Unsupervised Clustering Identifies High-Risk Phenotypic Subgroup in Crescentic Glomerulonephritis Patients

Background: The current classification of crescentic glomerulonephritis (cGN) primarily focuses on renal outcomes, its ability to predict overall patient survival remains controversial. This study aims to develop a risk stratification system for overall survival and validate a renal risk score for predicting end-stage renal disease (ESRD) in cGN. Methods: We retrospectively analyzed 224 cGN patients from the First Affiliated Hospital of Sun-Yat Sen University, with a median follow-up of 87.10 (interquartile range: 32.35, 153.97) months. The primary outcome was all-cause mortality and the secondary outcome was ESRD. K means clustering, Principal component analysis (PCA), and decision tree analysis were used to identify patient groups. Results: Traditional classification showed that Type III cGN patients had the poorest overall survival rates, whereas Type I had the poorest renal survival. K-means clustering stratified patients into high-risk (n=142) and low-risk (n=82) groups with 10 year survival rates of 63.1 % versus 89.6 % (P = 0.004). High‑risk status independently predicted mortality (HR=3.28, 95% CI 1.56 ~ 6.87, P=0.002). The crescentic glomerulonephritis kidney risk score (cGN-KS) consisting of serum creatinine levels, the percentage of normal glomeruli, and tubular atrophy/interstitial fibrosis independently predicted ESRD (P<0.001, HR=1.10, 95% CI 1.06–1.13). A decision tree based on age, serum C3, serum creatinine, albumin, 24-h proteinuria, and fibrous crescents percentage accurately identified high-risk patients in the development and validation sets. Conclusions: We identified two survival risk groups and developed a renal risk score in cGN patients, providing an informative and comprehensive risk stratification to facilitate the clinical management.