PURPOSE: To classify and describe clinically meaningful classes of color vision defects using pretreatment Farnsworth-Munsell 100-hue results from the Early Treatment Diabetic Retinopathy Study (ETDRS) patients using standard statistical techniques. DESIGN: The ETDRS was a randomized trial investigating retinal photocoagulation and oral aspirin in diabetic retinopathy. METHODS: Farnsworth-Munsell (FM) 100-hue test was successfully administered before initiation of study treatment in each eye of 2701 of the 3711 ETDRS patients. Test results were converted into a Fourier series, classified by cluster analysis in the deferred-treatment group of eyes, and verified in the immediate-treatment group of eyes as separate samples. RESULTS: Cluster analysis uncovered thirteen distinct patterns. Pattern A (51% or 1366 of the eyes) showed unimpaired hue discrimination and was comprised of younger patients with no or little macular edema. Pattern B eyes (10% or 262) showed generalized impairment of hue discrimination with no main axis defect. Patterns C (C1, C2, C3), comprising 26% (or 698) of the eyes, showed increasing severity of the yellow-blue diabetic retinopathy defect, associated with increasing mean age and increasing macular edema severity. Patterns D (D1, D2), comprising 6% (or 164) of the eyes, were similar to the C patterns but showed a stronger yellow-blue defect. Patterns E (E1, E2, E3), or approximately 2% (or 38) of the eyes and predominantly male, exhibited the expected pattern for congenital protan defect. Patterns F, G, and H, approximately 6% (or 153) of the eyes, showed distinct patterns of one-sided axes. The nomenclature is arbitrary. CONCLUSIONS: Cluster analysis of FM 100-hue test results has found 13 patterns of impaired hue discrimination, helpful in understanding color vision defects in diabetes mellitus. (C) 2004 by Elsevier Inc. All rights reserved.
