Abstract
OBJECTIVE: Lupus nephritis (LN) is a prevalent renal manifestation in patients with SLE, with kidney biopsy remaining the gold standard for evaluating disease activity. However, the invasive nature of biopsy and associated risks highlight the need for a non-invasive predictive tool. This study aimed to construct and validate predictive models for the activity index (AI) and tubulointerstitial lesions (TIL) in patients with LN as an alternative tool to assist clinicians in decision-making when kidney biopsy is not feasible. METHODS: We enrolled 266 patients with LN diagnosed by kidney biopsy from three centres, divided into a training cohort (n=213) and a validation cohort (n=53). Patients were stratified by AI and TIL scores: high AI (AI >4), low AI (AI ≤4), high TIL (TIL >4) and low TIL (TIL ≤4). Clinicopathological data were systematically collected. Multivariate logistic regression was employed to identify significant risk factors for high AI and TIL, and nomograms for individualised assessment were constructed. Model performance was evaluated using receiver operating characteristic curves, decision curve analysis, calibration plots and the Hosmer-Lemeshow test. RESULTS: The key independent risk factors for high AI included lymphocyte count, haematuria, albumin, serum creatinine, complement 4 and antihistone antibodies. For high TIL, the risk factors included age, haemoglobin, platelet count, blood urea nitrogen and antiribosomal P antibodies. Both nomograms demonstrated favourable performance in terms of discrimination and calibration across both cohorts. CONCLUSION: The developed nomograms provide reliable, non-invasive tools for identifying patients with LN with high AI and TIL, which can improve clinical risk assessment and help guide more personalised management strategies.