BMI and Mortality: the J-Shaped Curve

By Longevity Lab · Published: August 23, 2026

Based on peer-reviewed research — full source list at the end of this article. This is educational information, not medical advice.

Plotting body mass index against death rates produces a J: all-cause mortality is lowest in a flat band between roughly 20 and 25, climbs steadily above it, and rises again — more shallowly, and for different reasons — below it. The largest analysis, pooling 239 prospective studies and 10.6 million adults, found each 5 kg/m² above 25 associated with about 31% higher all-cause mortality in never-smokers free of disease at baseline. Understanding what distorts the two arms of that curve is what separates reading it correctly from over-reading it.

What the largest analyses found

The 2016 Global BMI Mortality Collaboration restricted its main analysis to never-smokers with no chronic disease at baseline, and excluded the first five years of follow-up. Both restrictions matter, and both push in the same direction: they remove people who are thin because they are ill or smoking. In that cleaner sample the minimum sat at 20.0–25.0, and the hazard ratio per 5 kg/m² above 25 was 1.39 in Europe and 1.29 in North America.

The earlier Prospective Studies Collaboration analysis of nearly 900,000 adults found essentially the same thing: above 25, each 5 kg/m² was associated with about 30% higher all-cause mortality, with the largest contributions from vascular, diabetic, renal and hepatic causes.

A 2018 UK cohort of 3.6 million adults reproduced the J and added detail on what changes along it. Above the minimum, deaths from cardiovascular disease, diabetes, kidney and liver disease climb. Below it, deaths from respiratory disease, dementia and some cancers climb. It is not one risk rising and falling — it is two different sets of risks meeting in the middle.

Why the left arm is the contested one

Three confounders inflate mortality at low BMI, and untangling them is most of the methodological argument in this field.

Reverse causation. Undiagnosed cancer, COPD, heart failure and dementia all cause weight loss, often for years before diagnosis. The low BMI is a symptom, not a cause. Excluding early follow-up years reduces this, but never entirely.

Smoking. Smokers weigh less on average and die more, which manufactures an association between thinness and death that has nothing to do with body mass. Restricting to never-smokers is the standard fix.

Frailty and muscle loss. In older adults especially, a low BMI often means low Sarcopenia-relevant lean mass rather than low fat, and that carries genuine risk. This one is not an artifact — it is a real mechanism that BMI happens to capture badly.

After correcting for the first two, the left arm flattens considerably but does not disappear. Very low body mass remains associated with worse outcomes, chiefly through infection, injury and reduced reserve.

What the right arm is made of

The excess mortality above 25 is not evenly distributed across causes or across people. It is concentrated in cardiovascular and metabolic disease, it rises steeply above 30, and it is heavily modified by fat distribution and fitness.

That last point deserves emphasis. Cohort studies that measure cardiorespiratory fitness alongside BMI consistently find fitness to be the stronger predictor. Being fit at a BMI of 28 is associated with better survival than being unfit at 23 in several large cohorts — which does not make excess weight harmless, but does mean the index is capturing something that other variables capture better.

Distribution does the same work. Two people at BMI 31 with waists of 92 cm and 118 cm are not on the same part of any curve, and BMI — body mass index alone cannot separate them.

How much life the curve corresponds to

These are population averages carrying wide uncertainty, and they are associations rather than causal estimates — the Body Mass Index (BMI) tool reports the same idea as a mortality-risk ratio relative to the normal band, which is easier to read than a table of years. They are useful for one purpose: putting the size of the effect in perspective. Moving from 26 to 24 is not on this table because the difference is too small to measure. Moving from 42 to 32 is.

What to take from it personally

Very little, directly. A curve built from ten million people tells you what happens on average, and you are not an average. What it does justify is a hierarchy of attention: if your BMI is above 32 or below 18.5, the index is probably telling you something real. Between 20 and 30, it is telling you almost nothing that waist, fitness and blood markers do not tell you better.

If you want an individual estimate rather than a population curve, the blood-based clocks do that job properly. PhenoAge Calculator — biological age from a blood test converts nine standard laboratory values into a biological age that predicts mortality substantially better than any body-size measure, and it costs one blood draw. Our BMI calculator is the free first step, not the answer.

Frequently asked questions

Sources

Disclaimer. This article is for information only and does not replace medical advice. Talk to a qualified clinician before changing anything about your health.