Your muscle can get fatter while the scale stays flat

Marcus is 47, a project manager, and by every number he tracks he is doing fine. His weight has not moved in three years. His doctor calls his BMI "borderline but acceptable." He walks the dog, he does not smoke, he eats roughly what he ate at 30. The only thing that has quietly changed is his grocery cart: more grab-and-go bars, more frozen entrees, more of the stuff that survives a week in a desk drawer. If you put Marcus in an MRI machine and looked at his thighs, you might find that the muscle there is not quite muscle anymore. Some of it has turned to fat, threaded right through the tissue, and the scale never said a word.

That is the uncomfortable picture from a study published in Radiology this April, and it is worth sitting with before you go back to your inbox.

TL;DR (too long, didn't read)

• A 2026 study in *Radiology* (DOI 10.1148/radiol.251129, published April 14, 2026) used thigh MRI on 615 adults (mean age 59.5, mean BMI 27) from the Osteoarthritis Initiative to measure fat infiltrating the muscle itself, not the fat you can pinch.
• The more ultra-processed food people ate, the more intramuscular fat they had. Each standard-deviation increase in UPF proportion tracked with higher muscle fat across all thigh muscles (beta = 0.108, 95% CI 0.029 to 0.186, P = .007).
• This held after adjusting for total daily calories, fat intake, physical activity, smoking, depression, and either BMI or waist size. In plain terms, it was not just "they ate more" or "they weighed more."
• It is observational and cross-sectional. It shows a strong association, not proof that cutting UPF will melt the marbling out of your muscle. Causation is not established.
• The honest takeaway for a busy over-40 reader: a normal weight is not a clean bill of muscle health. Diet quality, protein, and resistance training defend the tissue the scale cannot see.

Why fat inside the muscle is a different problem

When most people picture body fat, they picture the layer under the skin or the kind that wraps the organs. Intramuscular fat is neither. It is fat that has taken up residence inside and between the muscle fibers themselves, and radiologists have a name for the graded version of it: myosteatosis. Think of a well-marbled steak. Marbling is lovely on a plate and quietly bad news in your own quadriceps, because the fat is occupying space that contractile tissue used to hold. The muscle can look the same size on the outside while doing less work on the inside.

This matters more after 40, when muscle loss starts at 35 and accelerates with each decade. You can lose muscle quantity, which is the story most people know. You can also lose muscle quality, which is the quieter story, and fat infiltration is one of the clearest markers of it. A leg that is half-marbled is a leg that fatigues sooner on the stairs, recovers slower from a stumble, and gives out earlier in the decades when giving out gets dangerous.

The study, told straight

Researchers pulled 615 participants from the Osteoarthritis Initiative, a long-running cohort of adults at risk for knee osteoarthritis but without diagnosed radiographic disease. The group skewed middle-aged (mean 59.5 years), leaned slightly into the overweight range (mean BMI 27), and was 340 women. On average, about 41% of the food they reported eating was ultra-processed. That is the kind of share a lot of ordinary modern diets land in, not a fringe junk-food extreme.

To measure the muscle, they used axial T1-weighted MRI of both thighs and scored fat infiltration with Goutallier grades across the flexors, extensors, and adductors. Goutallier is a radiology grading scale for how much fat has crept into a muscle group, and summing it across muscles gives a single number for how marbled a person's thighs are.

The pattern was consistent. Greater UPF consumption was significantly associated with higher muscle fat across all thigh muscle groups. Per standard-deviation increase in UPF proportion, the all-muscle estimate was beta = 0.108 (95% CI 0.029 to 0.186, P = .007). The flexors came in at beta = 0.111 (95% CI 0.030 to 0.192, P = .008) and the adductors at beta = 0.122 (95% CI 0.040 to 0.204, P = .004). As the lead coverage put it, the more ultra-processed foods people consumed, the more intramuscular fat they had in their thigh muscles, regardless of caloric intake.

The press materials offered a face for the numbers. One participant eating 87.1% ultra-processed food had a Goutallier grade sum of 38. Another at 29.5% UPF had a grade sum of 25. Two people, very different muscle, and the difference tracked the food.

The number everyone is about to misread

Here is where it gets interesting, and where most of the internet will get it wrong. The association held after the researchers adjusted for total daily calories, fat consumption, physical activity, sociodemographic factors, smoking, depression, and either BMI or abdominal circumference. That last part is the whole game. The relationship between ultra-processed food and marbled muscle did not disappear when they accounted for body weight or waist size.

Translation: this is not simply "people who eat junk are heavier, and heavier people have fatter muscle." Two people at the same weight, eating the same number of calories, can have meaningfully different muscle quality, and the one eating more ultra-processed food tends to be on the wrong side of it. The thing degrading the tissue appears to be something about the food itself, not just the energy it carries. There is a plausible mechanistic story here about systemic inflammation and lipid handling, though this particular study did not measure those biomarkers, so treat the mechanism as inference rather than demonstrated fact.

This is the reframe that should land on the desk worker who has made peace with a stable weight. The scale is a measure of mass. It is not a measure of what your muscle is made of. You can hold a "normal" number for a decade and still be slowly swapping working tissue for fat, one convenient dinner at a time. It is the same logic behind why your protein total is lying to you after 40: the headline number reassures you while the thing that actually matters drifts.

The honest limitations

Now the part the press release downplays. This is an observational, cross-sectional study. It photographs a moment in time and finds that high-UPF eaters have more marbled muscle than low-UPF eaters. It does not, and cannot, prove that ultra-processed food caused the marbling, and it certainly does not prove that cutting your UPF intake will pull the fat back out of your muscle. That experiment, a randomized trial, has not been run here.

A few more caveats worth keeping honest:

• The cohort was middle-aged adults at risk for knee osteoarthritis, mean age around 60. It does not automatically generalize to a lean 28-year-old athlete or, frankly, to a 41-year-old with no joint issues. The biology may run the same direction, but this study did not test it there.
• The study used an aggregate UPF proportion. It did not separate soda from chips from processed deli meat, so it cannot tell you which offenders matter most.
• Unmeasured confounding is always possible in this design. The researchers controlled for a lot, which is genuinely impressive, but "a lot" is not "everything."

The authors themselves were measured about it. Their conclusion was that diet quality deserves more attention, and that weight-loss plans should account for what you eat, not only how much, given the link between ultra-processed food and the kind of intramuscular fat associated with knee osteoarthritis risk. That is a reasonable read. It is not "UPF rots your legs." Anyone selling you the second version is selling you something.

The broader evidence is mixed rather than settled. A separate NHANES analysis found the highest quartile of UPF intake carried 60% higher odds of low muscle mass in younger adults (OR 1.60, 95% CI 1.13 to 2.26), but a prospective Framingham cohort found no significant link between UPF and full-blown frailty over roughly a decade, only faint signals in gait speed and, in men, grip strength. The picture that emerges is of a slow, subclinical drift, not a cliff.

What to actually do about it

You do not need to fear your pantry or turn dinner into a moral exam. The practical moves here are the same boring ones that defend muscle in general, which is convenient, because they also happen to be doable on a real schedule.

Set a protein floor and hit it per meal, not just per day. Aging muscle is harder to stimulate, so spreading protein across meals beats backloading it all at dinner. This is the most effective single habit for someone over 40 trying to protect tissue quality, and it pairs directly with the leucine-threshold logic in the protein post above.

Train against resistance two or three times a week. Nothing reverses or resists fat infiltration like asking the muscle to do hard work it cannot phone in. It does not require a fancy gym. Two well-chosen sessions of squats, hinges, presses, and rows will do more for muscle quality than any supplement.

Swap the worst UPF offenders for whole-food versions, gradually. You are not banning anything. You are upgrading the highest-frequency items in your cart: the daily bar becomes Greek yogurt and fruit, the frozen entree becomes a sheet-pan thing you batch on Sunday. Ultra-processed food also tends to crowd out fiber and whole foods, which is its own quiet cost, and worth reading about in ultra-processed food displaces the fiber your gut needs.

Track an output you can actually feel. You will never see your own myosteatosis, but you can watch a proxy. Grip strength as the early warning signal is cheap, fast, and falls when muscle quality does. If your numbers on a few core lifts are holding or climbing year over year, your muscle is probably winning the quality fight even if the scale never moves.

The point of all this is not guilt. It is that the scale has been a flattering, incomplete witness, and the real story is happening one layer deeper. A normal weight bought you nothing if the tissue underneath is marbling.

This is where an AI coach earns its place. When you log a meal, the system can flag how much of it is ultra-processed and nudge the swap before it becomes a decade-long habit, instead of after it shows up on an MRI. It tracks your protein per meal against the leucine threshold your age actually needs, not just a daily total that lets you backload everything at dinner. It programs your two or three weekly resistance sessions around the week you really have, the travel, the kid pickups, the late meetings, and it progresses the load so the muscle keeps getting asked to do work it cannot fake. HERMES, our research engine, surfaced this Radiology study the week it published, which is the only reason your coaching can move with the science instead of trailing it by a year. The muscle you cannot see is exactly the kind of thing that decays when life gets busy and nobody is keeping score. The job of the system is to keep score.

---

Get Chiron in your pocket — $29.99/month

This is what Chiron, our AI head coach, does on every meal and workout you log: catches the small wrong detail before it costs you years. HERMES — our research engine — surfaces new science the morning it publishes, so your coaching moves with the literature instead of trailing it by quarters. You log; we adapt your plan that day. No PDF reprints, no static plan that ages out the day your shift changes or you have a kid.

$29.99/month  ·  $299.99/year  ·  cancel anytime  ·  no enrollment fee  ·  no contract

Start coaching →  ·  See the full app →

Jake Long built it after losing 112 lbs working hospital night shifts — when no human coach could keep up with his schedule. He wanted the system he wished he'd had at 308. Now you can use it too.