BMD-Related Genetic Risk Scores Predict Site-Specific Fractures as Well as Trabecular and Cortical Bone Microstructure

Abstract

Context: It is important to identify patients at highest risk of fractures.

Objective: To compare the separate and combined performances of bone-related genetic risk scores (GRSs) for prediction of forearm, hip and vertebral fractures separately, as well as of trabecular and cortical bone microstructure parameters separately.

Design, setting, and participants: Using 1103 single nucleotide polymorphisms (SNPs) independently associated with estimated bone mineral density of the heel (eBMD), we developed a weighted GRS for eBMD and determined its contribution to fracture prediction beyond 2 previously developed GRSs for femur neck BMD (49 SNPs) and lumbar spine BMD (48 SNPs). Associations between these GRSs and forearm (ncases = 1020; ncontrols = 2838), hip (ncases = 1123; ncontrols = 2630) and vertebral (ncases = 288; ncontrols = 1187) fractures were evaluated in 3 Swedish cohorts. Associations between the GRSs and trabecular and cortical bone microstructure parameters (n = 426) were evaluated in the MrOS Sweden cohort.

Results: We found that eBMDGRS was the only significant independent predictor of forearm and vertebral fractures while both FN-BMDGRS and eBMDGRS were significant independent predictors of hip fractures. The eBMDGRS was the major GRS contributing to prediction of trabecular bone microstructure parameters while both FN-BMDGRS and eBMDGRS contributed information for prediction of cortical bone microstructure parameters.

Conclusions: The eBMDGRS independently predicts forearm and vertebral fractures while both FN-BMDGRS and eBMDGRS contribute independent information for prediction of hip fractures. We propose that eBMDGRS captures unique information about trabecular bone microstructure useful for prediction of forearm and vertebral fractures. These findings may facilitate personalized medicine to predict site-specific fractures as well as cortical and trabecular bone microstructure separately.

Keywords: bone microstructure; bone mineral density; cortical; fractures; genetic risk scores; trabecular.

Figure 1.

Associations between 3 GRSs and risk of forearm, hip, and vertebral fractures, evaluated in separate logistic regression models. Models are adjusted for age, sex, height, weight, and MrOS site. Odds ratios (OR) and 95% confidence intervals given per SD increase of the genetic risk score (GRS) from inverse variance weighted meta-analysis of significant independent associations. The association between the eBMD_GRS and forearm fractures was significantly stronger than the corresponding associations for the FN-BMD_GRS (P = 0.002) and the LS-BMD_GRS (P = 0.001). N = total number of subjects/fracture cases. Abbreviations: eBMD, estimated bone mineral density analysed by ultrasound; FN-BMD, femoral neck bone mineral density analysed by dual-energy absorptiometry; LS-BMD, lumbar spine bone mineral density analysed by dual-energy absorptiometry.

Figure 2.

GRSs independently associated with risk of forearm, hip and vertebral fractures. The independently associated genetic risk scores (GRSs) were selected by forward stepwise selection in logistic regression models starting from a fixed base model including age, sex, height, weight, and MrOS site. We then validated that the final models, including either 1 or 2 independently associated GRSs, also resulted in the lowest AIC. Odds ratios (OR) and 95% confidence intervals given per SD increase in GRS from inverse variance weighted meta-analysis of significant independent associations. N = total number of subjects/fracture cases. Abbreviations: eBMD, estimated bone mineral density analysed by ultrasound; FN-BMD, femoral neck bone mineral density analysed by dual-energy absorptiometry.

Figure 3.

The variance explained (R²) of different bone parameters in MrOS Sweden by eBMD_GRS and FN-BMD_GRS evaluated using stepwise selection in combined linear regression analyses. Black bars = variance explained independently by eBMD_GRS. White bar = variance explained independently by FN-BMD_GRS. Grey bar = variance explained shared by eBMD_GRS and FN-BMD_GRS. Trabecular thickness, cortical bone area and cortical volumetric BMD (cortical vBMD) were analysed by high resolution peripheral quantitative computed tomography. Abbreviations: eBMD, estimated bone mineral density analysed by ultrasound; FN-BMD, femoral neck bone mineral density analysed by dual-energy absorptiometry; GRS, genetic risk score.

Figure 4.

Summary of the separate and combined prediction of fracture types and trabecular and cortical bone microstructure by eBMD_GRS and FN-BMD_GRS. eBMD_GRS but not FN-BMD_GRS independently predicted trabecular bone thickness as well as vertebral and forearm fractures. In contrast, both FN-BMD_GRS and eBMD_GRS independently predicted cortical bone mass parameters (cortical area and cortical thickness) and hip fractures. We propose that eBMD_GRS captures unique information of trabecular bone microstructure useful for the prediction of forearm and vertebral fractures. In contrast, both FN-BMD_GRS and eBMD_GRS capture unique information of cortical bone mass useful for the prediction of hip fractures. Abbreviations: eBMD, estimated bone mineral density analysed by ultrasound; FN-BMD, femoral neck bone mineral density analysed by dual-energy absorptiometry.