Digital pathology approach for the classification of focal and segmental glomerulosclerosis

Focal and segmental glomerulosclerosis (FSGS) refers to various diseases with different pathophysiological mechanisms and treatments. Primary FSGS (pFSGS) is dependent on immunological factors and is treated with immunosuppressants. However, the histological lesions by light microscopy are not specific and do not reflect a precise etiology. The aim of this study is to categorize FSGS biopsies into etiological groups using digital pathology combined with clinical parameters. This was a retrospective, observational, multicentre study. It involved three university hospitals in Paris and Bordeaux. A total of 58 patients with pFSGS and maladaptive FSGS (mFSGS) were included in a selected ‘ultra-phenotyped cohort’ based on kidney biopsies. Secondly, there was a ‘clinical cohort’ of 89 patients with non-selected FSGS. First, we evaluated the ability of histological morphometric parameters to discriminate pFSGS and maladaptive FSGS (mFSGS) on kidney biopsies in an ultraphenotyped cohort of normative pFSGS and mFSGS patients ( n = 58). We performed a multivariate analysis to identify variables associated with the type of FSGS, leading to the identification of the Mean Interglomerular Area (MIA) as a marker to distinguish pFSGS from mFSGS. Secondly, we evaluated the discriminative value of this parameter in a clinical cohort ( n = 89) including primary, maladaptive, and undetermined FSGS. High MIA classified mFSGS versus. pFSGS in the ultraphenotyped cohort, even after stratification by age, eGFR and proteinuria. In the clinical cohort, MIA was also significantly lower in primary FSGS. Although MIA was strongly correlated with serum albumin, a classical marker of pFSGS, they conflicted in 30% of patients. In this subset, relying solely on serum albumin frequently misclassified patients by FSGS subtype (27%), highlighting the usefulness of combining MIA with serum albumin to adjudicate FSGS subtypes. FSGS of uncertain origin segregated with either pFSGS or mFSGS cases on an MIA by albumin graph, suggesting that they do not constitute a different pathological entity. Digital pathology, in conjunction with simple biological features, is a useful tool for FSGS classification, suggesting that it might improve the design of future therapeutic and mechanistic FSGS studies.