Reproducibility of volume-adjusted bone mineral density of spine and hip from dual X-ray absorptiometry

Skeletal size has a confounding effect on areal bone mineral density (aBMD) related to differences in skeletal volume. Several methods have been proposed for calculating volumetric BMD (vBMD), but in vivo precision data are limited for the spine and have not been published for the hip. We prospectively performed duplicate dual X-ray absorptiometry measurements of the anteroposterior spine and hip (n = 121) in a diverse female population referred for initial clinical BMD testing. Each scan pair was performed and analyzed by two different technologists (mean interval of 4 d) to obtain standard aBMD. Scan data were reprocessed at a later date to calculate vBMD for the lumbar spine (L2-L4), femoral neck, and total hip in the 87 spines and 82 hips for which we had complete analyzable scan data. We found much worse precision in femoral neck volume (5.2% coefficient of variation [CV]) than in spine volume (2.6% CV; p < 0.003). This contributed to greater error in femoral neck vBMD (3.9% CV) than aBMD (2.3= CV; p < 10(-6)). Total hip aBMD had better precision than vBMD (1.0 and 1.3-2.5% CV; p < 10(-5)). The reverse pattern was seen in the spine with slightly better precision for vBMD than aBMD (1.1 and 1.5% CV; p < 0.002). Volumetric measures of lumbar spine density can be obtained with high precision. Because of poor reproducibility in the femoral neck, the total hip region may be preferable for measuring volumetric bone density in the proximal femur.