BACKGROUND AND AIMS: Weight measured by dual-energy X-ray (DXA) was shown to be increasingly underestimated in subjects over 75 kg compared to an electronic scale. This study compares body weight and composition measured by balance beam scale and three DXA acquisition modes in obese subjects. METHODS: In 39 obese, body weight was measured by balance beam scale, and body weight and composition by DXA Hologic QDR4500A in normal (NPM) and high power mode (HPM) (Enhanced v8.26 and v8.26* software) and DXA GE-Lunar Prodigy (v6.5 software). To ensure linearity of body weight and composition measured by the different DXA acquisitions, we also measured 13 women with a body mass index (BMI) of 25-30 kg/m(2). RESULTS: While QDR4500A HPM overestimates scale weight by about 2 kg over the whole BMI spectrum, QDR4500A NPM underestimates scale weight as a weight-dependent response (-1.7+/-1.8 kg overall, -4.1+/-1.6 kg in morbidly obese women). These results suggest switching from one mode to the other at a specific threshold, i.e. in our study a weight of 90 kg or a BMI of 34 kg/m(2). Prodigy gives weight about similar to scale (+0.5+/-0.8 kg). Both Hologic acquisition modes underestimate fat mass but overestimate lean body mass compared to Prodigy. CONCLUSIONS: The QDR4500A NPM is inappropriate in women over 90 kg. Unfortunately, the QDR4500A HPM overestimates body weight in the range of 90-150 kg. The difference between scale and Prodigy weight remains stable throughout weight ranges. To better assess their accuracies in terms of body composition, QDR4500A NPM, HPM and Prodigy should be tested against phantoms or in vivo multi-compartment models.