Why the Minnesota Starvation Experiment Doesn’t Generalize to Modern Dieting

I saw this tweet going viral about the Minnesota Starvation Experiment:

Online, the Minnesota Starvation Experiment is often cited as “proof” dieting works. Few would dispute calorie restriction can work, at least in the short term, as Dr. Ancel Keys showed in his infamous experiment. Every year, millions of Americans resolve to lose weight, and many do, at least initially. Typically this entails “eating less and moving more”. The issue is that despite following this advice, the vast majority of dieters regain the weight they lost, as calorie restriction proves unsustainable in the long term. Hunger takes over and people succumb to cravings. Even with a plethora of fitness and dieting apps and other trackers, only GLP-1 drugs have finally put a dent in the otherwise unrelenting upward trend of obesity.

To recap, between November 19, 1944 and December 20, 1945, 36 conscientious objectors were intentionally restricted food to test the physiological effects of calorie restriction. This was towards the end of the Second World War, and the U.S. government sought to understand how restricting rations, such as in the context of wartime when supply lines may be disrupted, would affect physical and mental performance. The participants started at around 3,300 kcal/day, and eventually had their food reduced to a low of 1,570 kcal/day, losing around 25% of their starting weight. The experiment was overseen by Dr. Ancel Keys, who was something a sadist in his own right who presumably saw this as perfectly ethically acceptable, who would go on to be a major but polarizing figure in nutrition research for positing a (dubious) link between saturated fat and cholesterol.

It didn’t come as a surprise that after months of dieting, the participants became agitated, physically weak, and preoccupied with food. 36 weeks later, by the final phase of the study, when allowed unrestricted food, subjects were ravenous and gained weight–and in many cases–even more weight than they had originally lost. Clearly, it was not a viable strategy, and the study has gone down as more of a cautionary tale than something to be emulated.

Despite ultimately regaining the weight, the participants of the Minnesota Starvation Experiment did lose significant weight, as the photos make clear. By weeks 30–36, when their intake had been reduced to a low of around 1,570 kcal/day, they were visibly wasting away. On average, they stabilized at body weights in the mid-to-low 100-pound range, with BMIs in the teens, compared to an initial average of about 153 lbs and initial BMI of 23—placing them solidly in the “underweight” category:

By the final week of the 24-week semistarvation period (1945), participants in the Minnesota Starvation Experiment saw their Body Mass Index (BMI) drop from an average of ~22 down to ~16.4.

But fast-forward to the 2000s and there seems to be the opposite situation. People are dieting, but the weight does not come off as effortlessly or the weight loss isn’t as drastic compared to the Minnesota Starvation Experiment, even when the calories are restricted to a similar degree as Keys enforced on his subjects. On Reddit, there are thousands of people who struggle to lose weight and are still overweight or obese on diets of 1,500 or even fewer calories per day, let alone become emaciated as Keys’ subjects became. There are by estimate hundreds of subreddits devoted to every conceivable diet or approach to dieting, but the overarching theme is the same: significant weight loss proves elusive for all but some. Or weight loss stalls out long before the sought thinness is achieved, even when calories drop to as few as 1,200/day, which would make even Keys wince.

True, these weight loss communities are mostly composed of women; men obviously have higher calorie needs, but evidently men struggle with the same problem. There are tons of stories–like on various fitness communities, blogs, Reddit, and on Twitter–of men who stall out around 170–200 lbs despite consuming similar calories as Keys’ subjects. They use fitness trackers and calorie-counting apps, in some cases weighing their food, and like Keys’ subjects, also engage in lots of cardio, but the weight just refuses to come off. Many of these men were formerly obese, and having lost enough weight to no longer be obese, find that they stall at an overweight or borderline overweight BMI (around 25–28) while eating relatively few calories, making further calorie restriction untenable. All the effects of calorie restriction are present–reduced energy, reduced libido, difficulty focusing–but without the sought leanness to show for it.

A Google search shows tons of reports of men who are unable to lose weight despite consuming between 1,600–2,000 kcal/day. Here is a detailed report from the user “TeoSenpaii”:

He used to weigh 106 kg in 2020 at 5’8″, putting him in the obese BMI category. He lost 33 kg, but now finds that at 81 kg (178 lbs) he has stalled at 1,800 kcal/day despite doing cardio, going to the gym, and tracking his food. Note: the heaviest Minnesota Starvation Experiment subject started at 170 lbs, and the fat melted off well before his food was restricted to 1,800 kcal/day. A second user “Snakeyb” who responded reported a similar situation, starting at 120 kg and stalling at 79 kg despite tracking calories and cardio.

This is evidently quite common, as it turns out. There are many similar examples on Twitter/X too…guys who weigh between 170–200 lbs who stall around 1,600–2,000 kcal/day at around a 15–25% body fat. From my post, “Bulking, Metabolism, and Selection Effects,” which investigates this finding further:

The above individual–despite only consuming 1,800 kcal/day, doing 14,500 steps daily, cardio, and resistance training–is still unable to get any leaner than 17% body fat. His stall weight of 74 kg is still heavier than the initial starting weight of all but one of the Minnesota Starvation Experiment participants.

Insufficient weight loss is often blamed on calorie miscounting. Keys, a perfectionist, meticulously tracked the calories of his subjects. Understandably, regular people cannot be expected to track their food as closely as he had done. But even under controlled conditions, when ingoing and outgoing calories are precisely being measured in laboratory conditions, such as by tracking CO2 and weighing the food by macro composition, the majority of obese people are not eating that much to begin with, or a lot less than commonly assumed.

A 1992 study, “Discrepancy between Self-Reported and Actual Caloric Intake and Exercise in Obese Subjects,” is often cited as evidence that obese people underestimate their calorie consumption and hence are unable to lose weight. But the underreporting of calories is orthogonal to how many are consumed. It’s possible to underreport and still not be eating that many calories. That study showed that obese participants who failed to track and didn’t lose weight were only consuming around 2,081 kcal/day, which isn’t that much. From the study:

The energy intake reported by the subjects in group 1 during the 14-day study period was 1028±148 kcal per day, whereas their actual energy intake was 2081±522 kcal per day. Thus, these subjects significantly (P<0.05) underreported their energy intake by a group mean of 1053 kcal per day, or a mean for individual subjects of 47±16 percent (Fig. 2). These subjects underreported their energy intake even more if their total energy expenditure (2468 kcal per day) represents the long-term energy intake required to maintain a stable weight.

Note: the ’92 study tracked participants with BMIs between 33–40, so the “2,468 kcal per day” is just to remain weight-stable at an obese BMI. This was in laboratory conditions, with the food intake closely measured, so no cheating. For the group that had stalled, 2,000–2,400 kcal/day is hardly “lots of food”. That is about the same as the ‘official’ FDA guidelines when establishing serving sizes, presumably for a healthy-weight person. Yet the participants in both groups of the study were still obese and/or had stalled. This does not leave much room to cut calories if someone is 50+ pounds overweight yet needs to get to 2,000–2,400 or fewer calories/day just to begin to actually start losing weight, let alone not be overweight. It’s little wonder the failure rate of dieting is so high when there isn’t much margin of error. Keys’ subjects, by contrast, were already ‘normal weight’ on diets closer to 3,300 kcal/day at induction, only to become rail-thin at closer to 1,800 kcal/day months into the study. Big difference.

There is also the National Weight Control Registry (NWCR), informally recognized as the largest database of long-term successful dieters. Success in this case is loosely defined as a long-term loss of 13 kg or more of initial body weight for at least a year. These are ‘successful losers,’ meaning they kept weight off for longer than a year, who are tracked as part of an ongoing observational study and required to periodically submit reports showing their progress, so it’s reasonable to assume they would know how to count calories more accurately than the typical dieter, and that the self-reported values are reasonably accurate.

A 2005 meta-analysis of the NWCR found similar results as Reddit, with participants typically consuming around 1,400–1,800 kcal per day:

Registry members reported eating 1381 kcal/d, with 24% of calories from fat. In interpreting their data, it is important to recognize that 55% of registry members report that they are still trying to lose weight and to consider that dietary intake is typically underestimated by 20–30%. Thus, registry members are probably eating closer to 1,800 kcal/d. However, even with this adjustment, it is apparent that registry members maintain their weight loss by continuing to eat a low-calorie, low-fat diet.

A 2013 study found similar results (emphasis in bold):

Participants averaged (mean ± SD) 46.8 ±12.4 years of age; 95.5% were Caucasian and the majority were married (64.4%), and highly educated (60.7% have a college degree or higher). BMI reported at registry entry was 25 ±4.6 and maximum (pre weight loss) BMI was 36.3 ±8.1. Weight loss reported on registry entry was 32.3 ± 16.7 kg and weight maintenance duration of the 13.6 kg minimum weight loss was 5.7±7.4 years. Average weekly energy expenditure on registry entry was 2621 ± 2252 kcal/wk and average daily energy intake on registry entry was 1400±584 kcals/day.

The 1,400 kcal/day figure was corroborated by a 2012 study which divided a subset of NWCR subjects into four categories. All the categories consumed between 1,350–1,450 kcal/day on average. The most successful category, 3, had an average BMI of 23.3 and had maintained this lower weight for a mean of 133 months, making them outliers even among NWCR participants. Group 3 is also close to a 50–50 split between men and women, so optimistically, we can infer 1,200 kcal/day for women and 1,600 kcal/day for men. The reported 1,420 daily kcal figure does seem legitimate in the sense they had been weight-stable for so long that it cannot be explained by trying to lose weight, but rather that is to maintain homeostasis. By comparison, Keys’ subjects started at a BMI of 23 whilst eating a little over 3,300 kcal/day.

The typical NWCR subject has a reported BMI of 23–26, so this is again consistent with Reddit and other social media reports: around 1,600–2,000 kcal/day for a formerly obese male to maintain a high-normal to slightly overweight BMI. Obviously, not 1,800 for all, but probably for a significant percentage as modeled by a bell curve with a mean of 1,800 and a variance of 200. This is pretty disappointing and hardly sounds sustainable for most dieters, even if it works for some. It is also consistent with the 1992 study: if subjects were still obese at 2,000–2,400 kcal/day, presumably they would have to cut to much lower, likely around 1,200–1,400 kcal/day for women to be normal weight. This can explain the high failure rate of dieting. For men at least, this means having to recreate the Minnesota Starvation Experiment for life. I dunno about you, but this sounds unsustainable for all but the most dedicated of dieters. And indeed, NWCR participants represent only a tiny minority of overall dieters, who fail.

As discussed earlier, nutrition and dieting studies of normal-weight subjects, or subjects who were never obese, may not generalize to presently obese subjects, or those who used to be obese. For higher BMIs (like above 35), the discrepancy between expected and actual results may be especially great when compared to non-obese controls. Indeed, Dr. Keys specifically recruited non-overweight or obese subjects. They were also selected for signifiers of high IQ, such as being much better educated than the general population at the time, as they were required to follow instructions closely and keep detailed records. Applicants who failed to meet Keys’ standards were rejected, leaving only 36 out of 400 who applied.

For the era, Keys’ participants were exceptional. So why would we expect 20-year-olds who were never obese or overweight in the 1940s to be a facsimile of obese people today? Despite not starting obese, Keys’ subjects would make perfect or model subjects for any modern weight loss or coaching program, whereas many of today’s obese dieters, such as on Reddit, struggle to lose much of anything. This is why so many online dieting or nutrition coaches have to cycle through so many clients until someone eventually loses enough weight for the putative “before and after” picture. Because their clients started out obese, we cannot expect the same rapid weight loss as Keys’ subjects saw. How can we possibly expect the exact same results when we’re talking about totally different circumstances?

This goes to show the power of survivorship bias: thousands of people apply popular ‘commonsense’ weight loss advice (“e.g. just cut sugar…move more”). The vast majority fail. The few who succeed tout their success as ‘evidence’ that said advice works for the general population. If you see a diet coach displaying an impressive weight loss transformation of a client, the relevant question to ask is how many clients he has (it’s probably a lot).

For obvious ethical reasons, the Minnesota Starvation Experiment will never be reproduced today, but there is a huge sample size of dieters today, such as on social media and the NWCR. We can still learn from them even if they are not under the same exacting conditions as the Minnesota Starvation Experiment. But it’s clear Keys’ subjects are hardly representative of America’s obese or overweight, such as on Reddit, who struggle to lose weight on 1,500 kcal, let alone become emaciated. Maybe something has gone wrong in the past 100 years, or there are selection effects. More needs to be done to investigate this, maybe for a future blog post.