Four diets, 811 people, two years, no winner#
No ratio of carbohydrate to fat has ever been shown to beat another for weight loss when calories are matched. The best test of the question randomized 811 overweight adults to four diets — fat/protein/carbohydrate targets of 20/15/65, 20/25/55, 40/15/45 and 40/25/35 percent of energy — and followed them for two years with group and individual counseling. Everyone lost about 6 kg by six months and began regaining after twelve. At two years the four arms were separated by less than a kilogram, and every macronutrient comparison came back null: 3.0 vs 3.6 kg for low vs high protein, 3.3 kg for both fat levels, 2.9 vs 3.4 kg for the highest vs lowest carbohydrate, P>0.20 throughout1.
So the practical answer is: pick the split you can live with, get the calorie total right, and keep protein adequate — that last one being the single exception, and a smaller one than the internet implies. But the reason this question refuses to die is not stupidity. It is that the headline result above has a limitation its own investigators published, and that when researchers do enforce a split under laboratory control, differences appear. Both facts are real. Here is how they fit together.
The fine print: the four diets converged into roughly one diet#
The body-composition arm of that same trial measured 424 participants by DXA and 165 by CT, and reported something the headline never carries. The protocol aimed for contrasts of 10 percentage points of energy in protein, 20 in fat and 30 in carbohydrate. What people actually ate was nothing like that. Self-reported contrasts came to about 2% protein, 8% fat and 14% carbohydrate at six months — and by two years, 1%, 7% and 10%2. Urinary nitrogen backed this up: the protein arms differed by 11 g of protein a day, two or three percent of energy, not ten.
That matters enormously for how you read the null. A trial in which the 15%-protein group and the 25%-protein group both ended up eating about 19% protein has not demonstrated that protein doesn't matter. It has demonstrated that free-living people drift back toward their habitual eating pattern, which is a finding about adherence, not about metabolism. The comparison that survives is the fat–carbohydrate one, where a 7–10 point contrast did persist and still produced nothing: fat mass, lean mass, visceral fat and liver fat all moved identically across arms (P ≥ 0.23 at two years)2.
The same reading applies to the year-long low-fat versus low-carbohydrate comparison that reached the same verdict — the details are in how many carbs you actually need. Free-living trials are the right design for the question most people are asking ("which diet should I follow?") and the wrong design for the question the argument is about ("does macronutrient composition change physiology?").
Lock people in a metabolic ward and low-fat wins the short run#
For the physiological question you need control that outpatient trials cannot deliver. Nineteen adults with obesity lived in a metabolic ward, exercised daily, and ate a balanced baseline diet for five days before having exactly 30% of their energy removed for six days — either by cutting carbohydrate 60% or by cutting fat 85%. Each person did both, in randomized order, with fat loss computed from intake minus oxidation measured in a metabolic chamber.
Carbohydrate restriction did what its advocates predict: fat oxidation rose and stayed up. It still lost less fat. Carbohydrate restriction produced 53 ± 6 g of body fat loss per day; fat restriction produced 89 ± 6 g/day, and the difference was significant (p = 0.002)3.
That is the inverse of the folk belief, and it deserves the same skepticism as the belief it contradicts. Six days is six days. The authors' own modeling concluded that "the body acts to minimize body fat differences with prolonged isocaloric diets varying in carbohydrate and fat" — the short-run gap is a transient of shifting fuel use, and it closes. Which is precisely why the two-year outpatient trials find nothing: they are measuring after the transient has resolved.
The live dispute: does cutting carbs raise your energy expenditure?#
This is where the field genuinely splits, and it is worth understanding because it is the only version of the macro-ratio question with real stakes. The carbohydrate-insulin model holds that a high-carbohydrate diet drives insulin, insulin traps energy in fat cells, and the resulting fuel shortage lowers energy expenditure — meaning the same calories on different splits should not produce the same result.
The strongest test came from a 20-week feeding trial. After a run-in that took 164 adults down 12% of their body weight, they were randomized to maintenance diets of 60%, 40% or 20% carbohydrate with protein held constant. Total energy expenditure, by doubly labeled water, ran 209 kcal/day higher on low-carbohydrate than high-carbohydrate in intention-to-treat analysis (95% CI 91 to 326) and 278 kcal/day in per-protocol analysis (95% CI 144 to 411)4. Two hundred calories a day for free would be a genuinely large effect.
Then three researchers pulled the posted data and code and re-ran the analysis the trial had pre-registered. The disagreement is not vague, and it has exactly two nameable parts. First, the anchor: the registered protocol compared expenditure against each person's pre-weight-loss baseline, and the published paper instead used the post-weight-loss period as the reference, a change made after all subjects had finished. Re-run against the registered baseline, expenditure fell in every arm — 240 ± 64, 322 ± 66 and 356 ± 67 kcal/day for low, moderate and high carbohydrate — with p = 0.435. Second, and more damaging, the reanalysts found no significant diet difference in measured carbon dioxide production under either analysis plan. Doubly labeled water measures CO₂; converting it to calories requires an assumed respiratory quotient, which necessarily differs between a low-carb and a high-carb diet.
The expenditure gap was not in the measurement. It was in the conversion factor applied to the measurement.
Both sides have declared interests, and the same standard applies to each. The trial's senior author receives royalties from books recommending low-glycemic-load diets4; the lead reanalyst discloses a running series of public debates against him on this exact model5. Neither disclosure decides anything. What decides it, for now, is that the effect appears under one analysis plan and disappears under the plan that was registered in advance — and that the trial with the physiologically tightest control found the opposite sign.
Protein is the one macro that behaves differently#
Everything above concerns the carbohydrate-to-fat trade. Protein is a different variable, and treating it as a third slice of the same pie is the mistake that makes macro debates circular. Higher protein in a deficit does not create more weight loss so much as redirect it: at identical calories, the composition of what you lose shifts toward fat and away from lean mass. That case, its effect sizes, and how modest they really are belong to protein for fat loss.
Notice that this does not contradict the null above. The trial that found no protein effect never achieved a protein contrast worth the name — a one-percentage-point separation at two years cannot test a hypothesis about protein. The strongest supporting evidence comes from controlled-feeding trials where the difference was actually delivered.
So what should you actually set?#
| Decision | What the evidence supports |
|---|---|
| Total calories | The variable that determines weight change (calorie deficit) |
| Protein | Set it first and hold it — the one macro with a defensible composition effect |
| Carbohydrate vs fat | Free choice inside the official ranges; no split has beaten another at matched calories |
| Which split to pick | The one you'll still be eating in six months |
The two remaining constraints are floors, not ratios: carbohydrate has a level below which hard training suffers, and fat has a floor below which vitamin absorption and essential fatty acids are compromised. Stay above both and the space between them is genuinely yours. If a plateau has you reaching for a new ratio, the arithmetic worth checking first is in adjusting macros to break a plateau — and it is usually about the total, not the split.
The unsatisfying conclusion is the correct one. Two years of counseling moved four very different prescriptions to within a kilogram of each other, and attendance at sessions predicted weight loss at 0.2 kg per session attended1 — showing up mattered measurably more than the ratio anyone showed up to follow. The macro split is the part of a diet that feels most like a decision and does the least work. The definitions and official ranges behind all of it are in macronutrients explained.
FAQ#
Does a low-carb diet burn more calories than a low-fat one?#
The claim rests on one 20-week trial reporting 209–278 kcal/day higher expenditure on low-carbohydrate4. Reanalyzed under the trial's own pre-registered plan, the difference was not significant (p = 0.43), and no diet difference appeared in the underlying CO₂ measurement at all5. Treat it as unresolved, not established.
Is 40/30/30 better than any other macro split?#
No trial supports it over alternatives. Four prescribed splits ranging from 20% to 40% fat and 35% to 65% carbohydrate produced weight losses within a kilogram of each other after two years, with P>0.20 for every comparison1. Any ratio inside the official ranges is defensible; the numbers 40/30/30 have no special evidence behind them.
If macro ratios don't matter, why did cutting carbs work for me?#
Most likely because it cut your calories. Removing a whole food category removes eating occasions, and low-carbohydrate eating tends to displace refined, easy-to-overeat foods. That is a real and useful mechanism — it just runs through how much you ate, not through a metabolic advantage, which is why the effect vanishes in trials that hold calories constant.
Sources#
- Sacks FM, Bray GA, Carey VJ, et al. Comparison of weight-loss diets with different compositions of fat, protein, and carbohydrates. N Engl J Med. 2009;360(9):859-73.
- de Souza RJ, Bray GA, Carey VJ, et al. Effects of 4 weight-loss diets differing in fat, protein, and carbohydrate on fat mass, lean mass, visceral adipose tissue, and hepatic fat: results from the POUNDS LOST trial. Am J Clin Nutr. 2012;95(3):614-625.
- Hall KD, Bemis T, Brychta R, et al. Calorie for Calorie, Dietary Fat Restriction Results in More Body Fat Loss than Carbohydrate Restriction in People with Obesity. Cell Metab. 2015;22(3):427-36.
- Ebbeling CB, Feldman HA, Klein GL, et al. Effects of a low carbohydrate diet on energy expenditure during weight loss maintenance: randomized trial. BMJ. 2018;363:k4583. (Senior author declares book royalties for low-glycemic-load diets.)
- Hall KD, Guo J, Speakman JR. Do low-carbohydrate diets increase energy expenditure? Int J Obes (Lond). 2019;43(12):2350-2354. (Lead author declares an ongoing debate series opposing the carbohydrate-insulin model.)
