Patient Selection

The AREDS2 Ancillary Spectral Domain OCT Study ( ClinicalTrials.gov identifier NCT00734487 ) is an ancillary study to the prospective, multicenter AREDS2 ( ClinicalTrials.gov identifier NCT00345176 ) for the purpose of identifying and evaluating SD OCT characteristics associated with AMD disease progression. The study followed the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board at all respective institutions. Subjects from the Duke Eye Center and National Eye Institute underwent EDI-OCT on at least 1 study visit between 2011 and 2013, and were included in this cross-sectional analysis. Major inclusion and exclusion criteria for AREDS2 and control subjects, as well as criteria for AMD classification, have been previously described. Additionally, eyes with high myopia >6 diopters were also excluded from this study. All participants had best-corrected visual acuity (BCVA) measured as Early Treatment Diabetic Retinopathy Study (ETDRS) letter score and refractive error measured by a certified ophthalmic technician.

Image Acquisition

Enhanced-depth imaging was performed on both eyes from each subject using the Heidelberg Spectralis SD OCT (870 nm) device (Heidelberg Engineering, Heidelberg, Germany). EDI-OCT scans were obtained using the Spectralis EDI mode, a preset, software-driven algorithm that places the RPE near the zero-delay line while producing an upright enhanced choroidal image. In EDI-OCT mode, a single 30-degree horizontal line scan (approximately 8.9 mm) captures 1536 A-scans per B-scan with 40 averaged B-scans per image, with automatic averaging and eye-tracking features. Parameters for standard SD OCT volume scans have been described previously.

Choroidal Segmentation

CT was measured semi-automatically using the Duke Optical Coherence Tomography Retinal Analysis Program (DOCTRAP). This software, which employs automated segmentation algorithms based on graph theory and dynamic programming, also provides a custom graphical user interface (GUI) for manual adjustments to allow for improved segmentation precision. The inner boundary of the choroid is defined by the outer border of the hyperreflective layer corresponding to the Bruch membrane, while the outer boundary is defined by the outer border of the choroid stroma. The visibility of a hyporeflective band corresponding to the suprachoroidal layer was determined by 2 independent masked observers (G.Y. and P.A.P.) using methods described previously, with any disagreements resolved by a masked senior grader (C.A.T.). When visible, the suprachoroidal layer was not included in choroidal thickness measurements, as the relevance of the suprachoroidal layer to AMD pathogenesis has not been well established. Choroidal thickness measurements were recorded at the subfoveal location and at 0.5-mm intervals from 1.5 mm nasal to 1.5 mm temporal to the fovea ( Figure 1 ). Corresponding retinal thickness measurements were also taken at the fovea (CFT) and at 0.5-mm intervals from 1.5 mm nasal to 1.5 mm temporal to the fovea. To minimize the variability in measurements taken at single point locations, average choroidal and retinal thicknesses were also calculated, respectively, as the mean of choroidal and retinal thickness measurements from the central 3-mm segment (from 1.5 mm nasal to 1.5 mm temporal to the fovea), based on the 3-mm diameter of the AREDS maculopathy grading grid for color fundus photography grading.

Semi-automated segmentation of retinal and choroidal thickness on enhanced-depth imaging optical coherence tomography. Images show original (Left) and segmented (Right) optical coherence tomography images taken from eyes with no age-related macular degeneration (AMD) (Top), intermediate AMD (Middle), and advanced AMD (Bottom). The inner border of the retinal surface (blue), outer border of the retinal pigment epithelium (yellow), and outer border of the stromal choroid (green) are shown in the segmented images. The fovea is marked with a yellow asterisk.

Statistical Analyses

Statistical analyses were performed using SAS (version 9.3; SAS Institute Inc, Cary, North Carolina, USA). Differences in systemic (age, sex, and smoking status) and ocular (laterality, lens status, refractive error, BCVA) characteristics between control and eyes with AMD were compared using generalized estimating equations to account for up to 2 eyes measured per subject. Time of image acquisition was compared using generalized estimating equations. Univariate linear regression followed by multivariate linear regression modeling was employed to evaluate the association of choroidal thickness with systemic and ocular parameters, as well as retinal thickness (CFT or average retinal thickness). Differences in choroidal and retinal thickness measurements were compared using analysis of variance using generalized estimating equations. To adjust for choroidal thickness variation with age and refractive error, comparisons were made using analysis of covariance using generalized estimating equations. Fellow eye comparisons were made using paired-samples t tests.

Demographic and Clinical Characteristics

The 164 subjects (mean age 72.1 ± 8.6 years) included in this analysis consisted of 67 male (40.9%) and 97 female subjects (59.1%), with 151 (92.1%) white and 13 (7.9%) nonwhite ( Table 1 ). Among these, 81 (49.4%) were never smokers, 28 (17.1%) were former smokers (quit >1 year prior), and 55 (33.5%) were current smokers. Both eyes were included from 161 subjects; in the remaining 3 subjects, 1 eye was excluded owing to high myopia >6 diopters. Of the 325 eyes included in this study, 154 eyes (47.4%) were from control participants with no AMD, 109 eyes (33.5%) had intermediate AMD, and 62 eyes (19.1%) had advanced AMD. A total of 162 (49.8%) were right eyes and 256 (78.8%) were phakic at the time of imaging, with more pseudophakic eyes in AMD vs control eyes ( P < .001). Mean refractive error was +0.194 ± 2.06 diopters spherical equivalent, with no significant difference among the 3 groups ( P = .236). As expected, BCVA was significantly reduced in eyes with intermediate AMD ( P < .001) and advanced AMD ( P < .001) when compared with control eyes. Mean age of subjects was also greater with more advanced AMD status ( P < .001). There was no significant difference among the 3 groups with respect to the median time of day at which the EDI-OCT image was obtained ( P = .084).

Variables Associated With Choroidal Thickness

Univariate regression analyses of systemic and ocular variables revealed a significant association of subfoveal choroidal thickness with age (β = −4.58, P < .001), lens status ( P = .046), refractive error (β = 6.52, P = .045), and BCVA (β = 1.30, P = .001), but not with sex, smoking history, laterality, CFT, or average retinal thickness ( Table 2 ). Similarly, average choroidal thickness was also associated with age (β = −4.22, P < .001), lens status ( P = .041), refractive error (β = 6.45, P = .037), and BCVA (β = 1.10, P = .003), but not with other variables. Multivariate regression modeling using these variables ( Table 2 , Model 1) showed that only age ( P < .001) and refractive error ( P < .001), but not lens status ( P = .614–.641) or BCVA ( P = .143–.261), were significant predictors of subfoveal and average choroidal thickness. A repeat analysis using only age and refractive error ( Table 2 , Model 2) confirmed that both subfoveal and average choroidal thickness are associated with younger age and hyperopic refractive error ( P < .001 for all values). These results suggest that choroidal thickness decreases with advanced age and myopic refractive error across all eyes included in the study.

Choroidal and Retinal Thickness in Age-Related Macular Degeneration

Without adjusting for age and refractive error, mean average choroidal thickness was significantly lower in eyes with advanced AMD when compared to control eyes ( P = .008), with no significant differences between eyes with advanced and intermediate AMD ( P = .152) or between intermediate AMD and control eyes ( P = .098). Decreased choroidal thicknesses were also noted at the fovea (subfoveal choroidal thickness) and at all other locations from 1.5 mm nasal to 1.5 mm temporal to the fovea ( P < .05 at all locations) when comparing advanced AMD with control eyes ( Figure 2 ). However, after adjustment for age and refractive error, mean average choroidal thickness and all choroidal thickness measurements across the macula ( Table 3 , Figure 2 ) showed no significant differences between the 3 groups ( P > .05 at all locations). In contrast, average retinal thickness ( Table 3 ) and retinal thickness measurements across the macula ( Table 3 , Figure 2 ) were significantly greater in more advanced forms of AMD even after adjusting for age and refractive error ( P < .05 at all locations except 1.0 mm nasal to the fovea, where P = .054). A scatterplot showing the relationship between average choroidal thickness and age shows significant variability of choroidal thickness among individuals in each of the 3 groups, with no discernable mean reduction in average choroidal thickness among eyes with AMD ( Figure 3 ). Together, these results indicate that unlike differences in retinal thickness, the apparent choroidal thinning in advanced AMD may be attributed at least to age and refractive error as covariants. Variation in choroid-scleral junction appearance does not appear to affect these findings, as the proportion of eyes with a visible suprachoroidal layer is similar between the 3 groups ( P = .107) ( Table 3 ).

Unlike retinal thickness, choroidal thickness is not significantly associated with age-related macular degeneration (AMD) status after adjustment for age and refractive error. Line graphs compare mean choroidal thickness measured at 0.5-mm intervals from 1.5 mm nasal to 1.5 mm temporal to the fovea before (Top) and after (Middle) adjustment for age and refractive error (Rx), along with corresponding adjusted mean retinal thickness measured at the same locations (Bottom) for eyes with AMD and healthy controls. Asterisks indicate significant differences between the 3 groups ( P < .05).

Average choroidal thickness decreases with age in eyes with and without age-related macular degeneration (AMD). A scatterplot demonstrates the relationship of average choroidal thickness with age among eyes with no AMD (blue), intermediate AMD (red), and advanced AMD (green), along with the corresponding trend line.

Choroidal Thickness in Neovascular Age-Related Macular Degeneration or Geographic Atrophy

Among eyes with advanced AMD (n = 62), 48 eyes (14.8%) showed CNV and were classified as neovascular AMD, while 14 eyes (4.3%) had central GA. Although mean average choroidal thickness was lower in eyes with CNV (193.5 ± 101.7 μm) than in eyes with central GA (224.9 ± 108.9 μm), there was no significant difference between the 2 groups ( P = .431), even after adjustment for age and refractive error ( P = .374). Similarly, choroidal thickness measurements across all points in the macula showed no significant difference between eyes with CNV and central GA ( P > .05 at all points). Among eyes with neovascular AMD, the mean cumulative number of anti-VEGF treatments was 9.43 injections, with no clear association between total anti-VEGF injections received and average choroidal thickness ( P = .583) or subfoveal choroidal thickness ( P = .492).

Although the category of advanced AMD included only fovea-involving central GA based on AREDS2 categorization, many eyes with intermediate AMD also exhibited signs of noncentral GA on fundus examination or color fundus photographs. Among eyes with intermediate or advanced AMD (n = 171), 44 eyes (25.7%) had either central or noncentral GA. Interestingly, while eyes with either central or noncentral GA showed no significant reduction in average choroidal thickness or single choroidal thickness measurements when compared with eyes without GA ( P > .05 at all points), adjustment for age and refractive error revealed borderline thinning along the central 1-mm segment among eyes with either type of GA ( P = .05–.069). Moreover, subgroup analysis of 13 eyes with GA area measured on color fundus photographs showed a significant inverse association between average choroidal thickness and GA area (R ² = .906, P < .0001) after adjusting for age and refractive error ( Supplemental Figure , available at AJO.com ). These subgroup analyses suggest the possibility that choroidal thinning may be present in GA when larger areas of noncentral GA are also taken into account.

Fellow Eye Comparisons

To further assess choroidal thickness in AMD while minimizing confounding variables such as age and refractive error, choroidal thickness measurements were compared between fellow eyes in subjects with discordant stages of AMD. Twenty subjects had intermediate AMD in 1 eye and advanced AMD in the other (14 with neovascular AMD, 6 with GA). No subjects had intermediate or advanced AMD in only 1 eye with no ocular pathology in the fellow eye. Fellow eyes showed significant correlation in refractive error (R = 0.608, P = .004). CFT was significantly greater in eyes with neovascular AMD (353.2 ± 145.6 μm) relative to fellow eyes with intermediate AMD (271.4 ± 59.9 μm) ( P = .033). However, there was no significant difference in subfoveal choroidal thickness between eyes with neovascular AMD (215.6 ± 111.9 μm) and those with intermediate AMD (228.1 ± 106.1 μm) ( P = .360). Likewise, average choroidal thickness also showed no statistical difference between neovascular AMD (213.1 ± 107.7 μm) and intermediate AMD (221.7 ± 102.2 μm) ( P = .482). Eyes with central GA showed no significant difference from fellow eyes with intermediate AMD in CFT (186.3 ± 49.9 μm vs 201.3 ± 37.9 μm; P = .562), subfoveal choroidal thickness (229.7 ± 117.9 μm vs 225.8 ± 82.9 μm; P = .921), or average choroidal thickness (211.5 ± 105.2 μm vs 224.0 ± 86.4 μm; P = .684). These results further support our finding that advanced AMD is not associated with choroidal thinning when compared with fellow eyes with intermediate AMD.