Three obesity definitions, one population: prevalence ran from 17% to 38%

Every statement about how common obesity is depends on a definition, and there is more than one in active use. A Swiss analysis published in Nutrition Research did the obvious experiment: take a single population, apply three definitions to it, and see how much the answer moves.

It moves by a factor of two.

The three definitions

The researchers used the CoLaus|PsyCoLaus cohort in Lausanne, Switzerland, covering adults aged 35 to 75, with 6,733 participants at baseline and several rounds of follow-up [s1].

Each participant was classified three ways [s1]:

• WHO — body mass index alone.
• NIH — body mass index together with waist circumference.
• EASO (European Association for the Study of Obesity) — body mass index or waist-to-height ratio, plus the presence of an obesity-related complication such as hypertension, type 2 diabetes, cardiovascular disease, or chronic kidney disease.

The three are not refinements of one another. WHO asks about a ratio of weight to height. NIH adds fat distribution. EASO adds whether the body is showing damage.

What happened to the numbers

Averaged across all follow-up rounds, obesity prevalence was 17.4% under WHO, 33.9% under NIH, and 37.5% under EASO [s1].

The combined prevalence of overweight and obesity did not change [s1]. What changed was where the line between the two categories fell. Overweight prevalence ran 38.7%, 22.2% and 18.6% under the three definitions respectively [s1] — the obesity category grew almost entirely at the expense of the overweight category.

The shift was sharpest at older ages. Among participants over 75 at the second follow-up, obesity prevalence was 19.2% by WHO, 48.9% by NIH, and 55.9% by EASO [s1]. That is the same people, the same day, and a 36-percentage-point spread.

Incidence moves too

Prevalence is a snapshot; incidence is the flow. Among the 2,334 participants without obesity at baseline who were followed prospectively — 34.7% of the baseline cohort — the incidence of obesity over a median 14.5 years was 5.8% under WHO, 16.8% under NIH, and 20.6% under EASO [s1].

Under the EASO definition, the authors note, obesity ended up more common than overweight [s1]: 20.6% against 18.3%. A category originally conceived as the tail of a distribution becomes the larger group.

Why anyone would broaden the definition

The case for the newer definitions is that BMI is a poor proxy for what clinicians actually care about. It does not distinguish fat from muscle, and it does not say where the fat is. Evidence that distribution matters keeps accumulating: a separate Dutch cohort analysis of 3,873 adults reported that when the same population was categorised by waist circumference and by BMI, the waist-based categories showed generally larger effect sizes in their association with cognitive performance [s2].

EASO's addition — requiring an obesity-related complication — goes further still, and moves the definition from a measurement toward a diagnosis. That is the direction several bodies in the field have been pushing.

What broadening costs

The authors of the Swiss analysis are direct about the trade-off: broader definitions identify more people at potential risk, but substantially expand the obesity category, with implications for case classification, clinical management, and resource allocation [s1].

Three concrete consequences follow.

First, prevalence figures from different sources are not comparable unless the definition is stated. A country that "has more obesity" than another may simply be counting differently.

Second, trend data breaks at the point a definition changes. An apparent surge can be a reclassification.

Third, eligibility follows definition. Treatment thresholds, insurance coverage, and clinical trial entry criteria are written in terms of specific cut-offs, and moving the cut-off moves millions of people across a line without anything about them changing.

What the study does not say

It does not say which definition is correct. It is a classification exercise in one Swiss cohort, not a comparison of how well each definition predicts death, disease, or treatment response — which is the test that would actually settle the question. The paper measures disagreement, not accuracy.

That test is the one the field still owes itself. Until it is run, the honest reading of any obesity prevalence figure is that it is a statement about a definition as much as about a population.

This article is informational and is not medical advice.

Sources

• Obesity prevalence varies markedly by definition: multiple cross-sectional and prospective studies — Nutrition Research, 2026-06-16
• Association of body mass index, waist circumference, (abdominal) overweight and obesity with sex- and age-specific cognitive function over time — the Doetinchem Cohort Study — International Journal of Obesity, 2026-03-17