Differentiating Multifocal Choroiditis and Punctate Inner Choroidopathy: A Cluster Analysis Approach.

PURPOSE: To develop a robust approach to clinical phenotyping of multifocal choroiditis (MFC) and punctate inner choroidopathy (PIC). DESIGN: Cross-sectional and longitudinal observational study. METHODS: This multicenter study included sites in the United Kingdom and Israel. The study population included 343 eyes of 185 subjects with hospital record diagnoses of MFC or PIC. Eyes were observed over a period of 5 years for clinically relevant characteristics, including demographics and multimodal imaging features, by observers masked to the original diagnoses. Multivariate 2-step cluster analysis was used to identify clusters of eyes in the database with similar clinical phenotypes, which were then analyzed for between-group differences. The primary outcome measure was the difference between clinical phenotype clusters identified using clinical criteria from the multivariate cluster analysis. RESULTS: Subjects ranged from 11 to 89 years of age, with a baseline best-corrected visual acuity of 2.3 to -0.2 logarithm of minimal angle of resolution. Eighty-two percent of eyes were from females, 74% were myopic with a refractive error of +3.00 to -17.00 diopters (spherical equivalent). Cluster analysis prioritized clinical criteria of chorioretinal lesion location and intraocular inflammation and identified 2 distinct phenotype clusters resembling the original descriptions of MFC and PIC. During the 5-year period of observation, the initial clinical diagnosis remained stable for most eyes and only 1 eye (0.3%) changed diagnosis from PIC to MFC because of newly developed peripheral lesions. There were significant between-group differences in clinical characteristics, for example, in choroidal neovascular membrane development and treatment received. CONCLUSIONS: Cluster analysis of this large cohort of eyes identified peripheral lesions and intraocular inflammation as distinct clinical phenotypes of MFC and PIC. The initial diagnosis remained stable for most eyes. This methodology could be useful for future uveitis classification and management.