Objective Skeletal fragility is a major comorbidity in systemic lupus erythematosus (SLE), yet the accuracy of fracture risk algorithms in this population remains uncertain. We compared the discriminative ability of Fracture Risk Assessment Tool (FRAX) and the Italian FRAX-derived tool (DeFRA) for fractures in SLE. Methods This is a secondary analysis of the multicentre osteoporosis and FRagility fracture Among SLE patients (FRAIL - NCT05590390 ) cohort that included patients with SLE who underwent dual-energy X-ray absorptiometry with vertebral fracture assessment (VFA). Vertebral fractures were confirmed by radiography. For each patient, 10-year major osteoporotic fracture probability was calculated using FRAX and DeFRA. Discrimination was assessed with receiver operating characteristic curves and DeLong’s test. Operational thresholds (FRAX ≥20%, DeFRA ≥20% and 15%) were evaluated for sensitivity, specificity, predictive values and number needed to scan (NNS=1/positive predictive value). Robustness of estimates was tested using 2000 bootstrap resamples with out-of-bag evaluation. Results 106 patients with SLE were included in the study. Mean age was 53.6 years, 88.7% were female and 41.5% were on glucocorticoids. Morphometric vertebral fractures were identified in 23 patients (21.7%), including 15 previously unrecognised. For newly detected fractures, area under the curve (AUC) was higher for DeFRA (0.834, 95% CI 0.725 to 0.944) than FRAX (0.681, 95% CI 0.495 to 0.867; p=0.022). Similar results were observed when considering any vertebral fracture (AUC 0.902 vs 0.770; p=0.013). At operational thresholds, DeFRA ≥20% identified 9/15 new fractures (NNS=1.78) versus 7 with FRAX ≥20%, while lowering the cut-off to DeFRA ≥15% increased sensitivity (10/15 fractures, NNS=2.0) without loss of specificity. Bootstrap validation confirmed the robustness of rank ordering. Conclusion In SLE, DeFRA outperforms FRAX in detecting vertebral fractures and offers a clinically efficient threshold for guiding targeted VFA, with potential implications for optimising imaging strategies and glucocorticoid management.
Aldegheri et al. (Thu,) studied this question.