Improved fracture prediction using different fracture risk assessment tool adjustments in HIV-infected women

Abstract

Objectives: A fracture risk assessment tool (FRAX) using clinical risk factors (CRFs) alone underestimates fracture risk in HIV-infected men. Our objective was to determine whether accuracy of FRAX would be improved by considering HIV as a cause of secondary osteoporosis, and further improved with addition of dual-energy X-ray absorptiometry parameters in HIV-infected women.

Design: Subgroup analysis of Women's Interagency HIV Study.

Methods: We included 1148 women (900 HIV-infected and 248 uninfected) over age 40 with data to approximate FRAX CRFs and 10-year observational data for incident fragility fractures; 181 (20%) HIV-infected women had dual-energy X-ray absorptiometry data. Accuracy of FRAX was evaluated by the observed/estimated ratios of fracture in four models: CRFs alone; CRFs with HIV included as a cause of secondary osteoporosis; CRFs and femoral neck bone mineral density (FN BMD); and CRFs, FN BMD and trabecular bone score.

Results: FRAX using CRFs were less accurate in HIV-infected than uninfected women for major osteoporotic (observed/estimated ratio: 5.05 vs. 3.26, P < 0.001) and hip fractures (observed/estimated ratio: 19.78 vs. 7.94, P < 0.001), but improved when HIV was included as a cause of secondary osteoporosis. Among HIV-infected women, FRAX accuracy improved further with addition of FN BMD (observed/estimated ratio: 4.00) for hip fractures, but no further with trabecular bone score.

Conclusion: FRAX using CRFs alone underestimated fracture risk more in older HIV-infected women than otherwise similar uninfected women. Accuracy is improved when including HIV as a cause of secondary osteoporosis for both major osteoporotic and hip fractures, whereas addition of FN BMD only improved accuracy for hip fracture.

Fig. 1.

Flowchart of the Women’s Interagency HIV study cohort.

Fig. 2.

Comparison of predictive accuracy of fracture risk assessment tool adjustments in HIV-infected women with clinical risk factor data (left of the dotted line, N=900) and in the subset with dual-energy X-ray absorptiometry data (right of the dotted line, N=181).