Weight change follows energy balance. Take in less energy than you spend and you lose tissue; take in more and you store it. That part is settled. What is not settled is whether anyone can measure either side of that equation accurately enough to run their life on it. Self-reported intake is off by hundreds of calories a day. Food labels are legally allowed to understate. The number your body spends moves while you diet, and it moves differently in different people.
The version of this article that stood here for years handed readers a formula, an activity multiplier and a target, and implied the rest was willpower. That was the wrong shape of advice.
The arithmetic is right; the accounting is not
Energy balance is not a theory about dieting. It is bookkeeping, and it holds. The trouble is that both entries in the ledger are estimates, and both are biased in the same direction.
The cleanest demonstration is old. In a 1992 study in the New England Journal of Medicine, Lichtman and colleagues studied 224 adults seeking obesity treatment and examined a subgroup of 10 who reported eating under 1,200 calories a day without losing weight. Measured against actual expenditure, that group under-reported food intake by 47 percent and over-reported physical activity by 51 percent. Their resting metabolic rates were within 5 percent of predicted. Nothing was wrong with their metabolisms. The records were.
This was not a quirk of one small sample. A 2024 analysis in Nature Food built a predictive equation from 6,497 doubly labelled water measurements and applied it to two large national dietary surveys. Roughly 27 percent of dietary reports fell outside a plausible range, mostly through under-reporting, and the discrepancy widened as body size increased.
Four places the numbers go wrong
Under-reporting is the largest error, but it is not the only one, and the others compound it.
Label tolerance. Under 21 CFR 101.9, a packaged food is misbranded for calories only when the measured content is more than 20 percent in excess of the declared value. A bar labeled 200 calories can legally contain 240. Restaurants are looser still: a 2011 research letter in JAMA laboratory-tested 269 items from 42 chain restaurants and found stated calories were close to accurate on average, but individual items scattered widely — 19 percent contained more than 100 calories per portion above the stated figure.
Metabolic adaptation. Energy expenditure falls during weight loss by more than the lost tissue alone predicts. A review in Proceedings of the Nutrition Society put the typical figure at a 20 to 25 percent drop in total expenditure after 10 percent weight loss, of which 10 to 15 percentage points exceed what body composition change accounts for — commonly 100 to 300 calories a day.
NEAT. Non-exercise activity thermogenesis is everything you burn that is not sleeping, eating or deliberate exercise. In a 1999 Science study, Levine and colleagues overfed 16 adults by 1,000 calories a day for eight weeks. The change in NEAT ranged from minus 98 to plus 692 calories a day, and fat gain varied roughly tenfold across the group on an identical surplus. Nobody chose those responses.
| Source of error | Typical magnitude | Direction |
|---|---|---|
| Self-reported intake | Hundreds of calories a day; up to 47% in one studied subgroup | Understates intake |
| Packaged food labels | Up to 20% above the declared calories, legally | Understates intake |
| Metabolic adaptation | Roughly 100–300 kcal/day after meaningful loss | Overstates the deficit |
| NEAT variation | Spanned about 790 kcal/day between individuals under identical overfeeding | Both |
What the longest trials found
If diet composition were the lever, controlled trials would show it. They mostly do not.
POUNDS LOST, published in the New England Journal of Medicine in 2009, randomized 811 adults to four diets varying fat (20 or 40 percent) and protein (15 or 25 percent) and followed them for two years. Average loss was about 6 kg at six months and 3.0 to 3.6 kg at two years, with no significant difference between any pair of diets. What did predict the result was attendance: about 0.2 kg of loss per group session attended. Participants also never reached their assigned targets, missing by an average of 14 percentage points on carbohydrate at six months.
The largest synthesis reached the same place. A 2020 network meta-analysis in The BMJ pooled 121 randomized trials and 21,942 participants across 14 named diet programs. Low-carbohydrate and low-fat patterns each produced about 4 to 5 kg more loss than a usual diet at six months. By 12 months that weight loss had diminished across every pattern and most cardiovascular improvements had disappeared, with a modest LDL reduction persisting for the Mediterranean pattern. The authors described the differences between diets as trivial to small.
Earlier, a 2005 trial in JAMA assigned 160 adults to Atkins, Ornish, Weight Watchers or the Zone for one year. Mean loss was modest and similar across all four, and the variable that predicted an individual’s result was their adherence score, not which diet they had been given. That finding has held up for twenty years and is the reason this blog treats the structural reasons diets fail as a separate subject from which diet you pick.
Composition matters less than adherence — but food form matters
Saying composition matters less than adherence is not the same as saying food choice is irrelevant. It changes how much you eat before any counting begins.
In a 2019 inpatient crossover trial in Cell Metabolism, Hall and colleagues fed 20 adults ultra-processed and unprocessed diets for two weeks each, matched for presented calories, energy density, macronutrients, sugar, sodium and fiber, and let them eat freely. On the ultra-processed diet participants ate 508 calories a day more and gained 0.9 kg; on the unprocessed diet they lost 0.9 kg. Nobody was tracking anything. The food did the work.
The 3,500-calorie rule was always wrong
The rule that a pound of fat equals 3,500 calories, so a 500-calorie daily deficit yields a pound a week indefinitely, treats the deficit as fixed. It is not. In a 2011 modelling paper in The Lancet, Hall and colleagues showed that a sustained 2,000-calorie daily reduction predicts about 22 kg of loss in the first year under the static rule but roughly half that under a dynamic model that accounts for adaptation.
Their replacement rule of thumb is less satisfying and more accurate: a sustained change of about 100 kilojoules a day produces an eventual weight change of roughly 1 kg, with about half arriving within a year and most of the rest over three. Plateaus are not failures of discipline. They are the arithmetic arriving on schedule.
Where this evidence is thin
The metabolic adaptation literature is weaker than its cultural footprint suggests. The most-cited demonstration, Fothergill and colleagues in Obesity in 2016, followed 14 former competitors on a televised weight-loss program and found resting metabolic rate about 500 calories a day below prediction six years later. Fourteen people, an extreme intervention, no control group. It should not carry the weight placed on it.
Others have argued the effect is partly an artifact of when it is measured. A 2020 analysis in the American Journal of Clinical Nutrition contended that measurable metabolic adaptation appears mainly while people are in active negative energy balance and largely resolves once weight stabilizes. That reading is contested, and the honest position is that the effect is real during dieting, smaller than folklore suggests, and of uncertain permanence.
The trials have their own limits. Intake in almost all of them is self-reported, which is the exact instrument shown above to be unreliable. Group means also hide enormous individual spread: within a single arm of a well-run trial, some participants lose a great deal and others gain. Very few studies of any design follow people beyond two years.
What to do with all this
Counting has one honest use: calibration. A week or two of careful weighing tells most people something they did not know about portion size, cooking oil or liquid calories. Keeping it up for years does not add much information and, for some people, adds harm.
Beyond that, the levers with the most support are unglamorous: meals built around foods that are hard to overeat, enough protein to get through the afternoon, resistance training in the week so that what is lost skews toward fat, and a measure other than the scale, such as waist-to-height ratio, which moves for reasons the scale does not explain. And note that exercise does not simply add to your deficit — appetite and spontaneous movement adjust around it.
One caution worth stating plainly: if tracking food has ever become something you find difficult to stop, or is accompanied by guilt about eating, that is a reason to put the app down and speak to a physician or registered dietitian. Counting is a tool, and it is the wrong tool for a substantial number of people.
Common questions
Do calories still matter if I only eat whole foods?
Yes, energy balance still governs weight change. What changes is that whole foods make it much harder to overshoot without noticing, which is what the 2019 Cell Metabolism crossover trial demonstrated. You are not exempt from the arithmetic; you are just less likely to lose track of it.
Have I damaged my metabolism by dieting?
Almost certainly not in the way the phrase implies. Energy expenditure does fall during and after weight loss, typically by 100 to 300 calories a day beyond what lost tissue predicts, and there is active scientific disagreement about how long that persists. That is a headwind, not a broken system.
Why did my weight loss stall around three months?
Because the deficit shrank while you were in it. A smaller body costs less to run, and reported intake tends to drift upward over time even when effort feels constant. The dynamic model in The Lancet predicts exactly this shape of curve.
Is a 500-calorie daily deficit the right target?
It is a reasonable starting estimate and a poor prescription, because intake and expenditure cannot be measured to that precision. Most people do better with a way of eating they can hold for a year, adjusted from what the scale and tape measure actually do over months.
Where this leaves the topic
Energy balance explains weight change. Calorie counting is one attempt to operationalize it, and it is a lossy one — good for a short calibration, unreliable as a long-term system, and actively unhelpful for some people. Building meals you would eat anyway tends to outperform arithmetic you cannot sustain.
Greens Plus makes a green superfood powder. It is food, not a weight-management product, and it is best thought of as one way to widen the range of plants in a day rather than as anything that changes the equation above; you can read what is in it at greensplus.com.
About the Author
