Table 1_Exploratory cluster analysis of self-reported adverse events associated with antidepressants using Gower distance and partitioning around medoids: a single-center cross-sectional study.docx

Associated adverse events (AEs) represent the primary reason for treatment discontinuation. Existing studies mostly limit analysis to simple stratification by organ system or severity, while exploratory subtyping grounded in multidimensional clinical features of AEs remains scarce.</p>Objective<p>This study aims to overcome the limitations of traditional single-dimensional analyses by integrating multidimensional data—including demographic and sociological​ characteristics, clinical disease severity, and AE-related information—to conduct exploratory cluster subtyping at the patient level.</p>Methods<p>This study is an exploratory secondary analysis of a single-center, retrospective, recall-based cross-sectional survey on antidepressant adverse events registered with the Chinese Clinical Trial Registry (ChiCTR2500111836).

It included 500 patients (905 AE episodes) self-reporting AEs within the past year at Beijing Anding Hospital (April 2025–January 2026). Unlike the original survey’s descriptive aim, this analysis identified heterogeneous patient subtypes: one AE per patient ID ensured independence; the Gower-PAM algorithm clustered 23 variables; and bootstrap-derived Adjusted Rand Index (ARI) values plus an all-event sensitivity analysis confirmed robust subtyping stability.</p>Results<p>K = 2 was the optimal clustering solution (silhouette coefficient = 0.219; mean ARI = 0.916).