Calorie restriction and ageing: what the human trials measured, and what they were built to avoid

CALERIE randomised 218 non-obese adults to two years of a 25% target and achieved 11.7%. It moved thyroid hormone, inflammation markers, liver enzymes and body composition. It also moved bone mineral density down at three sites and fat-free mass down by 2.2 kilograms.

goodWell supported, with the usual caveats about effect size.Revised 13 Oct 2026

A white plate of green vegetables on a wooden table with an old cream kitchen scale and a folded linen cloth behind it, warm late-afternoon window light

Restriction is the only intervention in this field with a randomized human trial behind it. The trial also produced a bone mineral density loss at the lumbar spine, the hip and the femoral neck.

Calorie restriction is the only thing in the entire ageing field that has ever been shown to extend lifespan in mammals and then tested in a randomized controlled trial in humans. There is one such trial. It ran for two years, in 218 people, it achieved about half of its target, and it produced a results table with benefits and costs printed on the same page.

That is rare, and it should be read that way.

CALERIE, the design

The two-year randomized controlled trial published in 2015 in the Journal of Gerontology: Series A enrolled 218 people aged 21 to 51, with a mean body mass index of 25.1. One hundred and forty-three were assigned to a 25% calorie restriction target, 75 to eat ad libitum. The primary outcomes were resting metabolic rate adjusted for weight change and core temperature — measures chosen precisely because they are the classic lifespan predictors in rodents. Secondary were plasma triiodothyronine and tumour necrosis factor alpha.

The feasibility result is the part usually skipped: the restriction arm achieved 11.7 ± 0.7% actual restriction and maintained a 10.4 ± 0.4% weight loss, and 82% of that group completed the protocol against 95% of the control. Weight change in the control was negligible. So: a target of 25%, an achievement of under 12%, and two-thirds of the people stuck with it for two full years. Most people reading about this for the first time assume the trial failed adherence. It did not.

What moved, and what did not

Thyroid hormone fell. Plasma T3 decreased more in the restriction arm at both 12 and 24 months (p < 0.001). In rodent gerontology, a lower T3 is one of the most consistent correlates of extended lifespan, and this is the canonical mouse finding reproduced in humans.

Inflammation fell. Tumour necrosis factor alpha decreased significantly more, but only at 24 months (p = 0.02). Two years to move a single inflammatory marker, and it moved.

Resting metabolic rate fell — at 12 months, then not at 24. The residual metabolic rate, adjusted for weight change, decreased significantly more in the restriction arm at 12 months (p = 0.04) and the difference was no longer significant at 24 months. Core temperature differed little between groups.

Cardiometabolic risk factors and energy expenditure fell more, adjusted for weight change, with no adverse effect on quality of life.

That last clause deserves more attention than it usually gets. The feared consequence of eating less for two years is a diminished life, and the trial looked for it. It did not find it. That is a genuine and under-reported result.

Liver markers moved selectively, in a separate analysis published in European Journal of Nutrition. At 24 months, relative to the control group, alkaline phosphatase was down 7 ± 1 units per litre, gamma-glutamyl transferase down 0.11 log units, and bilirubin up 0.21 log. There was no between-group difference in ALT or AST (−1 ± 1 and +2 ± 2 units per litre, p > 0.99 and p = 0.68). And there were sex-by-treatment-by-time interactions, with the reductions in ALT and gamma-glutamyl transferase and the increase in AST appearing in men only.

So the honest version is: two of five liver markers moved, and the two that moved most were the two with the least dramatic clinical meaning, while the transaminases — the ones you would actually look at clinically — did not.

The costs, which are in the same trial

The bone analysis, published in the Journal of Bone Mineral Research, is the paper in this literature that most deserves to be read before anyone starts a diet.

Same 218 people, two years. Body weight −7.5 ± 0.4 kg in the restriction arm against +0.1 ± 0.5 kg in the control. Fat mass −5.3 ± 0.3 against +0.4 ± 0.4. Fat-free mass −2.2 ± 0.2 against −0.2 ± 0.2. All between-group differences p < 0.001.

Bone mineral density at 24 months:

lumbar spine: −0.013 ± 0.003 against +0.007 ± 0.004, p < 0.001
total hip: −0.017 ± 0.002 against +0.001 ± 0.003, p < 0.001
femoral neck: −0.015 ± 0.003 against −0.005 ± 0.004, p = 0.03

And bone turnover markers moved at 12 months in the wrong direction: C-telopeptide 0.098 ± 0.012 against 0.025 ± 0.015; tartrate-resistant acid phosphatase 0.4 ± 0.1 against 0.2 ± 0.1; bone-specific alkaline phosphatase −1.4 ± 0.4 against −0.3 ± 0.5. At 24 months only bone-specific alkaline phosphatase still differed.

The authors also note that the restriction group had lower levels of physical activity, and that multiple regression explained about 31% of the variance in the bone changes from body composition, hormones, nutrients and activity. Their conclusion: bone loss at clinically important fracture sites is a potential limitation of prolonged calorie restriction for extending lifespan, and further work is needed on whether it translates into fracture risk and whether exercise can prevent it.

Note the effect sizes, because they matter: those are fractions of a percent of bone density per year, in adults whose mean age was 37.9. A person who already has osteopenia is not in the population this was measured in, and a person who already has osteoporosis should not read this and think a decade of mild restriction is a reasonable trade.

Why restriction makes muscle harder to build

A 2020 study in the European Journal of Applied Physiology measured the endocrine response directly, and it is the mechanistic explanation for the fat-free mass loss above.

Participants completed three-day conditions: energy restriction with post-exercise carbohydrate, energy restriction with post-exercise protein, and an energy-balance control with post-exercise carbohydrate. After five sets of five back squats on day three, serial blood draws.

Growth hormone peaked at 2.6 ± 0.4 and 2.5 ± 0.9 times the control peak in the two restriction conditions. And yet insulin-like growth factor-1 concentrations fell 18.3 ± 3.4% and 27.2 ± 3.8% over the following 24 hours, against 7.6 ± 3.6% in the control. Sclerostin, the protein that inhibits bone formation, rose 19.2 ± 5.6% and 21.8 ± 6.2% in the first two days of restriction against 13.4 ± 5.9% in control — although those increases were attenuated and no longer significant after the squat session.

The authors’ reading: during caloric restriction there is considerable endocrine anabolic resistance to a single bout of resistance exercise, and it persists in the presence of post-exercise protein.

That is a short, three-day study of an acute response, and it cannot establish anything about chronic outcomes. What it does establish is that the growth hormone response is intact and the downstream response is not — which is a coherent explanation for losing 2.2 kg of fat-free mass while training.

For the older population specifically, the 2025 review in Obesity reaches similar conclusions from randomized trials in adults with obesity: lifestyle-based weight loss gives clinically meaningful reductions, lean mass loss is significant though fat loss is preferential, bone loss is also significant, and resistance training may attenuate the bone loss.

The mouse study that says the obvious thing nobody says

The 2024 paper in Nature is the best-designed animal study in this field and it contains a finding that gets lost in every summary: improving health and extending lifespan are not synonymous.

960 genetically diverse female mice, four interventions: 20% and 40% calorie restriction, and intermittent fasting of one and two days a week. Findings:

Both restriction and intermittent fasting extended lifespan in proportion to the degree of restriction.
Lifespan was heritable, and genetics had a larger influence on lifespan than the dietary intervention.
40% restriction gave the strongest lifespan extension — and produced loss of lean mass and changes in the immune repertoire that could confer susceptibility to infection.
Intermittent fasting did not extend the lifespan of mice that started with high body weight, and two-day intermittent fasting was associated with disruption of erythroid cell populations.
Health effects differed between interventions and showed inconsistent relationships with lifespan extension.
And the finding that should end a hundred arguments: reduced adiposity and lower fasting glucose — the classic metabolic improvements — were not associated with increased lifespan. The authors conclude that dietary restriction does more than simply counteract the negative effects of obesity.

That last line is the whole field in one sentence. If the metabolic improvements did not predict the lifespan gain, then a fasting app that lowers your glucose is optimising a marker that the best longevity data says is not the thing.

The strongest trait associations with lifespan in that study were also not metabolic: retention of body weight through periods of handling, an indicator of stress resilience; high lymphocyte proportion; low red blood cell distribution width; and high adiposity in late life.

The mechanism, in one paper

A 2024 paper in Nature Cell Biology isolates a specific molecular route, and it is the most likely candidate for the non-metabolic effect.

Spermidine levels rose across distinct regimens of fasting or calorie restriction in yeast, flies, mice and human volunteers. Blocking endogenous spermidine synthesis, genetically or pharmacologically, reduced fasting-induced autophagy in yeast, nematodes and human cells. Perturbing the polyamine pathway in vivo abrogated the lifespan and healthspan effects, and the cardioprotective and anti-arthritic consequences of fasting. Mechanistically, spermidine mediated these effects through autophagy induction and the hypusination of the translation regulator eIF5A.

The authors describe the polyamine-hypusination axis as a phylogenetically conserved metabolic control hub for fasting-mediated autophagy and longevity. It is a good mechanism and it is consistent across four species. It is also a mechanism, and no human trial has yet shown that blocking or restoring it changes a human outcome.

Meanwhile the CALERIE trial has produced a genomic data resource — whole-genome genotypes plus three-timepoint longitudinal DNA methylation, mRNA and small RNA from blood, skeletal muscle and adipose tissue across all 218 participants, 2,327 samples in total. That is a genuinely valuable resource, and the molecular signature of restriction in humans is still being worked out in it.

What these trials do not show

CALERIE did not measure lifespan, and cannot. It is a two-year trial with surrogate and predictor endpoints. A lower T3 and a lower resting metabolic rate are correlates of longevity in mice. In humans they are correlates of being in a deficit, and whether the human relationship survives contact with a real human life is the question the field has not answered.

The trial achieved 11.7% restriction against a 25% target, so every effect reported is an effect of roughly half the intervention, and it may underestimate the benefit or the cost. Either way, the headline number people quote — “25% calorie restriction extends healthspan” — describes an intervention nobody in the trial actually performed.

The participants were 21 to 51 years old, not obese, and mean body mass index 25.1. Nothing in this trial applies to someone with obesity, for whom the same intervention is indicated on entirely different grounds — and the 2025 review of that population reaches different conclusions about costs.

Two years is long for a biomarker trial and short for an ageing intervention. Fracture risk, muscle mass trajectory over a decade, and what happens after stopping are all unmeasured. The bone deficit at 24 months has no follow-up in the literature.

No human trial in this field has a clinical endpoint. No fractures, no falls, no hospital admissions, no disability-free life years, no mortality. Every number is a biomarker or a body composition measure.

The mouse result cuts both ways and the people citing it usually quote one half. 40% restriction extended lifespan the most and cost the most lean mass and immune function. Intermittent fasting failed entirely in the mice that started heavy. Genetics beat diet. Metabolic improvements did not predict lifespan. None of that supports the version of the story that reaches a mainstream app.

The weight regain question is unaddressed. The 2025 review notes that fat mass regain after intervention cessation is common, while physical performance gains appear to be maintained. So the commonest real-world outcome of a two-year restriction trial is not the two years.

The defensible summary: in healthy non-obese adults, two years of an achieved 11.7% calorie restriction lowered thyroid hormone, lowered one inflammatory marker at the end, lowered two liver markers, reduced body weight by 7.5 kg of which 2.2 kg was fat-free mass, reduced bone mineral density at three fracture-relevant sites, lowered physical activity, and did not reduce quality of life. Everything in the ageing literature that promises you a lifespan from a diet is a promise about the first column of that table. The second column is why nutritionists mostly suggest eating less of the wrong things instead.