Our study found that hip geometric phenotypes derived from hip structural analysis (HSA) are highly heritable, with genes accounting for 40–84% of the residual variation in these phenotypes after accounting for the effects of age and sex. These heritabilities were found to be largely independent of BMD suggesting distinct genetic influences on these fracture risk factor traits. We found suggestive evidence of linkage to chromosome 17q11.2–12 for NN_CSA for men and women age 50 years and below, confirming a previous linkage reported by Xiong (20
) for cortical thickness, a trait that they found highly correlated (0.9) to CSA. A candidate gene within this region is sclerostin (SOST
), the gene that is abnormal in the sclerosing bone disorder sclerosteosis [33
In addition, suggestive evidence for linkage (LOD ≥ 2.3) was found at 10 other sites on chromosomes 1, 2, 5, 6, 11, 12, 14 and 15. Although none of these suggestive linkages confirmed previous reports, five linkages (at 1p35.5, 1p36, 2p11.2, 4q11 and 6q22–23) were reported in the Framingham population with different but related phenotypes [21
]. For example, at 1p35.5, our linkage was with NN_ID whereas the previous report was for NN_OD. Although not identical, these phenotypes are related in that NN_OD is composed of NN_ID plus 2 times cortical thickness [21
]. Similarly, our linkage at 4q11 was for S_Z, whereas the previous report was for NN_Z [21
], both related to bending strength but at different areas of the proximal femur. Among our suggestive linkages reported here, those at chromosomes 1p36 and 14q23 were previously reported by us as being suggestive regions of linkage for femoral neck BMD [24
]. However, in our previous report, the 1p36 linkage was found only in women, whereas in the current study, our findings are for men. Our suggestive linkage at 14q23 to NN_Z was found for men but previously was found for BMD for men and women age 50 years and under [4
]. The paucity of confirmatory linkages among studies may reflect the relatively low power in individual studies, differences in study design or genetic heterogeneity [34
Linkage analysis in our population as a whole yielded 2 suggestive linkages with LOD scores ≥ 2.3 (at 1p36 and 4q11). In subgroup analysis, 8 additional suggestive linkages were found, 2 for women only (6q22–23, 11q22.1), 2 for those > 50 (2p11.2, 12q24.2) and 4 for those ≤ 50 (1p35.5, 1q23–23, 2p25, 5q24–30). Finding different linkages in subgroups suggests that different genes may affect hip geometric phenotypes in men vs women and at peak bone mass vs those in the bone loss phase of life. Evidence for sex-specific QTLs related to hip geometric phenotypes has been found in other studies as well, but at different locations than we found. In linkage studies on sister pairs and brother pairs from Indiana, three out of seven suggestive autosomal linkages to hip structural phenotypes (at 2p, 5p and 4p) were male-specific (14
). In addition, in a large population of nuclear families of European descent living in Nebraska, suggestive or significant evidence for QTLs related to hip structural phenotypes was found to be male-specific at one site (7q21) and female-specific at 4 sites (2p14, 3q26, 15q21, 3q26,) (20
). Of linkage studies reported to date on hip structural phenotypes, we are aware of only one that has reported linkages that reached genome-wide significance, for regions of 20q12 and Xq25 (for cortical thickness, LODs 4.28 and 3.90 respectively) (20
). It is notable that this study included a huge number (3998) of participants (20
). Since none of our linkages reached genome-wide significance, confirmation will be required to determine their importance.
We found the greatest similarities with our linkage results when compared to results reported from the Framingham study, as described above. One reason for this could be that the Framingham hip geometric phenotypes were obtained using the same HSA program that we used. Two additional groups have generated hip geometry phenotypes using methods different from HSA, hip radiographs in one (17
) and DXA-derived phenotypes but not the same HSA analysis as used here in the other (19
). There was no overlap between linkages detected from these populations and those detected in ours. Methodological differences are one possible explanation for the differences in linkage results.
Data derived from Hip Structural Analysis (HSA) provide different information about hip strength than BMD and could help to explain why some patients with low trauma hip fractures have normal BMD. Previous studies have shown that hip structural phenotypes vary with race, age, gender, body weight and activity and that these changes are not necessarily concordant with changes in BMD [7
]. As found in other studies, men in our study had higher values for bone geometry measures than women except for buckling ratio which was higher in women, consistent with greater strength in male femurs. In addition, with age in both men and women, BR increased with age whereas Z, CSA, ID and OD decreased in the narrow neck. The effects of a change in geometry phenotype with age can be to decrease the deleterious reduction in BMD. For example, the relatively small reductions in section modulus with age, compared to larger BMD decreases, show that hip strength is more stable with age than BMD alone would indicate [8
]. This, in turn, may partly explain why hip fracture is less universal in the elderly than reductions in BMD. Thus, variability in HSA phenotypes with age may explain some of the fracture risk that cannot be attributed to low BMD.
Previous studies have shown that hip geometry-derived measures of bone strength are associated with fracture risk and that this association is at least partly independent of differences in BMD [9
]. Our data showed in the Amish support this. Previous data from HSA have shown that military recruits who suffered tibial stress fractures during basic training had reductions in section modulus and bone cross-sectional areas (CSA), relative to body mass, compared to those without fractures [4
]. Our data revealing that section modulus and CSA are highly heritable could help to explain in part why these measures were lower in some recruits than in others.
We have previously observed the hip fracture rate to be lower in the Amish compared to non-Amish Caucasians [23
]. It is possible that the Amish, perhaps by virtue of their active lifestyle, might also have a more favorable profile for other geometry-derived parameters of hip strength than the non-Amish. However, comparisons of these hip strength parameters across studies are difficult since HSA analysis is sensitive to the procedures used to measure BMD by DXA. A previous report has shown, however, that physical exercise increases hip CSA and section modulus [39
]. We speculate that beneficial effects from high physical activity on geometric measures of hip strength in the Amish could be a contributing factor to their reduced risk of hip fracture.
In summary, we have found high heritability of hip geometry measured by hip structural analysis in the Amish and found suggestive evidence for QTLs affecting proximal femur geometric measures of strength in 11 different chromosomal locations. Our linkage on chromosome 17q, confirms a previous report. The 10 additional linkages showed very modest evidence for QTLs affecting geometric hip strength measures on chromosomes 1, 2, 5, 6, 11, 12, 14 and 15. Continuing to study the genetics of these geometric phenotypes related to bone strength might further our understanding of the heredity of osteoporosis and hip fracture.