Response Regarding FRAX Performance Across Race/Ethnicity and Genetic Risk in the Women's Health Initiative.

Abstract

The Fracture Risk Assessment Tool estimates 10-year fracture probability, but its performance may vary across populations. We respond to correspondence concerning our Women's Health Initiative analysis by clarifying prior findings and addressing questions about age structure, differential follow-up, construction of clinical inputs, polygenic-score portability, comparator choice, and model performance. Observed fracture probability was estimated using cumulative incidence with death as a competing event. Additional age-stratified analyses showed a similar qualitative pattern of overestimation in Black and Hispanic/Latina women. Using the medium polygenic-score group as the reference, associations with major osteoporotic fracture remained directionally consistent; inverse probability of censoring-weighted estimates were similar for major osteoporotic fracture but less stable for hip fracture. Adding the polygenic score produced only modest gains in time-dependent discrimination. We acknowledge that informative censoring that differs across racial and ethnic groups, sparse subgroup events, multiple comparisons, and limited cross-ancestry portability remain important limitations. These considerations support cautious interpretation and further validation of fracture-risk tools in diverse populations.