The performance of genetic-enhanced DXA-BMD predicting models trained in UK biobank varies across diverse ethnic and geographical populations
Description
Background Osteoporosis presents a significant global health challenge, compromising bone quality and elevating fracture susceptibility. While dual-energy x-ray absorptiometry (DXA) stands as the gold standard for bone mineral density (BMD) assessment and osteoporosis diagnosis, its costliness and complexity impede widespread screening adoption. Predictive modeling of BMD, leveraging genetic and clinical data, emerges as a more viable and cost-effective approach for osteoporosis and fracture risk evaluation. Methods and Findings We developed BMD prediction models for the femoral neck (FNK) and lumbar spine (SPN) using various methods within a UK Biobank (UKBB) training set comprising 17,964 individuals from the British white population. Models based on Regression with Least Absolute Shrinkage and Selection Operator (LASSO), selected based on the coefficient of determination (R2) from a model selection subset of 5,973 individuals from the British white population, underwent testing on five UKBB test sets and 12 independent cohorts of diverse ancestries, totaling over 15,000 individuals. Furthermore, we assessed the correlation of predicted BMDs with fragility fractures in a distinct case-control set of over 287,000 participants lacking DXA-BMDs in the UKBB of the European white population. Incorporating genetic factors marginally improved predictions, capturing an additional 2.3% variation for FNK-BMD and 3% for SPN-BMD over clinical factors alone. Predicted BMDs exhibited significant associations with fragility fracture risk in the European white population. Nonetheless, the predictive model's performance varied between the UKBB population of other ethnic groups and the independent cohorts. Conclusions Our study yields novel insights into predicting osteoporosis and fracture risk. Genetic factors enhance BMD predictive performance beyond clinical factors alone. Adjusting inclusion thresholds for genetic variants (e.g., 5×10^(-6) or 5×10^(-7)) rather than solely considering genome-wide association study (GWAS)-significant variants may further refine the model's explanatory power for BMD variations. This study also underscores the imperative for training models on diverse population to bolster predictive performance across various ethnic and geographical populations.