
BMI alone missed who was at risk among 844,000 adults
- Observational study
- 844,108 adults in four countries
BMI, your weight measured against your height, is the number most people get at a checkup. In 2025 a Lancet Diabetes & Endocrinology Commission proposed a different way to define obesity. A Nature study published 11 November put it to the test using health data from 844,108 adults in China, the UK, the US and Spain.
The framework first checks that excess fat is really there, pairing BMI with a waist measurement or waist-to-height ratio. Then it splits obesity in two. 'Clinical' obesity means excess fat is already harming organs or getting in the way of daily life. 'Preclinical' obesity means the excess fat is present without that harm so far.
Measured against the framework, BMI on its own erred in both directions. Depending on the cohort, it missed obesity in 5.2–8.6% of people and labeled it wrongly in 1.2–3.1%.
The two categories also tracked with very different outcomes. Clinical obesity was linked to about 69% higher mortality. Preclinical obesity carried a much smaller increase, about 19%. Obesity defined by BMI alone showed no excess mortality at all. Clinical obesity made up only about 23–31% of obesity cases, and over 10 years, 77–91% of people with preclinical obesity did not progress to it.
These are associations in observational data, so they do not show that fat caused the deaths. Information on organ problems came from health records rather than new examinations. The four cohorts differ in age, ethnicity and health systems, and the US cohort, All of Us, is made up of volunteers. The biggest open question is practical: no one has yet tested whether using this framework in clinics leads to better care or better outcomes. That would take trials.
Get the next issue by email
Our weekly newsletter on what is new in health research, explained for curious readers and linked to the original studies. One email a week.
Sources
Source titles are preserved as published.
Reference Range Weekly is for general information. It does not provide medical advice, diagnosis, or treatment.