2026-09-09 · What Studies Say · 1438 words · Lab-Verified Data

What Studies Say: Low Carb and Energy Expenditure (BMJ 2018)

What the Study Asked

Most diet trials compare plans that produce different amounts of weight loss, which makes it impossible to separate the effect of the food from the effect of the pounds. This trial removed that confound deliberately. Published in The BMJ in November 2018 (Ebbeling et al., BMJ 2018;363:k4583, doi:10.1136/bmj.k4583), it asked a narrower question: once weight loss is held equal, does the amount of carbohydrate in the diet change how many calories the body burns?

The question comes from the carbohydrate-insulin model, which proposes that diets high in fast-digesting carbohydrate push fuel into storage and lower energy expenditure, and that the effect is largest in people who secrete more insulin. The model had been argued about for years mostly through indirect evidence; this trial was designed to test it with the most precise tools available.

How It Was Built

The design has three features that make it worth reading closely. First, all three diets were fed to participants, with protein fixed at 20% of energy, so nobody was guessing about what they ate. Second, weight was clamped: after a run-in phase in which participants lost about 12% of body weight, calories were individually adjusted throughout the 20-week test phase to keep weight within 2 kg. That is what makes the comparison fair — everyone was maintaining, not losing. Third, the primary outcome was total energy expenditure measured by doubly labeled water, a technique that tracks carbon dioxide production over days and is the accepted gold standard for free-living energy burn.

164 adults aged 18–65 with a body mass index of 25 or higher were randomized across two US sites between August 2014 and May 2017: 54 to a high-carbohydrate diet (60% of energy from carbohydrate), 53 to a moderate one (40%), and 57 to a low-carbohydrate one (20%). Fat filled the gap as carbohydrate dropped, so the low-carb arm was a high-fat diet. The sample was also split into thirds by pre-weight-loss insulin secretion, which allowed the researchers to test the model's specific prediction about insulin response.

The Numbers

The headline result is a difference in daily calories burned, at identical body weight.

Two honest notes about this paper. The journal published a correction notice in November 2020, which is linked from the article page — we flag it because citing a corrected paper without mentioning the correction is the kind of omission that erodes trust, and readers should always be able to find it. And the primary finding is a group average: 209 kcal/day is not a promise about any individual's metabolism, and a 95% confidence interval spanning 91 to 326 is a reminder of how much variation sits inside that average.

What It Means If You Eat Low-Carb

Read narrowly, the trial says something specific and useful: during weight maintenance, after a period of weight loss, a lower-carbohydrate diet was associated with a higher daily energy expenditure than a high-carbohydrate diet of the same calorie content and the same protein. It does not say low-carb diets cause weight loss by themselves, because the study was not designed to test weight loss — nobody was losing weight during the 20-week test phase by design. It also does not say the effect is permanent; 20 weeks is a medium-term window, not a life.

The insulin-secretion finding is the part that most interests us as a data team, because it is the first place in the paper where the model makes a falsifiable prediction and the data line up with it. Participants who secreted the most insulin after a glucose challenge showed the largest expenditure gap. That is a subgroup result from a modest sample, so it should be treated as a lead rather than a settled rule — but it is a lead with a plausible mechanism attached, which is more than most diet debates get.

For readers deciding between keto and low-carb, this paper is a reason to treat them as a spectrum rather than two camps. The 20% carbohydrate arm is closer to what most people mean by "low-carb" than to strict keto; the paper's own framing is about the carbohydrate-to-fat ratio, not about ketosis as a threshold. Our keto versus low-carb explainer covers where those lines fall and why our verdicts carry two badges instead of one, and the DIETFITS trial read covers the other major question — weight loss itself — where the two approaches came out even over twelve months.

What 20% of Energy From Carbohydrate Looks Like

Percentages hide portions, so we priced it against our own verified library. A 2,000 kcal day at 20% carbohydrate is about 400 kcal, or roughly 100 g of carbohydrate for the whole day. Against that budget: 100 g of potato costs 15.39 g net carbs (FDC #170026), 100 g of raw sweet corn 16.7 g (FDC #169998), and 100 g of raw cauliflower 2.97 g (FDC #169986). Spend the entire 100 g on potato and you get six and a half 100-gram servings for the day, with nothing left for vegetables, dairy or nuts. Spend it on cauliflower and the arithmetic stops being restrictive at all — the same 100 g buys about 55 servings, which is 5.5 kg of the stuff. That contrast, not the percentage, is what a low-carbohydrate day actually feels like on a plate.

Source Box

Every figure above appears in the published abstract or the USDA record of the paper; subgroup results are labelled as subgroup results, and confidence intervals are quoted where the paper reports them. Evidence grade: randomized controlled trial with controlled feeding and an objective primary outcome, 20 weeks, 164 randomized. This article is a plain-language summary of published research by a data editorial team — it is not medical advice.

Did this study show that low-carb diets burn more calories?

It showed that, among 162 adults maintaining a 12% weight loss under fully controlled feeding, those assigned to a 20%-carbohydrate diet expended about 209 kcal/day more than those on a 60%-carbohydrate diet (95% CI 91 to 326), with a per-protocol gap of 278 kcal/day. The difference was measured by doubly labeled water, the gold-standard method. It is a group average over 20 weeks in adults with a BMI of 25 or higher — not a guarantee about any individual, and not a weight-loss result, since participants were held at stable weight by design.

Does the result depend on how much insulin you secrete?

Partly, in this trial. The largest low-versus-high carbohydrate gap appeared in the third of participants with the highest insulin secretion 30 minutes after a glucose challenge: 308 kcal/day in the intention-to-treat analysis and 478 kcal/day per protocol. That is a subgroup finding from 164 participants, so it is a hypothesis-strengthening result rather than a rule you can apply to yourself without measuring anything. Source: Ebbeling et al., BMJ 2018;363:k4583.

Was the study corrected after publication?

Yes — the journal published a correction notice on November 3, 2020, linked from the article page. We mention it because readers deserve to know when a cited paper has been corrected, and we quote the corrected published figures above. Anyone relying on this trial should read the correction alongside the paper. Source: BMJ 2018;363:k4583, correction notice November 2020.

Related reading

Filed under researchweight-loss