Race-Neutral FRAX without BMD outperforms FRAX without BMD in a multiethnic cohort with Diabetes or Hypertension

Abstract

FRAX is poorly validated in Black and Hispanic people in the United States. We recently demonstrated that FRAX underestimated fractures in Black and Hispanic patients with diabetes (DM), but it is not known if this is a diabetes-specific issue. We sought to compare the performance of the original FRAX (oFRAX) without BMD and a hypothetical race-neutral FRAX (rn-FRAX) in a retrospective multiethnic cohort of patients with DM and control patients with hypertension (HTN). We specifically sought to compare performance across race. Patients were included if they were followed for DM or HTN in the health system and averaged at least 1 visit per year. Incident major osteoporotic fractures (MOFs) and FRAX inputs were collected using diagnosis codes or medication data. A rn-FRAX was created using U.S. census data. Performance was compared using measures of discrimination (area under the curve, AUC) and calibration (observed-to-predicted ratios, O/P). There were 17,400 Black (45.4% DM), 9,865 White (35.3% DM), and 1,196 Hispanic (57.9% DM) patients included. Discrimination was similar for oFRAX and rn-FRAX (AUC 0.72-0.73). In DM patients, the O/P was 1.4 for oFRAX and 1.0 for rn-FRAX. By race, we found good performance for both calculators in White patients (O/P 0.8-1.2). In Black patients, oFRAX underestimated risk (O/P 1.3-2.3). Conversely, rn-FRAX overestimated risk in Black women (O/P 0.7-0.8) but performed well in Black men (O/P 0.9-1.1). In Hispanic patients, oFRAX substantially underestimated fractures (O/P 1.8-2.4), and this was only partially alleviated by rn-FRAX (O/P 1.2-1.6). In conclusion, we found that the current FRAX without BMD significantly underestimated major osteoporotic fractures in Black and Hispanic patients, while a hypothetical race-neutral FRAX improved performance overall. Our findings suggest the current adjustments for race in U.S. FRAX may not be necessary or, if maintained, need to be updated.