Ensemble clustering of organ involvement patterns in ANCA-associated vasculitis identified four reproducible clinical phenotypes with distinct prognostic and therapeutic profiles, suggesting that integrating phenotypic patterns with ANCA serotype may enhance risk stratification and support more individualized management strategies.
Key Findings
Results
Ensemble clustering of organ involvement data identified four reproducible clinical phenotypes of AAV across two independent Japanese cohorts.
The study included 726 patients with newly diagnosed MPA or GPA from two nationwide Japanese registries (J-CANVAS as derivation cohort; JPVAS as validation cohort).
The four clusters were: (a) renal-dominant without interstitial lung disease (ILD); (b) renal-dominant with ILD; (c) systemic multi-organ; and (d) ear, nose and throat (ENT)-dominant.
Clustering was applied to high-dimensional organ involvement data derived from the Birmingham Vasculitis Activity Score and additional clinically relevant manifestations.
Classification and regression tree (CART) analysis was used to replicate and validate the clusters across both cohorts, demonstrating good concordance.
Results
Overall survival differed significantly across the four identified clinical phenotype clusters.
Survival outcomes were evaluated across all four clusters in both derivation and validation cohorts.
The ENT-dominant cluster demonstrated favourable overall survival compared to other clusters.
Renal-dominant and systemic multi-organ clusters were associated with worse survival outcomes.
These survival differences were observed after applying the CART-based cluster classification to both cohorts.
Results
Relapse incidence differed across the four clinical phenotype clusters, with the ENT-dominant cluster showing a higher relapse risk despite favourable survival.
The ENT-dominant cluster demonstrated favourable overall survival but a higher relapse risk relative to other clusters.
Relapse incidence was evaluated across all four clusters in both derivation and validation cohorts.
The combination of good survival but higher relapse risk in the ENT-dominant cluster highlights a clinically distinct trajectory for this phenotype.
These relapse patterns were reproducible across the two independent cohorts.
Results
The ENT-dominant cluster showed a trend towards improved relapse-free survival with rituximab compared with cyclophosphamide (CYC), including among MPO-ANCA-positive patients.
This trend toward rituximab benefit was observed specifically in the ENT-dominant phenotype cluster.
Notably, this association was present even among MPO-ANCA-positive patients in the ENT-dominant cluster, a subgroup not conventionally prioritized for rituximab therapy.
The finding suggests that organ involvement phenotype, not just ANCA serotype, may influence therapeutic response.
The study framed this as a 'trend' rather than a statistically definitive finding, indicating the need for further confirmatory research.
Methods
The study was conducted as a multicentre retrospective cohort study using two nationwide Japanese registries to enable derivation and independent validation of clusters.
J-CANVAS served as the derivation cohort and JPVAS served as the validation cohort.
A total of 726 patients with newly diagnosed MPA or GPA were included across both cohorts.
Ensemble clustering was applied to derive phenotypes, and CART was used for cluster replication and validation.
The retrospective multicentre design across two independent registries strengthens the reproducibility and generalizability of the identified phenotypes within the Japanese population.
Background
Conventional disease subtypes (MPA vs. GPA) and ANCA serotypes do not fully capture the clinical heterogeneity of AAV, motivating data-driven phenotyping approaches.
The authors state that 'clinical heterogeneity in ANCA-associated vasculitis (AAV) is not fully captured by conventional disease subtypes or ANCA serotypes.'
The study aimed to identify 'data-driven, clinically meaningful disease phenotypes based on patterns of organ involvement.'
The authors propose that 'integrating phenotypic patterns with ANCA serotype may enhance risk stratification and support more individualized management strategies in AAV.'
The identification of MPO-ANCA-positive patients in the ENT-dominant cluster potentially benefiting from rituximab exemplifies how organ-involvement phenotypes can provide therapeutic insights beyond serotype classification alone.
What This Means
This research suggests that patients with ANCA-associated vasculitis (AAV), an autoimmune disease that causes inflammation of blood vessels, can be grouped into four distinct clinical types based on which organs are affected. Using advanced computational clustering methods applied to data from 726 Japanese patients across two independent national registries, researchers identified groups characterized by: kidney involvement without lung scarring, kidney involvement with lung scarring, widespread multi-organ involvement, and predominant involvement of the ear, nose, and throat (ENT). These groupings were reproducible across both patient databases, lending confidence to their validity.
The four groups had meaningfully different outcomes. Patients in the ENT-dominant group had better overall survival chances, but were more likely to experience disease relapse (flare-ups) compared to patients in the other groups. Importantly, this research suggests that patients in the ENT-dominant group may respond better to a drug called rituximab compared to cyclophosphamide for preventing relapses — and this potential benefit appeared even among patients with a specific antibody type (MPO-ANCA positive) who are not conventionally considered the primary candidates for rituximab therapy.
This research matters because current approaches to classifying and treating AAV rely heavily on the type of antibody a patient has (their 'ANCA serotype') and their clinical diagnosis (MPA or GPA). This study suggests that the specific pattern of organ involvement provides additional, clinically important information that could help doctors better predict a patient's prognosis and choose the most appropriate treatment. Combining organ involvement patterns with antibody type may allow for more personalized treatment decisions in the future.
Inoue H, Kida T, Fujioka K, Omura S, Yanagida T, Shimada Y, et al.. (2026). Clinical phenotypes of ANCA-associated vasculitis using ensemble clustering: prognostic and therapeutic insights from two Japanese cohorts.. Rheumatology (Oxford, England). https://doi.org/10.1093/rheumatology/keag483