Each NHANES participant had nonmydriatic 45° photographs taken of the macula and optic disc of both eyes (CR6-45NM; Canon USA, Melville, NY, USA). Initial grading of the photographs, including cup-to-disc ratio (CDR), was performed at the University of Wisconsin Fundus Photograph Reading Center.
13 The photographs were also evaluated for the presence of macular disease including macular edema, panretinal photocoagulation, focal photocoagulation, artery or vein occlusion, diabetic retinopathy, age-related macular degeneration, chorioretinal abnormalities, macular hole, and retinal detachment. All images with a CDR ≥ 0.6 on initial grading (1201 images of 1073 eyes from 549 participants) were reviewed and regraded by three glaucoma specialists (PG, MVB, and DSF) at the Wilmer Eye Institute of the Johns Hopkins University School of Medicine. For this analysis, we assumed that participants with CDR < 0.6 in both eyes did not have glaucoma (this value is close to the optimal cutoff point for defining glaucomatous optic neuropathy in population-based glaucoma risk factor analysis).
14 As a consequence, all glaucoma cases in the analysis were derived from participants who had at least one eye with CDR ≥ 0.6 on initial grading.
The nonstereo color images of participants with CDR ≥ 0.6 as determined by the reading center were transferred to a tablet-based review system (TruthMarker, IDx, LLC; Iowa City, IA, USA), and three glaucoma specialists regraded each image to determine image quality (excellent, good, fair, poor, ungradable), vertical CDR (0.0–1.0 in increments of 0.1), notching of the neuroretinal rim (none, inferior, superior, both), excavation of the optic cup (no, maybe, yes, unable), optic disc hemorrhage (no, maybe, yes, unable), tilting of the disc (no, yes), and relative disc size (small, average, large). Each glaucoma specialist then determined likelihood of glaucoma (no, possible, probable, definite, unable) based on clinical judgment of all features of disc appearance. The results were adjudicated where necessary.
To evaluate the reliability of the CDR determined by the reading center, 180 participants with CDR < 0.6 in both eyes (423 images of 360 eyes of 180 participants) were randomly sampled. Among the 180 participants, 2 participants had missing grading on CDR; 157 participants had CDR < 0.6 in both eyes in both the initial and second readings; and 21 participants were classified as having a CDR ≥ 0.6 in at least one eye on the reading at Johns Hopkins University. Among all regraded participants, the kappa for agreement between the Johns Hopkins University and University of Wisconsin readings on CDR categorization was 0.76.