Repeated dual-energy x-ray absorptiometry (DEXA) measurements are often performed both in clinical work and in research studies. The aims of the present study were to investigate the repeatability of DEXA total body measurements, to clarify the effect of the scanning positioning of the subject, and to compare the reliability of DEXA measurements of the extremities between automatically and manually defined regions of interest (ROIs). Three DEXA measurements of the total body composition, that is, fat tissue mass, lean tissue mass (LM), and bone mineral content, were performed on 30 male volunteers (mean age, 45.2 years) in addition to measurements of bone mineral density. Using a narrow fan-beam Lunar Prodigy densitometer (GE Lunar, Madison, WI), 3 DEXA scans (2 supine and 1 prone) of the total body were performed. For regional measurements of the right arm and leg, ROIs were set automatically and manually in the supine-supine and supine-prone positions. Repeatability of total body DEXA measurements was excellent for bone mineral content (r = 0.99), LM (r = 0.99), fat tissue mass (r = 1.00), and bone mineral density (r = 0.98) in supine scanning. Change of position from supine to prone slightly decreased the reproducibility of total body measurements. Reproducibility of regional measurements was inferior to total body results; especially in the upper extremity, the repeated automatic LM measurements in supine-supine positions produced r values as low as 0.74 but increased to 0.93 after manual adjustment of the ROIs. To obtain maximal reliability of the composition measurements, we recommend manual checking of machine-made ROIs and, if needed, manual adjustment to avoid measurement errors.