Dual-energy X-ray absorptiometry (DXA) is a novel, non-invasive technique for the measurement of gross body composition in small animals. In the present study the absolute accuracy of the Hologic QDR-1000W scanner was assessed by comparison with direct analysis in twelve rats with a range of body fat and bone mineral content (BMC) values. Fat masses measured by DXA and petroleum-ether extraction were significantly different (P < 0.0023). The DXA technique consistently overestimated fat mass by approximately one third of the measured fat content. BMC derived from the measurement of Ca in ash gave a mean of 8.26 (range 1.57-15.71)g. BMC measured by DXA was not significantly different for the group as a whole. However, there was a trend for DXA to overestimate BMC in animals with low BMC and underestimate in those with higher BMC, compared with direct analysis, such that the 95% limits of agreement for the two techniques were +2.73 to -2.58 g. These results suggest that the present small-animal software developed for use with currently available Hologic machines does not give an accurate measure of gross body composition compared with the results from classical direct analysis.