Longitudinal precision of dual-energy x-ray absorptiometry in a multicenter study. The Nafarelin/Bone Study Group.

Reproducibility is a key issue in both clinical and research applications of bone mineral density (BMD) measurements. To examine the longitudinal precision of dual-energy x-ray absorptiometry (DEXA) for the measurement of mineral density in vivo and in vitro, the performance of a group of instruments in the course of a multicenter longitudinal clinical trial was monitored. Measures were performed on eight identical machines (Hologic QDR1000) and analyzed using the same automated software program. Short-term precision was good in vitro, [anthropomorphic spine phantoms; mean intrasite coefficient of variation (CV) 0.42 +/- 0.1% (SD)] and in vivo (lumbar spine; CV 1.1 +/- 0.5%). Intersite measures of a single spine phantom (specified mineral content -57.8 g) revealed a range of 57.3-58.4 g (CV 0.7%). In two subjects intersite CV in vivo were 3.7 and 2.1% (spine) and 1.8 and 3.2% (femoral neck). At five sites frequent phantom measures were performed over a 1 year period (mean number of measures 196) and revealed a mean all-point CV of 0.43% (range 0.35-0.53%). Longitudinal precision in vivo was somewhat less (mean CV of spinal measures 1.1%, femoral neck 1.2%, trochanter 1.3%, and Ward's area 2.4%). At one additional site large variations in phantom measures heralded repeated mechanical failures that eventually required machine replacement. In summary, DEXA demonstrates good in vitro and in vivo longitudinal precision, providing the basis for expanded clinical and research usefulness. Nevertheless, stringent quality assurance measures are required to detect and respond to system malfunctions.