Risk factors for conversion to neovascular age-related macular degeneration based on longitudinal morphologic and...

To use longitudinal quantitative morphologic and visual acuity (VA) data to investigate the risk of choroidal neovascularization (CNV) event occurrence in eyes with dry age-related macular degeneration (AMD).

Prospective observational study.

A total of 513 participants (844 eyes) followed longitudinally in one center enrolled in the Age-Related Eye Disease Study (AREDS) or the Prophylactic Treatment of AMD Study (PTAMD).

We assessed images of previously obtained fundus photographs for the presence of macular pigmentation, drusen area, and drusen distribution (number and size), and fellow eye CNV status at baseline. Early Treatment Diabetic Retinopathy Study (ETDRS) best-corrected visual acuity at each visit and the age of each subject were obtained. We used a longitudinal logistic mixed-effects model with random intercepts fitted to event occurrences to assess risk on a per eye basis.

Odds ratios for CNV event.

Thirty-one subjects (6.0%) had events. Only VA changes over time and follow-up interval showed statistically significant effects. Several statistical models that included VA at the previous visit were used. In 2 models, 3 categories of VA were used: ≤ 75 letters, >75 and ≤ 85 letters, and >85 letters. Two categories were used for follow-up: ≤ 3 years versus >3 years or ≤ 1 year versus >1 year. In the first model with categorization at 3 years, a decrease in acuity from the >85 letter category to ≤ 75 letters increased the odds of CNV by 16.9 times (P = 0.022). In the model with categorization at 1 year, a decrease in acuity from the >85-letter category to ≤ 75 letters increased the odds of CNV by 21.4 times (P = 0.0175). Differences between the follow-up intervals were significant (P = 0.043) and indicated a more than 7-fold increase in the odds. Changes in morphologic features of the macula did not show significant effects.

A decrease in VA to ≤ 75 ETDRS letters in an eye with an initial ETDRS baseline acuity of >85 letters increases the likelihood of CNV by approximately 20-fold. This likelihood also increases with aging.