heart-2 Phase 1b interim results. May 25, 2026.
Ornish D, Scherwitz LW, Billings JH, et al. Intensive lifestyle changes for reversal of coronary heart disease. JAMA. 1998;280(23):2001-2007.

Legal Medical Disclaimer

Everyone Measures It But Us

Jake Kelly MD — Thu, 04 Jun 2026 12:03:40 GMT

Gary’s lipid panel is the kind most cardiologists dream to review with their patients. Total cholesterol in the 120s. Triglycerides under 60. HDL 51. LDL 61. ApoB 52. All well below the threshold where most of us even start a conversation about medication. By every conventional measure of cardiovascular risk, his circulation is excellent.

He was moderately athletic growing up, playing a variety of sports, was above average, and enjoyed them. But as a busy professional in his late forties, his athletic history has faded into the background. When I asked Gary about his daily activity, the clinical picture sharpened. As part of his emerging health journey, he’d recently purchased an Apple Watch. His average: about 5,200 steps per day. He used to jog two miles on the weekends, but with a six-year-old and a three-year-old, weekends had become stroller walks in the park, one to two miles at a casual pace. During the week, his step count often fell below 5,000. The weekends were pulling his daily step count average up.

Gary has a family history of coronary artery disease. His father and uncles developed coronary artery disease in their late sixties and early seventies. He and his wife had started their family a little later in life, beginning in their early forties. Now, at 48, with two young children counting on him, he was thinking differently about what it meant to be healthy. Gary didn’t want to just survive long enough to meet his grandchildren. He wanted to live well and be present for decades to come. Mobile, capable to hold, carry and play with his grandchildren. Gary had found and started reading the growing body of longevity and fitness writing and it led him to consult with me.

We talked through my framework, the Cardiovascular Longevity Triangle. Circulation, metabolism and cardiorespiratory fitness. Two of Gary’s three legs were outstanding. His lipids were enviable. His metabolic markers were clean. But the third leg, capacity, was unmeasured. I agreed to measure what he asked for: a cardiopulmonary exercise test.

His peak VO₂ came back at 31 ml/kg/min.

For a 48-year-old man, that places Gary between the 30th to 35th percentile according to the FRIEND registry, the national reference standard for directly measured cardiorespiratory fitness. Below the 50th percentile. Below average.

Gary’s lipid panel said he was fine. His fitness test said something else. One of the strongest predictors of whether my patients live or die isn’t on their lab panel. It’s a number most of them have never been told.

Gary decided to focus on the one leg of the Triangle that needed work. And that decision, not a medication, or a procedure, but a measurement followed by a plan, is the most important thing I did for him.

Here’s the problem: almost no one is doing what we did for Gary.

The data we keep ignoring

In 2009, Kodama and colleagues published a meta-analysis in JAMA of 33 studies spanning 102,980 healthy adults. The finding was simple and staggering: for every one-MET increase in cardiorespiratory fitness, all-cause mortality dropped 13 percent.

A MET, or metabolic equivalent of task, is a unit of energy expenditure. One MET equals the oxygen your body consumes at rest, about 3.5 ml of oxygen per kilogram of body weight per minute. Sitting quietly is 1 MET. A slow walk is about 2. A brisk walk is 3 to 4. Jogging is 7 to 8. Running hard is 10 METs or above. When we say someone achieved “9 METs” on a treadmill test, it means their body was consuming nine times the oxygen it uses at rest.

So one MET higher. That’s the difference between a slow walk and a brisk walk. A 13 percent reduction in dying.

In 2018, Mandsager and colleagues at the Cleveland Clinic published exercise treadmill data on 122,007 consecutive patients in JAMA Network Open. They stratified patients from low fitness to elite fitness and measured long-term mortality. The result: a five-fold difference in death between the least fit and the most fit. Being unfit carried the same mortality risk as smoking. And diabetes.

Your patient who can’t walk briskly up a hill carries the same mortality risk as the one who still smokes a pack a day.

In 2022, Kokkinos and colleagues extended this to 750,302 US veterans, published in JACC. The conclusion was even more direct: low cardiorespiratory fitness was a stronger predictor of death than any single traditional cardiac risk factor. More than hypertension. Diabetes. Coronary artery disease. And there was no upper limit of benefit. The fittest patients lived the longest. Period.

By 2024, Lang and colleagues synthesized the entire field: 20.9 million observations across 199 cohorts, published in the British Journal of Sports Medicine. High fitness versus low fitness: 53% lower all-cause mortality. For heart failure incidence specifically: 69% lower.

It’s not just your baseline. Kokkinos showed in a 2023 follow-up study of 93,060 veterans with serial exercise testing that each one-MET improvement over time was associated with a 15 percent reduction in mortality. The number moves and the trajectory matters.

The American Heart Association recognized all of this in 2016, publishing a scientific statement in Circulation titled, “Importance of Assessing Cardiorespiratory Fitness in Clinical Practice: A Case for Fitness as a Clinical Vital Sign.” They called CRF a vital sign, in those exact words, and recommended it be assessed routinely in all healthcare settings.

That was ten years ago.

Almost nothing has changed.

So who actually measures it?

I started making a list of every institution, organization and profession that formally measures cardiorespiratory fitness. It got long.

NASA measures VO₂ max on every astronaut with a metabolic cart, the gold standard, before flight, monthly during spaceflight, and after return. The requirement: at least ~ 35 ml/kg/min for space station missions. At least 40 for Mars. If the number drops, the mission changes.

The FAA requires every pilot flagged with a cardiac condition to complete the Bruce Protocol treadmill test. Minimum: Stage III, at least nine minutes, approximately nine METs, 100% of predicted maximum heart rate. Fail and you don’t fly.

Every branch of the US military tests aerobic fitness annually. The Army’s ACFT includes a two-mile run, a direct proxy for cardiorespiratory fitness. Research shows that VO₂ max explains 38 percent of the variance in total ACFT scores. For Special Operations forces, including Navy SEALs, Army Rangers, British SAS, German KSK, and Israeli Sayeret Matkal, selection itself is a cardiorespiratory fitness test. You either have the aerobic capacity or you’re out.

Fire departments have what may be the most consequential standard of all. For decades, the National Fire Protection Association mandated a flat, absolute aerobic floor of 12 METs (42 ml/kg/min) for all personnel. But in a massive shift to align with modern clinical data, the updated standard (NFPA 1580) now utilizes strict, age- and sex-adjusted population percentiles. Fall below the 35th percentile and you are restricted from essential firefighting duties. Land between the 35th and 50th percentile and you are placed on a mandatory, career-altering cardiorespiratory exercise program. These aren't suggestions. They are operational standards.

Why? Because cardiac arrest is the number one killer of firefighters in the line of duty. They started measuring fitness because their people were dying.

The NFL Combine runs an intensive cardiac workup on every prospective player: ECG, echocardiogram, stress testing when warranted. Two to three players are red-flagged every year. The NBA goes further: mandatory annual ECG and stress echocardiogram for every active player. Not triggered by symptoms. Not conditional. Mandatory. Administered in partnership with Columbia University, results stored in a centralized database. Basketball has a higher rate of sudden cardiac death than other professional sports. So they measure.

FIFA requires a Pre-Competition Medical Assessment for all international competitions, mandatory since 2006, with ECG plus additional testing when indicated. Italy stands alone globally. Since 1982, Italian law has mandated both a resting ECG and an exercise stress test for every competitive athlete at every level. Not just professionals. All of them. And the data suggests it’s working. Italy has documented a reduction in sudden cardiac death from hypertrophic cardiomyopathy since implementing the program.

Tour de France cycling teams perform full cardiopulmonary exercise testing with gas exchange analysis, including VO₂ max, lactate thresholds and ventilatory efficiency, on every rider. These are the most physiologically measured humans on earth.

Surgeons have their own version. Before major noncardiac surgery, functional capacity is assessed, formally or informally. Can the patient climb two flights of stairs without stopping? That’s roughly four METs. Below four METs, surgical risk increases substantially. Increasingly, CPET is used for formal preoperative risk stratification: an anaerobic threshold below 9 to 11 ml/kg/min predicts ICU admission and mortality.

Life insurance companies require a treadmill stress test for applicants over 50 seeking high-value policies. They’re pricing your survival. They want the number.

Cardiac rehabilitation programs test exercise capacity at entry and at exit. The AHA and AACVPR set a goal of at least 15 percent improvement in peak VO₂ over the course of rehabilitation.

In heart failure, CPET isn’t optional. It’s mandatory. In 1991, Mancini and colleagues published the landmark study in Circulation establishing that peak VO₂ at or below 14 ml/kg/min identifies patients who benefit from heart transplantation. In 1995, Lynne Warner Stevenson refined the approach in a study of 737 consecutive advanced heart failure patients in JACC, showing that peak VO₂ should be measured on optimal medical therapy and that an improvement of at least 2 ml/kg/min predicted better survival, sometimes enough to defer transplant entirely. Today, the ISHLT guidelines make maximal CPET a Class I recommendation for transplant evaluation.

In advanced heart failure, this test decides whether you live or die.

The gap

And then there’s the 55-year-old sitting in my prevention clinic.

Family history of heart disease. Desk job. Resting heart rate of 78. Lipid panel looks adequate. Blood pressure is controlled. The standard risk calculator puts him at intermediate risk.

Nobody has measured his fitness. Nobody has even asked.

I could estimate his cardiorespiratory fitness in 30 seconds using the Duke Activity Status Index, a simple set of questions about exercise tolerance. I could get a rough MET estimate from his age, weight, resting heart rate and activity level using a validated nonexercise equation. I could refer him for a treadmill test or a full CPET.

But Medicare will not reimburse an exercise stress test for an asymptomatic patient. The CPT codes exist, 93015 through 93018 and they’re used every day for chest pain evaluation, post-revascularization monitoring, arrhythmia assessment, valvular disease management, preoperative clearance and transplant candidacy. The test exists. The infrastructure exists. Cardiologists know how to order it, read it and act on it.

The exclusion is specific: asymptomatic screening. The one population where the mortality data is strongest.

The AHA called cardiorespiratory fitness a clinical vital sign. The ACC/AHA ASCVD risk calculator does not include it. The USPSTF has actually downgraded exercise stress testing in low-risk asymptomatic adults. The field that identified the relationship between fitness and mortality has not built a system to measure fitness in the people most at risk of dying.

We measure cholesterol. We measure blood pressure. We measure glucose, hemoglobin A1c, BMI and resting heart rate. We do not routinely measure the single variable that predicts death better than all of them.

What you can do about it

The metric debate, VO₂ max versus peak METs versus wearable estimates, is real and I’ll address it along with my personal thoughts in detail in a future post. Topol is right that 99 percent of the mortality data comes from estimated peak METs on treadmill tests, not from gas-exchange VO₂ max. Attia is right that the signal is powerful regardless of how you label it. But the argument about which metric to use has become a distraction from the more important question: are you (we?) measuring it at all?

Here’s what I want you to take from this.

If you’re a patient: Ask your doctor a simple question at your next visit: what is my estimated cardiorespiratory fitness? If they’ve never assessed it, that’s a gap worth closing. Know the threshold: the Kodama data shows that 7.9 METs is an inflection point for mortality. Can you walk briskly up two flights of stairs without stopping? That’s roughly four METs. Can you jog for ten minutes? Roughly seven to eight. If you wear a Garmin or Apple Watch, use the VO₂ max estimate for trend, not for the absolute number. If it’s declining over six months, pay attention. Your trajectory matters more than any single reading.

If you’re a clinician: Start estimating CRF at every visit. It takes 30 seconds. A simple question about exercise tolerance gives you a MET estimate that predicts your patient’s survival better than their lipid panel. The AHA called this a clinical vital sign a decade ago. We owe our patients at least an estimate.

If you’re in the fitness community: The debate between Topol, Attia and Magness about whether we’ve created a counterproductive “VO₂ max craze” is interesting. But don’t miss the forest. The data across 20.9 million observations says the same thing: raise your fitness and you lower your risk of dying. The UK Biobank data from Kim and colleagues shows that cardiorespiratory fitness plus grip strength together reduces mortality more than either alone. Train both.

The strongest predictor of survival we have is free to estimate, inexpensive to measure and modifiable at any age. After nine months of endurance training, adults in their sixties increased their VO₂ max by 19 to 22 percent. An 80-year-old man holds a recorded VO₂ max of 50 ml/kg/min, equivalent to an inactive 25-year-old. The number moves.

We just have to start measuring it.

If you know someone whose doctor has never asked about their fitness level, send them this.

If this changed how you think about cardiorespiratory fitness and survival, share it with someone who needs to hear that the strongest predictor of longevity isn’t on their lab panel.

For individuals who want this integrated approach, where cardiorespiratory fitness, circulation, and metabolic health are treated as a single system: jakekellymd.com.

Legal Medical Disclaimer

The information provided on this site including articles, newsletters, social-media posts, videos, and downloadable resources is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment. Your use of this content does not establish a physician–patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement. Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.

References

Ross R, Blair SN, Arena R, et al. Importance of assessing cardiorespiratory fitness in clinical practice: a case for fitness as a clinical vital sign: a scientific statement from the American Heart Association. Circulation. 2016;134(24):e653-e699.
Ross R, Arena R, Myers J, Kokkinos P, Kaminsky LA. Update to the 2016 American Heart Association cardiorespiratory fitness statement. Prog Cardiovasc Dis. 2024;83:10-15.
Kodama S, Saito K, Tanaka S, et al. Cardiorespiratory fitness as a quantitative predictor of all-cause mortality and cardiovascular events in healthy men and women: a meta-analysis. JAMA. 2009;301(19):2024-2035.
Mandsager K, Harb S, Cremer P, et al. Association of cardiorespiratory fitness with long-term mortality among adults undergoing exercise treadmill testing. JAMA Netw Open. 2018;1(6):e183605.
Kokkinos P, Faselis C, Samuel IBH, et al. Cardiorespiratory fitness and mortality risk across the spectra of age, race, and sex. J Am Coll Cardiol. 2022;80(6):598-609.
Kokkinos P, Faselis C, Samuel IBH, et al. Changes in cardiorespiratory fitness and survival in patients with or without cardiovascular disease. J Am Coll Cardiol. 2023;81(12):1137-1147.
Lang JJ, Prince SA, Merucci K, et al. Cardiorespiratory fitness is a strong and consistent predictor of morbidity and mortality among adults: an overview of meta-analyses representing over 20.9 million observations from 199 unique cohort studies. Br J Sports Med. 2024;58(10):556-566.
Mancini DM, Eisen H, Kussmaul W, Mull R, Edmunds LH Jr, Wilson JR. Value of peak exercise oxygen consumption for optimal timing of cardiac transplantation in ambulatory patients with heart failure. Circulation. 1991;83(3):778-786.
Stevenson LW, Stevenson WG, Middlekauff HR, et al. Improving survival for patients with advanced heart failure: a study of 737 consecutive patients. J Am Coll Cardiol. 1995;26(6):1417-1423.
Heidenreich PA, Bozkurt B, Aguilar D, et al. 2022 AHA/ACC/HFSA guideline for the management of heart failure: a report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. J Am Coll Cardiol. 2022;79(17):e263-e421.
Kaminsky LA, Imboden MT, Arena R, Myers J. Reference standards for cardiorespiratory fitness measured with cardiopulmonary exercise testing using cycle ergometry: data from the Fitness Registry and the Importance of Exercise National Database (FRIEND) Registry. Mayo Clin Proc. 2017;92(2):228-233.
Topol E. The flawed VO₂ max craze. Ground Truths (Substack). February 23, 2026.
Kim Y, White T, Wijndaele K, et al. The combination of cardiorespiratory fitness and muscle strength, and mortality risk. Eur J Epidemiol. 2018;33(10):953-964.

What Your Handshake Tells Me About Your Heart

Jake Kelly MD — Wed, 27 May 2026 11:00:25 GMT

Your handshake just told me more about your heart than your cholesterol panel.

This may sound too good to be true. It isn’t.

It Started With My Own Hands

Confession: I love doing pull-ups.

In medical school, I purchased a door hang to learn to do pull-ups. At that time, I could not perform a single one. Now, on a good day, I can do 30 pull-ups. I hex bar deadlift regularly. I run or bicycle daily and routinely do push-ups and planks. By most measures of fitness, I’m doing well.

When I tested my grip strength, my number came back average. Not above average. Not strong. Average. For someone who focuses on exercise as consistently as I do, this metric sat wrong.

I’m a cardiologist. I spend my days measuring things that matter: ApoB, coronary artery plaque & calcium, cardiorespiratory fitness (VO2 max), blood pressure. I’ve building a clinical practice focused on detecting cardiovascular risk years before it becomes cardiovascular disease.

Here was my grip strength, the one metric I’ve never taken seriously, telling me something I didn’t want to hear.

This finding sent me searching into the literature. What I learned changed how I think about strength, risk, and what we should be measuring in every patient who walks through the door.

The Study That Changed How I Think About Strength

In 2015, a study in The Lancet forced a rethinking of what we measure and why.

The PURE study followed 142,861 adults across 17 countries. The investigators measured grip strength with a handheld dynamometer, then tracked who had heart attacks, strokes and who died over four years.

For every 5 kg reduction in grip strength, cardiovascular death increased by 17%. All-cause mortality by 16%. Myocardial infarction by 7%. Stroke by 9%. All after adjusting for age, sex, smoking, diabetes, BMI, prior heart disease and a dozen other variables.

When the investigators compared grip strength head-to-head against systolic blood pressure, grip was the stronger predictor of death. Per standard deviation: grip HR 1.37, systolic blood pressure HR 1.15. For cardiovascular death: grip HR 1.45, blood pressure HR 1.15.

Blood pressure was better at predicting who would develop a new cardiovascular event. Grip strength was better at predicting who would die from one.

Low grip doesn’t just flag who’s accumulating plaque. It flags who lacks the physiological reserve to survive a cardiovascular crisis when it arrives.

This Isn’t One Study

The UK Biobank replicated the finding in over 500,000 adults. Per 5 kg lower grip, cardiovascular mortality increased 19 to 22 percent depending on sex. The effect was actually stronger in younger adults, which means this isn’t just a geriatric marker. If you’re 45, your grip strength may be telling you more than if you’re 75.

A meta-analysis of heart failure patients found that low grip strength doubled the risk of death (RR 2.04) with essentially zero heterogeneity across studies.

One finding from the UK Biobank deserves special attention: obese individuals with high grip strength had lower mortality than non-obese individuals with low grip strength. Strength modified the mortality risk of obesity.

A 2026 systematic review of 94 cohort studies confirmed it across every outcome examined: cardiovascular disease, diabetes, dementia, disability. The highest grip strength was associated with 27% lower odds of cardiovascular disease.

The cumulative sample size across these studies exceeds 3.5 million people, spanning all 6 inhabited continents.

This is not a signal that needs replication. It needs translation into practice.

A Window Into the Whole System

Grip strength isn’t just about your forearms. It’s a window into total-body physiological capacity and understanding why requires thinking about what muscle does beyond moving bones.

Metabolic engine: Skeletal muscle is the largest insulin-sensitive organ in the body and the primary site of glucose disposal. Higher grip strength correlates with lower fasting glucose, lower insulin resistance, lower triglycerides and higher HDL. The entire cardiometabolic profile that drives atherosclerosis.

Vascular health: Grip strength is significantly correlated with central arterial stiffness, a direct measure of vascular aging. Each 5 kg decrease in grip is associated with a measurable increase in aortic pulse wave velocity. Stiffer arteries. Less compliant vasculature.

Inflammatory burden: Stronger people have lower CRP, lower TNF-alpha, lower leptin. This relationship appears as early as childhood and persists across the lifespan. Muscle is an endocrine organ. It secretes myokines during contraction that actively suppress systemic inflammation. Weak muscle doesn’t just fail to produce these signals. Its absence creates an inflammatory vacuum.

Mitochondrial reserve: Skeletal muscle is the body’s largest reservoir of mitochondria. Grip strength, as a proxy for total muscle quality, reflects mitochondrial density and function: the capacity to generate ATP under stress. When a cardiovascular event occurs, survival depends on the body’s ability to maintain cardiac output, perfuse tissues and tolerate metabolic stress. This capacity lives in the mitochondria. The mitochondria live in the muscle.

Neuromuscular integrity: Grip is not a pure measure of muscle mass. The European consensus on sarcopenia (EWGSOP2) prioritizes grip strength over muscle mass as the first diagnostic step because strength is more predictive of adverse outcomes than size. Grip reflects motor unit recruitment, neuromuscular junction quality and coordination. It’s a neural signal as much as a muscular one.

This is why grip strength’s association with cardiovascular mortality was twice as strong as its association with incident cardiovascular events. Low grip doesn’t just predict who develops disease. It predicts who can’t survive it.

Physiological reserve: This is the concept. Grip strength is its most accessible measurement. In the UK Biobank, CRF and grip strength were independently associated with heart failure (per 1 SD: CRF reduced risk by 50%, grip by 35%), confirming these are overlapping but distinct systems converging on the same truth: capacity protects.

But Is It Causal?

Maybe weak grip is just a marker of being sick. Reverse causation dressed up as prediction.

The Mendelian randomization data settled this for me. MR uses genetic variants as instruments to test causality. Because genes are assigned at conception, randomly with respect to confounders, they function as a natural experiment.

Multiple independent analyses converge: genetically determined higher grip strength is protective against coronary artery disease (OR 0.69), myocardial infarction and atrial fibrillation (OR 0.75). The effect is not mediated through traditional risk factors like LDL, BMI or blood glucose.

The decisive finding: bi-directional MR tested causality in both directions. Low grip strength causally increased cardiovascular risk. Cardiovascular disease did not causally reduce grip strength. The arrow points one way.

Grip likely functions as both cause and barometer. It predicts mortality from nearly every cause, not just cardiovascular disease. But a measurement that out predicts blood pressure, survives exhaustive confounder adjustment, shows dose-response across 3.5 million people and passes Mendelian randomization deserves a place in routine assessment.

The Trajectory Problem

A single measurement is valuable. Serial measurements are more valuable still. The Hertfordshire Ageing Study found an 8-fold difference in death rates between adults with high, stable grip and those with low, declining grip.

Grip peaks in the 30s and 40s. After 50, men lose roughly 0.5 kg per year, women 0.3 kg. After 70, the decline accelerates to nearly 9% per year in men. This creates a clinical window: between 40 and 65, you can measure, establish a trajectory and intervene before the decline compresses physiological reserve.

We don’t wait for heart failure to measure VO2 max. Grip strength deserves the same upstream approach.

Where Do You Stand?

The dose-response relationship between grip strength and mortality is close to linear between 26 and 50 kg, with no clear floor below which the benefit disappears. Here are the reference ranges:

The EWGSOP2 thresholds for probable sarcopenia (below 27 kg in men, 16 kg in women) were derived from elderly populations and may underperform for younger adults. What matters most isn’t the absolute number. It’s where you sit relative to your age-matched peers and whether that number is stable, climbing, or falling.

How I Measure It

I use a calibrated Jamar hydraulic dynamometer in the second handle position. Seated, shoulder neutral, elbow at 90 degrees, forearm in neutral rotation. Three attempts per hand, 30 to 60 seconds of rest between trials. I record the maximum value.

The test takes less than two minutes. No blood draw, no imaging, no fasting.

No major cardiovascular guideline (AHA, ACC, ESC) currently includes grip strength in formal risk assessment. The AHA’s 2023 Scientific Statement on Resistance Exercise acknowledges the prognostic value but stops short of a recommendation. Adding grip to existing risk scores improves discrimination modestly, comparable to other novel biomarkers.

That gap between evidence and guideline adoption is exactly the kind of space where prevention-oriented clinicians can move first.

Building What You’ve Measured: The Grip Training Toolkit

After seeing my own numbers, I started training grip deliberately.

Resistance training reduces all-cause mortality by 15 to 21 percent and cardiovascular mortality by 17 to 19 percent in meta-analyses. The MR data support the biological plausibility that building strength is genuinely protective, not just associated with protection.

The forearm contains over 20 muscles. Different tools stress different aspects of grip: crushing strength, pinch strength, support strength, wrist stability, and rotational control. Here’s what I’ve learned from the evidence and my own hands.

Your Main Lifts Already Build Grip (If You Let Them)

The simplest way to improve grip is to stop bypassing it during the lifts you’re already doing.

Deadlifts, conventional or hex bar, are the foundation. Every heavy pull is a grip endurance test. My approach: pull strapless until grip fails, then strap up for remaining volume (the last 1-2 sets).

Rows and pull-ups load the grip in a flexed position under sustained tension. Thirty pull-ups demand that your fingers maintain force output through fatigue, exactly the kind of grip endurance that correlates with the mortality data (consider using chalk extend and further challenge your grip endurance).

Farmer carries changed my grip the most. Pick up heavy dumbbells or a trap bar and walk. Maintain crushing grip while your core stabilizes, your cardiovascular system works to perfuse loaded muscles and your entire posterior chain sustains an isometric contraction. A full-system stress test disguised as a walk across the gym. I program these at 50 to 75 percent of bodyweight per hand, 40 to 100 meters, twice a week when possible.

I train in Alaska. Farmer carries are easier to program in summer, when longer distances and heavier outdoor carries become practical. Winter pushes grip work indoors. Seasonality matters when you’re designing a training year.

Any dumbbell or kettlebell exercise where you hold the weight builds grip as a byproduct. Goblet squats, dumbbell lunges, single-arm rows, kettlebell swings. If the weight is in your hand and gravity is pulling it away, your forearm is working. Choosing dumbbells over machines for accessory work turns every set into grip training without adding a single minute to your session.

Grip Tools: Old School and New School

Hand grippers (Captains of Crush, etc.) train crushing grip with clear progressive overload. Simple, portable, effective. The limitation: one dimension of grip through a limited range of motion.

Thick bar training (Fat Gripz or a towel wrapped around the bar) forces the hand open wider, recruiting more motor units per rep. Old school and effective.

Twist grip trainers stress the forearm through pronation and supination, not just flexion. This mimics real-world grip demands (opening a jar, turning a wrench) and loads muscles that standard grippers miss.

Wrist curls and reverse curls address the flexor-extensor imbalance most people carry. Useful for elbow health and joint balance.

Finger extensors and band work target the opposing muscle group. The research focuses on flexion, but extensor health matters for injury prevention and long-term hand function.

Digital dynamometers double as training tools and measurement devices. Real-time force feedback turns every session into a data point.

How I Program Grip Work

These aren’t separate workouts. They’re finishers: the last set on lifting days, an intermediate exercise between harder movements, or 5 to 10 minutes at the end of a cardio session. The barrier to entry is close to zero.

Dead hangs. Hang from a pull-up bar for time. Start with 30 seconds. Progress to 60, then add weight. Trains support grip and decompresses the spine. The simplest and one of the most effective grip exercises.

Plate pinches. Two weight plates together, smooth sides out, held between thumb and fingers for time. Pinch grip is often the weakest pattern. Start light.

Dead bar holds. Load a barbell to your working deadlift weight. Lift to lockout. Hold. This bridges grip training and your main lifts.

Towel hangs. Drape a towel over a pull-up bar and hang from the towel ends. Dramatically increases grip demand over a standard hang and builds thick-grip strength without any extra equipment.

Rice bucket digs. Fill a 5-gallon bucket with rice. Plunge your hands in and open, close, twist, extend against resistance. This is old-school rehabilitation work that trains every grip pattern simultaneously, including finger extension and wrist rotation. Low-tech, surprisingly taxing, and excellent for joint health.

The minimum effective dose: 2 to 3 of these finishers per week, layered on top of compound lifts performed without straps. Pick two exercises, rotate them and track your progress.

The Sixth Vital Sign

Blood pressure, heart rate, weight, temperature, respiratory rate. Five vital signs, unchanged for decades.

Grip strength predicts cardiovascular death more strongly than the most commonly measured of those five. It integrates metabolic, inflammatory, vascular and neuromuscular health into a single number. It captures something no blood test or imaging study currently quantifies: the body’s total physiological reserve.

I started measuring my own grip because my number bothered me. The literature convinced me to measure everyone’s. Training it made me a better clinician, because I now understand, in my own hands, the gap between being fit and being strong.

Your handshake is telling you something.

Test your grip. Know your number. Then build it.

I’d like to hear where your number lies. If you’ve tested your grip strength, share your number and your age in the comments. If you’re a clinician, measure 10 patients this week and tell me what surprised you.

If this changed how you think about strength and cardiovascular risk, share it with someone who could use the signal.

For individuals who want this integrated approach, where grip strength, metabolic architecture and longevity cardiology are treated as a single system: jakekellymd.com.

Jake Kelly, MD, MHS, FACC is a board-certified performance and preventive cardiologist, ACC Governor for Alaska and founder of ROOL Health, a cardiometabolic longevity practice in Anchorage.

Legal Medical Disclaimer

The information provided on this site including articles, newsletters, social-media posts, videos, and downloadable resources is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment. Your use of this content does not establish a physician–patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement. Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.

References

Leong, D. P., Teo, K. K., Rangarajan, S., Lopez-Jaramillo, P., Avezum, A., Jr, Orlandini, A., Seron, P., Ahmed, S. H., Rosengren, A., Kelishadi, R., Rahman, O., Swaminathan, S., Iqbal, R., Gupta, R., Lear, S. A., Oguz, A., Yusoff, K., Zatonska, K., Chifamba, J., ... Prospective Urban Rural Epidemiology (PURE) Study investigators. (2015). Prognostic value of grip strength: findings from the Prospective Urban Rural Epidemiology (PURE) study. The Lancet, 386(9990), 266–273.
Celis-Morales, C. A., Welsh, P., Lyall, D. M., Steell, L., Petermann, F., Anderson, J., Iliodromiti, S., Sillars, A., Graham, N., Mackay, D. F., Pell, J. P., Gill, J. M. R., Sattar, N., & Gray, S. R. (2018). Associations of grip strength with cardiovascular, respiratory, and cancer outcomes and all cause mortality: prospective cohort study of half a million UK Biobank participants. BMJ, 361, k1651.
Wang, Y., Pu, X., Zhu, Z., & Chen, B. (2023). Handgrip strength and the prognosis of patients with heart failure: a meta-analysis. Clinical Cardiology, 46(12), 1474–1483.
Kim, Y., Wijndaele, K., Lee, D. C., & Sharp, S. J. (2017). Independent and joint associations of grip strength and adiposity with all-cause and cardiovascular disease mortality in 403,199 adults: the UK Biobank study. The American Journal of Clinical Nutrition, 106(3), 773–781.
Marín-Jiménez, N., Bizzozero-Peroni, B., Molina-Garcia, P., Ortega, F. B., et al. (2026). Clinical importance of simple muscular fitness tests to predict long-term health conditions: a systematic review and meta-analysis of 94 cohort studies. British Journal of Sports Medicine, 60(4).
König, M., Buchmann, N., Seeland, U., Spira, D., Demuth, I., Steinhagen-Thiessen, E., & Norman, K. (2021). Low muscle strength and increased arterial stiffness go hand in hand. Scientific Reports, 11(1), 13559.
Cruz-Jentoft, A. J., Bahat, G., Bauer, J., Boirie, Y., Bruyère, O., Cederholm, T., Cooper, C., Landi, F., Rolland, Y., Sayer, A. A., Schneider, S. M., Sieber, C. C., Topinkova, E., Vandewoude, M., Visser, M., Zamboni, M., & Writing Group for the European Working Group on Sarcopenia in Older People 2 (EWGSOP2). (2019). Sarcopenia: revised European consensus on definition and diagnosis. Age and Ageing, 48(1), 16–31.
Sillars, A., Celis-Morales, C. A., Ho, F. K., Petermann, F., Welsh, P., Gill, J. M. R., Sattar, N., & Gray, S. R. (2019). Association of fitness and grip strength with heart failure: findings from the UK Biobank population-based study. Mayo Clinic Proceedings, 94(11), 2230–2240.
Tikkanen, E., Gustafsson, S., Amar, D., Shcherbina, A., Waggott, D., Ashley, E. A., & Ingelsson, E. (2018). Biological insights into muscular strength: genetic findings in the UK Biobank. Scientific Reports, 8(1), 6451.
Liu, X., Wang, Y., Wang, Z., & Zhang, Y. (2023). Association between sarcopenia-related traits and cardiovascular diseases: a bi-directional Mendelian randomization study. Frontiers in Endocrinology, 14, 1138816.
Syddall, H. E., Westbury, L. D., Dodds, R., Dennison, E. M., Cooper, C., & Sayer, A. A. (2017). Mortality in the Hertfordshire Ageing Study: association with level and loss of hand grip strength in later life. Age and Ageing, 46(3), 407–412.
Lopez-Bueno, R., Andersen, L. L., Koyanagi, A., Casaña, J., Smith, L., Borrallo-Riego, Á., del Pozo Cruz, J., & del Pozo Cruz, B. (2022). Thresholds of handgrip strength for all-cause, cancer, and cardiovascular mortality: a systematic review with dose-response meta-analysis. Ageing Research Reviews, 80, 101737.

The 16-Hour Runway

Jake Kelly MD — Wed, 20 May 2026 12:01:55 GMT

Two weeks I wrote about the biology that defends your weight. This week: what it looks like when you change the inputs.

Meet Liam

Liam works on the slope. Two weeks on, two weeks off. On rotation: 6 a.m. to 6 p.m., one hour for lunch. When I first saw him, his day looked like this.

Up at 5:30. Straight to the cafeteria. Eggs, toast, hash browns, coffee. Twelve hours of work with modest walking, about 4,000 steps total. Sedentary lunch in the break room. Off shift at 6, dinner by 6:30. Then TV, phone calls to his wife and kids, internet scrolling, a few snacks taken back to his room. Lights out around 11. Up at 5:30. Repeat for 14 days.

When he came home for his two weeks off, the schedule barely changed. Less walking. No exercise. He enjoyed taking his kids to school but otherwise he meandered through days that looked a lot like his evenings on the slope.

His labs and basic vitals:

Fasting insulin: 17 (early insulin resistance), hba1c 5.8% (pre-diabetes)
Total cholesterol: 220, Triglycerides 220, HDL 37, LDL 142, VLDL 44, Non-HDL 184, Lp(a) 15 nmol/L (reassuringly low)
Blood pressure 139/83, Weight 210 lbs at 5’10” (BMI 30.1)

Every number heading the wrong direction. Not unwell but sliding.

A few years ago I would have started with the following: statin conversation, maybe metformin, dietary counseling. I was not wrong. But now, I focus on the behaviors and habits that set the foundation of health. When I asked Liam how he was sleeping, he paused. Not great. He never feels rested. He has an energy crash every afternoon. Five and a half hours of sleep most nights. His nightlife shows in every line of his lab panel.

I no longer start with Liam’s cholesterol. I start with his day.

5:30 a.m. The Clock Starts

Your body runs two systems that produce sleep. The first is your circadian clock, a 24-hour rhythm governed by the suprachiasmatic nucleus and calibrated primarily by light. The second is sleep pressure: adenosine, a molecule that accumulates in your brain every hour you’re awake. By evening, if you’ve built enough pressure and your circadian clock is properly set, sleep comes. If you haven’t, it doesn’t.

Liam’s first problem was light. On the slope in winter, he walked from his sleeping bunk to the cafeteria and then to the work site in darkness. There is no bright (or natural) light exposure in the first hour of waking. His circadian clock was never properly anchored. A 2025 AHA scientific statement on circadian health now formally recognizes morning bright light as a tool for metabolic regulation: it synchronizes circadian rhythms, promotes daytime alertness and is associated with lower body fat and appetite.

The fix was simple. On clear mornings: 10 to 15 minutes outside before or after breakfast, no sunglasses. On dark mornings: a 10,000-lux light therapy lamp at his desk as soon as he wakes up. The goal isn’t duration of light. It’s timing. Bright light in the first hour anchors the circadian clock that will make sleep possible 16 hours later.

The international expert consensus on indoor light (2022) now uses a metric called melanopic EDI, melanopic equivalent daylight illuminance, as the standard for how light affects the circadian system. The practical translation: most indoor lighting is too dim in the morning and too bright at night. Your office at 300 lux isn’t enough to properly set the clock. Evening scrolling your Iphone at 100 lux in bed is enough to delay it.

6:00 a.m. Breakfast is a Metabolic Decision

Liam’s cafeteria breakfast was carbohydrate-heavy: toast, hash browns, juice and coffee. He was hungry by 9:30 and crashing by 2 p.m. We restructured breakfast around protein.

30+ grams of protein at breakfast. Eggs. We added Greek yogurt or a protein shake on the side. Carbs stayed too, but moved to whole grains and fruit. The fiber target: 30 grams or more across the day. These choices are not about caloric restriction. It is about satiety. Feelin full. Protein and fiber slow gastric emptying, blunt glucose spikes and sustain energy through the morning.

The caffeine conversation. Liam drank coffee at breakfast (fine) and another cup mid-afternoon (not fine). Caffeine’s half-life averages 2.5 to 4.5 hours. A 2023 meta-analysis of 24 studies showed that a single cup of coffee needs about 8.8 hours before bed to avoid measurable reductions in total sleep time. A pre-workout supplement at 200 mg? Closer to 13 hours. These are averages. CYP1A2 polymorphisms, oral contraceptive use and smoking can prolong caffeine’s half-life. Some people are slow metabolizers. A 2 p.m. coffee for a 10 p.m. bedtime was costing Liam sleep he didn’t know he was losing.

We moved his cutoff to one (or two) cup(s) at breakfast. No caffeine after 8 a.m.

12:00 p.m. The Midday Reset

Liam’s lunch was sedentary. Break room, food, phone, back to work. This is where most people miss an opportunity to improve their health.

We made one change: eat first, then walk for 15 to 20 minutes outside (or in the compound). Not a full workout. A walk.

Walking mid-day does three things simultaneously. It builds adenosine (sleep pressure accumulates faster with physical activity than with sitting). It clears the cortisol that’s been building through the morning. It also provides a second dose of bright light that reinforces the circadian signal set at dawn.

Liam added hourly standing breaks during the afternoon: 2 to 5 minutes of movement each hour. Bathroom, water, a short walk. This is NEAT, non-exercise activity thermogenesis. The energy burned through daily movement that isn’t structured exercise. NEAT varies by up to 2,000 calories per day between individuals, doesn’t trigger compensatory hunger and accumulates without requiring willpower or equipment.

For Liam, these small additions moved his daily steps from 4,000 to 8,000 over the first month. No gym required. Organized movement through a day that previously was spent sitting.

4:30 p.m. The Training Window

After his shift, Liam had roughly 90 minutes before dinner. We put a gym session here: 20 minutes cardio, 15 minutes strength. He started at 3 days per week. Eventually this habit turned into a daily routine most days (6/7 on average).

The timing matters. Exercise raises core body temperature. You need that temperature to drop for sleep onset. Training in the late afternoon gives 3 to 5 hours for the body to cool before bed, which is ideal. Training at 9 p.m. can work against the runway we were building.

Strength training wasn’t optional. During any caloric deficit, roughly 25% of weight lost is lean tissue unless you actively maintain it. For a man Liam’s age, 2 to 3 sessions per week with moderate resistance preserves muscle, maintains metabolic rate and supports the glucose disposal that his insulin resistance needed.

6:30 p.m. The Kitchen Closes

Dinner by 6:30. Then the kitchen is closed.

Liam’s old pattern: dinner, then snacks through the evening. Chips, crackers, whatever was in his room. Those calories weren’t feeding hunger. They were feeding boredom and they were wrecking his sleep.

Eating late raises core body temperature at exactly the wrong time. It blunts melatonin onset. It forces the gut into active digestion when the body should be entering recovery mode. A three-hour buffer between the last bite and lights out gives the system time to transition.

This was also where Liam’s evening chronotype was hurting him. The 2026 ADA Standards of Care now recognize evening chronotype as an independent risk factor for type 2 diabetes and is associated with 2.5-fold higher odds of T2DM independent of sleep duration. Liam wasn’t a natural night owl. He’d trained himself into one through a decade of late-night screen time and snacking. Moving dinner earlier and closing the kitchen after was the first step in pulling his sleep time back.

9:00 p.m. The Phone Goes Down

This is the hour most patients resist. Liam’s evenings on the slope were built around his phone: calling his family, watching videos and scrolling news. It was his only connection to home. I wasn’t going to take that away. We reorganized it.

Family calls moved to 7:30, right after dinner. The scrolling, the news, the videos: done by 9. We also moved the phone and his charger to the corner of the room, not the nightstand (as above).

The reason isn’t only blue light, though even ~100 lux of typical indoor lighting can suppress melatonin and delay the circadian clock by up to 1.5 hours. The bigger problem is content. Scrolling activates the sympathetic nervous system. News spikes cortisol. Social media delays sleep onset in ways that aren’t fully explained by light exposure alone.

Liam’s last hour before bed became quiet. He read. He wrote a few notes about the day. Some evenings he just sat. The last hour isn’t empty space. It’s the runway. And what you do in it determines how you land.

We also addressed the room itself. I told Liam what I tell every patient: treat sleep with reverence. The condition, the ritual, the environment. You wouldn’t perform surgery in a noisy room with the lights on. Sleep is a performance too. Room temperature: 64 to 68 degrees. Quiet. Dark. Move the phone across the room, not within arm’s reach.

10:00 p.m. Lights Out

Liam’s old bedtime was 11. His new bedtime: 10. Up at 5:30. That’s 7.5 hours in bed.

This single change, moving bedtime back one hour, may have been the most consequential. Not only is there extra sleep, but the extra hour contains something even more regenerative.

Sleep cycles through stages every ~90 minutes. Deep sleep (N3) concentrates in the first half of the night: this is when the immune system primes, growth hormone releases, tissue repairs and the glymphatic system clears amyloid-beta and tau from the brain. The 2024 AHA scientific statement on sleep and brain health placed glymphatic clearance at the center of dementia prevention. Without adequate deep sleep, the waste products that become Alzheimer’s pathology over decades aren’t cleared.

REM sleep back-loads toward morning: memory consolidation, emotional processing. Go to bed late and you compress the front-loaded deep sleep. Wake too early and you lose the REM cycles you haven’t completed. Liam’s 5.5-hour nights were stealing from both ends.

The Home Schedule: Holding the Line

Two weeks off with no externally imposed structure is where most patients lose everything they have built. Liam’s approach at home lost routine: later bedtimes, no movement, no schedule.

Our solution: same architecture, two adjustments.

30 minutes later to bed (10:30, aligned with his spouse)
30 minutes later to rise (6:00)
Same 7.5-hour window. Same circadian rhythm within 30 minutes of the slope schedule.
Morning: kids to school, then walk or gym.
Same meal structure, same protein targets.
Evening: same runway.

The 30-minute shift matters because of what I consider the most important sleep finding of the past five years.

A 2024 UK Biobank study tracked 60,977 people with over 10 million hours of wrist-worn accelerometry. It found that sleep regularity predicted mortality better than sleep duration. People in the most consistent quintile had 20 to 48% lower all-cause mortality compared to the least regular. Adding sleep duration to the model didn’t significantly improve prediction. The 5th percentile of regularity carried a 53% higher mortality risk than the median.

The 2025 AHA scientific statement on multidimensional sleep health now lists regularity as a distinct dimension with independent cardiometabolic implications.

The practical message: going to bed at the same time every night may matter more than how long you sleep. Liam’s two-week rotation had been creating a cycle of circadian disruption every 14 days, a kind of self-imposed jet lag. By keeping his schedule within a 30-minute window between on-slope and off-slope, we eliminated the re-adaptation his biology had been fighting through every month.

Six Months Later

Liam started tracking with a WHOOP band. Within four weeks: higher sleep efficiency, more time in restorative stages, more consistent overnight heart rate variability. He reported feeling rested for the first time in years. The afternoon crash disappeared.

A note on wearables: a 2025 meta-analysis of 24 validation studies found that consumer wrist-worn devices overestimate total sleep time by about 17 minutes and underestimate wake-after-sleep-onset by about 13 minutes. Sleep/wake detection is reasonably accurate (above 95% sensitivity). Sleep stage classification is not. In head-to-head PSG validation, Oura Ring and WHOOP had the highest multi-state agreement with polysomnography. What to track: weekly average total sleep time, bedtime consistency, resting heart rate trends. What to mostly ignore: night-to-night deep sleep and REM percentages.

Liam tracked what mattered: consistency and duration. Both improved.

His daily steps held at 8,000. His weight moved from 210 to 190. Twenty pounds. A 9.5% reduction.

We repeated labs. Roughly 30% improvement across the board:

Fasting insulin: 17 to 12
HbA1c: 5.8 to 5.4
Triglycerides: 220 to 155
HDL: 37 to 44
LDL: 142 to 118
Non-HDL: 184 to 131
Blood pressure: 139/83 to 124/76

No statin. No metformin. No GLP-1 agonist. A restructured day built around sleep.

The connection between his sleep and his insulin is not speculative. A 2024 randomized crossover trial found that just six weeks of 1.5 hours less sleep per night raised fasting insulin and HOMA-IR without any change in body weight. The 2026 ADA Standards of Care now formally name short sleep as a modifiable risk factor for type 2 diabetes, with up to 50% increased risk at the extremes. Liam’s fasting insulin of 17 wasn’t just about diet. It was about 5.5 hours of sleep, 14 consecutive nights, repeating every month, for years.

His spouse noticed. She adopted a similar schedule. Her sleep improved. She started joining him at the gym on his weeks home. The cascade moved through the household. Well done, Liam.

When the Day Isn’t Enough

Not every patient is Liam. Some patients have optimized the day and still can’t sleep. Two things to rule out.

Chronic insomnia. The first-line treatment is not a sleeping pill. It’s cognitive behavioral therapy for insomnia (CBT-I). A 2024 network meta-analysis confirmed CBT-I produces better long-term remission than pharmacotherapy. About 50% of patients reach remission after 6 to 8 weeks. Digital CBT-I programs (Sleepio, SHUTi, Somryst) now have effect sizes comparable to in-person therapy and outperform medication in head-to-head comparisons. If you’ve been on a sleep medication for years, ask about CBT-I.

Obstructive sleep apnea. If you snore, wake unrefreshed, have a thick neck, or your partner has watched you stop breathing: get evaluated. The STOP-BANG questionnaire is the most sensitive primary care screen. The landscape here changed in 2024: tirzepatide is now FDA-approved for moderate-to-severe OSA in adults with obesity. In the SURMOUNT-OSA trials, it reduced the apnea-hypopnea index by 50 to 59%, and roughly half of patients reached levels where CPAP would no longer be indicated. A 2026 network meta-analysis supports a combined approach: CPAP for airway control, GLP-1 receptor agonists for the underlying metabolic driver. This is a paradigm shift. We’re no longer just treating the airway. We’re treating the disease.

The Prescription

I write sleep prescriptions on a real prescription pad. Every patient gets one (almost).

Rx: Sleep. Dose: 7 to 8 hours. Frequency: Nightly. Refills: Unlimited. Expiration: None.

The data from 2024 through 2026 has made this non-negotiable. Short sleep carries a 14% higher all-cause mortality. Objectively measured sleep under 7 hours raises the risk of heart attack, heart failure, and atrial fibrillation and Mendelian randomization confirms the relationship is causal, not just correlated. Insomnia is associated with a 53% increase in dementia risk. Sleep regularity may predict mortality better than duration.

Sleep is something you work for throughout the day. It is also something that benefits you throughout the day. The work and the reward are the same thing.

Same seriousness as a statin. Larger upside. Cadence is medicine.

If you found this article valuable, share it with someone who is running out of runwayor someone whose health could be transformed by shifting their day.

For individuals who want this integrated approach: where sleep, metabolic architecture and longevity cardiology are treated as a single system visit jakekellymd.com.

Jake Kelly, MD, MHS, FACC is a board-certified performance and preventive cardiologist, ACC Governor for Alaska and founder of ROOL Health, a cardiometabolic longevity practice in Anchorage.

Legal Medical Disclaimer

The information provided on this site including articles, newsletters, social-media posts, videos, and downloadable resources is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment. Your use of this content does not establish a physician–patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement. Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.

References

American Diabetes Association. (2026). Standards of Care in Diabetes: 2026. Diabetes Care, 49(Suppl 1), S1–S312.
Brown, T. M., Brainard, G. C., Cajochen, C., Czeisler, C. A., Hanifin, J. P., Lockley, S. W. & Wright, K. P. (2022). Recommendations for daytime, evening, and nighttime indoor light exposure. PLoS Biology, 20(3), e3001571.
Cribb, L., Sha, R., Yiallourou, S., Grima, N. A., Cavuoto, M., Baril, A. A., & Pase, M. P. (2023). Sleep regularity and mortality: A prospective analysis in the UK Biobank. eLife, 12, RP89975.
Gardiner, C., Weakley, J., Burke, L. M., Roach, G. D., Sargent, C., Maniar, N., ... & Halson, S. L. (2023). The effect of caffeine on subsequent sleep: A systematic review and meta-analysis. Sleep Medicine Reviews, 69, 101764.
Gottesman, R. F., et al. (2024). Impact of sleep disorders and disturbed sleep on brain health: A scientific statement from the American Heart Association. Stroke, 55(3), e89–e103.
Knutson, K. L., et al. (2025). Role of circadian health in cardiometabolic health and disease risk: A scientific statement from the American Heart Association. Circulation, 151(4), e45–e61.
Malhotra, A., Bednarik, J., Fietze, I., Heinzer, R., Jenkins, M. R., Mansukhani, M. P., ... & SURMOUNT-OSA Investigators. (2024). Tirzepatide for the treatment of obstructive sleep apnea and obesity. New England Journal of Medicine, 391(1), 21–31.
Morin, C. M., & Buysse, D. J. (2024). Management of insomnia. New England Journal of Medicine, 390(12), 1111–1121.
Ungvari, Z., et al. (2025). Imbalanced sleep increases mortality risk by 14-34%: A meta-analysis. GeroScience, 47(3), 4545–4566.
Windred, D. P., Burns, A. C., Lane, J. M., Saxena, R., Rutter, M. K., Cain, S. W., & Phillips, A. J. K. (2024). Sleep regularity is a stronger predictor of mortality risk than sleep duration: A prospective cohort study. Sleep, 47(1), zsad253.
Zuraikat, F. M., Laferrère, B., Cheng, B., Scaccia, S. E., Cui, Z., Aggarwal, B., Jelic, S., & St-Onge, M. P. (2024). Chronic insufficient sleep in women impairs insulin sensitivity independent of adiposity changes: Results of a randomized trial. Diabetes Care, 47(1), 117–125.

Rx: Sleep

Jake Kelly MD — Tue, 12 May 2026 15:53:59 GMT

Joan came in this week for a murmur evaluation. She was doing well: exercising, eating thoughtfully, in a strong marriage. As we were wrapping up the visit, I asked the question I ask every patient: how is your sleep?

Her answer: terrible.

This is what I hear frequently from seemingly dialed in patients. Patients have thoughtful health forward plans on diet, exercise and all the visible aspects of their lie. Sleep is where the truth comes out and it is the part of health maintenance that I am coming to care most about, because sleep sits upstream of nearly everything I treat: metabolic, cardiorespiratory fitness, circulation and cognitive health.

I told Joan about the two questions I ask myself every morning:

Did I have a good night’s sleep?
How do I make sure tonight is as good, or better?

This is my sleep framework. Last night’s sleep is the only outcome that matters. Tonight is the only thing you can change. And the answer to question #2 almost never happens at bedtime. It happens hours earlier, in the choices most people don’t realize they’re making throughout their daily lives.

We walked through her pattern. She was disciplined on caffeine: none after lunch. But she was eating right up to bedtime and scrolling on her phone in bed. Two simple adjustments. No medication, no supplements. This is where most patients are losing ground on good (better?) sleep.

I’ve written about sleep before. This is what I’d say now.

Rx: Sleep Dose: 7–8 hours Frequency: Nightly Refills: Unlimited Expiration: None

I write that on a real prescription pad. Every patient gets one. I do it because the data on sleep, 2024 through 2026, has become so strong it feels like malpractice not to take it seriously. The AHA elevated sleep into Life’s Essential 8 in 2022, then released two more scientific statements (2024), 2025 on cardiometabolic health explicitly recognizing not just sleep duration but sleep regularity as independent predictors of disease and death. The FDA approved tirzepatide for obstructive sleep apnea last year. The 2026 ADA Standards of Care now name short sleep as a modifiable risk factor for type 2 diabetes.

Sleep is the most underprescribed intervention in medicine. Let’s fix that.

I Call Sleep a Religious Experience

I mean it. The condition. The ritual. The reverence.

You wouldn’t expect to perform surgery in a noisy room with the lights on and your phone buzzing. You shouldn’t expect to perform sleep that way either. Mood, setting and pattern do most of the work towards improving and having good sleep. The drug part, the actual physiology, only kicks in when you’ve built the runway.

The 6+ / 3 / 2 / 1 Sleep Runway

Most of what people get wrong about sleep happens in the hours before they get into bed. Think of bedtime like landing a plane. You don’t pull the nose up at the last second. You set up a runway.

6+ hours before bed: No caffeine. This is the rule I get the most pushback on. The pharmacology is unforgiving. Caffeine’s half-life in adults averages 2.5–4.5 hours, but a 2023 dose-response meta-analysis showed that a standard cup of coffee (~107 mg) needs about 8.8 hours before bed to avoid measurable reductions in total sleep time. A 200 mg pre-workout? Closer to 13 hours. CYP1A2 genetics, oral contraceptives and smoking status shift the pharmacodynamics further back. A 2 p.m. coffee for a 10 p.m. bedtime is reducing your evening sleep quality.

3 hours before bed: No food. Eating late raises core body temperature exactly when it needs to drop, blunts melatonin onset and forces the gut into digestion when the body should be entering recovery mode. Joan was eating dinner past 8. We moved it earlier. The cleaner the runway, the easier and deeper we move into slow-wave sleep.

2 hours before bed: No water. The bladder doesn’t care how tired you are. Each nocturnal wake-up fragments sleep architecture and fragmentation, not just total sleep time, is what predicts cognitive decline and amyloid deposition.

1 hour before bed: No screens. Partly about blue light (even ~100 lux of typical indoor lighting can meaningfully suppress melatonin) and partly about content. The phone is not neutral. Doomscrolling spikes cortisol. News, email and social media delay sleep onset in ways that aren’t fully explained by light alone.

That last hour is not empty space. It’s where you read, talk to your partner, stretch, sit with your kids or write. The runway clears the obstacles. What you do in the hour is the landing.

The Other Half: Building Sleep Pressure During the Day

I saw a different patient earlier this week with a more significant sleep problem. We didn’t change much about her evening routine. We changed her day.

Most people think about sleep as a 60-minute project that starts at 10 p.m. It isn’t. Sleep is a 16-hour project that starts the moment you wake up.

Here’s why: adenosine, the molecule that makes you sleepy, accumulates in your brain every hour you’re awake. That accumulation is the engine of sleep. If you nap too long, sit in dim light all day, or stay sedentary, you don’t build enough pressure for sleep to come easily. You arrive at bedtime with an empty tank, then wonder why you can’t sleep.

This is the part that’s fascinated me more as I’ve aged: sleep is something you work for throughout the day and it’s also something that benefits you throughout the day. The work and the reward are the same thing. The investment is the return.

The day-side of the protocol:

Morning light within an hour of waking. 10–30 minutes outside, no sunglasses if you can manage it. This sets your circadian clock and starts the adenosine engine.
Move. Walking, training, anything that gets your nervous system out of low-power mode. Movement is fuel for sleep.
Skip the long afternoon nap. Keep naps under 20 minutes if you nap at all. Longer naps blow off your sleep pressure.
Front-load your eating window. Time-restricted eating that closes by 7 p.m. supports both metabolism and sleep.

How Sleep Actually Works

Sleep isn’t passive. Two systems govern it.

Process C (Circadian). Your suprachiasmatic nucleus runs a roughly 24-hour clock that controls alertness, body temperature and hormone release. It’s calibrated by light. Morning bright light advances the clock; evening blue light delays it. The 2022 international consensus on indoor light now uses melanopic equivalent daylight illuminance (melanopic EDI) as the standard. The practical translation hasn’t changed: get bright light in your eyes within an hour of waking, dim the lights after sunset.

Process S (Sleep pressure). Adenosine builds while you’re awake. Sleep clears it. Caffeine blocks adenosine receptors. It doesn’t reduce the pressure, it masks it. When caffeine wears off, the accumulated adenosine is still there waiting.

Inside that 7- to 8-hour window, you cycle through stages every ~90 minutes:

N1, N2: Light sleep, transition phases.
N3 (slow-wave / deep sleep): Front-loaded in the first half of the night. Where the immune system primes, growth hormone releases, tissue repairs, and, critically, the glymphatic system clears amyloid-β and tau from the brain. The 2024 AHA scientific statement on sleep and brain health placed glymphatic clearance at the center of dementia prevention.
REM: Back-loaded toward morning. Memory consolidation, emotional processing.

Cut the night short by going to bed late and you preferentially lose REM. Wake at 4 a.m. and can’t get back to sleep and you lose the deep sleep cycles you haven’t completed. Both are important. Neither is replaceable.

The Evidence

Mortality. A 2025 meta-analysis of 79 cohort studies confirmed the U-shape: short sleep (<7 hours) carries a 14% higher all-cause mortality risk; long sleep (≥9 hours) carries 34% higher risk, more pronounced in women. UK Biobank accelerometry (n = 88,282), using objective measurement rather than self-report, found a 29% higher mortality risk at 5 hours per night. Mendelian randomization suggests the short-sleep association is causal; the long-sleep association is likely confounded by underlying illness.

Cardiovascular disease. Objectively measured sleep under 7 hours raises the risk of incident CVD, MI, heart failure, and atrial fibrillation, with a dose-response curve that steepens below 7 hours. Genetic studies confirm short sleep causes hypertension and coronary artery disease. It isn’t just correlated with them.

Dementia. Insomnia is associated with a 53% increase in dementia risk. Both short and long sleep duration have been linked to at least a doubling of risk. Without adequate deep sleep, the glymphatic system fails to clear the amyloid and tau that, over decades, become Alzheimer’s pathology. A 2024 JAMA Neurology study showed disrupted 24-hour activity rhythms predicted subsequent amyloid-β deposition, especially in APOE4 carriers.

Metabolism. A 2024 randomized crossover trial in women found that just six weeks of 1.5 hours less sleep per night raised fasting insulin and HOMA-IR without any change in body weight. The 2026 ADA Standards of Care now formally recognize sleep duration as a modifiable risk factor for type 2 diabetes, with up to 50% increased risk at the extremes.

The New Story: Sleep Regularity

This is the finding I want every patient to hear. The 2024 Windred et al. UK Biobank study (n = 60,977, >10 million hours of wrist-worn accelerometry) showed that how consistent your sleep timing is may predict mortality better than how long you sleep.

People in the top quintile of sleep regularity had 20–48% lower all-cause mortality, 16–39% lower cancer mortality, and 22–57% lower cardiometabolic mortality compared to the most irregular quintile. Adding sleep duration to a regularity-based model didn’t significantly improve prediction. The 5th percentile of regularity carried a 53% higher mortality risk than the median.

The 2025 AHA scientific statement on multidimensional sleep health now lists regularity as a distinct sleep dimension with independent cardiometabolic implications.

The practical message: going to bed at midnight every night may beat the pattern of 10 p.m. some nights and 2 a.m. others, even if the average is the same. Shift workers and college students living through “social jet lag” should hear this loudly. So should anyone whose weekend schedule looks nothing like their weekday one.

The body wants rhythm. Cadence is medicine.

Wearables: What They Get Right, What They Don’t

A 2025 meta-analysis of 24 validation studies found that consumer wrist-worn devices systematically overestimate total sleep time by about 17 minutes and underestimate wake-after-sleep-onset by about 13 minutes. Sleep/wake detection is reasonably accurate (>95% sensitivity for sleep). Sleep stage classification still isn’t.

Wearables validated against polysomnography in head-to-head 2024 data:

Oura Ring Gen3: Best overall stage agreement; no significant difference from PSG for any sleep stage duration.
WHOOP: Comparable multi-state agreement to Oura.
Apple Watch Series 8: High REM sensitivity but underestimates deep sleep by ~43 minutes.

What to track: weekly average total sleep time, bedtime and wake-time consistency (your SRI proxy), resting heart rate trends. What to mostly ignore: night-to-night REM and deep sleep percentages.

Sleep Apnea: The Tirzepatide Era

If you snore, wake unrefreshed, have a thick neck or morning headaches, or your partner has watched you stop breathing, get evaluated. The STOP-BANG questionnaire is the most sensitive primary care screen.

The biggest change in the OSA landscape: tirzepatide is now FDA-approved for moderate-to-severe OSA in adults with obesity. In the SURMOUNT-OSA trials, it reduced the apnea-hypopnea index by 50–59%, and 42–50% of patients reached levels where CPAP would no longer be indicated. CPAP remains first-line for respiratory control, but a 2025 review supports a combined airway-metabolic approach: CPAP for the airway, GLP-1 for the underlying obesity that drives it.

When the Runway Doesn’t Work: CBT-I

For chronic insomnia, the first-line treatment is not a sleeping pill. It’s cognitive behavioral therapy for insomnia (CBT-I). A 2024 network meta-analysis confirmed CBT-I produces better long-term remission than pharmacotherapy. About 50% of patients reach remission after 6–8 weeks; 40–45% sustain it at one year.

Digital CBT-I (Sleepio, SHUTi, Somryst) has matured into a real option, with effect sizes comparable to in-person therapy and outperforming medication in head-to-head comparisons. If you’ve been on Ambien or trazodone for years, ask your physician about CBT-I.

The Playbook

Ask the morning questions. Did I sleep well? How do I make tonight better?
Build pressure during the day. Morning light. Movement. Skip the long nap.
Run the 6+ / 3 / 2 / 1 runway.
Sleep regularity matters as much as duration. Same bedtime, same wake time, weekends included.
Cool the cave. 64–68 °F.
Move your caffeine cutoff to noon. Earlier if you’re a slow metabolizer.
Screen for OSA if you snore. STOP-BANG → sleep study → treatment.
CBT-I before pills for chronic insomnia.

Joan left with a sleep prescription. Two adjustments: dinner earlier, phone out of the bedroom. That’s where most of the work happens. The runway, not the bed.

Sleep is something you work for throughout the day. It is also something that benefits you throughout the day. The work and the reward are the same thing. That’s why I treat it as a religious experience, and why I write it on a real prescription pad. Same seriousness as a statin. Larger upside.

Seven to eight hours. Same time, every night. The most underprescribed intervention in medicine.

References

Ungvari Z, et al. Imbalanced sleep increases mortality risk by 14–34%: a meta-analysis. GeroScience. 2025.
Saint-Maurice PF, et al. Associations between actigraphy-measured sleep duration, continuity, and timing with mortality in the UK Biobank. Sleep. 2024.
Ai S, et al. Causal associations of short and long sleep durations with 12 cardiovascular diseases: linear and nonlinear Mendelian randomization analyses in UK Biobank. Eur Heart J. 2021.
Windred DP, et al. Sleep regularity is a stronger predictor of mortality risk than sleep duration: a prospective cohort study. Sleep. 2024.
Gottesman RF, et al. Impact of sleep disorders and disturbed sleep on brain health: a scientific statement from the American Heart Association. Stroke. 2024.
St-Onge MP, et al. Multidimensional sleep health: definitions and implications for cardiometabolic health. Circ Cardiovasc Qual Outcomes. 2025.
Zuraikat FM, et al. Chronic insufficient sleep in women impairs insulin sensitivity independent of adiposity changes. Diabetes Care. 2024.
American Diabetes Association. Standards of Care in Diabetes, 2026. Diabetes Care. 2026.
Malhotra A, et al. Tirzepatide for the treatment of obstructive sleep apnea and obesity. NEJM. 2024.
Gardiner C, et al. The effect of caffeine on subsequent sleep: a systematic review and meta-analysis. Sleep Med Rev. 2023.
Morin CM, Buysse DJ. Management of insomnia. NEJM. 2024.
Lee YJ, et al. Performance of consumer wrist-worn sleep tracking devices compared to polysomnography: a meta-analysis. J Clin Sleep Med. 2025.
Brown TM, et al. Recommendations for daytime, evening, and nighttime indoor light exposure to best support physiology, sleep, and wakefulness in healthy adults. PLoS Biol. 2022.
Lastra AC, et al. Diagnosis and treatment of obstructive sleep apnea. JAMA Intern Med. 2025.
Nguyen Ho PT, et al. Sleep, 24-hour activity rhythms, and subsequent amyloid-β pathology. JAMA Neurol. 2024.
Lloyd-Jones DM, et al. Life’s Essential 8: updating and enhancing the American Heart Association’s construct of cardiovascular health. Circulation. 2022.
Robbins R, et al. Accuracy of three commercial wearable devices for sleep tracking in healthy adults. Sensors. 2024.

Patient names and identifying details have been changed to protect privacy.

If you found this article valuable, share it with someone whose sleep could be better.

For patients who want this kind of thinking applied to their own cardiovascular health, longevity and performance, visit jakekellymd.com.

Jake Kelly, MD, MHS, FACC is a board-certified performance and preventive cardiologist, ACC Governor for Alaska and founder of ROOL Health, a cardiometabolic longevity practice in Anchorage.

Legal Medical Disclaimer

The information provided on this site including articles, newsletters, social-media posts, videos, and downloadable resources is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment. Your use of this content does not establish a physician–patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement. Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.

The Thermostat in Your Brain: Why Weight Loss is a Biological Fight

Jake Kelly MD — Wed, 06 May 2026 15:37:37 GMT

In 2016, a group of researchers published one of the most unsettling papers in obesity science.

They followed 14 contestants from Season 8 of The Biggest Loser, six years after the competition ended. The average contestant had lost 128 pounds on the show. Six years later, they had regained 90 of them.

This wasn’t the disturbing part. Weight regain after extreme dieting is common. Everyone expected the contestants lowest weight would trend up somewhat.

The disturbing part was what happened to their metabolism.

Six years later, after regaining most of the weight, their resting metabolic rate was still suppressed by almost 500 calories per day below what it should have been for their body size. The contestants bodies were burning 500 fewer calories daily than a person of the same weight who had never dieted.

The metabolic adaptation triggered by the weight loss didn’t resolve. It persisted. Even after the weight came back.

Their bodies weren’t just returning to the old weight. They were defending it.

The Thermostat You Didn’t Know You Had

This is the weight setpoint, a concept that’s been debated in physiology for decades but is now increasingly difficult to dismiss.

The basic idea: your brain, primarily the hypothalamus, maintains a defended range of body weight through a feedback loop involving hormones, appetite circuits, and metabolic rate adjustments. When your fat stores change, signals cascade through the system to push you back toward the defended range.

The key player is leptin, a hormone produced by fat cells in direct proportion to fat mass. When you lose weight, leptin drops. The hypothalamus reads that drop as a starvation signal. Two things happen simultaneously: orexigenic neurons (the ones that drive hunger) fire up, and anorexigenic neurons (the ones that signal fullness) quiet down. Hunger increases. Metabolic rate decreases. Your biology tilts the entire playing field toward weight regain.

In 2011, Priya Sumithran and colleagues published a landmark study in the New England Journal of Medicine that quantified the setpoint hypothesis. After a 10-week diet producing significant weight loss, they measured circulating levels of leptin, ghrelin, peptide YY, amylin, cholecystokinin, and insulin. All of these hormones moved in the direction of increased hunger and decreased satiety.

Then the researchers measured the hormone levels again over a year later at 62 weeks.

The hormonal changes hadn’t resolved. Every signal was still pushing toward regain. The body was still fighting to restore what it had lost, a full year after the diet ended.

This isn’t willpower. This is endocrinology.

Three Models, One Reality

The field hasn’t fully settled on how to describe this system, but three frameworks dominate the debate.

1. The Classic Setpoint: The Thermostat

The hypothalamus defends a specific weight like a thermostat defends a temperature. Drop below it, and active mechanisms (hormonal, metabolic, behavioral) drive you back up. Leptin is the primary signal. The arcuate nucleus is the control center. The defended weight is encoded neurologically.

Strength: Explains the persistent hormonal counterattack after weight loss.

Weakness: Struggles to explain why population weight has shifted so dramatically since 1980 if the setpoint is genetically fixed.

2. The Settling Point: The Passive Equilibrium

There’s no active thermostat. Weight simply settles wherever energy intake and expenditure reach equilibrium in a given environment. Change the environment (food access, activity level) and weight drifts to a new equilibrium without active resistance.

Strength: Explains the obesity epidemic well. Our genes didn’t change, but our food environment did.

Weakness: Can’t explain the aggressive hormonal counterattack documented after weight loss. If the system were truly passive, Sumithran’s data wouldn’t exist.

3. The Dual Intervention Point: The Zone Model

Proposed by John Speakman and refined with Kevin Hall, this may be the most clinically useful framework. It proposes a zone of indifference bounded by two thresholds: a lower intervention point and an upper intervention point. Within the zone, weight drifts based on environment. Settling-point behavior. But cross either boundary, and active physiological regulation kicks in.

The lower intervention point is tightly defended. Evolutionary pressure against starvation made sure of that. The upper intervention point is loosely defended and genetically variable. Some people have a high ceiling. Others don’t. In a food-abundant environment, everyone drifts upward until they hit their individual upper boundary. This explains why obesity prevalence rises across populations as food environments change, but not everyone reaches the same weight.

Strength: Reconciles both camps. Explains why weight loss is asymmetrically hard (you’re pushing toward the strongly defended lower point) while weight gain in a permissive environment faces comparatively little resistance.

The models differ in mechanism. But they converge on a clinical reality: the body does not treat weight loss as a neutral event. It treats it as a threat and responds accordingly.

The Proof of Concept We Didn’t Expect

For decades, the setpoint was an elegant theory with strong observational support but no pharmacologic confirmation. We could describe the system but we couldn’t override it cleanly enough to prove how it worked.

Then came the GLP-1 receptor agonists.

GLP-1 (glucagon-like peptide-1) is an incretin hormone released by the gut after eating. It signals the brain that food has arrived. The synthetic versions (semaglutide, tirzepatide) do the same thing, but louder and longer.

Here’s what makes them mechanistically revealing: GLP-1 receptors sit directly on POMC/CART neurons in the arcuate nucleus of the hypothalamus. The same neurons that leptin acts on. The same neurons that the setpoint system uses to defend body weight. GLP-1 RAs don’t just slow gastric emptying or reduce appetite through peripheral discomfort. They act centrally, on the core circuitry of weight defense.

They effectively amplify a satiety signal that the defended-weight system is actively suppressing after weight loss.

The weight regain data after discontinuation is the most compelling evidence that the setpoint is real.

The STEP 4 trial randomized patients who’d lost weight on semaglutide to either continue or switch to placebo. The placebo group regained toward their pre-treatment weight within a year. SURMOUNT 4 showed the same pattern with tirzepatide. A 2025 systematic review and meta-analysis in eClinicalMedicine confirmed it across studies: discontinuation is consistently followed by regain toward baseline. Earlier this year, a second meta-regression in the same journal mapped the trajectory of that regain, showing it begins within weeks and accelerates through the first year off therapy.

The field now calls this disease recurrence, not relapse. Not failure. Recurrence. The same language we use for cancer, for heart failure, for any chronic condition with biological underpinnings.

That framing matters here for understanding long-term successful treatment. Because if weight regain after stopping a medication reflects the re-engagement of a biological defense system, then the weight was never a behavioral problem that got “fixed.” It was a physiological state being pharmacologically managed. Remove the management and the physiology reasserts itself.

GLP-1 RAs didn’t just give us a treatment. They gave us a diagnosis. The setpoint is real. It’s hypothalamic. It’s hormonal. It persists for years. And it’s not a character flaw.

Working With Biology, Not Against It

This week I’m giving a talk to a local TOPS chapter. TOPS stands for Taking Off Pounds Sensibly. It’s a community I’ve been looking forward to speaking to, because the science I’ve been reading this year keeps circling back to a word they built their entire organization around: sensibly.

If the system actively defends weight, then the strategy can’t be to overpower it. Extreme caloric restriction crosses the lower intervention point and triggers the full starvation response: metabolic adaptation, hormonal counterattack, relentless hunger. The Biggest Loser data shows where that leads.

The sensible strategy is to stay within the zone where the defenses don’t fully activate and to shift the inputs the system uses to calculate its defended range.

A modest caloric deficit: 250 to 500 calories per day. This produces half a pound to one pound of weight loss per week. Slow. Unsexy. But it stays below the threshold that triggers maximal counterregulation. The body notices. It doesn’t panic.

Increase NEAT (non-exercise activity thermogenesis). This is the energy you burn through daily movement that isn’t structured exercise: standing, walking, gardening, fidgeting, taking stairs. The variation in NEAT between individuals can reach 2,000 calories per day. Critically, NEAT doesn’t provoke the compensatory hunger response that structured exercise often does. The reason is physiological: intense exercise acutely spikes cortisol and drops blood glucose, triggering the hypothalamus to read the session as a metabolic emergency and compensate with increased appetite afterward. NEAT stays under that radar. The movements are too mild to trip the alarm. You burn more without your appetite fully compensating. Small movements, accumulated across a day, can add 200 to 350 calories of expenditure.

Defend lean mass. Without resistance training, roughly 25% of weight lost is lean tissue: muscle, bone density, metabolic machinery. That loss drops your resting metabolic rate further, making the defended weight harder to escape. Even light resistance training (body weight exercises, 2 to 5 pound dumbbells, resistance bands) preserves muscle during caloric restriction. For older adults, this is non-negotiable.

Prioritize protein. During weight loss, protein requirements increase. Older adults face an additional challenge: anabolic resistance, meaning muscles respond less efficiently to the protein signal. The evidence supports 1.0 to 1.3 grams per kilogram per day, distributed across meals in 25 to 35 gram doses.

Guard sleep. Sleep deprivation independently raises ghrelin (hunger) and lowers leptin (satiety), the exact hormonal profile that the setpoint system uses to drive regain. In one clinical trial, extending sleep by just one hour in habitual short sleepers reduced caloric intake by approximately 270 calories per day, with no dietary instruction at all. Sleep is a metabolic intervention.

None of these strategies are dramatic. None of it will produce 128 pounds of weight loss in 30 weeks. That’s the point. The dramatic approach activates the defense system you’re trying to avoid. The modest approach works within the biology.

This what sensible actually means. Not timid. Not settling for less. Working with the thermostat instead of fighting it.

Trajectory Over Threshold

The weight setpoint isn’t destiny. It’s biology and biology responds to sustained inputs over time.

The evidence suggests that the defended range can shift. Long-term weight maintenance, holding a lower weight for months to years, gradually recalibrates the system. The hormonal adaptations documented by Sumithran may attenuate (though likely never fully resolve). The settling point within the zone of indifference can drift downward with durable environmental changes.

But that recalibration requires consistency, not intensity. It requires the kind of sustained behavioral support that structured programs provide: community accountability, weekly engagement, multicomponent intervention. It’s why the people I’ll be speaking to this week, who show up every week for each other, are doing something the evidence actually supports.

The question was never “How much willpower do you have?” The question is: “Can you sustain a modest, evidence-based approach long enough for your biology to adjust?”

That’s a different challenge. And it’s one that’s actually winnable.

If this changed how you think about weight loss, share it with someone who's been told they just need more willpower. Next week in Part II: The system that actually shifts the thermostat.

For patients who want this kind of thinking applied to their own cardiovascular health, longevity and performance, visit jakekellymd.com.

Jake Kelly, MD, MHS, FACC is a board-certified performance and preventive cardiologist, ACC Governor for Alaska and founder of ROOL Health, a cardiometabolic longevity practice in Anchorage.

Legal Medical Disclaimer

The information provided on this site including articles, newsletters, social-media posts, videos, and downloadable resources is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment.

Your use of this content does not establish a physician–patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement.

Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.

References

Fothergill E, et al. Persistent metabolic adaptation 6 years after “The Biggest Loser” competition. Obesity. 2016;24(8):1612-1619.
Sumithran P, et al. Long-term persistence of hormonal adaptations to weight loss. N Engl J Med. 2011;365:1597-1604.
Speakman JR, Hall KD. Models of body weight and fatness regulation. Phil Trans R Soc B. 2023;378:20220231.
Lincoff AM, et al. Semaglutide and cardiovascular outcomes in obesity without diabetes (SELECT). N Engl J Med. 2023;389:2221-2232.
Deanfield J, et al. Cardiovascular benefit of semaglutide independent of weight loss (SELECT prespecified analysis). Lancet. 2025.
Metabolic rebound after GLP-1 receptor agonist discontinuation: a systematic review and meta-analysis. eClinicalMedicine. 2025.
Speakman JR. Set points, settling points and some alternative models. Dis Model Mech. 2011;4:733-745.
Levine JA. Non-exercise activity thermogenesis (NEAT). Arterioscler Thromb Vasc Biol. 2006;26:729-736.
Tasali E, et al. Effect of sleep extension on objectively assessed energy intake among adults with overweight in real-life settings. JAMA Intern Med. 2022;182(4):365-374.

Your Lipid Panel Is Missing Something

Jake Kelly MD — Thu, 30 Apr 2026 14:02:44 GMT

This article exists because of you.

Last week I posted something simple: I check Lp(a) on every new patient. Once. It’s genetically determined, largely non-modifiable and one of the strongest independent cardiovascular risk predictors we have. If it’s elevated, it changes my threshold for treating all other risk factors.

The response was unlike anything I expected. Readers and commenters shared stories of family members with heart attacks before 60, Lp(a) discovered only after a second MI, insurance companies denying coverage for a $30 test and numbers dropping after menopause treatment. Physicians wrote in about how it’s changed, or hasn’t changed, their practice. The questions were sharp, specific and deeply personal.

This article is a summary of and an attempt to answer to all of them.

What Is Lipoprotein(a)?

Lipoprotein(a) or Lp(a) is an LDL-like particle with a single apolipoprotein(a) [apo(a)] strand covalently bound to its apoB-100 component. The extra protein is the problem.

Apo(a) is a highly polymorphic glycoprotein characterized by variable numbers of kringle IV type 2 (KIV-2) repeats: individuals can carry anywhere from roughly 4 to more than 40 copies. Smaller apo(a) isoforms (fewer KIV-2 repeats) are produced more efficiently by hepatocytes, resulting in higher circulating Lp(a) concentrations. This is why Lp(a) levels vary so widely between individuals as it’s encoded in the structure of the protein itself.

Lp(a) is the only apoB-containing lipoprotein that carries oxidized phospholipids (OxPLs), which are central to its unique pathogenicity. Apo(a) is also structurally similar to plasminogen, a key protein in clot dissolution. This means Lp(a) sits at the intersection of multiple dangerous pathways simultaneously. It promotes atherosclerosis through four mechanisms:

Atherogenesis: The apoB moiety enters the arterial intima like LDL, but Lp(a) is selectively retained through apo(a) binding to extracellular matrix proteins making it approximately 5–6 times more atherogenic than LDL on an equimolar basis.
Vascular inflammation: OxPLs carried by Lp(a) are danger-associated molecular patterns (DAMPs) recognized by the innate immune system, upregulating endothelial adhesion molecules, facilitating monocyte transmigration and driving arterial wall inflammation (confirmed on FDG-PET imaging).
Calcification: Lp(a) components: apo(a), OxPL and autotaxin colocalize in diseased tissue and upregulate procalcific and osteogenic genes, leading to hydroxyapatite mineral deposition.
Thrombosis: Apo(a) kringles attach to fibrin and competitively inhibit plasminogen activation, impairing fibrinolysis and promoting thrombus formation.

It is both proatherogenic and prothrombotic. This combination of mechanisms is part of what makes it so dangerous and what distinguishes it from ordinary LDL.

What makes Lp(a) different from nearly every other cardiovascular risk factor? It is primarily determined by genetics. Specifically, variation in the LPA gene on chromosome 6 accounts for roughly 90% of the variation in circulating Lp(a) levels. Diet does not meaningfully lower it. Exercise does not meaningfully lower it. Statins do not lower it. Your Lp(a) at 30 is, with few exceptions, your Lp(a) at 70.

This is why you only need to check it once. One blood draw. A lifetime of context.

Why Does Lp(a) Still Exist?

If Lp(a) is so dangerous, why do roughly 20% of people carry elevated levels?

Because it probably wasn’t always dangerous. Apolipoprotein(a) evolved independently twice: once in the European hedgehog and once in Old World monkeys, apes and humans. This dual arrival suggests it confers a genuine survival advantage in young mammals. The leading hypotheses center on wound healing and hemostasis (Lp(a)’s ability to inhibit fibrinolysis would promote clot stability during trauma, childbirth, or predator encounters), cholesterol delivery for tissue repair at sites of arterial injury and innate immune defense (OxPLs on Lp(a) activate innate immune responses that may have been protective against infection).

The pathological consequences of atherosclerosis and aortic stenosis predominantly materialize after age 50, well beyond reproductive age in ancestral populations. Natural selection never had reason to act against high Lp(a) levels. This is a classic example of antagonistic pleiotropy: a trait that benefits survival early in life but causes harm later (Thanks for letting me indulge in my undergraduate studies of evolutionary biology).

Lp(a) is not a genetic defect. It’s a mismatch between biology and the modern timeline.

The 2026 Guidelines: What Changed

In March 2026, the ACC/AHA and nine co-sponsoring societies released updated dyslipidemia management guidelines, which is the first comprehensive update since 2018. For Lp(a), it was a new chapter.

Universal Lp(a) screening is now a Class I recommendation.

For the first time in a US guideline, the recommendation is clear: measure Lp(a) at least once in every adult. This carries the highest level of guideline endorsement: Class I, which means there is strong consensus that the benefit of testing outweighs any risk or cost.

The guidelines define risk thresholds:

≥125 nmol/L (≥50 mg/dL), ~80th percentile: Risk-enhancing factor, associated with approximately 1.4-fold increased ASCVD risk
≥250 nmol/L (≥100 mg/dL), ~95th percentile: Associated with ≥2-fold higher estimated risk
≥430 nmol/L (≥180 mg/dL), ~99th percentile: Approximately 4-fold increased risk, which is comparable to heterozygous familial hypercholesterolemia

Lp(a) is genetically determined and stable over a lifetime; the test does not need to be repeated. The guidelines also note that since Lp(a) is not routinely captured on standard lipid panels, it must be specifically ordered.

Beyond Lp(a), the 2026 guidelines restored LDL-C and non-HDL-C treatment goals based on risk. This is a significant about face after the 2013 guidelines moved away from specific numeric targets. The new guidelines also incorporated selective apolipoprotein B (ApoB) measurement to improve risk assessment, particularly in patients with elevated triglycerides, diabetes or low achieved LDL-C.

How Lp(a) Fits With ApoB and Non-HDL Cholesterol

Lp(a) particles contain apolipoprotein B and contribute to your total ApoB count, your calculated LDL-C and non-HDL cholesterol. If your Lp(a) is very high, a meaningful portion of your “LDL-C” is actually cholesterol carried on Lp(a) particles but not conventional LDL. Your true LDL-C may be lower than your lab report suggests, but the risk from that Lp(a)-carried cholesterol still “counts” and is greater per particle than conventional LDL.

This is why the 2026 guidelines endorse ApoB measurement. ApoB gives you a direct count of all atherogenic particles: LDL, VLDL remnants and Lp(a) combined. When LDL-C, non-HDL-C and ApoB tell different stories, ApoB is often the most accurate reflection of true atherogenic burden. The 2026 guideline ApoB goals align with LDL-C targets:

Very high risk (secondary prevention): ApoB < 55 mg/dL, LDL-C < 55 mg/dL
High risk: ApoB < 70 mg/dL, LDL-C < 70 mg/dL
Moderate risk: ApoB < 90 mg/dL, LDL-C < 100 mg/dL

Risk Stratification: The Cut Points I Use

For clinical decision-making, I use the following cut points (consistent with EAS, NLA and the 2026 ACC/AHA guidelines):

< 75 nmol/L (< 30 mg/dL): Low risk
75–125 nmol/L (30–50 mg/dL): Borderline / intermediate
> 125 nmol/L (> 50 mg/dL): High risk

An important nuance: the relationship between Lp(a) and cardiovascular risk is continuous and log-linear above approximately 30 mg/dL. The cut points are clinically useful for risk stratification and trial enrollment, but biologically, there is no meaningful cliff between 120 and 130 nmol/L. It is a gradient of risk, not a binary switch.

One useful framing I reviewed from a VA analysis: Lp(a) is “specific but not sensitive.” It doesn’t broadly improve ASCVD prediction when added to standard risk calculators across the full concentration range. But when elevated, it reliably identifies the 10–20% of patients with significantly enhanced risk. When Lp(a) is high, it correctly reclassifies approximately 23% of individuals who go on to have a first MI into a higher risk category… with no erroneous downward reclassification of event cases. Lp(a) provides a powerful signal for a single blood test.

A quick note on units: nmol/L and mg/dL are not directly interconvertible for Lp(a) because of variability in apo(a) isoform size. This is a real source of confusion for patients and clinicians (I review the units and reference tables each time I interpret Lp(a)). If your lab reports in mg/dL, the general approximation is to multiply by roughly 2.5 to estimate nmol/L, but this is imprecise. Whenever possible, use nmol/L, which is the preferred reporting standard.

How Elevated Lp(a) Shifts My Clinical Approach

Lp(a) is not currently a direct treatment target (that may change in the near future). But it functions as a powerful risk modifier. When Lp(a) is high, I lower my threshold for intervening on all other cardiovascular risk factors. Here’s my approach in primary prevention when Lp(a) is > 125 nmol/L (> 50 mg/dL):

Lipid management: I target lower ApoB and LDL-C: typically < 80 mg/dL and < 70 mg/dL respectively starting with lifestyle and considering medications sooner through shared decision-making. The presence of elevated Lp(a) strengthens the case for pharmacotherapy even when conventional LDL-C is only modestly elevated.

Blood pressure: I shift the target lower. Instead of the standard < 130/80, I aim for < 120-125/70-75 in appropriate patients.

Imaging: I consider coronary artery calcium (CAC) scoring earlier at age 40 instead of the typical 50–55 and in select cases, coronary CT angiography (CCTA) with plaque analysis. Elevated Lp(a) in the setting of any detectable coronary calcium is a combination I aggressively target with lifestyle and earlier pharmaceutical intervention.

Metabolic health: I focus intensely on insulin resistance. Elevated Lp(a) in the presence of metabolic dysfunction compounds risk multiplicatively. Triglyceride-to-HDL ratio, fasting insulin, waist circumference: all of these signals get closer attention.

Cardiorespiratory fitness: I double down on building physiologic reserve. VO₂ max, zone 2 training capacity and functional strength are not just fitness metrics: they are independent predictors of cardiovascular and all-cause mortality. In someone with an unmodifiable genetic risk factor, modifiable capacity matters even more.

Aspirin: Lp(a) is prothrombotic. In select patients with elevated Lp(a) and evidence of subclinical atherosclerosis (e.g., CAC > 0 or plaque on CCTA), low-dose aspirin may enter the shared decision-making conversation, particularly if bleeding risk is low. The data here are still evolving, but the prothrombotic mechanism of Lp(a) provides biological rationale beyond what we’d consider for conventional lipid-driven risk alone.

For secondary prevention, all of the above should already be in motion. Elevated Lp(a) in someone with established cardiovascular disease simply reinforces the urgency.

Beyond Atherosclerosis: Aortic Valve Stenosis

This is one of the most underappreciated aspects of Lp(a).

Lp(a) is the only lipoprotein with robust evidence, including Mendelian randomization data, supporting a causal role in calcific aortic valve stenosis (CAVS). The data are compelling and growing:

A meta-analysis of 44 studies encompassing over 163,000 subjects found Lp(a) levels were 22.6 nmol/L higher in patients with aortic stenosis, with genetic variants at the LPA locus associated with significantly elevated risk. The Copenhagen General Population Study demonstrated that individuals with Lp(a) ≥70 mg/dL had approximately a 2-fold higher risk of aortic valve stenosis, independent of CRP levels. A 2024 JAMA Cardiology meta-analysis showed that patients in the highest Lp(a) tertile had 41% faster hemodynamic progression of aortic stenosis compared to the lowest tertile. Approximately one-third of aortic stenosis cases are associated with highly elevated Lp(a) and OxPL-related pathways.

The mechanism involves the same OxPLs that drive Lp(a)’s atherogenicity colocalized in diseased valve tissue, upregulating procalcific genes and driving hydroxyapatite mineral deposition. Patients with elevated Lp(a) demonstrate enhanced ¹⁸F-NaF aortic valve uptake on PET (active calcification) and faster progression on serial CT.

Aortic stenosis is common, progressive and largely asymptomatic until advanced stages. In a patient with elevated Lp(a), I have a lower threshold for echocardiographic surveillance. It also means the Lp(a)-lowering trials underway could have implications beyond coronary disease. If reducing Lp(a) slows valve calcification, the impact could be significant.

What Changes Lp(a) and What Doesn’t

Because Lp(a) is ~90% genetically determined, the list of things that meaningfully modify it is short. This is important to understand because it prevents both false hope and misguided interventions.

What does NOT lower Lp(a):

Statins do not lower Lp(a). Meta-analyses suggest they may slightly raise it, on the order of 10–20%, though the clinical significance of this increase is unclear and is far outweighed by the LDL-lowering benefit. Ezetimibe does not lower Lp(a). Bempedoic acid does not lower Lp(a). Diet and exercise, broadly, do not meaningfully lower Lp(a). Supplements marketed as Lp(a) reducers lack rigorous evidence.

What modestly lowers Lp(a):

PCSK9 inhibitors (evolocumab, alirocumab) reduce Lp(a) by approximately 20–30%, an effect that is modest but consistent across trials. Inclisiran, the siRNA-based PCSK9 inhibitor, shows a comparable ~15–25% reduction. These agents are primarily prescribed for LDL-C lowering, but the Lp(a) reduction is a meaningful secondary benefit, particularly in patients with both elevated LDL-C and elevated Lp(a).

Lipoprotein apheresis: For the most severely affected patients such as those with heterozygous FH, documented CAD/PAD, Lp(a) ≥60 mg/dL and LDL-C ≥100 mg/dL despite maximally tolerated therapy, lipoprotein apheresis is FDA-approved. It physically removes Lp(a) and LDL from the blood. This is a highly selective intervention for a small subset of patients, but it exists and should be mentioned.

Hormonal influences: the menopause connection:

This is a clinically important and underappreciated topic. Estrogen has a suppressive effect on hepatic apo(a) production. During the menopausal transition, as estrogen levels decline, Lp(a) can rise approximately 20%. Sometimes more. I see this frequently in practice.

Estrogen-containing HRT can reduce Lp(a), sometimes meaningfully: it’s one of the few interventions that consistently moves the number and several readers shared exactly that experience. HRT decisions are complex and individualized, but for women entering perimenopause with elevated Lp(a), it’s an additional factor in the risk-benefit discussion that deserves to be part of the conversation.

Weight loss, inflammation, and assay variability:

Some readers shared meaningful Lp(a) drops with weight loss and endurance training. I’ve seen this in practice, not universally, but in a subset. The mechanism potentially involves reduced systemic inflammation (Lp(a) is a mild acute-phase reactant), decreased hepatic apo(a) output with visceral fat reduction and zone 2 training (via lower IL-6) and inherent assay variability (~8–15% coefficient of variation). If your Lp(a) drops 15–20% with sustained lifestyle change, some is real biology, some is probably measurement noise and the result is likely closer to your true biological set point.

Other modulators: Thyroid dysfunction (hypothyroidism raises Lp(a)), testosterone, kidney function and severe insulin resistance can all influence Lp(a) to varying degrees. These should be evaluated and optimized, but are unlikely to dramatically alter Lp(a) in most patients.

Lp(a)-Lowering Clinical Trials

This is where the field is heading and why the next few years in lipidology are potentially very exciting. Several therapies specifically targeting Lp(a) are in Phase 2 and Phase 3 clinical trials. The central question: does directly lowering Lp(a) reduce cardiovascular events?

We don’t have the answer yet. But we’re close to some early answers.

A few things worth noting:

Pelacarsen is first to the finish line. Lp(a)HORIZON is the first cardiovascular outcomes trial ever conducted for an Lp(a)-lowering therapy. Its readout expected in the first half of 2026 (may be further delayed) will define the field. If positive, it validates decades of genetic and epidemiologic evidence suggesting Lp(a) is causal and modifiable. The trial was delayed by approximately one year because fewer cardiovascular events occurred than anticipated, which actually is an interesting signal in itself.

Olpasiran follows closely. OCEAN(a)-Outcomes, run with the TIMI Study Group, is expected in late 2026 or early 2027. Two independent outcomes trials reporting in close succession would provide powerful confirmatory evidence.

Muvalaplin is the only oral option. Every other agent requires subcutaneous injection. If a daily pill shows outcomes benefit, the accessibility advantage over quarterly injections could be transformative for global adoption.

Zerlasiran’s future is uncertain. Despite strong Phase 2 data (>80% Lp(a) reduction), Silence Therapeutics paused Phase 3 to seek a development partner.

How to Get Tested

This should be simple. It often isn’t.

Lp(a) is a standard blood test available at any commercial laboratory. It does not require fasting. A single measurement is sufficient because levels are genetically determined and stable over a lifetime.

The cost: Typically $24–50 out of pocket if not covered by insurance.

The insurance problem: Some insurers including, as one physician reader reported, certain Anthem plans are still denying coverage, claiming the test is “investigational.” This is increasingly indefensible given the 2026 Class I recommendation for universal screening, but it remains a real barrier in practice. If your insurer denies coverage, push back. The guideline endorsement is now unambiguous.

A free alternative: The Family Heart Foundation offers free at-home Lp(a) testing through their Cholesterol Connect program at familyheart.org/cholesterol-connect. No insurance required. No physician order required. One kit, one finger stick, mailed results.

Family cascade testing: If your Lp(a) is elevated, first-degree relatives have approximately a 50% chance of carrying the same variant. One test in one family member can unlock risk awareness for an entire family. One reader described exactly this: a brother’s second heart attack led to Lp(a) testing, which led to their own discovery of elevated levels, after decades of an apparently healthy lifestyle.

What Lp(a) Is Not

A few clarifications worth stating directly:

Lp(a) is not LDL. It’s a modified LDL particle, but it requires a separate, specific test. Your standard lipid panel does not measure it.

Lp(a) is not something you can “optimize” with supplements, fasting protocols or sauna (and I love sauna). If someone is selling you an Lp(a)-lowering stack, be skeptical.

Lp(a) is not a reason to take niacin. For years, niacin was the go-to recommendation for elevated Lp(a): it does lower the number, sometimes meaningfully. But the HPS2-THRIVE and AIM-HIGH trials demonstrated that niacin did not improve cardiovascular outcomes and caused harm (increased bleeding, infections, diabetes, gastrointestinal side effects). Some older-school practitioners still suggest it. The evidence is clear: niacin lowers the Lp(a) number but does not lower the risk. A biomarker reduction without outcomes benefit is not treatment. It is a cosmetic change to a lab report.

Lp(a) is not a death sentence. Approximately 20% of the global population has elevated Lp(a). Many of these individuals will live long lives without cardiovascular disease, if other risk factors are well managed. Elevated Lp(a) is an important risk factor. It is not a guarantee of disease. I have family members with elevated Lp(a) living long, healthy lives with excellent lifestyle habits and well-managed modifiable risk factors.

Lp(a) is not a reason to panic. It is a reason to measure, understand and act… with precision and proportion.

The Bottom Line

Currently available lipid lowering therapies do not significantly lower Lp(a). This may change soon with the clinical trials underway representing the most important evidence generation in lipidology in decades. If Lp(a)HORIZON confirms what genetics and epidemiology have long suggested, it will open an entirely new therapeutic frontier.

Until then, the path is clear: measure your Lp(a). Understand what it means. Use it to sharpen every other decision.

One blood draw. A lifetime of context.

Do you know your Lp(a)?

Jake

If you found this article valuable, share it with someone who should know their number. And if you haven’t had your Lp(a) checked, ask your doctor or visit familyheart.org/cholesterol-connect for a free at-home test.

For patients who want this kind of thinking applied to their own cardiovascular health, longevity and performance, visit jakekellymd.com.

Jake Kelly, MD, MHS, FACC is a board-certified performance and preventive cardiologist, ACC Governor for Alaska and founder of ROOL Health, a cardiometabolic longevity practice in Anchorage.

Legal Medical Disclaimer

The information provided on this site including articles, newsletters, social-media posts, videos, and downloadable resources is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment.

Your use of this content does not establish a physician–patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement.

Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.

References

2026 ACC/AHA/AACVPR/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Dyslipidemia. Circulation. 2026.
Nordestgaard BG, Langsted A. Lipoprotein(a) and Cardiovascular Disease. Lancet. 2024.
Reyes-Soffer G, Ginsberg HN, Berglund L, et al. Lipoprotein(a): A Genetically Determined, Causal, and Prevalent Risk Factor for ASCVD: A Scientific Statement From the AHA. Arterioscler Thromb Vasc Biol. 2022.
Tsimikas S. A Test in Context: Lipoprotein(a): Diagnosis, Prognosis, Controversies, and Emerging Therapies. JACC. 2017;69(6):692-711.
Arsenault BJ, et al. Lipoprotein(a) and Calcific Aortic Valve Stenosis Progression: A Systematic Review and Meta-Analysis. JAMA Cardiol. 2024.
Koschinsky ML, et al. A Focused Update to the 2019 NLA Scientific Statement on Use of Lipoprotein(a) in Clinical Practice. J Clin Lipidol. 2024.
O’Donoghue ML, et al. Design and Rationale of Lp(a)HORIZON Trial. Am Heart J. 2025.
Nissen SE, et al. Lepodisiran, an Extended-Duration Short Interfering RNA Targeting Lipoprotein(a). JAMA. 2023;330(21):2075-2083.
Nicholls SJ, et al. Oral Muvalaplin for Lowering of Lipoprotein(a) (KRAKEN). JAMA. 2024.
Burgess S, et al. Association of LPA Variants With Risk of Coronary Disease: A Mendelian Randomization Analysis. JAMA Cardiol. 2018;3(7):619-627.

Riding For The People Who Run Toward The Fire

Jake Kelly MD — Sun, 26 Apr 2026 17:22:37 GMT

I’m a little embarrassed to say this. I’ve never raised money for cancer before. I’ve donated from time to time. As a kid, I loved the buzz American Heart Association jump rope-a-thons in elementary school. Cancer has touched my life as it has almost everyone’s, but I always thought that going to work as a cardiologist was enough. Now I know that it isn’t.

This September, I will ride 343 miles from Boston to New York City over five days, alongside firefighters, survivors and people who love them, to honor first responders fighting cancer. The ride is called the First Responder Ride for Resilience. The number is not random. It is the count of firefighters lost on 9/11.

As a cardiologist and a cyclist, I spend a lot of time thinking about what the body can endure and how it breaks down over time. As someone who has benefited from the protection of people in uniform for my entire life, I have thought less than I should about what that protection costs them physiologically. This is my attempt to think about it out loud.

Who we are talking about

Firefighters. Paramedics. Police. Military. Nurses. Emergency response workers. Coast Guard. Search and rescue. The people whose job description includes going toward the thing everyone else is fleeing. The ride focuses on firefighters, so that is where the data lives, but the honor belongs to anyone who trades personal safety for the safety of others.

These everyday heroes are not, as a group, unhealthy people. To become a structural firefighter, one must pass the Candidate Physical Ability Test, the CPAT. In 10 minutes and 20 seconds, wearing a 50-pound weighted vest, you climb stairs at 60 steps per minute, drag charged hose lines, raise ladders and haul a 165-pound rescue mannequin 70 feet across a course. This is the entrance exam.

Wildland firefighters train to a different but equally demanding standard: a 3-mile hike with a 45-pound pack in under 45 minutes, then deployment to fires that burn for days or weeks, often with no respiratory protection at all.

When researchers compare overall firefighter mortality to the general working population, firefighters actually die at a lower rate. Epidemiologists call it the healthy worker effect. These are self-selected, physically tested, fitness-oriented people. They are still being killed by their jobs.

The physiologic cost of service

A firefighter who looks healthy on a standard risk calculator, low Framingham score, normal blood pressure, non-smoker, may already be carrying significant subclinical coronary artery disease. The occupation itself is the exposure. Not their diet or their habits. What these heroes are doing for a living. Every day.

What they carry home

Every fire generates the same toxic mixture: polycyclic aromatic hydrocarbons, benzene, formaldehyde, particulate matter, asbestos, heavy metals, diesel exhaust. Exposure happens through inhalation, skin, and ingestion and does not stop when the flames do. The post-fire overhaul phase, when crews inspect a structure, is one of the highest-exposure windows, precisely when gear comes off. Wearing PPE more than 4 hours after an incident carries a 2.3-fold increase in cancer risk.

Wildland firefighters face a compounding version of this problem. Per-incident smoke concentrations may be lower than structural fires, but these brave individuals work fires for days or weeks, often through an entire season, with no respiratory protection. Exposure accumulates across rotations and does not fully clear between them.

Then there are PFAS. Per- and polyfluoroalkyl substances, the “forever chemicals,” are embedded in turnout gear and firefighting foam. They do not degrade. They accumulate in tissue. But PFAS is increasingly a general population story too. A 2025 study estimated that PFAS in U.S. drinking water contributes to roughly 6,800 cancer cases per year. PFAS has been detected in approximately 45% of U.S. drinking water supplies, with communities exceeding recommended limits showing up to 33% higher incidence of certain cancers. Firefighters get the highest doses. The rest of us are not untouched.

Add shift work into the mix. Circadian disruption is independently associated with elevated cancer risk through its effects on immune surveillance, DNA repair and cortisol regulation. For emergency responders, night shifts are not a scheduling quirk. They are the job.

Where my framework applies

I think about health through what I call the Triangle: Circulation, Metabolism and Cardiorespiratory Fitness, with Physiologic Reserve at the center. Reserve is what determines whether a cancer diagnosis becomes a detour or a dead end. It is what makes treatment survivable.

Firefighters enter the profession with exceptional fitness. The CPAT selects for it. What depletes their reserve is invisible: accumulated carcinogen exposure, years of circadian disruption, subclinical vascular disease building quietly behind normal-looking risk scores. By the time that reserve is needed most, it has often been spent.

The screening gaps

No major medical society has issued firefighter-specific cancer screening guidelines. The NFPA 1582 occupational standard goes further than general population recommendations in several areas, and still sees alarming non-compliance. Volunteer firefighters make up two-thirds of the U.S. fire service. Most departments do not have dedicated health and safety officers, and screening uptake reflects it.

What this ride means to me

I ride bikes because of what they do to the body and the mind. Long hours in the saddle have a way of clarifying things. You think about what matters. You notice your own physical capacity and feel gratitude for it. You become, for a few hours, very aware that the ability to move freely through the world is a privilege.

I have spent my career trying to help people hold onto that privilege longer. That is what cardiology is, at its best: not just treating disease, but expanding the window of healthy life. I have done my job safely under the protection of people who chose, voluntarily, to absorb risk on my behalf. Police officers. Firefighters. Nurses on night shifts. Soldiers. Coast Guard crews. Emergency medical technicians. People whose occupational biology I have not thought nearly enough about.

This ride is, in part, a chance to correct that. Not in a grandiose way. Just to pedal 343 miles with intention. To think about what it means that the people keeping the rest of us safe are being quietly harmed by that work. To ask, as a physician and a fellow human being, what I can do to keep them a little healthier, a little better cared for and a little more seen.

Two small things, if you’re willing.

Say thank you to a first responder you encounter this week. A nurse, a firefighter, a paramedic, a police officer, a soldier in an airport. The words land more than people realize.

And do one thing today that builds your own reserve. A walk, a workout, a real meal or an earlier bedtime. The body keeps score for all of us.

Both matter more than we tend to think.

Jake

For patients who want this kind of thinking applied to their own cardiovascular health, longevity, and performance, visit jakekellymd.com.

Jake Kelly, MD, MHS, FACC is a board-certified performance and preventive cardiologist, ACC Governor for Alaska and founder of ROOL Health, a cardiometabolic longevity practice in Anchorage.

References

Teras LR, et al. Occupation as a firefighter and cancer mortality in a population-based cohort in the United States. Int J Epidemiol. 2025;54(4):dyaf104.
Kales SN, et al. Emergency duties and deaths from heart disease among firefighters in the United States. N Engl J Med. 2007;356(12):1207-1215.
Ashen MD, et al. Coronary calcium scanning and cardiovascular risk assessment among firefighters. Am J Prev Med. 2022;62(6):870-877.
IARC Monographs Vol 132. Carcinogenicity of occupational exposure as a firefighter. International Agency for Research on Cancer; 2022.
Wolffe TAM, et al. Cancer incidence amongst UK firefighters. Sci Rep. 2023;13:5290.
Cleven KL, et al. Performance of risk factor-based guidelines for lung cancer screening in WTC-exposed firefighters. Chest. 2021;160(2):560-570.
Li S, et al. Associations between PFAS and county-level cancer incidence in the United States. J Expo Sci Environ Epidemiol. 2025;35(3):425-436.
CDC/NIOSH. A deeper look into protecting wildland firefighter safety and health. NIOSH Science Blog; April 2024.
Legal Medical Disclaimer
The information provided on this site including articles, newsletters, social-media posts, videos, and downloadable resources is intended for general educational purposes only and does not constitute medical advice, diagnosis, or treatment.

Your use of this content does not establish a physician–patient relationship with me or any contributor. Always seek the guidance of your own physician or another qualified health-care provider before beginning an exercise program, undergoing diagnostic testing, making lifestyle changes, or starting or stopping any medication or supplement.

Never disregard professional medical advice or delay seeking care because of something you have read here. If you think you may have a medical emergency, call 911 (or your local emergency service) immediately.