Supplementary Material for: Evaluation of Biopsy-Based Molecular Risk Prediction in Crescentic Glomerulonephritis

Novel molecular tools have the potential to improve current clinical and histology-based risk classification systems for various medical renal diseases including glomerulonephritis (GN). We aimed to assess the utility of gene expression for improving biopsy-based risk prediction in patients with GN with and without crescent formation. This retrospective case-control study used NanoString nCounter to measure the expression of 54 previously described inflammation, nephron injury, endothelium, and crescent-related genes in 335 archival, formalin-fixed paraffin-embedded native kidney biopsies, including a 288-biopsy discovery cohort representing a broad spectrum of crescentic GN subtypes, and an independent 47-biopsy validation cohort focused on ANCA-associated crescentic GN. Clinical, histologic, and gene expression data were compared. Discovery cohort analysis demonstrated increased expression of 13 genes in crescentic GN cases that developed end-stage renal disease (ESRD) versus those that did not (FDR<0.05). Within the 75-biopsy subset of ANCA-associated crescentic GN cases in the discovery cohort, this 13-gene set was found to be independently predictive of ESRD in multivariate Cox proportional hazards regression analysis (p=0.015), with significant differentiation of high and low risk patients in Kaplan-Meier renal survival analysis (log rank test, p=0.002). However, validation cohort analysis did not demonstrate significant improvement in risk stratification with the 13-gene set when compared with established clinicopathologic models. These results suggest that biopsy-based gene expression may provide the opportunity for improved risk stratification in crescentic GN; however, the genes evaluated in this study appear to have limited added clinical utility over existing risk scores.