An independent evaluation of the Age-Related Eye Disease Study (AREDS) cataract grading system

To assess intra- and interobserver reproducibility of the Age-Related Eye Disease Study (AREDS) system for grading lens opacities and to provide data on its capacity to reliably detect changes in lens status.

Independent and replicate grading of 40 sets of lens photographs (one slit-lamp and two retroillumination photographs) were performed by three experienced observers. Patients were participants in the Collaborative Italian-American Clinical Trial of Nutritional Supplements which is testing the effect of a mineral-multivitamin supplement on age-related cataract (CTNS). Scatterplots and intraclass correlation were used to assess measurement error.

Analysis revealed good intra- and interobserver reproducibility of the system. Greatest intraobserver measurement error showed 100% of pairs within 10% areal difference for cortical cataract, 97.5% within 15% areal difference for posterior subcapsular cataract, and 100% within 1 density unit difference for nuclear opacity. Greatest interobserver measurement error showed 95% of pairs within 10% areal difference for cortical cataract, 97.5% within 15% areal difference for posterior subcapsular cataract, and 97.5% within 1.5 density unit difference for nuclear opacity.

The AREDS lens opacities grading system appears to be sufficiently reliable to detect changes of at least 10% areal involvement for cortical, 15% areal involvement for posterior subcapsular, and 1.0 units for nuclear opacities. It therefore seems sufficiently sensitive to adequately monitor progression of lens opacities in a longitudinal study of patients with early cataract. Its applicability in a population with advanced or complex mixed opacities must await further testing.