These results are consistent with correlation coefficients
(R 2 = 0.5–0.94) determined for the cross-manufacturer forearm DXA standardization effort commissioned by the International Committee of Standards in Bone Measurement [19] as well as with the results reported for similar algorithms developed for QCT [25]. True vBMD was less well correlated to aBMDdxa. This is not surprising given the size dependence inherent to projectional BMD measures. It follows that simulation of the projection process does significantly improve prediction of DXA-based BMD values. It is important to note that the standard VOI for a clinical HR-pQCT acquisition (9.02 mm in length) is shorter than the standard ultra-distal ROI prescribed by DXA manufacturers (20 and 15 mm in length for Lunar and Hologic, respectively). Furthermore, selleck screening library each manufacturer uses different anatomical landmarks to localize the ROI. These two facts may partly explain the discrepancy in the coefficients of determination for aBMDsim compared to Lunar and Hologic
(R 2 = 0.87 vs. R 2 = 0.82) and the difference in the regression intercept (0.04 vs. 0.11 g/cm2). As expected, the aBMDsim better predicted PF-6463922 Lunar aBMDdxa values, where the ROI is more similar with respect to the longitudinal placement compared to the Hologic ROI. The difference in the correlation coefficients also likely reflects the relative variability in the patient cohorts scanned on either device. As expected, aBMDsim Tacrolimus (FK506) and aBMDdxa of the UD radius were poor to moderate predictors of aBMD at axial skeletal sites (lumbar spine and proximal femur). Despite the significantly smaller analysis ROI, aBMDsim had an equivalent degree of predictive power for DXA aBMD in the lumbar spine and proximal femur. The magnitude of the predictive power for the Lunar cohort was
similar to previous studies comparing intersite BMD relations [26, 27]. This group spanned a larger age and BMD range, compared to the Hologic cohort, which was comprised exclusively of osteopenic women with a narrow range of aBMD values at axial skeletal sites. An important limitation is that this simulation technique is limited to anatomical sites that may be imaged by HR-pQCT. In this study, we have applied the technique to the distal radius, as this is a routine site for clinical densitometry and a common site of Selleckchem LEE011 osteoporotic fracture (Colles’ fracture). This technique could also be applied to the distal tibia, which is routinely imaged during clinical HR-pQCT exams, and of interest as a load-bearing site. On the other hand, the proximal femur and lumbar spine—critical sites of osteoporotic fracture—are not accessible by HR-pQCT.