Classification of Primary Angle Closure Disease by Hierarchical Cluster Analysis of Ocular Biometrics in the Dark and Light

Abstract

ABSTRACTPurposeTo investigate the classification of eyes with primary angle closure disease (PACD) based on hierarchical cluster analysis of ocular biometrics measured in the dark and light using anterior segment OCT (AS-OCT).MethodsParticipants of the Chinese American Study received complete eye exams, including gonioscopy and AS-OCT imaging, to identify primary angle closure suspects (PACS) and primary angle closure without/with glaucoma (PAC/G). Biometric parameters analyzed included angle opening distance (AOD750) and trabecular iris space area (TISA750), iris area (IA), iris thickness (IT750), iris curvature (IC), lens vault (LV), anterior chamber width (ACW) and anterior chamber depth (ACD). Hierarchical cluster analysis was performed using Ward’s method and Euclidean distance.ResultsAnalysis of 159 eyes with PACS or PAC/G produced 2 clusters in both dark and light. In both analyses, the primary cluster (N=132 in the dark, N=126 in the light) was characterized by smaller AOD750 and TISA750, greater IC, and greater LV (p<0.001). The proportion of PACS to PAC/PACG eyes was significantly different between clusters in the light (p=0.02) but not the dark cluster analysis (p=0.08). On multivariable logistic regression analysis, smaller TISA750 (OR=0.84 per 0.01μm2) and AOD750 (OR=0.93 per 0.01mm) in the light and smaller TISA750 (OR=0.86 per 0.01μm2) in the dark were significantly associated (p≤ 0.02) with higher odds of PAC/G.ConclusionCluster analysis of ocular biometrics can classify PACD eyes by disease severity. Ocular biometrics appear equally if not more strongly predictive of disease severity when measured in the light than dark. Clustering of biometric measurements obtained in the light could provide a novel method to risk-stratify patients for more severe PACD.