Award Date

August 2023

Degree Type


Degree Name

Doctor of Philosophy (PhD)


Epidemiology and Biostatistics

First Committee Member

Qing Wu

Second Committee Member

Ann Vuong

Third Committee Member

Soumya Upadhyay

Fourth Committee Member

Fatma Nasoz

Number of Pages



Introduction: Early identification of individuals at high-risk for osteoporotic fractures who may benefit from preventive intervention is essential. However, the predictive accuracy of the currently used fracture risk assessment tool remains suboptimal. The first aim of this research is to construct genome-wide polygenic scores for the femoral neck (PGS_FNBMDidpred) and total body BMD (PGS_TBBMDidpred) and to estimate their potential in identifying individuals with a high risk of osteoporotic fractures. The second aim is to validate the predictive performance of two previously established PGSs (PGS_FNBMDidpred and PGS_TBBMDidpred) in an external cohort of 9,000 postmenopausal women of European ancestry. The third aim is to develop and evaluate a novel approach called metaPGS, which combines genetic information from multiple fracture-related traits to further improve the predictive accuracy of genetic information in fracture risk assessment.

Methods: The first manuscript constructed genome-wide PGS for femoral neck and total body BMD. We externally tested the PGSs, both by themselves and in combination with available clinical risk factors, in 455,663 European ancestry individuals from the UK Biobank. The predictive accuracy of the developed genome-wide PGS was also compared with previously published restricted PGS employed in fracture risk assessment. The PGSs developed in the first study were then externally validated in the second study using the Women's Health Initiative (WHI) study data. The magnitude of the association between each PGS and Major Osteoporotic Fractures (MOF)/Hip Fractures (HF) risk was assessed by using the Cox Proportional Hazard Model. To investigate whether adding PGS would improve the predictive ability of FRAX, we formulated four models: (1) Base model: FRAX risk factors; (2) Base model + PGS_FNBMDidpred; (3) Base model + PGS_TBBMDidpred; (4) Base model + metaPGS. The reclassification ability of models with PGS was further assessed using the Net Reclassification Improvement (NRI) and the Integrated discrimination improvement (IDI). In the third study, to develop the novel metaPGS combining PGSs of multiple fracture-related traits/diseases, we first derived individual PGS from genome-wide association studies of 16 fracture-related traits. Then, we employed an elastic-net logistic regression model to examine the association between the 16 PGSs and fractures while controlling for covariates such as age, sex, and the first four principal components. The optimal metaPGS model was chosen based on the highest area under the receiving-operating characteristic curve (AUC). The metaPGS was constructed by combining the 11 most significant individual PGSs selected using the elastic regularized regression model. We evaluated the predictive power of the metaPGS alone and in combination with clinical risk factors recommended by guidelines. The ability of the models to reclassify fracture risk was also assessed using NRI and IDI.

Results: In the first study, for each unit decrease in PGSs, the genome-wide PGSs were associated with up to 1.17-fold increased fracture risk. The genome-wide total body PGS (PGS_TBBMDidpred) (HR: 1.17; 95%CI 1.15-1.19, p<0.0001) showed a significantly higher association with fractures compared to the restricted total body BMD (PGS_TBBMD81) (HR: 1.03; 95%CI 1.01-1.05, p=0.001). In the reclassification analysis, compared to the FRAX base model, the models with PGS_FNBMD63, PGS_TBBMD81, PGS_FNBMDidpred, and PGS_TBBMDidpred improved the reclassification of fracture by 1.2% (95% CI, 1.0% to 1.3%), 0.2% (95% CI, 0.1% to 0.3%), 1.4% (95% CI, 1.3% to 1.5%), and 2.2% (95% CI, 2.1% to 2.4%), respectively. The second study failed to validate the findings discovered in the first study. The results showed that these PGSs were not significantly correlated with MOF or HF in the WHI cohort. Additionally, incorporating genetic information into the FRAX tool showed minimal improvement in the predicted probabilities of hip fracture risk for elderly Caucasian women. In the third study, the metaPGS had a significant association with incident fractures (HR: 1.22, 95% CI: 1.191.27), which was stronger than previously developed bone mineral density (BMD)-related individual PGSs. The metaPGS had comparable predictive power to established risk factors such as age, body weight, and early menopause. The association between the metaPGS and incident fractures remained significant after adjusting for clinical risk factors, indicating added predictive value beyond established clinical risk factors. Adding the metaPGS to the FRAX model improved the discrimination of fractures from non-fracture cases. Conclusions: The findings indicate that integrating PGS data into clinical diagnosis has the potential to enhance the efficacy of screening programs at a population level. The metaPGS approach shows promise for stratifying fracture risk in the European population, as it combines genetic data from various fracture-related traits, improving fracture risk prediction.


Disease and disorders of/related to bone; fracture risk assessment; genetic research; Human association studies; metaPGS; osteoporosis



File Format


File Size

3660 KB

Degree Grantor

University of Nevada, Las Vegas




IN COPYRIGHT. For more information about this rights statement, please visit

Included in

Biostatistics Commons