Detection of new-onset choroidal neovascularization using optical coherence tomography: the AMD DOC Study

Abstract

Purpose: To determine the sensitivity of time domain optical coherence tomography (OCT) in detecting conversion to neovascular age-related macular degeneration (AMD) in eyes at high risk for choroidal neovascularization (CNV), compared with detection using fluorescein angiography (FA) as the gold standard.

Design: Prospective, multicenter, observational study.

Participants: Individuals aged ≥50 years with nonneovascular AMD at high risk of progressing to CNV in the study eye and evidence of neovascular AMD in the fellow eye.

Methods: At study entry and every 3 months through 2 years, participants underwent best-corrected visual acuity, supervised Amsler grid testing, preferential hyperacuity perimetry (PHP) testing, stereoscopic digital fundus photographs with FA, and OCT imaging. A central Reading Center graded all images.

Main outcomes measures: The sensitivity of OCT in detecting conversion to neovascular AMD by 2 years, using FA as the reference standard. Secondary outcomes included comparison of sensitivity, specificity, positive predictive value, and negative predictive value of OCT, PHP, and supervised Amsler grid relative to FA for detecting incident CNV.

Results: A total of 98 participants were enrolled; 87 (89%) of these individuals either completed the 24-month visit or exited the study after developing CNV. Fifteen (17%) study eyes had incident CNV confirmed on FA by the Reading Center. The sensitivity of each modality for detecting CNV was: OCT 0.40 (95% confidence interval [CI], 0.16-0.68), supervised Amsler grid 0.42 (95% CI, 0.15-0.72), and PHP 0.50 (95% CI, 0.23-0.77). Treatment for incident CNV was recommended by the study investigator in 13 study eyes. Sensitivity of the testing modalities for detection of CNV in these 13 eyes was 0.69 (95% CI, 0.39-0.91) for OCT, 0.50 (95% CI, 0.19-0.81) for supervised Amsler grid, and 0.70 (95% CI, 0.35-0.93) for PHP. Specificity of the OCT was higher than that of the Amsler grid and PHP.

Conclusions: Time-domain OCT, supervised Amsler grid, and PHP have low to moderate sensitivity for detection of new-onset CNV compared with FA. Optical coherence tomography has greater specificity than Amsler grid or PHP. Among fellow eyes of individuals with unilateral CNV, FA remains the best method to detect new-onset CNV.