Evaluating clinical heterogeneity in relapsing polychondritis through unsupervised cluster analysis

As a rare autoimmune disease, relapsing polychondritis (RP) exhibits individual variations in clinical manifestations, treatment response and prognosis. To date, no biomarkers have been implemented to clinical practice.This study aimed to apply cluster analysis to identify the clinical phenotype, clarify their prognosis predictor, and monitor treatment decisions to improve the outcomes of patients.

A total of 135 RP patients hospitalized in two centers from January 2005 to December 2024 were involved in the study.The K-means clustering algorithm was used as the core clustering tool. The optimal number of clusters(K = 3)was determined through comprehensive evaluation using the Elbow Method and Silhouette Analysis.

Cluster analysis identified 3 RP phenotypes.Cluster 1 comprised 45 cases(33.3%),characterized by a hyperinflammatory state and predominant respiratory involvement; Cluster 2 included 11 cases (8.1%),defined by severe and multi-organ involvement Cluster 3 which accounted for 79 cases (58.52%), more than half of the cohort, was predominantly male and exhibited the most favorable prognosis.In contrast, Clusters 1 and 2 were associated with the highest rates of clinical deterioration.

This study confirms that relapsing polychondritis is not a single-dimensional disease, but a clinical syndrome with significant heterogeneity. Random forest analysis further identifies that systemic inflammation markers (CRP, CAR, ESR) and sensory organ involvement (hearing impairment and inner ear dysfunction) as the most critical driving factors for distinguishing these subtypes.