Why processed food makes us fat
It's probably not why you think
This post isn’t going to be about seed oils, or high fructose corn syrup, or Big Twinkie, or farm subsidies, or lithium, any of the usual rants.
What I aim to do in this post is give the background on nutritional labels and results from several studies to show why and how processing food liberates more calories, thereby increasing calories overall, driving people fatwards.
Behold the nutritional label:
Labels like these are driven by the Atwater System. Wilbur Atwater was a Yale educated chemist, who in the late 1800’s studied human nutrition. He’s the “father of human nutrition,” and it’s thanks to him that we have the 4/4/9 carb / protein / fat calories-per-gram numbers.
He came up with these by burning various foodstuffs in bomb calorimeters, including pure fats, proteins, and carbs, and came up with equations to calculate the total energy.
This was great to know, and he was doing yeoman’s work. But he was also doing this to a pretty approximate degree, and it is precisely the holes and averages and approximations which obscure some of the differences between processed and unprocessed food.
For example, even in bomb calorimeters, the amount of heat liberated from various sources of fats, proteins, and carbs varies, not to mention in individual foods.
Many people interested in nutrition and diet have pointed out that we’re not bomb calorimeters, and that human digestion tends to work differently than combustion.
Atwater wasn’t dumb - he knew this. He explicitly measured digestibility coefficients for the various macros, and ended up averaging them to reach the 4/4/9 convention. In particular, his overall digestibility coefficent averages out to a <10% discount when considering “digested” vs “combusted” food.
Amazing! I certainly wouldn’t have bet that we were ~90% of the way to bomb calorimeters in terms of digestion efficiency.
Other people have known about these discrepancies, and have come up with various fixes. But ultimately you need something like a matrix of every individual food’s combustion value, plus every single individual food’s digestibility coefficients, and even THEN you need to assume there’s no interactions between individual foods when mixed in the intestines, which we know to be false (in the short term, fiber impairs absorbability and slows glucose uptake, in the long term your microbiome specializes).
“This system relies on having measured heats of combustion of a wide range of isolated proteins, fats and carbohydrates. It also depends on data from digestibility studies, where individual foods have been substituted for basal diets in order to measure the apparent digestibility coefficients for those foods. This approach is based on the assumption that there are no interactions between foods in a mixture in the intestine, and from a practical view point, such studies with humans are difficult to control with the required accuracy.”
So people more or less throw up their hands and say “whatever, 4/4/9 is fine.”
Let’s pause here and look at processing itself.
You want a fun one? K. Oka, Food Texture Differences affect Energy Metabolism in Rats (2003).
Split 20 rats into two populations. Feed one population regular rat chow. Feed the other one the exact same rat chow with more air in the pellets to make them softer and easier to eat, otherwise the exact same rat food. They both ate the same amounts of food, calorie wise.1
The "softer" rats ended up at 6% heavier in weight total, and with 30% more abdominal fat. The ONLY difference was that their food was "softer," ie more processed.
A similar study in pythons fed them raw meat, ground raw meat, cooked meat, and cooked ground meat. Grinding adds about 10% to absorbable calories, and cooking adds about 10%.2
And this is in pythons, who specialize in eating foods whole, they're not like us who have coevolved with cooking (cooking in humans makes foods between 10-50% more absorbable). Making food smaller / finer, or cooking a food, makes a real difference to how many calories you get out of a food. Processing matters.
And what about my 10-50% more absorbable cooking point right there?
From Wrangham’s Catching Fire:
“The percentage of cooked starch that has been digested by the time it reaches the end of the ileum is at least 95 percent in oats, wheat, potatoes, plantains, bananas, cornflakes, white bread, and the typical European or American diet (a mixture of starchy foods, dairy products, and meat). A few foods have lower digestibility: starch in home-cooked kidney beans and flaked barley has an ileal digestibility of only around 84 percent.”
“Comparable measurements of the ileal digestibility of raw starch are much lower. Ileal digestibility is 71 percent for wheat starch, 51 percent for potatoes, and a measly 48 percent for raw starch in plantains and cooking bananas.”
It’s worth noting here, a bomb calorimeter sees zero difference between cooked and raw foods. Not that many people are chowing down on a lot of raw starch, but people do eat raw fruits and vegetables, and occasionally rarer meat.3
In the book, Wrangham has a whole (multiple study-supported) section about “raw food” diets and how the people undertaking them are super skinny and clearly not getting the labeled calories for those raw foods.
Do you want to know the difference that cooking their feed at a higher temperature (going from 140-158F to 248F) and puffing the pellets made for farmed fish?
“During the 1990s, the average weight of harvested fish rose from 2.5 kilograms (5.5 pounds) to almost 4 kilograms (8.8 pounds), survival rose from about 60 percent to 90 percent, and production costs fell.”
Another fun one from the book, you remember Sly Stallone drinking raw eggs in Rocky? Or a young Arnold recommending you drink a thermos full of raw eggs and cream every day?
They would have done better soft boiling or scrambling and then blending them:
“When the eggs were cooked, the proportion of protein digested averaged 91 percent to 94 percent. This high figure was much as expected given that egg protein is known to be an excellent food. However, in the ileostomy patients, digestibility of raw eggs was measured at a meager 51 percent. It was a little higher, 65 percent, in the healthy volunteers whose protein digestion was estimated by the appearance of stable isotopes in the breath. The results showed that 35 percent to 49 percent of the ingested protein was leaving”
Let’s hit one more study:
KD Hall et al. Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain: An Inpatient Randomized Controlled Trial of Ad Libitum Food Intake (2019)
Take 20 adults in their early thirties, 10 male, 10 female. Feed them unprocessed food for 2 weeks, then processed food for 2 weeks. For half of them flip it, and feed them the other one first. People ate ad libitum - ie whatever they chose and as much as they wanted.
When on the processed food portion, people ate ~500 more calories per day on average.
People gained ~0.9kg on the processed diet, and lost ~0.9kg on the unprocessed diet, regardless of order. The unprocessed diet cost $150 a week, and the processed one $100 a week, which is a pretty noticeable delta.
An interesting tidbit, when people on the processed diet were eating more, they ate more carbs and fat, and that's where the extra calories came from, but didn’t eat more protein.4
Let’s recap
So the Atwater system uses individual ingredients and fat / carb / protein ratios to calculate a calories-per-food item. This explicitly does NOT take into account complete food item digestibility, particle size or processing, in vivo food mixing, or anything else like that.
To calculate that Twinkie nutritional label, all they did was measure the average carbs, protein, and fat going into the Twinkie, and 4/4/9-d it.
But when we look into it with more rigor, we see bonuses as large as 10% just for grinding or air puffing an identical foodstuff, and bonuses as large as 10-50% for cooking.
Processed food generally has smaller particle size and higher temperatures and more homogenous cooking than unprocessed food, and indeed, we see that rats, pythons, fish, and humans eating processed food gain weight, often significant weight.
What’s that you say? Duh, the people in the KD Hall study were eating 500 calories more per day, of course they gained weight! This isn’t just some 10% rounding error problem, they were stuffing their faces!
Yes, this is actually another major factor driving the obesity crisis.
If you feed lab rats a standard, nutritionally balanced lab diet of chow and water, they will maintain a healthy weight indefinitely. But offer them a “cafeteria” diet of typical Western foods, with lots of tasty options, and they will inevitably overeat and get fat. Since the initial finding with rats, researchers have shown the same phenomenon in a range of species, from monkeys to elephants, and, unsurprisingly, in humans.5
So yeah, Hall’s study participants were stuffing their faces, because they were given a cafeteria-style selection of processed food, and were eating ad libitum. This is a problem above and beyond processing - but the processing delta is pernicious, because even people with discipline, who look at nutritional labels and try to count calories, can still be eating and absorbing noticeably more calories than they thought they were.
Our modern environment of a wide selection of processed foods is basically tailor-made to make everyone fat, and indeed, 75% of Americans are overweight or obese now, and this is bad because obesity is a permanent, one-way ratchet.
How far off could the Atwater system really be?
I mean, if the difference between literal bomb calorimeters burning the food and our digestion is only 10%, isn’t the most incremental calories you could be getting something like 10%?
Not exactly - remember, that 10% was reached at by looking at the “digestibility” of various things,6 but he was actually measuring the digestibility of fats, carbs, and proteins. There was variation between those 3 macros in the measured diets he was feeding people, but it was small enough in the samples he was considering he decided to ignore it. And one fairly large omission, he largely ignored fiber.
But the whole point of showing all these processed food studies is that the variation is NOT minimal. For a given largely carbohydrate rat chow, with the only difference air puffing the exact same chow pellets, the processed food rats gained 6% more weight and had 30% more abdominal fat! For simply grinding meat vs whole meat, the difference was 10%! Fish with puffed and higher-temperature cooked fish meal gained 1.6x the weight and had a 1.5x improvement in survival! These aren’t trivial differences.
In humans, cooking alone makes between a 10%-50% absorbability difference, and we haven’t systematically explored or measured the impacts of grinding to various fineness, or cooking in various methods and temperatures.
More, in many of these examples, the weight gain between puffed, ground, or cooked food vs unprocessed was driven by the fact that the animals eating the meals expended fewer calories afterwards digesting the food - their calories out are less, as measured by their post-prandial thermogenesis being less.
My point is, that there is a hidden ceiling here in less processed food, such that largeness vs fineness of particles, fiber, the way it’s cooked, the energy the body spends digesting it, and whatever else could be routinely *discounting* non-processed foods from their labeled calories, so the incremental headroom available for being off on calories isn’t limited to 10% - if unprocessed foods are routinely 10-15% under labels due to some combination of these things, then there’s a potential incremental 25% delta in caloric availability from labels, for example.
And indeed, we find something very much like that - looking at high fiber diets, the available calories are 4-11% lower than the labels.7 And it’s worth noting, this is JUST an impact from fiber, and we know there are other potentially significant factors.
One interesting fact that stuck with me after reading Herman Pontzer’s Burn:
Westerners gain about half a pound a year on average, and that’s only 5 incremental calories per day. Another way to put this is that even an overweight or obese person’s weight homeostat operates at 99.8% fidelity.
Even for people gaining a pound a year, or a kilo a year, that’s just 10-25 calories a day, and using a (probably inaccurate for people gaining that much) 2,000 calorie per day average, that’s <= 1%. That tiny 0.2-1% daily variance is driving a literal worldwide obesity crisis!
Could you be eating an extra 5-25 calories per day in the error margins in the Atwater system used in all nutritional labels? It’s pretty much a certainty.
In fact, it’s such a certainty that there is almost certainly some larger problem.
I suspect the real problem with the obesity crisis is something like processed food or junk or fast food wrecking our “weight homeostats.” There've been some studies showing hypothylamic inflammation and changes in microbiome and inflammation-mediated increases in liposaccharides in mice and humans who eat high fat diets,8 but the writing isn’t quite on the wall yet.
Really, worrying about nutritional label inaccuracy is kind of funny to begin with - what percent of people even looks at them, much less counts calories? 5%? Maybe as low as 2%?
Obviously, people got by just fine in the ~2M years of hominin life in the EEA, and also just fine in the ~10-12k years since agriculture was invented, without Atwater there to measure stuff and tell us about 4/4/9 and without nutritional labels entirely - human weight homeostats operated just fine for all that time, and to a better precision than 99.8%, because nobody was fat.
But that’s just more evidence that something IS seriously wrong, to my mind. We’re not sure whether it’s processed food, junk food, fast food, plastics, lack of exercise, or something else (my money’s on some hellish, multiply-confounded and feedback-looping mixture of these things), but we do know it’s *something.*
My second post on processed food, where I go into some of the plausible smoking guns around that, is here.
But regardless of the mechanism, you know the solution.
Prepare and eat real foods your great-grandma would recognize, and move and exert your body regularly.
This lets you avoid any incremental calories from processing, AND gives you lower calories overall AND healthier ingredients AND more fiber AND probably a bunch of other benefits.
You also avoid all the stuff people love to rant about - seed oils, HFCS, pufas, fupas, whatever.
Not to mention the fact that even when controlled for lack of fiber and macros (fat / sugar / protein proportions), the top quarter of ultra processed food eaters (by proportion of diet) have between 26-69% higher all cause mortality than the lowest quarter, and this follows a direct dose-response curve.9 I’m going to do a followup post digging into some of the (fairly extensive) literature around ultra-processed foods pretty soon, but thought I’d tease that one statistic.
It takes too much time to cook and eat real food? Preach.
Do meal prep - then you spend a few hours on a Sunday preparing all your meals for the week, and you save 5-8 hours overall per week! All that time usually spent on deciding and choosing and cooking and eating and cleaning up (or deciding and driving to the restaurant and choosing and waiting and eating and driving home), suddenly free! Every day! Every meal! It’s not a trivial savings, it’s quite significant. It’s actually one of my go-to bits of advice to young founders when mentoring - your time budget becomes a real issue in a startup, and food takes a LOT of time.
I write quite a bit about fitness here - if you’re interested in fitness, eating is one of the major legs of the tripod, just as important as sleeping or your actual workouts. And meal prep is THE key to reliably having the right macros and caloric deficits or surpluses to gain or lose to your target weight.
I think making a choice that improves your health, saves money, lets you meet your macro and weight goals, and saves 5-8 hours a week is a slam dunk…but you know, we all make choices.
Another important point to note:
If you DO count calories, this whole dynamic shouldn’t be much of a problem. This is because if you’re rigorous enough to count calories, you should be measuring your weight on a daily or weekly basis, so you have a simple feedback loop.
Even if your label calories are under-or-over estimating by 10-20%, it doesn’t much matter because you’ll see you’re not moving in weight over a week or so, and reduce (or increase if bulking) your intake by another 200 calories a day or whatever. Apps have made this so much easier than the bad old days, too - My Fitness Pal and similar apps have basically anything in there, from any restaurant or recipe. GPT and Claude do decent enough jobs too. It’s never been a better time to be serious about what you eat!
For the CICO folk out there, the difference is indeed Calories Out - the rats that ate the harder pellets had greater post-prandial thermogenesis, from their bodies working harder to digest the harder pellets.
The air puffed rats didn’t need to work as hard to digest the pellets, so they absorbed more calories overall.
Boback, Secor et al. Cooking and grinding reduces the cost of meat digestion (2007)
Rare vs well cooked steak has no effect on digestibility in young people, but a study in 70+ year olds found about a 30% buff in bloodstream amino acids with well done + minced versus rare.
As Hall says in the paper - “The remarkable stability of absolute protein intake between the diets, along with the slight reduction in overall protein provided in the ultra-processed versus the unprocessed diet (14% versus 15.6% of calories, respectively) (Table 1), suggests that the protein leverage hypothesis could partially explain the increase in energy intake with the ultra-processed diet in an attempt to maintain a constant protein intake.”
Later they calculate and find that it couldn’t account for more than about 50% of the overage in calories, though.
One thing I like about Kevin Hall - he’s a physicist with a Phd in mathematical modeling, who just sort of stumbled into obesity and nutritional research, and now is cranking out some of the best studies about them.
Yes, the sample size is low - but that’s because paying people to literally take a month off and come live in a lab with a diet totally controlled by the experimenters is expensive. Most nutritional studies rely on self-reports, food diaries, and surveys, all of which are absolute hot garbage and lies - just the error margins in those methods would have swamped this very significant 500 calories differential - Hall has avoided this and has a completely controlled and measured diet where you can actually detect things like this caloric difference, and that’s genuinely hard.
Rats: “A. Sclafani and D, Springer (1976). “Dietary obesity in adult rats: Similarities to hypothalamic and human obesity syndromes.” Physiol. Behav. 17 (3): 461–71.”
Humans: R. Rising et al. (1992). “Food intake measured by an automated food-selection system: Relationship to energy expenditure.” Am. J. Clin. Nutr. 55 (2): 343–49.”
First by literally measuring the outputs from people with ileostomy bags (most nutrients we can use are absorbed in the small intestine, the large intestine basically feeds your microbiome, so short circuiting at the end of the small intestine is mostly fine for this), later by measuring nitrogen content on breath.
When compared against more rigorous measurements for a “high cereal fiber” diet and a low-fat, high fiber diet, Atwater was off by -4% for the cereal fiber diet, and -11% for the low-fat, high fiber diet.
Zou et al. Accuracy of the Atwater factors and related food energy conversion factors with low-fat, high-fiber diets when energy intake is reduced spontaneously. (2007)
This is actually one of the expected results, given that bomb calorimeters burn fiber as calories, but our digestive systems don’t.
Hadza hunter gatherers, incidentally, eat 5x the amount of fiber as Westerners every day.
Inflammation and hypothalamic impacts from high fat diets in humans:
https://pmc.ncbi.nlm.nih.gov/articles/pmid/22201683/
Microbiome and liposaccharides in humans:
https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2020.594150/full
From: SJ Dicken, R Batterham, The Role of Diet Quality in Mediating the Association between Ultra-Processed Food Intake, Obesity and Health-Related Outcomes: A Review of Prospective Cohort Studies (2021)
n=~106k Adjusting for French dietary guideline adherence or for both French dietary guideline adherence and for Western dietary pattern still resulted in each 10% increment in UPF intake being associated with a 14% (1.04, 1.27) or 19% (1.05, 1.35) increased risk, respectively, of all-cause mortality.
n=105k Srour et al. reported a 12% (1.05, 1.20), 13% (1.02, 1.24) and 11% (1.01, 1.21) increased risk of CVD, coronary heart disease (CHD) and cerebrovascular disease, respectively, per 10% increase in UPF in the diet [117]. Multiple dietary adjustments did not alter these risk estimates.
n=105k Fiolet et al. reported a 12% (1.06, 1.18) and 11% (1.02, 1.22) increased risk of all cancer and breast cancer, respectively, per 10% increase in UPF in the diet [126]. Adjustment for lipids (including fat), sodium and carbohydrate intake had no impact on the risk of all cancer (HR: 1.12 (1.07, 1.18)) or breast cancer (HR: 1.11 (1.01, 1.21)) per 10% increase in UPF in the diet, respectively [126]. Adjustment instead for Western dietary pattern also did not change the 12% (1.06, 1.18) and 11% (1.02, 1.22) increased risks
n=~20k Four studies within the Seguimiento Universidad de Navarra (SUN) cohort have adjusted for fat, added sugar and sodium intake, or for dietary pattern. Rico-Campa et al. demonstrated that the highest vs. lowest quartile of UPF intake was associated with a 62% (1.13, 2.33) increased risk of all-cause mortality [103]. Adjustment for saturated and trans fats, added sugar and sodium intake still resulted in a 69% (1.12, 2.56) increased risk of all-cause mortality. A 58% (1.10, 2.28) increased risk still remained after adjusting for Mediterranean diet pattern adherence instead.
n=~12k In the US Third National Health and Nutrition Examination Survey (NHANES III) cohort, there was a 31% (1.09, 1.58) increased risk of all-cause mortality in the highest vs. lowest quartile of UPF intake, which remained significant after further adjustment for dietary quality score using the HEI-2000 (p-trend = 0.001)
n=22k In the Italian Moli-sani cohort, the highest vs. lowest quartile of UPF intake was associated with a 32% (1.15, 1.53) higher risk of all-cause mortality, 65% (1.29, 2.11) higher risk of CVD mortality, and a 63% (1.19, 2.25) higher risk of ischemic heart disease (IHD)/cerebrovascular mortality [105]. Adjusting for saturated fat, sugar, sodium and dietary cholesterol intake resulted in a 28% (1.09, 1.49), 56% (1.19, 2.03) and 33% (0.94, 1.90) increased risk of all-cause, CVD and IHD/cerebrovascular mortality, respectively, in the highest vs. lowest quartile of UPF intake.
n=3k In the Framingham Offspring cohort, each additional serving of UPF per day was associated with a 5% (1.02, 1.08), 9% (1.02, 1.16), 7% (1.03, 1.12) and 9% (1.04, 1.15) increased risk of overall CVD, CVD mortality, hard CVD and hard coronary heart disease, respectively [118]. Further adjustment for diet quality using the DGAI-2010 still resulted in a 4% (1.01, 1.07), 9% (1.02, 1.16), 6% (1.02, 1.11) and 9% (1.03, 1.15) increased risk of overall CVD, CVD mortality, hard CVD and hard coronary heart disease
n=~98k In the US Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial cohort, the highest vs. lowest quintile of UPF intake was associated with a 50% (1.36, 1.64) increased risk of CVD mortality, and a 68% (1.50, 1.87) increased risk of heart disease mortality [119]. Multiple dietary adjustments did not alter this risk; adjustment for saturated fat, added sugar and sodium resulted in a 48% (1.34, 1.63) and 65% (1.47, 1.85) increased risk of CVD mortality and heart disease mortality, adjustment for diet quality using HEI-2005 resulted in a 48% (1.35, 1.63) and 67% (1.49, 1.86) increased risk of CVD mortality and heart disease mortality, and adjustment instead for red meat, processed meat, whole grains, fruit, vegetables, fibre and dairy intake also still resulted in a 49% (1.35, 1.64) and 66% (1.48, 1.86) increased risk of CVD mortality and heart disease mortality, respectively