Ultra-Processed Food and Fat Inside the Thigh

# Ultra-Processed Food and Fat Inside the Thigh

Cole texted me before my shift. The group chat already ordered a nacho tray for the 7 p.m. Central watch. Red Sox and Yankees, Wild Card Game 1. First pitch of the day is earlier. Phillies and Braves are listed at 2 p.m. Eastern on the network Wild Card schedule, which is 1 p.m. Central. His calorie app is open. The note under it says he will be good tomorrow.

I know that note. I wrote versions of it on overnight 12s, next to a vending machine, before this post. I am 40. I am a day-shift hospital security supervisor. I went from 308 to 196 over 9.5 months, starting mid-2025. The tray is not what took the weight off. A promise before sunrise did not either.

A calorie app counts the tray. It does not read the thigh.

Quick take

• A 2026 paper in Radiology scanned 615 adults at risk for knee osteoarthritis who did not yet have it on imaging. A higher share of ultra-processed food in the prior year lined up with more fat inside the thigh muscles.
• That link stayed after the model accounted for calories, activity, and either body mass index or waist size. The meals in the data are from an older cohort. Tonight's nachos are not a new experiment.
• The coefficient is small and standardized. Do not turn it into a percent of muscle that turned to fat.
• Swap one packaged item already in the drawer. Leave the rest of the week alone.

The note was the actual risk

Cole is about my age. He sits most of the day. He is a friend in a group chat, not a client I am using as proof. He wants the tray, and he wants a stricter self that starts after the series.

I have given that speech. It did not take the 112 pounds off. The boring version did. Same house. Smaller clocks. I kept them when the week got ugly.

He asked if a lunch walk makes the tray fine. I told him the model already held activity constant. Then I opened the meal log I actually use. Chiron, the AI head coach, only sees the meal I enter. That sunrise note is invisible to him. That is the point. A plan that believes the note will bless the tray.

What the scan actually found

Zehra Akkaya and colleagues published this in Radiology, 2026, volume 319, number 1, article e251129 (doi:10.1148/radiol.251129). It is a cross-sectional look at baseline visits from the Osteoarthritis Initiative, February 2004 through October 2015. Not a trial run last week. The press write-up circulated in mid-September 2026 (the society's release). The meals in the file are older than Cole's group chat.

They kept 615 people at risk for knee osteoarthritis, with a Kellgren-Lawrence grade of 1 or lower, and no pain in either knee or hip. Mean age was 59.5 years, standard deviation 9. There were 340 women. Mean body mass index was 27, standard deviation 4.3. Mean abdominal circumference was 98.9 centimeters, standard deviation 12.1. Ultra-processed foods were a mean 41.4 percent of the diet, standard deviation 13.4, from a food frequency questionnaire about the prior 12 months. The groups follow Monteiro and colleagues in Public Health Nutrition, 2019 (doi:10.1017/S1368980018003762).

Fat inside the muscle was graded on magnetic resonance imaging of the thigh, on the Goutallier scale. Grade 0 is no fatty streaks. Grade 4 is more than half the signal coming back as fat. They summed those grades across both legs, for the flexors, the extensors, the adductors, and all thigh muscles, then standardized the sums.

Cole heard "more packaged food, fattier thigh" and opened a notes app to ban his personality. I stopped him. Then I opened my own desk drawer, because the text had made me smug about his tray and quiet about my breakfast.

The bar in there has a paragraph of ingredients. I had already counted it as the good choice. The nachos were going to be his sin. That is the trick the calorie app loves.

I gave him one number, not a spreadsheet. After body mass index, Akkaya's standardized coefficient for all thigh muscles was 0.108. The 95 percent confidence interval ran from 0.029 to 0.186. P was 0.007. Flexors were 0.111. Adductors were 0.122. When the model used abdominal circumference instead, all muscles moved to 0.134, and the extensors joined at 0.097. Sex did not change the relationship.

He asked if 0.108 meant a tenth of his thigh had turned to fat. I told him to close that caption. One standard deviation more ultra-processed share lined up with about a tenth of a standard deviation on the fat-infiltration score. The paper's picture of two women, 29.5 percent versus 87.1 percent, Goutallier sums 25 versus 38, is an illustration. The higher-share woman also had a waist at or above 88 centimeters. Do not steal that pair and call it the result.

The scan still saw the fat after they subtracted the calories.

Cole went quiet in a way I trust more than a ban. The walk does not buy the tray. The bar does not get a halo because the wrapper says protein.

What I am not going to sell you

This does not prove the nacho tray at 7 p.m. Central changes a thigh by Wednesday. A food frequency questionnaire is a year of memory. Cole's Tuesday is one tray. Red Sox at Yankees is listed at 8 p.m. Eastern on that same schedule, which is 7 p.m. Central. The clock is a broadcast fact. It is not a dose.

Physical activity was already in the model. The association did not leave. So a lunch loop is not a receipt. The paper also did not test whether squats melt fat already inside the muscle. I still lift. I will not invent that trial to make the lift sound smarter.

If you want the small-clock version of eating, the half-serving piece is an add, not a ban: five extra minutes and half a vegetable serving. Sitting has a separate bill: an extra hour of long sitting. Do not mash them into one slogan.

Before first pitch

The swap was not his tray. It was my bar.

I ate the apple that had been rolling in the same drawer, and I logged the bar as the thing it is. Chiron, reading that entry, does not get to call a paragraph of ingredients a virtue because I walked at lunch. HealthKit can see the steps. It does not get to pardon the year.

One swap before first pitch is a protocol. A cleanse until October is a mood.

Cole deleted the ban. His sticky note names whatever is already in his drawer. Mine names the bar. The tray can still show up at 7. The year-long share in this sample sat at 41.4 percent. That number is made of Tuesdays nobody photographs.

Forge does not get a new split out of this either. The session I already had stays. Volume can come down if the night runs long. Thigh work stays if the knees allow it. A playoff is not a reason to install a personality, and it is not a reason to install a program.

If your job is a chair and your evening is a screen, open the drawer before you open the group chat. I learned the vending-machine version of this lie on overnights. Tomorrow's stricter self was always the one who refused to log tonight.

When the tray still happens

It will. Someone will bring a second bag. Cole will eat some. If he then decides the apple was pointless, he missed the paper.

The finding lived in a year of diet, not in a perfect Tuesday. A missed swap is not a reason to install October as a new person. Log the tray you ate. Do not log the tray you meant to skip. In-app meal logging only works if the entry matches the plate.

Activity was in the model. The association did not leave.

That is also why I will not tell a desk worker that a step challenge starting this morning erases the scan. Steps still help other clocks. They are not a pardon for the food pattern the model already held constant.

How the coaching uses the paper

A printed plan could not see the vending machine on those overnight years. That is why I built Legacy In Motion.

Chiron adjusts the next day off the meal you logged and the sleep you actually got. When HealthKit is connected, short sleep can pull volume down. He does not install a ban because a series started. He asks whether today's packaged share climbed against your own week, not against a stranger's clean-eating graphic.

Forge rewrites the session you already had. Thigh work stays if the knees allow it. Volume comes down if the night was short. No new split as a reward for playoff week. That same-day rewrite is the part a static app will not do when the group chat changes the evening.

HERMES keeps the Akkaya citation on the protocol so next month nobody rewrites this into "just eat clean." Clean is not a dose. The dose in the paper is the share of the diet.

You log the meal you ate, the same way I do. Chiron cannot see a pattern you fictionalize after the game.

The offer is $29.99 a month or $249.99 a year after a 7-day free trial. Cancel anytime. The free community argues these swaps in public at Discord.

If you want the same system on a week that includes a tray, start at Legacy In Motion.