A data-driven, weighted flare score that accurately predicts organ damage in systemic lupus erythematosus

Objectives: Flares are key determinants of outcome in systemic lupus erythematosus (SLE), yet the conventional mild/moderate vs severe classification inadequately reflects prognosis. We sought to refine flare classification to improve damage prediction and develop a data-driven flare score. Methods: We analysed a multicentre longitudinal cohort of 354 patients with active SLE requiring treatment at inclusion (median follow-up 66 months). Flares were adjudicated using a modified Safety of Estrogens in Lupus Erythematosus National Assessment-SLE Disease Activity Index Flare Index (SFI). Random Forests and modified flare definitions assessed the contribution of individual SFI variables to subsequent damage. We derived a flare score using bootstrap-validated mixed-effects models. This was validated in an independent cohort (n = 110). Results: Both mild/moderate and severe SFI flares contributed to elevated 12-month damage risk (incidence rate ratio [IRR]:1.42 and 1.54, respectively). Among SFI subcriteria, major organ involvement requiring high-dose glucocorticoids (GCs) ranked highest for damage prediction (weight = 6), followed by the isolated GCs escalation criteria (weights = 1.5-3), whereas minor organ involvement and immunosuppressants/biologics addition carried lower weights (1 each). A flare score based on these subcriteria demonstrated good predictive performance (Brier score = 0.07). A threshold of ≥ 3.5 identified a subset of high-risk flares (40% of all episodes) that displayed the strongest association with subsequent damage (IRR: 2.49, 95% CI: 1.51-4.10) and outperformed the original SFI definition in the validation cohort (area under the curve-receiver operating characteristic: 0.893 vs 0.816). Although both Definitions of Remission in SLE (DORIS) and lupus low disease activity state correlated with lower risk of SFI-defined flares, only sustained DORIS (>50% of time) reduced high-risk flares. Conclusions: We defined and validated a novel weighted SLE flare score that better predicts damage accrual and may support treat-to-target implementation and clinical trial design.

Objectives: Flares are key determinants of outcome in systemic lupus erythematosus (SLE), yet the conventional mild/moderate vs severe classification inadequately reflects prognosis. We sought to refine flare classification to improve damage prediction and develop a data-driven flare score. Methods: We analysed a multicentre longitudinal cohort of 354 patients with active SLE requiring treatment at inclusion (median follow-up 66 months). Flares were adjudicated using a modified Safety of Estrogens in Lupus Erythematosus National Assessment-SLE Disease Activity Index Flare Index (SFI). Random Forests and modified flare definitions assessed the contribution of individual SFI variables to subsequent damage. We derived a flare score using bootstrap-validated mixed-effects models. This was validated in an independent cohort (n = 110). Results: Both mild/moderate and severe SFI flares contributed to elevated 12-month damage risk (incidence rate ratio [IRR]:1.42 and 1.54, respectively). Among SFI subcriteria, major organ involvement requiring high-dose glucocorticoids (GCs) ranked highest for damage prediction (weight = 6), followed by the isolated GCs escalation criteria (weights = 1.5-3), whereas minor organ involvement and immunosuppressants/biologics addition carried lower weights (1 each). A flare score based on these subcriteria demonstrated good predictive performance (Brier score = 0.07). A threshold of ≥ 3.5 identified a subset of high-risk flares (40% of all episodes) that displayed the strongest association with subsequent damage (IRR: 2.49, 95% CI: 1.51-4.10) and outperformed the original SFI definition in the validation cohort (area under the curve–receiver operating characteristic: 0.893 vs 0.816). Although both Definitions of Remission in SLE (DORIS) and lupus low disease activity state correlated with lower risk of SFI-defined flares, only sustained DORIS (>50% of time) reduced high-risk flares. Conclusions: We defined and validated a novel weighted SLE flare score that better predicts damage accrual and may support treat-to-target implementation and clinical trial design.