Dual X-ray absorptiometry: cross-calibration of a new fan-beam system

The high precision and stable calibration of dual X-ray absorptiometry (DXA) scanners have led to their widespread use in longitudinal studies for research and the follow-up of individual patients who are receiving treatment for osteoporosis. However, difficulties in maintaining the continuity of the bone mineral density (BMD) calibration scale can arise when an old DXA system is replaced by a newer model. We report the results of an in vivo cross-calibration study performed when a GE-Lunar Prodigy fan-beam system replaced a DPX-L pencil-beam scanner. Lumbar spine and hip DXA scans were performed in 133 patients (104 female, 29 male) attending long-term BMD monitoring. On average, lumbar spine BMD measurements on the two systems agreed closely, with Prodigy values 1% lower than those on the DPX-L. However, after allowing for this difference, the root mean square error (RMSE) of 0.037 g/cm2 was larger than in previous cross-calibration studies reported in the literature, and was 3 times the value expected from the precision of the BMD measurements. Mean femoral neck BMD also agreed closely between the two systems, although for Prodigy, the spread of measurements was 10% smaller than that for the DPX-L. For the trochanter and Ward's triangle regions, mean BMD was 4% and 6% lower, respectively, on the Prodigy system, and the results were affected by a similar compression of the range of values. RMSE values were 0.037 g/cm2, 0.044 g/cm2, and 0.044 g/cm2, respectively, for the femoral neck, trochanter, and Ward's triangle sites. When the high value of the RMSE was investigated, it was found that for lumbar spine BMD, patient body weight and the difference between the two systems in the percentage fat reported in the soft tissue reference region explained 40% of the variance. This enabled equations to be developed that significantly improved the agreement between scans performed on the two systems. Smaller improvements were obtained for the femur BMD measurements.