Measuring and monitoring changes in bone mineral density (BMD) is usually done by dual-energy X-ray absorptiometry (DXA). Replacement of old devices is becoming increasingly frequent. To cross-calibrate two Hologic devices, a QDR 1000 and a QDR 4500A, we measured three phantoms - a Hologic spine phantom, a Hologic block phantom (without and with subregions analysis) and a European Spine Phantom - 20 times each without repositioning on both devices. The mean difference between BMD obtained on the two devices was 0.003, 0.033, 0.051 and -0.045 g/cm2 respectively. We also measured the spine and hip of 60 women aged 19-78 years twice on the same day on both devices. Another group of 30 women aged 52-83 years were measured twice on the QDR 4500 A device (15 days apart). We analyzed the data using Pearson's correlation coefficient, and Bland and Altman's method, and calculated the smallest detectable difference (SDD). Results on the two devices were highly correlated: r2 = 0.99, 0.95, 0.96 for spine, femoral neck and total hip BMD respectively. SDD was higher for scans done on different devices than for those done twice on the same device: the SDDs were 0.048, 0.046 and 0.047 g/cm2 for spine, femoral neck and total hip BMD respectively measured on two different devices, while the equivalent values were 0.034, 0.036 and 0.027 g/cm2 using a single device. The difference in BMD results was not dependent on BMD. Our results suggest that, although devices are properly cross-calibrated, differences among them great enough to be clinically relevant can be observed in vivo.