Clinical Characteristics, Prognosis, and Predictive Modeling in Class IV±V Lupus Nephritis

Anjing WangYunlong QinYan XingZixian YuLiuyifei HuangJinguo YuanYueqing HuiMei HanGuoshuang XuJin Zhao^*Shiren Sun^*

• Department of Nephrology, Xijing Hospital, Fourth Military Medical University, Xi’an, China

The objective of this study is to compare the clinical features and survival outcomes of class IV±V lupus nephritis (LN) patients, identify risk factors, and develop an accurate prognostic model. This study enrolled patients diagnosed with class IV±V LN by renal biopsy at Xijing Hospital from December 2013 to June 2023. The composite endpoint of the study was defined as a decline in the estimated glomerular filtration rate (eGFR) by more than 50%, progression to end stage renal disease, or death, whichever came first.ESRD is defined as an eGFR ＜15ml/min/1.73m2, necessitating the commencement of chronic dialysis (hemodialysis or peritoneal dialysis) or kidney transplantation. We compared the baseline features and survival prognosis between patients with class IV±V LN. The prognostic model was developed using machine learning algorithms and Cox regression. The model's performance was evaluated in terms of discrimination, calibration, and risk classification using the concordance index (C-index), integrated brier score (IBS), net reclassification index (NRI), and integrated discrimination improvement (IDI), respectively. Results: A total of 313 patients were enrolled for this study, including 156 class IV and 157 class IV+V LN. During the median follow-up period of 42.6 (17.0, 83.4) months, 35 (22.4%) class IV and 38 (24.2%) class IV+V LN patients experienced combined events. Class IV and class IV+V patients have similar clinical manifestations, treatment strategies, and long-term prognosis, despite class IV having a higher chronic index (CI) score (P < 0.001). Seven eligible variables (eGFR, CI, age, basophil percentage, red blood cell count, mean arterial blood pressure, and uric acid) were selected to develop the random survival forest (RSF) model. This model demonstrated the best performance with a C-index of 0.771 (0.667, 0.848) and an IBS of 0.144 (0.132, 0.154). The IDI and NRI in the testing set further confirmed that the RSF model exhibited superior risk classification and discrimination capabilities. Conclusion: Class IV±V LN was similar in clinical manifestations, treatment strategies, and longterm prognosis, despite differences in pathological features. The RSF model we established for class IV±V LN patients, incorporating seven risk factors, exhibits superior survival prediction and provides more precise prognostic stratification.