Objective: To determine the correlations between ophthalmoscopic estimations and the measurements with a semiautomated image-analysis device of the vertical cup-to-disc ratio (VCDR) in the human optic disc.
Design: Population-based, cross-sectional study.
Participants: All subjects 55 years of age and older from the population-based sample of 6777 ophthalmologically examined subjects from The Rotterdam Study of whom gradable optic disc transparencies of at least 1 eye and ophthalmoscopic data of the same eye were available.
Main outcome measures: Ophthalmoscopic assessment of the VCDR and semiautomated measurement of the VCDR.
Methods: Indirect and direct ophthalmoscopy were performed in mydriasis to assess the VCDR. Optic disc transparencies made with a simultaneous stereoscopic telecentric fundus camera were analyzed with a semiautomated measurement system (Topcon Imagenet).
Results: In 5143 subjects, the mean ophthalmoscopic VCDR was 0.30 (standard error [SE], 0.0021; range, 0.00, 1.00) compared with a semiautomatically measured VCDR of 0.49 (SE, 0.0019; range, 0.04, 0.86; difference, 0.19; P < 0.0001). The overall correlation between both methods was moderate (correlation coefficient, 0.61; SE, 0.11) and lower in small optic discs. Semiautomated optic disc measurements correctly identified 76% of the glaucoma cases (as defined using visual field data and ophthalmoscopic data about the optic disc).
Conclusion: Semiautomated measurements of the VCDR are larger than the ophthalmoscopic VCDR estimate with a moderate correlation. The interobserver variability using Imagenet was smaller compared with the ophthalmoscopic assessments, and Imagenet was better standardized, which is important for epidemiologic surveys and follow-up studies.