Large, multicentre clinical trials using DXA to monitor bone density following intervention are now common. At the same time, several different bone densitometers and calibration phantoms are currently in use. The aim of this study was to document the technical information required on cross-calibration of equipment, reproducibility and patient dose before commencing a multicentre clinical trial. To this end, we obtained an in vitro and in vivo cross-calibration of two machines (a Hologic QDR 2000 and a Lunar DPX-L) that were not significantly different. Interobserver and intraobserver precision, and radiation dose were also measured and three commonly used phantoms assessed for their usefulness in cross-calibration and quality assurance. Measured in vitro precision on the two machines (0.3-0.7%) was better than that specified by the manufacturers. In vivo precision was worse (1.4-2.1%), as might be expected in patients with reduced bone mass. Mean entrance skin radiation doses on each machine were 280 microSv for the QDR 2000 and 38 microSv for the DPX-L. No one phantom is ideal, but the European Spine Phantom or Lunar Aluminium Spine Phantom will provide an adequate cross-calibration for a clinical trial. This study demonstrates that an adequate cross-calibration can be obtained for use in groups of patients and that the equipment used is reproducible with a low radiation output.