Objective: To determine whether dual-energy X-ray absorptiometry (DXA) is a valid method for body composition assessment of obese and non-obese subjects. Design: Cross-sectional study. Subjects: Chinese women living in Hong Kong; 66 of 91 subjects had body mass index (BMI) of > 25 kg/m(2). Measurements: Anthropometrics, including body weight, body height, waist and hip girth. Percentage body fat (%BF) and fat-free mass (FFM) from DXA (Hologic 2000 plus, Enhanced Array Whole Body Version 5.63) were compared with that based on a tracer dose of deuterium oxide for the determination of total body water (TBW). Results: In both obese and non-obese subjects, FFMDXA was Similar to FFMTBW. The Bland and Altman-type analysis indicated that comparable between-methods differences (mean bias) and limits of agreement were obtained in obese and non-obese subjects for FFM (0.4, between -4.4 and 5.2 kg vs 0.5, between -3.1 and 4.1 kg) and %BF (-0.6, between -7.6 and 6.4% vs -1.2, between -8.6 and 6.2%). The %BF bias was independent of age, BMI, hip circumference, and waist-to-hip ratio, but correlated with waist girth (r = 0.24, P = 0.021). Conclusion: The sources of bias are methodological and anthropometric in nature. The between-methods differences, however, are small and clinically insignificant. DXA is a valid method for assessing the body composition of obese patients.