Abstract Accurate classification of obesity levels is of clinical importance for public health and decision making. Body mass index (BMI) is the most broadly used measure of obesity on a population scale. But, as with other diagnostic indicators, BMI can have potential pitfalls. We attempted to explore the diagnostic performance of BMI for assessing obesity. 330 healthy Caucasian women (native Greeks) (ages, 20 to 85 years) volunteers participated in the study. Dual-energy X-ray absorptiometry (DXA) was used to measure percentage of fat mass (%FM). The diagnostic performance of BMI was assessed using the widely used indicator for obesity of %FM > 35% in women. Receiver operating characteristic (ROC) analysis was performed to evaluate the ability of BMI to discriminate obesity from non obesity. The correlation between BMI and %FM, and lean mass (LM) and %LM, was also investigated. Logistic regression was performed to assess for associations between DXA-derived fat metrics and age. BMI-defined obesity (≥ 30 kg/m 2 ) was present in 40.3% of women, while %FM DXA-defined obesity was present in 72% of the subjects. The largest discrepancy between BMI- and DXA-defined obesity was noted in women over 80 years (p = 0.0002). A BMI ≥ 30 kg/m 2 had excellent sensitivity 99%, 95% confidence interval (CI), 95.8–99.9, but poor specificity 45%, 95% CI 38–52 to detect %FM-based obesity. BMI ≥ 25 kg/m 2 had high sensitivity (92%, 95% CI 87.5–94.8) and good specificity (79%, 95% CI 68.7–86.6). The optimal cut point for BMI in the entire cohort found to correspond closer to the criterion value of 35%FM was 26.32 kg/m 2 (p < 0.001). At this level, BMI had excellent sensitivity (96.7%, 95% CI 93.3–98.6) and moderate specificity (71.5%, 95% CI 62.4–79.5). In all women, BMI correlated better with %FM (r = 0.74 to 0.84, p < 0.05) than with %LM (r= −0.68 to −0.86, p < 0.05). Logistic regression analysis found age was a significant predictor of obesity (OR = 1.074, 95% CI 1.036–1.114, p < 0.05), increasing odds for excess adiposity by 7.4% each year. The diagnostic performance of BMI for assessing obesity is limited particularly in women with BMI ≥ 30 kg/m 2 . A BMI cut point ≥ 30 kg/m 2 has very high sensitivity, but misclassifies as obese 55% of normal women. At the BMI cut off ≥ 25 kg/m 2 sensitivity was high, but 21% of women were still misclassified as obese. A BMI cut point value of 26.32 kg/m 2 has higher sensitivity and falsely classifies as obese 28% of the women. In light of these findings, revision of current BMI criteria for obesity screening may be warranted in white women of Mediterranean origin such as Greek women, and perhaps other populations.
Theodorou et al. (Wed,) studied this question.