Abstract
Trabecular volumetric BMD (Tb.vBMD) and cortical surface BMD (Ct.sBMD) derived from three-dimensional modeling of hip DXA scans (3D-DXA) correlate strongly with areal BMD (aBMD). Such correlations have been interpreted as indicating a substantial overlap between 3D-DXA parameters and the information captured by aBMD. We investigated whether correlations of this magnitude are specific to DXA-based modeling by comparing 3D-DXA with quantitative computed tomography (QCT) in the same subjects. Two cohorts with paired hip DXA and QCT were analyzed: 37 Spanish adults and 128 postmenopausal women with osteoporosis from the United States (PaTH cohort), which had lower variability in age and bone density. Pearson correlations were computed between Ct.sBMD, Tb.vBMD, and aBMD for both imaging modalities, and between QCT- and 3D-DXA-derived parameters before and after adjustment for aBMD. QCT-derived and 3D-DXA-derived compartmental parameters correlated similarly with aBMD and with each other within each cohort (R=0.84–0.96 in the Spanish cohort, 0.44–0.82 in PaTH), with markedly weaker correlations in the narrower-range PaTH cohort across both modalities. QCT–3D-DXA correlations were 0.82–0.97 and remained significant after adjustment for aBMD (partial R=0.65–0.84, all p<0.001). Strong correlations with aBMD are observed for both QCT and 3D-DXA and vary with the population studied. Thus, strong correlations with aBMD should not, by themselves, be interpreted as evidence that 3D-DXA merely reproduces aBMD or that these correlations arise from the DXA-based modeling approach. 3D-DXA parameters remained substantially correlated with QCT-derived compartmental BMD after adjustment for aBMD, suggesting that their associations with compartment-specific bone density extend beyond the information captured by standard DXA.