Investigations › Frameworks
Investigation No. 20 · Frameworks
“Calories in and calories out is everything, or it is nothing”
Energy balance holds. What the argument is actually about is whether diet composition changes the out side, and on that two serious research groups disagree.
At a glance
- Verdict
- Misleading framing
- Certainty in the evidence
- Moderate, starting at High and moved by 1 domain
- Evidence base
- 3 randomised trials, 792 participants · 3 studies read · 792 people
- Last reviewed
- 19 Sept 2026
- Next review due
- Sept 2027
How people actually say it
- “It is just calories in, calories out.”true, and unhelpfully incomplete
- “Calories do not matter, hormones do.”false as stated, pointing at something real
Why this verdict and not the next one
Filed as misleading framing rather than true or false, because both stated positions are wrong in the same way. The law is not in dispute; what is in dispute is a narrower empirical question that neither slogan is capable of expressing.
Where the claim came from
From a real physical law being used to end a conversation it cannot end. Energy balance is arithmetic and nobody serious disputes it. The useful question is what determines the two sides of the equation, and a slogan in either direction is a way of avoiding that question.
What the studies did
Hall matched calories exactly in a metabolic ward and found cutting fat produced more fat loss than cutting carbohydrate, which is the opposite of the hormonal prediction. Gardner ran 609 people for a year and found a 0.7 kg difference between diets, with pre-specified tests of genotype and insulin secretion both coming back null. Ebbeling, after weight loss, found that lowering carbohydrate raised total energy expenditure measurably.
Why these papers and not others
These three were chosen because each was designed to test the mechanism rather than to compare adherence. Hundreds of diet-comparison trials exist and almost none can move this question, because they measure whether people stuck to the diet rather than what the diet did.
The full search and appraisal protocol, including what gets excluded and why, is set out in the method.
The evidence, study by study
| Study | Design | n | What it found | What it cannot say |
|---|---|---|---|---|
| Hall 2015 | Randomised crossover in a metabolic ward, calories matched, every meal controlled | 19 | Carbohydrate restriction produced 53 g/day of body fat loss; fat restriction produced 89 g/day (P = 0.002 favouring fat restriction). | What happens beyond six days, which the authors’ own modelling suggests would narrow the difference |
| Gardner 2018 | Randomised trial, healthy low-fat versus healthy low-carbohydrate, 12 months, free living | 609 | Weight change was -5.3 kg on low fat and -6.0 kg on low carbohydrate, a difference of 0.7 kg (95% CI -0.2 to 1.6). Neither genotype pattern nor insulin secretion predicted who did better. | People with diabetes, who were excluded |
| Ebbeling 2018 | Randomised trial of 60, 40 or 20 percent carbohydrate for 20 weeks after 12 percent weight loss | 164 | Total energy expenditure rose by 52 kcal/day for every 10 percent reduction in carbohydrate (95% CI 23 to 82), reaching 278 kcal/day per protocol at the low-carbohydrate extreme. | Whether this translates into weight over years, which was not measured |
How the certainty grade was reached
Certainty starts at a level set by the study design, then moves down for every domain that fails a written threshold and up for every domain that clears one. The arithmetic is shown rather than asserted.
What the evidence cannot say
Settle the maintenance question, which remains genuinely open. It also cannot tell you which diet you will stick to, and adherence is plausibly the dominant factor outside a trial.
One person, worked through
Two people arguing, one saying it is only calories and the other saying calories are irrelevant. Both are answering different questions. The first is right about the arithmetic and silent on what drives intake and expenditure. The second is wrong as stated and pointing at a real effect on the expenditure side that one good trial has measured.
Whether this applies to you
A verdict is a reading of a population. Whether it describes you depends on the things below, and any one of them can change what the right move is for you.
- Which question you are asking whether a diet loses more fat, and what happens to expenditure after loss, have different answers
- Whether you have diabetes since every trial here excluded it
- What you can actually sustain the variable none of these trials was designed to predict
What would change this entry
Independent replication of the Ebbeling expenditure result by a group not involved in the original, which would move the maintenance question from unsettled toward settled.
Read this before you act on it
This is an appraisal of published evidence, not medical or nutrition advice, and it is not a substitute for care from a clinician who knows your history, your medications and your labs. Nothing here is a diagnosis, a prescription or a treatment plan, and reading it does not create a clinician and patient relationship.
Do not start, stop or change a medication, a supplement or a diet on the strength of this page. If you are pregnant, breastfeeding, managing kidney or liver disease, diabetes, an eating disorder or any chronic condition, the general reading above may not describe you at all. Talk to your physician or a registered dietitian first.
The verdict above was true to the evidence on 19 Sept 2026. Evidence moves. Check the review date.
References
- Hall KD, 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):427-436. PMID 26278052.
- Gardner CD, et al. Effect of low-fat vs low-carbohydrate diet on 12-month weight loss: the DIETFITS randomized clinical trial. JAMA. 2018;319(7):667-679. PMID 29466592.
- Ebbeling CB, et al. Effects of a low carbohydrate diet on energy expenditure during weight loss maintenance: randomized trial. BMJ. 2018;363:k4583. PMID 30429127.
