The number 278 has had a remarkable life. It was born in a BMJ paper from David Ludwig’s research team at Harvard, became a ScienceDaily headline read by millions, and landed in YouTube lectures as settled proof that cutting carbs burns hundreds of extra calories. But when independent researchers downloaded the original raw data and reanalyzed it, the number got smaller.
A 164-person controlled-feeding trial published in the BMJ — every meal prepared, every calorie counted for 20 weeks — found that people eating low-carb burned 209 extra calories per day compared to those on high-carb. The per-protocol number was even higher: 278.
That 278 became a headline. A ScienceDaily article. A sound bite in YouTube lectures. A round number passed from creator to creator until it felt like settled science.
Then researchers at the NIH downloaded the same raw data. They applied a correction for a known bias in the measurement method — the tool used to estimate calories burned makes a fuel-mix assumption that shifts depending on what people eat. After that single correction: 139 calories per day. No longer a reliable finding.
They went further. Two participants had implausible energy accounting — more than 600 unaccounted calories per day. Removing them: 46 calories.
The number that launched a thousand keto clips had four independent fault lines. Each correction stripped one methodological assumption. The 278 didn’t fall because of one problem. It had four.
Which Ruler Changes the Answer
Two tools exist for estimating how many calories a person burns. One, doubly labelled water, works while people go about their normal lives but makes fuel-mix assumptions that shift with diet composition. Its measurement variability runs 8 to 15 percent.
The other tool, a sealed metabolic chamber, measures gas exchange directly. Its variability: 2 to 3 percent.
A separate 17-person metabolic ward study put both tools to work on the same question. Every participant locked in, every bite controlled, identical calories on both diets. The chamber found a 57-calorie-per-day increase on the ketogenic diet. The same participants, measured with doubly labelled water at the same time, showed 151.
Same bodies. Same diet. One tool said 57. The other said 151.
The entire gap between the two research camps — 209 versus 57 — starts to look less like a disagreement about biology and more like a choice of measuring tool.
The Gas Their Bodies Actually Exhaled
One more layer from the reanalysis.
The original trial registered an analysis plan before data collection began. Under that plan, the diet effect on energy expenditure was not reliable — the data could not reliably distinguish the diet effect from chance. The plan was changed after all 164 participants had finished the study. Under the new plan: the result held up.
This isn’t fabrication. The study was peer-reviewed and published in one of medicine’s most respected journals. But a result that only becomes significant after changing the analysis plan raises a question: what does the raw physical evidence show?
The answer was in the CO2 data. Every participant’s exhaled gas was measured continuously. CO2 production is the physical measurement underneath the calorie estimate — the thing the conversion formula turns into a number. The CO2 showed no significant difference between diets, regardless of which analysis plan was used.
The calorie gap between low-carb and high-carb existed in the conversion formula. Not in the lungs.
Probably Real, Definitely Small
After the shrinking number, the ruler problem, and the changed plan — is there anything left?
Probably, yes.
Even Kevin Hall — the NIH researcher behind both reanalyses — concedes he cannot exclude a real effect in the 50 to 140 calorie range. Both studies found some increase in energy expenditure on low-carb. The direction is consistent across every analysis. It’s the magnitude that’s in dispute.
So what does 50 to 150 extra calories per day actually look like? A banana. A small handful of nuts. Over a month, that’s a theoretical 0.2 to 0.6 kilograms of additional fat loss — less than your daily weight fluctuation from water and food.
In every trial we examined lasting longer than a few months, low-carb did not produce more actual fat loss. The largest diet-comparison trial (609 adults, 12 months) found individual variation within each diet was 57 times larger than the difference between diets.
Based on everything we examined: the edge probably exists. It’s too small and too uncertain to build a diet strategy around.
If you happen to prefer lower-carb eating, the evidence suggests a modest metabolic tailwind comes along for the ride. If balanced eating suits you better, the edge is a banana — not worth switching for.
Two Bets on One Hormone
The insulin question goes deeper than the population average. Within the 164-person trial, participants in the highest third for insulin response appeared to get a much bigger metabolic boost — 478 extra calories per day compared to 131 for the lowest third. A 3.6-fold variation based on one blood marker.
Intriguing. But unreplicated, measured with the same disputed tool, and contradicted by a separate 609-person trial that found insulin response did not predict who lost more weight on low-carb. An interesting signal, not a proven personalized strategy.
The theory behind both findings — the carbohydrate-insulin model — actually made two independent, testable predictions. One: carbs spike insulin, which drives hunger, which causes overeating. Two: carbs spike insulin, which suppresses energy expenditure, causing metabolic slowdown.
Prediction one was tested directly in a controlled metabolic ward. People on a high-carb diet spontaneously ate 689 fewer calories per day than those on keto — with identical hunger ratings. The hunger prediction didn’t just fail. It went the opposite direction.
Prediction two — the one this page just examined — partially holds. There’s probably a small metabolic effect. Magnitude disputed. Never translates to fat loss.
The model is half-right. And the half it’s right about is the half that matters least for your actual fat loss. The hunger side — the side that determines how much you eat day after day — is where the evidence went the opposite direction from what the model predicted.
That hunger evidence, and what it reveals about what actually controls how much ends up on your plate, is where the more consequential story unfolds.
50 to 150 extra calories per day is the energy in a banana to a small handful of nuts. Over a month, that adds up to a theoretical 0.2 to 0.6 kilograms of additional fat loss — less than your normal daily weight fluctuation from water and food. The translation is not "switch to low-carb for the metabolic edge." The translation is: if you're already eating lower-carb because it suits you, the evidence suggests a small metabolic tailwind exists. If you're eating balanced because that suits you better, the edge is too small and too uncertain to be worth switching for.