08/22/2026
Strive for healthy body composition not “weight loss”
A number on the scale is treated as the summary of a body. In 260,861 adults, the tape measure disagreed with it in more than a third of cases.
BMI is height and weight and nothing else. It cannot tell muscle from fat, and it cannot tell fat under the skin from fat packed around the organs. That second distinction is the one that matters, because visceral fat sits in the drainage of the portal vein and delivers free fatty acids and inflammatory signals straight to the liver. The idea that a tape measure captures what the scale misses is old. In 27,098 people across 52 countries, waist-to-hip ratio graded heart attack risk cleanly across every fifth of the distribution, while BMI's association collapsed to nothing once the ratio and other risk factors were accounted for.
Whether that translates into anything useful in a clinic has been genuinely contested. In 221,934 people across 58 cohorts, adding BMI, waist, or waist-to-hip ratio to a risk model that already contained blood pressure, lipids and diabetes history did not meaningfully improve discrimination, with C-index changes at the fourth decimal place. Waist also reproduces less reliably than BMI on repeat measurement. That analysis is the strongest argument against bothering. What has shifted the balance since is genetic evidence: a polygenic score for waist-to-hip ratio adjusted for BMI was associated with 1.77 times the odds of type 2 diabetes and 1.46 times the odds of coronary disease, alongside higher triglycerides, higher two-hour glucose and higher systolic pressure, which is what a causal relationship looks like rather than a marker travelling alongside one.
The new analysis pooled 15 cohorts followed a median of 20 years across nine cardiovascular outcomes. The reclassification was substantial in both directions. Among people whose BMI read normal, 18 percent had a high waist-to-hip ratio and 5 percent a high waist circumference. Among people with overweight, roughly 40 percent had one or the other. Among people whose BMI read obesity, 45 percent had a low waist-to-hip ratio and 9 percent a low waist. A high waist inside a normal or overweight BMI carried 15 to 50 percent greater risk across most outcomes.
Everything is observational, so nobody was assigned a body shape, and people who carry weight centrally differ in diet, alcohol, sleep and stress in ways no adjustment reaches. More particularly, this analysis reports how much better the tape classifies people, not whether measuring them changes what happens to them. No trial has randomised clinicians to measure waists. The thresholds used, 88 and 102 centimetres, were derived largely in European-ancestry populations and understate risk in South Asian and East Asian populations, where lower cut-points are recommended. And the reverse direction deserves as much attention as the forward one: among people with obesity, a low waist was not associated with different risk than normal weight for most outcomes, and was associated with lower all-cause mortality, though women with obesity and a low ratio kept elevated risk where men did not.
Among people with obesity, the share of cases attributable to a high waist circumference ran to 48.9 percent for heart failure and 46.2 percent for atrial fibrillation, against 28.8 percent for death from any cause. Nearly half the heart failure in that group tracked to where the fat sat rather than how much there was.
What is still missing is any demonstration that acting on the measurement helps. A consensus statement in 2020 argued waist circumference should be recorded as routinely as blood pressure, and it still generally is not.
A tape measure costs almost nothing and takes fifteen seconds, and in roughly one normal-weight adult in five it says something the scale did not. That is not a reason to discard BMI, which remains more reproducible and predicts perfectly well at a population level. It is a reason to record both, and to stop treating a normal BMI as an all-clear.
Dardari et al., J Am Coll Cardiol 2026;88(6):670-684 · PMID 42583989
Yusuf et al., Lancet 2005;366(9497):1640-9 · PMID 16271645
Wormser et al., Lancet 2011;377(9771):1085-95 · PMID 21397319
Emdin et al., JAMA 2017;317(6):626-634 · PMID 28196256
Ross et al., Nat Rev Endocrinol 2020;16(3):177-189 · PMID 32020062