A previous study showed that measurements of total-body bone mineral changes made with a Hologic QDR 1000W were unreliable when the subjects underwent weight change. The study has been extended to dual energy X-ray absorptiometry (DXA) apparatus from other manufacturers. Re-analysis of published results during weight loss using a Lunar DPX showed that they varied with the software used. Using the Extended mode, there was a 1% loss of bone mineral areal density (BMD), but no significant change in bone mineral content (BMC) or bone area (BA) following a weight loss of 16 kg, whereas the use of the Standard mode led to a larger fall of BMC and BMD. Similar findings arose from the consideration of two studies using Norland XR 26 HS absorptiometers. On the other hand, separation of two groups with similar weight changes from the population studied with a Hologic QDR 1000W confirmed that BMC changed directly with weight, but there was an inverse relationship for BMD, owing to an inappropriate change of BA. The use of Hologic Enhanced and Standard software modes led to significant differences in initial readings and measured changes. With each instrument there was a strong correlation between changes in BA and changes in BMC. When 6 kg of lard was wrapped around the limbs of volunteers or a semi-anthropomorphic phantom to simulate weight change, there were spurious increases of measured BMC and BA by about 5% with each instrument. There were no changes of BMD with Lunar, variable results with Norland, but decreases with Hologic. The results observed in vivo could be explained by the effects of fat changes, without there being any real change of bone mineral. Changes of BMD in the skeleton of the phantom were underestimated with all three brands. The anomaly observed with the Hologic QDR 1000W is less apparent with a Lunar DPX or a Norland XR 26, but there are sufficient uncertainties for all total-body measurements during weight change to be treated with suspicion.