Populations Where Standard BMI Misclassifies Risk: Athletes, Older Adults, Pregnancy, and More

Your doctor, an app, or a scale handed you a BMI number. What it means depends on who you are. Standard BMI misclassifies health risk for several specific populations. If you belong to one of them, the number you see is not a direct readout of your body fat or your cardiometabolic health. The phrase populations where BMI misclassifies risk describes exactly this gap, the groups where a population-level screening tool breaks down at the individual level.

Body mass index is screening arithmetic, not a diagnostic for an individual body. It divides weight in kilograms by height in metres squared and sorts adults into bands fixed by a World Health Organization committee in the 1990s using epidemiological cut-points, not physiological measurements. The thing newcomers most often get wrong is treating a BMI number as a direct readout of metabolic health or body fatness for a specific person. It was designed to spot statistical shifts in large groups. The Edmonton Obesity Staging System moves beyond BMI alone, adding measures of body fat distribution and metabolic health markers to stage obesity rather than relying on a single anthropometric number.

• WHO Adult BMI Cut-offs (1995/2000): Underweight <18.5, Normal 18.5-24.9, Overweight 25.0-29.9, Obese I 30.0-34.9, Obese II 35.0-39.9, Obese III ≥40.0. Source: WHO Technical Report Series 854.
• WHO Asian BMI Thresholds (2004): Underweight <18.5, Normal 18.5-22.9, Overweight 23.0-27.4, Obese ≥27.5. Source: WHO Expert Consultation, The Lancet, 2004.
• NICE Waist-to-Height Ratio (2022): Keep waist less than half your height (ratio 0.5). Source: NICE guideline CG189.
• NHLBI/NIH Waist Thresholds (1998): High risk: Men >102 cm, Women >88 cm. Source: Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults.
• BMI Sensitivity for Adiposity: Using BMI ≥30, sensitivity to detect excess body fat is approximately 50%; specificity is about 95% (Okorodudu et al. 2010 meta-analysis, N=31,968).

The central problem with BMI is that it cannot distinguish fat mass from lean mass. Two people can have the same Quetelet index and very different body compositions. This matters most for five population groups where the relationship between BMI and body fat percentage diverges most sharply.

Athletes and Highly Muscular Individuals

Standard BMI overestimates body fat in athletes because it cannot account for high lean mass. A rugby player at BMI 30 may have a body fat percentage in the normal range. Sensitivity for detecting obesity in male athletes using BMI ≥30 sits around 50-70%. Measure body fat percentage directly, using DXA as the reference method. BIA and skinfold measurement carry standard errors of ±3-5% compared to DXA, but they are more useful than BMI alone for this group.

Older Adults Over Age 65

Standard BMI underestimates body fat in older adults because of sarcopenia, the age-related loss of lean mass and bone density. An older person at BMI 23 may have body fat percentage in the obese range. The optimal mortality BMI range shifts upward to approximately 23.0-27.0 in some cohort studies, because the loss of muscle reduces BMI while fat mass stays the same or increases. Waist measurement and waist-to-height ratio are better markers of cardiometabolic risk in this group. DXA can quantify the fat-to-lean ratio directly, but for routine screening, NICE CG189 advises using the waist-to-height 0.5 cut-off.

Pregnant Women

BMI before pregnancy is a baseline. During pregnancy it is meaningless for assessing maternal body fat. Gestational weight gain follows its own trajectory, and the Institute of Medicine (IOM) pregnancy weight gain guidelines provide trimester-specific ranges by pre-pregnancy BMI category. Waist size changes throughout pregnancy and cannot be compared to non-pregnant limits. No single number replaces the IOM schedule, which is the authority for this period.

Children and Adolescents

Adult BMI cut-offs do not apply to children. Paediatric BMI is plotted against age- and sex-specific growth charts, the CDC 2000 growth charts in the United States, for example. A BMI centile describes where a child sits on a reference distribution, not whether they are healthy. Two children at the same 85th percentile can have very different body compositions, especially during puberty when lean mass and fat mass gain are uneven. The International Obesity Task Force (IOTF) cut-offs link child BMI values to adult limits, but these are research tools, not clinical diagnoses. Do not use a centile as a diagnosis of a health condition.

BMI overestimates body fat most reliably in individuals with high lean mass. This is the group that gets the wrong message most often. A strength athlete at BMI 28 may be told they are overweight, when their body fat percentage by DXA is 15% for a man or 22% for a woman, well within the healthy range. The same pattern applies to bodybuilders, powerlifters, and anyone whose training builds significant muscle mass. The scale does not lie, but the BMI formula cannot see the difference between muscle and fat. For this group, BMI misclassifies risk by flagging a cardiometabolic risk that is not there. Use body fat percentage from a method you can verify, DXA, or at minimum a controlled BIA reading taken in the morning after a fast. Ignore the BMI band entirely.

BMI also overestimates risk in certain ethnic groups. The WHO Asian BMI thresholds (2004) were introduced because people of Asian descent have a body fat percentage 3-5% higher than European populations at the same BMI. A South Asian person at BMI 23 may have the same diabetes risk as a European person at BMI 30. The NHS Health Survey for England data shows this clearly. If you are of South Asian, Chinese, or Japanese descent, use the WHO Asian cut-offs: normal up to 22.9, overweight from 23.0 to 27.4, obese at 27.5 and above. The standard WHO cut-offs will underestimate your risk.

BMI underestimates body fat most often in older adults, as described above, and in the phenomenon known as normal-weight obesity. In NHANES data, a significant fraction of normal-BMI adults have excess body fat by DXA, above 30% in women or above 23% in men. These individuals have elevated cardiometabolic risk factors despite a BMI in the healthy range. Waist measurement and waist-to-height ratio catch them. A normal-weight person with a waist-to-height ratio above 0.5 has central adiposity and the metabolic risk that goes with it, regardless of their BMI number. The Edmonton Obesity Staging System incorporates this distinction by adding metabolic health markers and waist size to the BMI band, so that a normal-weight person with high waist size and poor metabolic markers is staged as having obesity, while a muscular person with high BMI and normal waist size is not.

Sarcopenia also drives BMI underestimation in older adults. A 75-year-old woman with a BMI of 22 may have lost 20% of her lean mass and gained 10% of her fat mass, putting her body fat percentage in the obese range. Her waist size may also be elevated. The BMI number says healthy; her actual metabolic risk says otherwise. For this group, waist measurement taken at the superior border of the iliac crest using the NIH protocol, end-expiration, non-elastic tape, is the minimum additional measure. The NIH/NHLBI limit of 88 cm for women flags the risk that BMI misses.

No single alternative replaces BMI for every group. The right measure depends on what the individual BMI is missing.

Waist Measurement

Use the NIH/NHLBI limits (men >102 cm, women >88 cm) for white populations. For South Asian, Chinese, and Japanese populations, the International Diabetes Federation (IDF) limits apply: men ≥90 cm, women ≥80 cm. The measurement protocol matters. The NIH protocol uses the superior border of the iliac crest; the WHO protocol uses the midpoint between the lowest rib and the iliac crest. Inter-rater error is ±1.5-3.0 cm. Self-measurement is unreliable, errors exceed 3-5 cm, so have a trained observer do it.

Waist-to-Height Ratio

NICE CG189 advises keeping this below 0.5 as a population-level health signal. It removes the need for sex-specific cut-offs and adjusts for frame size. The Ashwell et al. 2012 meta-analysis of over 300,000 participants across 31 studies found waist-to-height ratio superior to BMI and waist measurement for discriminating cardiometabolic risk, with an AUC improvement of 0.01-0.03. It is the single most practical screening measure for most populations where BMI fails.

Body Fat Percentage

No universal healthy-range standard exists across all measurement methods and populations. DXA is the reference method, with precision ±0.5-1.0% body fat. BIA is widely available but has a standard error of ±3-5% compared to DXA and is highly sensitive to hydration status. Skinfold measurement accuracy depends entirely on technician skill. Use DXA if you can access it. Otherwise, treat any single BIA reading as a rough estimate and control for hydration: morning, fasted, no exercise for 12 hours.

Visceral Fat Measurement

Waist size correlates with CT-measured visceral adipose tissue at r=0.7-0.8, but it cannot distinguish subcutaneous from visceral fat. CT or MRI at the L4-L5 level is the reference standard. A single-slice area of ≥130 cm² is associated with elevated metabolic risk in Japanese criteria (Matsuzawa et al. 1995). For routine screening, waist measurement is sufficient; for clinical assessment, imaging is the only way to quantify visceral fat directly.

Edmonton Obesity Staging System

This framework moves beyond BMI alone by adding waist size and metabolic health markers: blood glucose, lipids, blood pressure, to stage obesity from 0 to 4. A person with high BMI but normal waist size and metabolic markers is stage 0, no obesity-related risk. A person with normal BMI but high waist size and poor metabolic markers is stage 2 or higher. It is the most complete framework for individual assessment, but it requires laboratory results, not just a tape measure.

Two populations where BMI misclassifies risk most sharply, athletes and older adults, are often discussed together because they represent opposite errors. BMI reads high for athletes, overestimating adiposity, and low for older adults, underestimating it. The practical takeaway: if you have high muscle mass, do not trust a BMI number that calls you overweight. If you are over 65, do not trust a BMI number that calls you normal weight. For both groups, add waist-to-height ratio as the minimum alternative. The 0.5 cut-off works for all ages and body compositions, and it catches the risk that BMI misses in both directions.

The NICE CG189 waist-to-height 0.5 rule is the best single population-level alternative to BMI. It requires a tape measure and your height. It adjusts for stature, so a tall person and a short person with the same abdominal diameter are evaluated fairly. It works for all adults, regardless of age, sex, or ethnicity, provided you use the correct measurement protocol. For Asian populations, use the lower IDF waist limits alongside the waist-to-height ratio to capture the earlier onset of cardiometabolic risk.

BMI alone has moderate sensitivity for cardiometabolic risk. The Emerging Risk Factors Collaboration (2011, N=221,934 across 17 countries) found that adding waist measurement or waist-to-hip ratio to conventional risk factors improved the C-statistic by only 0.003-0.005, a statistically significant but practically small improvement at the population level. At the individual level, the improvement can be much larger for someone who falls into one of the misclassified groups. The Prospective Studies Collaboration (2009, N=894,576) showed that the lowest all-cause mortality occurs at BMI 22.5-25.0, with each 5 kg/m² above 25.0 associated with 30% higher mortality. But this is a population curve, not an individual risk. For a muscular athlete at BMI 28, that 30% increase does not apply. For a sarcopenic older adult at BMI 22, the mortality risk may be higher than the curve suggests because their body fat percentage is elevated.

The Global BMI Mortality Collaboration (2016, N=10,625,411) confirmed the J-shaped curve with the nadir at BMI 20.0-25.0, but the hazard ratio per 5 kg/m² increase above 25.0 was 1.11, lower than the Prospective Studies estimate, because the larger dataset included more diverse populations. The shape of the curve shifts by age: the nadir moves upward by approximately 1 kg/m² per decade after age 50. A BMI of 26 may be optimal for a 70-year-old, while a BMI of 22 is optimal for a 30-year-old. If you are older, do not aim for a BMI below 23; the evidence does not support it.

If you belong to any of the populations described here, athlete, older adult, pregnant, child, or of Asian descent, do not use your BMI number as the final word on your health. Measure your waist using the NIH protocol, at the superior border of the iliac crest, end-expiration, with a non-elastic tape, and calculate your waist-to-height ratio. If your ratio is below 0.5, your central adiposity risk is within the population-level target, regardless of your BMI. If it is above 0.5, that is the signal to investigate further: a DXA scan for body composition, or blood tests for metabolic health markers. The Edmonton Obesity Staging System can then stage you accurately. Start with the tape measure. It is free, it is reliable, and it catches what BMI misses.