The objective of this study was to develop cut-off values and evaluate the accuracy of body mass index (BMI) in the definition of obesity in the Thai population. A cross-sectional, epidemiologic study in 340 men and 507 women aged 50 +/- 16 yr (mean +/- SD; range: 20-84 yr), were sampled by stratified clustering sampling method. Body composition, including percentage body fat (%BF), was measured by dual energy X-ray absorptiometry (GE Lunar Corp, Madison, WI). BMI was obtained by dividing weight (in kg) by height (in m2). The "golden standard "for defining obesity was %BF > or =25% in men and %BF > or =35% in women. The %BF-based prevalence of obesity in men and women was 18.8% and 39.5%, respectively. However, using the BMI cut-off of > or =30, only 2.9% of men and 8.9% of women were classified as obese. In the cubic regression model, BMI was a significant predictor of %BF, such that in men a BMI of 27 kg/m2 would predict a %BF of 25%, and in women a BMI of 25 kg/m2 would correspond to a %BF of 35%. The area under the receiver operating characteristic curve for BMI was approximately 0.87 (95% CI: 0.82-0.92) and 0.86 (95% CI: 0.83-0.90) in men and women, respectively. In conclusion, for the Thai population, BMI is a reasonably useful indicator of obesity; however, the cut-off values of BMI for diagnosing obesity should be lowered to 27 kg/m2 in men and 25 kg/m2 in women.