Variation of Retinal and Choroidal Vasculatures in Patients With Age-Related Macular Degeneration

Abstract

Purpose: To investigate variations in chorioretinal vasculatures in fellow eyes of patients with unilateral neovascular age-related macular degeneration (nAMD).

Methods: We included fellow eyes of consecutive patients with unilateral nAMD from swept source optical coherence tomography (SS-OCT) angiography database. Vascular and nonvascular indices were determined based on SS-OCT and SS-OCT angiography images. Variation of the vascular or nonvascular index was compared between fellow eyes with and without early AMD.

Results: In 146 fellow eyes, 88 (60.3%) had early AMD and 58 (39.7%) had a normal fundus. Vascular density (VD) values of the superficial and deep retinal capillary plexus and choriocapillaris were smaller in eyes with early AMD than in those without (P < 0.001, P = 0.001, P = 0.006, respectively). Flow void area in choriocapillaris was greater in eyes with early AMD than those without (P = 0.015). In 88 fellow eyes with early AMD, vascular indices of the retina were correlated with those of choroid while nonvascular indices were not. Fellow eyes of patients with classic exudative AMD had greater foveal avascular zone area and smaller VD values of the deep retinal capillary plexus than those with polypoidal choroidal vasculopathy (P < 0.001, P = 0.004).

Conclusions: In addition to vascular insufficiency in the choroid or choriocapillaris, retinal vascular alteration was apparent in eyes with early AMD. It may suggest that retinal vessels were involved in the pathogenesis of AMD even while changes in nonvascular components of the retina were not yet apparent.

Age-related macular degeneration (AMD) is the leading cause of blindness in patients aged 65 and over.¹ Clinical and histopathologic changes can occur in the aging macula, which range from normal aging to the development of more significant drusen and retinal pigment epithelium (RPE) changes that can impair visual function and are considered risk factors for the development of advanced AMD.² Histologic and pathologic changes in the aging eye with AMD have been vigorously studied to identify the pathophysiologic mechanism of AMD. Most of them focused on the outer retinal layer and RPE, Bruch's membrane, and the choriocapillaris.³ 

The macula is composed of two compartments: the retina and the choroid. While the two compartments have different developmental origins,⁴ the outer part of the retinal compartment depends on nutritional support from the choroid.⁵ Choroidal blood flow has been suggested to be involved in the pathogenesis of AMD. However, it is not easy to compare vascular components quantitatively between the retina and the choroid using fluorescein angiography (FA) or indocyanine green angiography (ICG). After administration of optical coherence tomography (OCT) and enhanced depth OCT, choroidal thickness (CT) was presented as a risk factor of early and late AMD.^6–8 In recent studies,^9,10 variation of retinal layers has been suggested. While the choroid comprises mostly vascular and stromal tissues, the retina is composed of several components including vascular and neuronal structural components.¹¹ However, it is not certain how these components in the two compartments or developmental origin could vary during aging and degeneration of macula. 

Recent advances in OCT angiography technologies make it possible to delineate vascular and nonvascular components in the retina and choriocapillaris layers.^12,13 OCT angiography also provides depth-resolved information and detailed images of choroidal neovascularization (CNV) in neovascular AMD (nAMD).¹⁴ In addition, choriocapillaris flow has been determined using OCT angiography.^15,16 OCT angiography can provide information about both retinal and choroidal vessels through simultaneously obtained images. Quantitative data about retinal or choroidal blood flow using OCT angiography can help us to understand the variation of vascular components in chorioretinal compartments. 

In the current study, we hypothesized that the vascular components of both retina and choroid may be involved in the pathogenesis of AMD. We investigated the variation of vascular components of the chorioretina in fellow eyes of patients with unilateral nAMD. 

Methods

The Institutional Review Board of Korea University approved this study, and all research and data collection were conducted in accordance with the tenets of the Declaration of Helsinki. 

Subjects

In this retrospective cross-sectional study, we included consecutive patients with unilateral nAMD from the swept source OCT (SS-OCT) database between March 2016 and July 2017, at Korea University Medical Center. We excluded patients who underwent laser photocoagulation or an intraocular surgery for their fellow eye. We also excluded patients who had epiretinal membrane, proliferative diabetic retinopathy, vitreomacular traction, macular hole, or geographic atrophy or CNV in their fellow eyes. All patients underwent FA and ICG (Spectralis HRA; Heidelberg Engineering, Heidelberg, Germany) and SS-OCT. OCT and OCT angiography images with poor quality (image quality score in arbitrary units less than 50) were excluded. Sound eye was defined as an eye with nAMD. Classification of neovascular lesions in sound eyes was performed independently by two retina specialists (B.L. and J.A.) who evaluated the presenting color photographs, OCT, FA, and ICG. The cases were classified into group 1 for polypoidal choroidal vasculopathy (PCV), group 2 for classic exudative AMD, and group 3 for retinal angiomatous proliferation (RAP). A third supervising grader (J.O.) evaluated the lesion type in the presence of significant discrepancies. Early AMD in fellow eyes was defined as the age-related eye disease study (AREDS) category 2 or 3.¹⁷ For group comparisons, normal eyes without AMD were included. We defined this normal control group as group 0. 

Measurement on SS-OCT

The SS-OCT instrument (DRI OCT Triton, software version 10.10; Topcon Corp., Tokyo, Japan) used in this study had a central wavelength of 1050 nm, a speed of 100,000 A-scans/second, a horizontal resolution of 20 μm, and an axial resolution of 7 μm. A 9-mm horizontal line scan centered on the fovea was obtained and averaged from 96 B-scans to improve the signal-to-noise ratio. 

Central macular thickness (CMT) was measured at the central 1 mm area of the early treatment of diabetic retinopathy study (ETDRS) chart.¹⁸ Ganglion cell layer to inner plexiform layer (GC-IPL) thickness was also measured at the central 1 mm area of the ETDRS. Outer layer thickness of the macula (OLT) was defined as retinal thickness outside the inner plexiform layer. It was calculated as CMT minus GC-IPL thickness in each patient. Subfoveal CT was measured from a line scan image centered at the fovea. Subfoveal CT was measured manually with a caliper tool built in an image viewer program of OCT. The CT was defined as the perpendicular distance from the inner surface of the RPE to the choroidoscleral interface (Fig. 1). The choroidal vascularity index (CVI), defined as the proportion of vascular area to total choroidal area, was measured on the SS-OCT images as previously described, with some modifications.^19–21 In brief, after uploading the images on ImageJ software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health), they were converted to 8-bit images to allow application of the auto-threshold. A 3000-μm linear line centered at the fovea was drawn. The choroidal block beneath the line was segmented, and total subfoveal choroidal area was computed. An auto local threshold was applied using the Niblack auto local threshold to demarcate the choroidal vascular area and stromal area. The image adjusted by auto local threshold was converted back to an RGB (red, green, blue) image, and the vascular area and stromal area were determined using the color threshold tool (Fig. 2). Two independent observers (B.L. and J.A.) masked to patient information measured parameters in SS-OCT images such as CT. The mean value of the two measurements was used for the final analysis. 

Figure 1

 

Analysis of OCT results. Measurement of subfoveal CT. The arrowhead indicates the choroidoscleral junction. CT was defined as the perpendicular length between the outer surface of the hyperreflective line of the RPE and the inner margin of the choroidoscleral junction (red double-headed arrow).

Figure 2

 

Representative image processing to obtain CVI and choroidal vascular area. (A) Original SS-OCT image. (B) A 3.0-mm segmentation block of the subfoveal choroidal area using the polygon selection tool. (C) Segmented OCT image using a modified image binarization approach. (D) The vascular area was highlighted by applying the color threshold. The ratio of the vascular area to the total choroidal area was termed the CVI.

Measurement on SS-OCT Angiography (SS-OCTA)

SS-OCTA images were obtained when collecting SS-OCT images. Scans were performed in a 3 × 3 mm area centered on the macula, with each cube consisting of 320 A-scans of four repeated B-scans centered on the fovea. En face SS-OCTA images of the choriocapillaris extending from Bruch's membrane to 10.4 μm inside Bruch's membrane were based on automated layer segmentation performed by the SS-OCT instrument software. The software provided a density map of the SS-OCTA image (Fig. 3). The density map was divided into nine levels depending on the density of blood vessels. 

Figure 3

 

Density maps of SS-OCTA with adjoining scale. The VD of a specific area was calculated using the area of the color and the VD corresponding to the color. The area per color was calculated by counting the number of pixels using ImageJ software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health). The mean VD of the superficial (left column) and DCP (middle column) of the retina and the choriocapillaris (right column) was measured within a 2.5-mm-diameter central circle. Red color means higher VD, while blue color means lower VD.

Foveal avascular zone (FAZ) area was manually calculated with superficial capillary plexus (SCP) and deep capillary plexus (DCP) images using the caliper tool built into the SS-OCT viewer program. The vascular density (VD) of a specific area was calculated using the area of the color and the VD corresponding to the color. The area per color was calculated by counting the number of pixels using ImageJ software. The VD of each color was considered a median value. The mean VD of the superficial and DCP of the retina and choriocapillaris was measured within a 2.5-mm-diameter central circle. Flow void area (FVA) in the choriocapillaris was defined as an area with a VD lower than the mean value (Fig. 4).²² 

Figure 4

 

Measurement of FVA, indicated in yellow and calculated as lower VD color pixel count / Total pixel count × 100 (%).

Statistical Analysis

Continuous variables were represented as the mean ± standard deviation, and categorical variables were expressed as count (%). Statistical analyses were performed with SPSS software version 20.0 for Windows (IBM Corp., Armonk, NY, USA). Linear correlations were analyzed with Pearson's correlation coefficient (r) for normally distributed continuous variables. Spearman's correlations coefficients (r_s) were used for FVA measurement. The baseline characteristics of the groups were compared using χ² tests for categorical variables. Analyses of covariance (ANCOVAs) with adjustments for baseline values were used to test for differences in the measurements between groups. Results were considered statistically significant at P values < 0.05. 

Results

General Characteristics

In this study, we included 146 fellow eyes of 146 patients. The mean age was 68.6 ± 8.5 years (Table 1). Among 146 patients, 91 (62.3%) were male and 55 (37.7%) were female. Age did not differ between genders (P = 0.982). Nonvascular and vascular indexes in the retina and choroid are shown in Table 2. CMT, GC-IPL thickness, OLT, and FAZ area in SCP and DCP were not correlated with age (P = 0.670, P = 0.574, P = 0.798, P = 0.170, P = 0.109, respectively). However, mean VD in SCP and DCP was correlated with age (r = −0.292, P < 0.001; r = −0.304, P < 0.001, respectively). Mean subfoveal CT, choroidal area, and choroid stromal area were correlated with age (r = −0.309, P < 0.001; r = −0.318, P < 0.001; r = −0.266, P = 0.001, respectively). VD and FVA in the choriocapillaris and CVI were also correlated with age (r = −0.508, P < 0.001; r_s = 0.464, P < 0.001; r = −0.165, P = 0.046, respectively). 

Table 1

 

General Characteristics of Fellow Eyes in 146 Patients With Unilateral nAMD

 

Table 2

 

Nonvascular and Vascular Indexes in the Retina and the Choroid of Fellow Eyes in 146 Patients With Unilateral nAMD

 

Normal Eyes Versus Early AMD Eyes in Fellow Eyes

Among 146 fellow eyes, 88 (60.3%) had early AMD and 58 (39.7%) had a normal fundus (Table 3). Age was greater in eyes with early AMD (70.1 ± 8.7) compared to eyes with a normal fundus (66.4 ± 7.7) (P = 0.009). Gender did not differ between eyes with and without early AMD (P = 0.179). When considering age, CMT, GC-IPL thickness, and OLT did not differ between eyes with and without early AMD. However, subfoveal CT was different between groups (P = 0.046). FAZ area in SCP and DCP was not different between eyes with and without early AMD. The VD values of the SCP, DCP, and choriocapillaris in eyes with early AMD were smaller than those without AMD (P < 0.001, P = 0.001, P = 0.006, respectively). Normal eyes had a smaller FVA than eyes with early AMD (P = 0.015). 

Table 3

 

Comparison of the Vascular and Nonvascular Indexes Between Fellow Eyes With and Without Early AMD in Patients With Unilateral nAMD in Their Sound Eyes

 

In 88 fellow eyes with early AMD, vascular indexes were correlated with nonvascular indexes within the retina or the choroid (Table 4). FAZ area in SCP and DCP was correlated with CMT (P < 0.001, P < 0.001), GC-IPL thickness (P < 0.001, P < 0.001), and OLT (P = 0.016, P = 0.008). VD in SCP was correlated with CMT (P < 0.001), GC-IPL thickness (P = 0.038), and OLT (P < 0.001). VD in DCP was also correlated with CMT (P = 0.001) and OLT (P < 0.001). However, VD in DCP was not correlated with GC-IPL thickness (P = 0.545). Choroidal vascular indexes were also correlated with choroidal nonvascular indexes (Table 4). VD of the choriocapillaris was correlated with subfoveal CT, choroidal area, and choroid stromal area (all P < 0.001). FVA was also correlated with subfoveal CT, choroidal area, and choroidal stromal area (all P < 0.001). Choroidal vascular area was correlated with subfoveal CT, choroidal area, and choroidal stromal area (all P < 0.001). CVI was not correlated with subfoveal CT, choroidal area, or choroidal stromal area. Nonvascular indexes were not correlated between retina and choroid (Supplementary Table S1). Retinal parameters of CMT, GC-IPL, and OLT were not correlated with choroidal parameters of subfoveal CT, choroidal area, and choroidal stromal area. However, vascular indexes were correlated between retina and choroid (Figs. 5, 6). Retinal vascular indexes were correlated with choroidal vascular indexes. FAZ area and VD in SCP was correlated with VD of the choriocapillaris (P = 0.045, P < 0.001) and DCP (P = 0.017, P < 0.001). The retinal vascular index, except FAZ area in SCP, was also correlated with FVA of the choriocapillaris. VD in SCP and DCP was correlated with choroidal vascular area (P < 0.001, P < 0.001). However, those patterns of correlation were less apparent in the 58 normal fellow eyes. Vascular or nonvascular indexes were not correlated between retina and choroid (Supplementary Table S2). 

Table 4

 

Correlation Between Vascular and Nonvascular Indexes in Chorioretina in 88 Eyes With Early AMD

 

Figure 5

 

Representative image of density maps from high VD (left) to low VD (right). Numbers represent VD.

Figure 6

 

Scatter plots of vascular indexes between the retina and the choroid in 88 fellow eyes with early AMD, adjusted for age.

Classic nAMD Versus PCV Versus RAP in Sound Eyes

In 85 normal eyes without AMD, the mean age was 63.8 ± 8.3 years; 41 (48.2%) were male and 44 (51.8%) were female. All parameters except FAZ area of SCP were significantly different between the normal subject group and the wet AMD in sound eyes group (all P < 0.05). Among 146 sound eyes, 45 (30.8%) were classified as group 1, 94 (64.4%) were classified as group 2, and 7 (5.8%) were classified as group 3. Age and gender did not differ between these groups (P = 0.094, P = 0.877, respectively). CMT, GC-IPL thickness, and OLT also did not differ between groups (P = 0.363, P = 0.149, P = 0.260). Group 3 had a smaller CT (143.9 ± 61.6) than group 1 (227.5 ± 83.1) and group 2 (226.9 ± 87.4) (P = 0.044, P = 0.037). The FAZ area of SCP did not differ among groups (P = 0.383) (Fig. 7). However, group 1 had a smaller DCP FAZ area (0.786 ± 0.272 mm²) than group 2 (0.969 ± 0.354 mm²) (P < 0.001). The VD of SCP was not different among groups (P = 0.182). However, the VD of DCP was significantly different among groups. The mean VD of DCP in group 1 (20.79 ± 5.36%) was greater than those of group 2 (18.03 ± 4.98%) and group 3 (14.00 ± 6.63%) (P = 0.004, P = 0.002). FVA was significantly different among groups. Group 1 had a significantly smaller choriocapillaris FVA than group 2 or group 3 (P = 0.001, P = 0.015). The mean VD of the choriocapillaris was different among groups with borderline significance (P = 0.089). Choroidal vascular area in group 3 (0.541 ± 0.277 mm²) was different from group 1 (0.857 ± 0.282 mm²) (P = 0.050). CVI did not differ among groups (P = 0.597). 

Figure 7

 

Comparison of the vascular index of normal control eyes (group 0, G0) and fellow eyes among different types of CNV in sound eyes of patients with unilateral nAMD. CNV was classified into three groups: group 1 (G1) for PCV, group 2 (G2) for classic exudative AMD, and group 3 (G3) for RAP. Vascular indexes of G0 were significantly different from those of G1, G2, and G3 (all P value < 0.01) except in FAZ in SCP.

In 45 patients in group 1, 94 in group 2, and 7 in group 3 fellow eyes, vascular indexes were correlated between retina and choroid (Supplementary Tables S3–S5). However, nonvascular indexes were not correlated between retina and choroid in either group. 

Interobserver Reproducibility

We assessed interobserver reproducibility of FAZ area and CT measurements with the intraclass correlation coefficient, which ranged from 0.806 to 0.992 with good agreement (Supplementary Table S6). 

Discussion

In this study of fellow eyes of patients with unilateral nAMD, we compared retinal structural thickness between eyes with and without early AMD. CMT, GC-IPL thickness, and OLT were smaller in eyes with AMD, but this was not statistically significant. Previous studies showed that macular GC-IPL thickness was significantly lower in eyes with dry AMD than in control eyes.^9,10,23 Our study does include different populations compared to previous studies. In our study, characteristics of choroidal vessels were significantly different between eyes with and without early AMD. Choriocapillaris vessels had deteriorated more in eyes with early AMD, which was consistent with previous observations.^3,24 In a previous histopathologic study, Seddon et al.²⁵ showed submacular choriocapillaris vessel dropout without RPE loss in all cases with early stage AMD. They demonstrated that choriocapillaris loss can occur in the absence of RPE atrophy in some eyes with early AMD. In addition, there were fewer retinal vessels in eyes with early AMD. In previous population-based studies, retinal arteriolar changes were inconsistently related to the incidence of AMD.^26–29 In our study with OCT angiography, retinal vascular changes were apparent in eyes with AMD and were related to retinal neuronal structural changes. Our findings support the hypothesis that retinal change can be accompanied by progression of AMD.^9,10,23 It could also suggest a role of retinal vessels in the progression of AMD. 

We investigated the vascular components of the retina and choroid and found that variation of vascular components in the retina was correlated with variation in the choroid. The vascular component of the retina, such as the VD in SCP or DCP, was correlated with the vascular component of the choroid such as the VD of the choriocapillaris, choroidal vascular area, or CVI. The correlation of the vascular component between the retina and the choroid was more prominent in eyes with early AMD than in those without. The presence of the correlation between retina and choroid could be explained in several ways. First, it might be a consequence of AMD. Borrelli et al.⁹ suggested the concept of postreceptor retinal neuronal loss as a contributor to retinal thinning in intermediate AMD. They showed a correlation between ganglion cell complex thinning and ellipsoid zone reflectivity to support their hypothesis. The results of our study suggest that postreceptor retinal neuronal loss could have induced retinal vascular change in AMD. Second, eyes of patients who had a vascular covariation pattern between the retina and the choroid could have been more vulnerable to early AMD.⁹ The choriocapillaris has been suggested as a key vascular component that maintains the integrity of the RPE and photoreceptors in the aging eye. This hypothesis depends on the observation of differences in oxygen saturation through the chorioretinal compartment.^30,31 However, photoreceptors receive oxygen from both the choroidal and retinal circulation. The choroidal circulation provides approximately 90% of the oxygen used in the dark,³² an estimate that is derived from the relative fluxes across the outer segment and outer nuclear layer.³³ Moreover, photoreceptors belong to cone or rod cells in the outer nuclear layer, which is a part of a columnar unit around a Müller cell.³⁴ The nutritional and oxygen supplies of this unit depend on retinal vascular systems. Therefore, the adult neural retina is supported by two distinct vascular systems, the retinal vessels and the choroidal vessels.⁴ Therefore, vascular variation between the retina and the choroid is crucial in maintaining homeostasis of the neural retina. Previous studies demonstrated that reduced blood flow in the choroid and retina causes chronic ischemia in Bruch's membrane, RPE, and the neuroretina.^35–37 Vascular deficits due to both reduced choroidal and retinal blood flow have been identified in early and late AMD using FA and Doppler imaging.^37–39 This might suggest that, in addition to impairment of choriocapillaris flow, decreased flow in retinal vessels might have contributed to the progression of AMD in our patients. However, we were not able to confirm whether the vascular covariation was only a consequence of AMD or whether it was also a risk factor of progression of AMD. Further studies are required to reveal a role of retinal vascular deficit in AMD eyes with decreased choroidal supply. 

In this study, we classified neovascular lesions into three groups: PCV, classic exudative AMD, and RAP. Characteristics of neovascular vessels have been suggested to be different among the three groups. Neovascular lesion in PCV is characterized as a branching vascular network and polypoidal lesion. In RAP, neovascular lesions originate in the retina, proliferate, and produce chorioretinal anastomosis. While these neovascular lesions in different groups induced similar clinical manifestations such as hemorrhage and exudation, both eyes of subjects usually had neovascular lesions in same group if they had neovascular lesions in both eyes. In a previous study of patients with unilateral nAMD, it was suggested that certain nonneovascular features of the fellow eye correlate with the neovascular lesion composition based on type, as anatomically classified utilizing both FA and OCT.⁴⁰ In our study, retinal and choroidal vascular characteristics in fellow eyes differed based on classifications of CNV in sound eyes. While previous studies showed that PCV or RAP had different choroidal characteristics from classic nAMD, based on the CT or medium and large size of choroidal vessels, we showed that characteristics in specific layers of the retina and choroid could be different among patients with different classifications of nAMD using SS-OCTA. In this study, FVA was significantly different among groups. In particular, group 1 had a significantly smaller choriocapillaris FVA than group 2 or 3. This could mean that flow in the choriocapillaris in patients with classic exudative AMD or RAP is more deteriorated than in patients with PCV. In addition, we found that the mean VD of DCP was decreased more in patients with classic exudative AMD or RAP than in patients with PCV. This suggests that retinal vascular changes differ among patients with different types of CNV. However, intercompartmental covariation between the retina and the choroid was similarly observed in both groups 1 and 2. Therefore, vascular covariation between the retina and the choroid could be a general phenomenon in eyes with AMD. 

Wei et al.⁴¹ showed that aging plays an important role in all three components (i.e., microstructure, microvasculature, and microcirculation) of the neuro-vascular-hemodynamic system in their study with normal subjects. In our study of AMD patients, the mean VD in the retina was correlated with age, which was consistent with previous studies.^41–43 However, FAZ area and retinal layer thickness were not correlated with age, which was inconsistent with previous studies.^41,44,45 This might be due to differences in age distribution between studies, or because we included eyes of patients with AMD. When considering that both the nonvascular and vascular indices in the choroidal layer were correlated with age in our study, choroidal change might continue and be prominent even in old age. 

This study has several limitations. In addition to the retrospective study design, the number of patients with RAP was small; thus, the results do not sufficiently reflect differences in patients with RAP. In this retrospective study with consecutive patients with unilateral nAMD, we were not able to include enough patients with unilateral RAP because the prevalence of RAP is relatively low and patients with RAP frequently had bilateral neovascular lesions at presentation. We expect that future studies using larger sample sizes with RAP will reveal additional and more reliable differences. We manually measured FAZ area, which might have influenced the result of this study. We also manually measured CT. Although the reproducibility of the manual measurement was high, it is possible that the manual measurement influenced the results. However, for the measurement of VD in each layer, we used a built-in density map, which could have provided more reproducibility than manual measurement. However, it is not free from the influence of image quality. Therefore, we excluded OCTA images with poor quality (image quality score in arbitrary units less than 50). 

In conclusion, in addition to vascular insufficiency in the choroid or choriocapillaris, retinal vascular alteration was apparent in eyes with early AMD. The vascular alteration was apparent while changes in nonvascular components of the chorioretina were not yet apparent. It may suggest that retinal vessels are involved in the pathogenesis of AMD. 

Acknowledgments

Supported by a grant from Korea University (K1620061). 

Disclosure: B. Lee, None; J. Ahn, None; C. Yun, None; S.-w. Kim, None; J. Oh, None