External bone size predicts different associations between bone mineral content and the proportion of cortical and trabecular bone
Michael A Casden, Robert W Goulet, Erin MR Bigelow, Bonnie T Nolan, Kerry Richards-McCullough, Steffenie Merillat, Carrie A. Karvonen-Gutierrez, Karl J Jepsen
- Abstract
- Study Goals: Osteoporotic hip fractures remain a major public health problem. Low areal bone mineral density (aBMD) measured by Dual-energy X-ray absorptiometry (DXA) is used to clinically identify those at increased risk of fracturing. The majority of fractures, however, occur in individuals outside the range considered to be at risk, which indicates the association between aBMD and fracture risk varies among individuals. By better understanding the inter-individual differences in how bone structure influences aBMD, we can begin to better predict who is at increased risk of fracturing. The goal of this study was to determine the association between the amount of cortical and trabecular bone from NanoCT with bone parameters from DXA and to test if this association differs between narrow and wide proximal femurs.
Methods: Femoral neck (FN) area, bone mineral content (BMC), and aBMD were measured using DXA for 22 cadaveric proximal femurs (white women, 42-89+ years old). High resolution 3D volumes of the proximal femurs were obtained using a NanoCT system at 27um voxel size. Using ORS Dragonfly, FN volumes of interest (VOI) corresponding to the DXA ROI were extracted from the NanoCT scans. Validated convolutional neural networks were used to segment the bone from background and the cortical bone from the trabecular bone. A custom plugin was used to count the number of cortical, trabecular, and total bone voxels in the direction of the 2D projection of the DXA scans. Samples were sorted based on FN area from the DXA scans into narrow and wide subgroups and linear regression analysis was performed.
Results: Linear regression analysis revealed strong correlations between the NanoCT and DXA FN regions for area (R2=0.91, p<0.0001), BMC (R2=0.91, p<0.0001), and aBMD (R2=0.89, p<0.0001). There was a strong association between BMC and the total number of bone voxels which did not differ between narrow (R2 =0.87, p<0.0001) and wide (R2 =0.93, p<0.0001) subgroups (ANCOVA: slope, p=0.199; y-int, p=0.559). The association between BMC and the number of cortical vs. trabecular voxels, however, differed between subgroups. While both subgroups showed significant associations between BMC and the number of cortical bone voxels (R2 =0.88, p<0.0001 and R2 =0.54, p=0.010, respectively), the slope was significantly greater for narrow bones (ANCOVA, slope p=0.002). In contrast, the wide (R2=0.88, p<0.0001) but not narrow (R2=0.008, p=0.79) subgroup showed a significant association between BMC and the number of trabecular voxels (ANCOVA, slope, p=0.001).
Discussion/Conclusion: The strong correlation between the NanoCT and DXA regions for area, BMC, and aBMD indicates successful alignment of the NanoCT and DXA VOIs. When separated into the narrow and wide subgroups, both demonstrated a strong correlation between BMC and the total number of bone voxels, which was expected because X-ray attenuation is based largely on the presence of bone. Importantly, the subgroup analysis revealed that for narrow bones, lower BMC values reflect lower amounts of cortical bone. Alternatively, for wide bones, lower BMC values reflect proportionally lower amounts of both cortical and trabecular bone. Therefore, this study demonstrates that FN BMC values from DXA reflect different proportions of cortical and trabecular tissues in narrow and wide proximal femurs. This outcome motivates future studies investigating how the relationship between changes in BMC and changes in bone strength differ between narrow and wide subgroups.
- Presented by
- Casden, Michael
- Institution
- University of Michigan, Department of Orthopaedic Surgery
- Other Affiliations
- University of Michigan Medical School; University of Michigan School of Public Health