Phenotype, serotype, and data-driven clustering reveal complementary dimensions of heterogeneity in ANCA-associated vasculitis: a multicentre Japanese cohort (J-CANVAS)

OBJECTIVE: To compare clinicopathological phenotype-based, anti-neutrophil cytoplasmic antibody (ANCA) serotype-based, and unsupervised data-driven classifications in relation to clinical outcomes and patient heterogeneity in a large Japanese cohort with ANCA-associated vasculitis (AAV). METHODS: This multicentre, retrospective cohort study analysed data from a nationwide Japanese registry of 729 newly diagnosed patients with granulomatosis with polyangiitis (GPA) or microscopic polyangiitis (MPA), all positive for myeloperoxidase (MPO)- or proteinase 3 (PR3)-ANCA. Patients were classified by phenotype, serotype, combined phenotype-serotype groupings, and data-driven clustering based on baseline clinical and laboratory features. Associations with clinical outcomes-including mortality, relapse, and response to rituximab (RTX) versus cyclophosphamide (IVCY)-were evaluated using inverse probability of treatment weighting (IPW). RESULTS: Phenotype-based classification more accurately distinguished all-cause mortality risk (MPA vs. GPA: hazard ratio [HR] 2.53, 95% CI 1.34-4.76). Combined phenotype-serotype analysis identified MPO-MPA patients with the highest mortality (HR 3.45, 95% CI 1.09-11.0, vs. PR3-GPA) and PR3-GPA with the highest severe relapse. Discordant groups, such as MPO-GPA, demonstrated unique clinical characteristics. After IPW adjustment, no significant difference in 24-week remission rates was observed between RTX and IVCY across classifications, both overall (RR 1.02, 95% CI 0.95-1.09) and within subgroups. Unsupervised clustering identified four distinct clinical subgroups, with limited concordance with conventional phenotype or serotype classifications. CONCLUSION: Phenotype and serotype classifications provide complementary, not competing, prognostic insights in Japanese patients with AAV. Data-driven clustering revealed additional clinical heterogeneity not captured by traditional systems, underscoring the need for integrated, multi-dimensional stratification approaches to improve personalised risk assessment and treatment strategies.