Data Collection

We retrospectively reviewed the charts and imaging data of 374 consecutive patients diagnosed with treatment-naïve neovascular AMD between October 1, 2005 and December 1, 2012. Treatments with ranibizumab (0.5 mg/0.05 mL; Lucentis, Genentech, San Francisco, California, USA), bevacizumab (1.25 mg/0.05 mL, Avastin; Genentech), or aflibercept (2.0 mg/0.05 mL, Eylea; Regeneron, Tarrytown, New York, USA) were administered by a single physician (K.B.F.).

Inclusion criteria were similar to the Minimally Classic/Occult Trial of the Anti-VEGF Antibody Ranibizumab in the Treatment of Neovascular Age-Related Macular Degeneration (MARINA) and Anti-VEGF Antibody for the Treatment of Predominantly Classic Choroidal Neovascularization in Age-Related Macular Degeneration (ANCHOR) study groups. All participants were older than 50 years with newly diagnosed treatment-naïve NV as evidenced by clinical examination and FA. Best-corrected visual acuity was 20/20–20/800 on a Snellen chart (differed from ANCHOR/MARINA, which included 20/40–20/320 on the Early Treatment Diabetic Retinopathy Study charts). Additionally, eyes in the study must have had OCT imaging (time-domain or spectral-domain) performed at the time of diagnosis.

Exclusion criteria were any of the following: previous treatments for CNV in the study eye, including photodynamic therapy (PDT), intravitreal steroids, intravitreal pegaptanib (Macugen; Valeant, Montreal, Quebec, Canada), or thermal laser and eyes with CNV lesions presenting with subfoveal fibrosis, central geographic atrophy (GA) at baseline, or retinal pigment epithelial tears, or composed of more than 50% hemorrhage. Eyes with CNV secondary to other maculopathies, including degenerative myopia, angioid streaks, presumed ocular histoplasmosis syndrome, or inflammatory maculopathies, were excluded.

Demographic information including age; sex; race; family history of AMD, smoking status (current, former, never), history of hypertension and diabetes, history of statin, aspirin, clopidogrel, and/or warfarin use, and history of glaucoma were collected for each patient.

FA images were obtained using a Topcon TRC 501x fundus camera (Topcon Imagenet, Tokyo, Japan). OCT imaging of all patients was performed with time-domain OCT (Stratus; Carl Zeiss Meditec Inc, Dublin, California, USA) or spectral-domain OCT (Spectralis; Heidelberg Engineering, Heidelberg, Germany; or 3-D OCT-2000; Topcon, Tokyo, Japan). OCT instrumentation was necessary for additional accurate identification of lesion subtype utilizing the anatomic classification of lesion subtype. Standard methods of image acquisition were employed for all imaging modalities.

Image Grading

The classification of neovascular lesions was made independently by 2 experienced retina specialists (S.M. and R.G.P.) who evaluated the presenting color photographs, FA, and OCT. First, all the color photographs and FA corresponding to the baseline diagnostic visit were analyzed. Neovascular lesions were subtyped according to the MPS criteria and the Digital Angiographic Reading Center (DARC) Reader’s Manual as occult or classic CNV, and RAP lesions were identified by criteria defined by Yannuzzi and associates and the DARC Reader’s Manual. Secondly, OCT images corresponding to the same diagnostic visit were reviewed, and each case was classified according to the guidelines provided by Freund and associates. The anatomic classification, which uses OCT in combination with FA, categorizes lesions as type 1 (sub–retinal pigment epithelium [RPE]), type 2 (subretinal), type 3 (intraretinal), or mixed NV. Eyes with PCV were considered to be a form of type 1 CNV. Type 1, 2, and 3 NVs corresponded to occult, classic, and RAP angiographic lesions, respectively. Cases with multiple lesion types were identified as mixed NV and each component was also recorded. Table 1 provides further details of the criteria of CNV subtype classification based on each imaging modality. Finally, in cases with disagreement between FA and OCT findings ( Figure 1 ), FA images were reanalyzed focusing on early frames to better recognize subtle angiographic findings, in particular those of RAP lesions. A third supervising grader (K.B.F.) evaluated the lesion type in the presence of significant discrepancies.

Color fundus photograph of the right eye of an 82-year-old white woman with age-related macular degeneration (Top left) showing a grayish choroidal neovascular membrane (CNV) with surrounding drusen. Fluorescein angiogram shows early well-demarcated intense hyperfluorescence (Top right) and intense late leakage (Bottom left) graded as classic CNV. The addition of spectral-domain optical coherence tomography (green arrows, Top right) shows both type 2 neovascularization (white arrow, upper image of Bottom right) and type 1 neovascularization (white arrow, lower image of Bottom right) with associated cystoid macular edema, constituting a mixed lesion.

Readers also graded the lesion location and overall size. FA was used to measure the greatest linear diameter (mm) and the total area of CNV lesion (mm²). Measurements were performed only on fundus camera images. The total area of CNV lesion was defined as the area of CNV leakage plus any contiguous areas of hemorrhage, blocked fluorescence, or serous PED that could be obscuring the boundaries of the CNV. The lesion location was defined as foveal (subfoveal or juxtafoveal) or extrafoveal, as determined according to the MPS terminology.

Statistical Analysis

Statistical analysis was performed using SPSS software Version 21 (SPSS, Inc, Chicago, Illinois, USA). The numbers of neovascular AMD lesions as identified by the FA classification system and by the anatomic (FA and OCT) classification system were recorded, including the breakdown of neovascular lesion components in the mixed NVs. A subgroup analysis was also performed to identify neovascular AMD lesions based on these same classification systems in the newly diagnosed neovascular AMD eyes that had baseline spectral-domain OCT (3-D OCT 2000; Topcon, Tokyo, Japan; and Spectralis; Heidelberg Engineering, Heidelberg, Germany). The κ statistic was performed to compare the agreement between both classification systems with the entire cohort and subgroup with initial spectral-domain OCT. This analysis expresses the extent to which the observed agreement exceeds that which would be expected by chance alone and is defined as follows: greater than 0.75 represents “excellent” agreement; 0.40–0.75 represents “fair” to “good” agreement; and less than 0.40 represents “poor” agreement.

The associations between FA CNV and anatomic NV classifications and each possible demographic variable were assessed individually by Fisher exact test, χ ² test, or independent Student t test. Similarly, the associations between lesion characteristics such as size, greatest linear diameter, and location and each possible demographic variable were also analyzed. The demographic factors that were shown to have a significant association were incorporated into regression analyses as covariates when examining the associations between lesion characteristics and neovascular lesion subtypes. Adjusting for demographic confounders ensured that the observed associations between clinical characteristics and NV types were real and not attributable to demographic confounders.