Dual X-ray absorptiometry (DXA) performs noninvasive assessment of bone and soft tissue with high precision. However, soft tissue algorithms assume that 73.2% of the lean body mass is water, a potential source of error in fluid retention. We evaluated DXA (model QDR-2000; Hologic Inc, Waltham, MA), bioelectrical impedance analysis (BIA). and simple anthropometry in 19 patients (9 women and 10 men, mean age 46 y) before and after hemodialysis, removing 0.9-4.3 L (x̄: 2.8 L) of ultrafiltrate. The reduction in fat-free mass (FFM) measured by DXA was highly correlated with the ultrafiltrate, as determined by the reduction in gravimetric weight (r = 0.975. P < 0.0001; SEE: 233 g), whereas BIA was considerably less accurate in assessing FFM reductions (r = 0.66, P < 0.01; SEE: 757 g). Lumbar bone mineral density (BMD) was unaffected by dialysis, as were whole-body fat and BMD. Whole-body bone mineral content, however, was estimated to be 0.6% lower after dialysis. None of the simple anthropometric measurements correlated significantly with the reduction in FFM. In an unmodified clinical selling, DXA appears to be superior to other simple noninvasive methods for determining body composition, particularly when the emphasis is on repeated measurements.