LAUSR

OBJECTIVE To develop a short-term renal outcome prediction model for acute kidney injury (AKI) in patients with lupus nephritis. METHODS Two lupus AKI cohorts from 2 independent centers during 2013-2019 were included, i.e., a derivation cohort from a rheumatology center and a validation cohort from a nephrology center. Clinical characteristics and renal histologic features were obtained. The outcome measurement was the recovery of kidney function within 12-month. Lasso regression was used for feature selection. Prediction models with or without pathology were built and nomogram was plotted. Model evaluation including calibration curve and decision curve analysis was performed. RESULTS 130 patients and 96 patients were included in the derivation and validation cohorts, of which 82 and 73 patients received renal biopsy, respectively. The prognostic nomogram model without pathology included determinants of SLE duration, days from AKI onset to treatment and baseline creatinine level (C-index 0.85 (95%CI 0.78∼0.91) and 0.79 (95%CI 0.70∼0.88) for the 2 cohorts). Combination of histologic interstitial tubular fibrosis in the nomogram gave an incremental predictive performance (C-index 0.93 vs 0.85, p = 0.039) in the derivation cohort, but failed to improve the performance in the validation cohort (C-index 0.81 vs 0.79, p = 0.78). Decision curve analysis suggested clinical benefit of the prediction models. CONCLUSION The predictive nomogram models might facilitate more accurate management for lupus patients with AKI.