Retinal Vessel Changes in Geographic Atrophy in AMD: Insights From Imaging and Histology

Abstract

Purpose

The purpose of this study was to investigate retinal vascular changes in geographic atrophy (GA) secondary to age-related macular degeneration (AMD) using swept-source optical coherence tomography angiography (SS-OCTA), and to correlate imaging findings with histology.

Methods

Sixty subjects were enrolled: 20 with GA, 20 with intermediate AMD, and 20 healthy controls. SS-OCTA imaging was used to quantify retinal perfusion density (PD) and vessel length density (VLD) in the superficial capillary plexus (SCP), deep capillary plexus (DCP), and full retina. A topographical analysis distinguished regions with and without retinal pigment epithelium (RPE) atrophy in GA eyes. Additionally, flat-mount immunohistochemistry was performed on a donor eye with GA to assess retinal vasculature. Main outcome measures included PD and VLD across SCP, DCP, and full retina, in regions with and without RPE atrophy.

Results

Retinal PD and VLD were significantly reduced in GA eyes compared with intermediate AMD eyes, particularly in the DCP. Topographical analysis revealed more pronounced vascular impairment in areas with RPE atrophy, whereas regions without RPE atrophy in GA eyes exhibited perfusion comparable to intermediate AMD and healthy controls. Histological analysis confirmed a substantial reduction in vascular density within atrophic regions.

Conclusions

Retinal vascular changes in GA predominantly occur within regions of RPE atrophy. The preservation of perfusion in regions without RPE atrophy suggests that vascular impairment is localized. These findings underscore the importance of regional analysis and histopathologic correlation in understanding vascular remodeling in GA. Future longitudinal OCTA studies are warranted to clarify the temporal progression of these vascular alterations in relation to RPE atrophy.

Age-related macular degeneration (AMD) is a leading cause of vision loss in the elderly, with geographic atrophy (GA) representing a severe late-stage manifestation of dry AMD.^1–3 Although the exact mechanisms underlying GA remain incompletely understood, research suggests that its development and progression are influenced by a combination of factors, including oxidative stress, environmental influences, and genetic predisposition.^4,5 These factors ultimately contribute to damage within the functional unit composed of photoreceptors, Bruch's membrane, the retinal pigment epithelium (RPE), and the choriocapillaris (CC).

Changes in the choroidal vasculature serve as an important biomarker of AMD and are closely linked to the development and progression of GA.^6–9 As a result, advances in imaging technologies have significantly enhanced our understanding of choroidal changes in both wet and dry AMD, particularly by improving our ability to visualize and characterize the choroid.¹⁰ Specifically, optical coherence tomography angiography (OCTA) has allowed detailed visualization of the choriocapillaris (CC), enabling relative quantification of perfusion in this layer.¹⁰ Studies using OCTA in eyes with early and intermediate AMD have demonstrated that the CC is affected in these stages,^11,12 which aligns with findings from earlier histopathological studies.^13–15 Notably, OCTA and histopathologic studies have consistently demonstrated that the CC is markedly compromised in eyes with GA. In particular, regions of complete RPE atrophy exhibit profound CC flow impairment, including capillary dropout and rarefaction, whereas areas adjacent to the atrophic border often show patchy flow deficits and reduced perfusion signal.^16–18 The latter aspect of CC being more severely affected within the GA region, with a lesser impact observed at the GA border, is also supported by histopathological evidence.^7–9 Furthermore, reduced CC perfusion at the GA border has been associated with faster longitudinal growth of GA, suggesting that CC dysfunction may play a direct role in disease evolution.^16–18

Although the choroid is widely recognized as the primary site of vascular involvement in AMD, growing evidence suggests that the retinal vasculature may also be secondarily affected in these patients. Previous studies using spectral domain OCTA (SD-OCTA) have shown reduced retinal perfusion in patients with intermediate AMD.^19,20 These changes were primarily observed in the superficial capillary plexus (SCP), whereas they were less pronounced in the deep capillary plexus (DCP).¹⁹ Moreover, retinal perfusion impairment was more prominent in intermediate AMD eyes with OCT biomarkers associated with an increased risk of GA development, such as nascent GA.²⁰ Subsequent OCTA studies have further confirmed these findings, consistently showing that retinal perfusion is impaired in intermediate AMD, with this impairment being more pronounced at the SCP level.^21–26 Using SD-OCTA, retinal perfusion has also been studied in eyes with GA and found to be reduced in both regions with and without RPE atrophy when compared with healthy controls without evidence of AMD.²⁷ However, the latter study did not compare regions without RPE atrophy to age-matched eyes with intermediate AMD, making it unclear whether the perfusion changes in these regions are more pronounced with increased disease severity. More importantly, the study did not provide a correlation with histological findings, which may limit the understanding of these changes. Compared to SD-OCTA, swept source OCTA (SS-OCTA) offers superior imaging of deeper vascular layers due to its longer wavelength and faster scanning speed, which help reduce signal attenuation and shadowing artifacts. This advantage is especially relevant in eyes with macular atrophy, where accurate visualization of the DCP and sub-RPE structures is critical.²⁸

Despite increasing interest in retinal vascular changes in AMD, few studies have addressed these alterations in GA specifically, and none have directly correlated OCTA findings with histologic validation. Moreover, most previous work used SD-OCTA, which is limited by shadowing artifacts and depth resolution. The present study addresses these gaps by combining high-resolution SS-OCTA with flat-mount histopathologic analysis in GA, enabling topographical validation of in vivo findings.

Using SS-OCTA technology, the primary aim of this study was to investigate regional quantitative differences in retinal perfusion—specifically in areas with and without RPE atrophy—in patients with GA, by comparing them to age-matched patients with intermediate AMD. Additionally, we analyzed and compared the OCTA findings with histopathologic assessments of retinal vessels in a donor eye with GA.

Methods

Study Participants in the In Vivo Analysis

This prospective cross-sectional study was approved by the University of Turin Ethics Committee (Protocol Number 0001159) and conducted in accordance with the principles of the 1964 Declaration of Helsinki and its subsequent revisions. Informed consent was obtained from all participants prior to their inclusion in the study.

Consecutive patients with GA secondary to AMD were enrolled at the Medical Retina Unit of the City of Health and Science Hospital – University of Turin. The diagnosis of GA was confirmed through a comprehensive assessment, including fundus ophthalmoscopy, fundus autofluorescence (FAF), structural OCT, and OCTA. Specifically, the diagnosis of GA was made by the treating physician (author E.B.) and was based on the presence of a hypopigmented area with visible choroidal vessels on fundus ophthalmoscopy, corresponding to a hypoautofluorescent region on FAF, and associated RPE atrophy visualized on structural OCT, as previously described.²⁹ The following exclusion criteria were applied to the study eye: (i) history or evidence of macular neovascularization (MNV), including non-exudative cases; (ii) prior complex cataract or vitreoretinal surgeries, including anti-vascular endothelial growth factor (VEGF) injections; and (iii) history or evidence of other retinal or optic nerve disorders.

A group of age- and gender-matched subjects with intermediate AMD in both eyes was included for comparison. Intermediate AMD was defined by the presence of large drusen (≥125 µm) and/or pigmentary abnormalities, without signs of nascent GA or incomplete RPE and outer retinal atrophy (iRORA), based on multimodal imaging. These criteria are aligned with the classification proposed by Ferris et al.³⁰

Additionally, a group of age- and gender-matched healthy subjects, with no signs of retinal or optic nerve disorders and no history of diabetes, was also included. In these two control groups, a single eye was selected for analysis, with the selection made randomly.

Regarding patients with intermediate AMD, the same exclusion criteria applied to patients with GA who were used for this group. Additionally, these patients did not show evidence of nascent GA or iRORA, as these OCT findings could prelude the development of GA and potentially confound the analysis.^31,32

All participants, including both healthy controls and individuals with AMD, underwent a comprehensive ophthalmologic examination comprising visual acuity testing and advanced retinal imaging. The SS-OCT and OCT angiography were performed using the PLEX Elite 9000 platform (Carl Zeiss Meditec Inc., Dublin, CA, USA), a system that utilizes a 1050-nm wavelength light source and acquires volumetric data at a rate of 100,000 A-scans per second. The device provides high-resolution imaging, with an axial resolution of approximately 5 µm and a lateral resolution near 14 µm at the retinal surface.

Macular scans were acquired over a 3 × 3 mm area centered on the fovea, using a scan density of 300 A-scans by 300 B-scans. Images with a signal strength index below 7 (on a 0–10 scale) were excluded, in accordance with manufacturer recommendations and prior methodological standards.^33,34 The volumetric data obtained were used to reconstruct various en face images, including:

• –En face OCTA image of the SCP: this image was obtained by setting the inner boundary at the inner limiting membrane (ILM) level and the outer boundary at the inner plexiform layer (IPL) level.
• –En face OCTA image of the DCP: this image was obtained by setting the outer boundary at the IPL level and the outer boundary at the outer plexiform layer (OPL) level.
• –En face OCTA image of the full retina: this image was obtained by setting the inner boundary at the ILM level and the outer boundary at the inner margin of the RPE.
• –En face structural OCT image of the sub-RPE slab: generated by placing the segmentation boundaries at 64 µm and 400 µm beneath Bruch's membrane, as previously described.³⁵

Although the PLEX Elite 9000 device features an automated segmentation algorithm, eyes with AMD, and particularly those with GA, may exhibit segmentation artifacts (i.e. the boundaries used to generate the slab for the en face image may not accurately follow the retinal layers, as disease can affect this process). Therefore, all images were carefully reviewed, and the segmentation boundaries were manually adjusted in all B-scans and for each slab to improve accuracy. This process was carried out by a single grader (author C.O.), and the slabs were subsequently reviewed and, if necessary, further adjusted by the senior author (E.B.).

Images Processing

The obtained en face images of SCP, DCP, full retina, and sub-RPE slab were exported in .jpg format and then imported into Fiji distribution of ImageJ (software version 2.14.0; National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html) for processing and analysis.³⁶

A “complex” (or “multistep”) binarization thresholding method was applied to the SCP, DCP, and full retina OCTA images in healthy control, GA, and intermediate AMD eyes. This process combines a global threshold with a local methodology.^37–39 Specifically, each image was duplicated, and the two resulting images were processed using two different approaches: (i) one image was processed with a “Hessian” filter and then binarized using the “Huang’s fuzzy” global thresholding method; (ii) the other image was binarized using the “median local” threshold method. Finally, a combined final image was generated, including only those pixels present in both binarized images (Fig. 1).

Figure 1.

Representation of the algorithm used to investigate the retinal perfusion. Images of the full retina, superficial capillary plexus (SCP), and deep capillary plexus (DCP) were imported into ImageJ software (National Institutes of Health; http://imagej.nih.gov/ij/). A thresholding algorithm was applied to generate binarized versions of these images, which were then used to assess perfusion density in the full retina, SCP, and DCP. Perfusion density was calculated as the proportion of white pixels (i.e. those above the threshold) relative to the total number of pixels within the analyzed scan area, and is expressed as a unitless value. To evaluate vessel length density, the binarized images were further processed using skeletonization, producing representations where vessels appear as 1-pixel-wide traces. Vessel length density was defined as the total length of perfused vasculature divided by the total number of pixels in the scan area. The sub-RPE OCT image was used to delineate the area of geographic atrophy, marked by an orange dashed line in the figure. This delineation enabled a topographical analysis comparing regions with and without atrophy.

For each resulting image, perfusion density (PD) was measured as a unitless proportion of the number of pixels above the threshold divided by the total number of pixels within a region of interest (ROI) centered on the fovea, which was defined as a circle with a 2.5-mm diameter. The binarized images were also skeletonized, this process resulting in vessels with a width of one pixel, to calculate vessel length density (VLD; see Fig. 1), defined as the ratio between the total number of pixels above the threshold in the skeletonized image and the total number of pixels in the ROI. These OCTA metrics — PD and VLD — were calculated after excluding the foveal avascular zone (FAZ) from the analysis. The FAZ was manually outlined in the original full retina scans, as previously described,⁴⁰ and the remainder of the procedure was automated using an ImageJ software macro.

In patients with GA, the en face structural OCT image of the sub-RPE slab was used to delineate the borders of macular atrophy based on the presence of hypertransmission defects (hyperTDs), following criteria established in previous studies.^35,41–44 HyperTDs ware identified as well-defined bright regions visible on the en face image. The grader included the entire area displaying hyperTDs after reviewing the corresponding OCT B-scans, as previously suggested. Using an ImageJ software macro, the OCTA metrics were then measured separately in areas with RPE atrophy versus areas without RPE atrophy (i.e. within the ROI), based on the presence of hyperTDs. In both intermediate AMD and healthy control subjects, all analyses were conducted within the circular ROI after excluding the FAZ, without distinguishing based on the presence of drusen in patients with intermediate AMD.

Donor Eye and Tissue Preparation for Histopathologic Analysis

The eyes from a 93-year-old deidentified Caucasian female with GA were obtained with consent from National Disease Research Interchange (NDRI) and shipped on wet ice, opened at the limbus, and gross macroscopic images taken of the posterior eyecup to document the disease state. The cause of death was unknown and the death to preservation time was 28 hours. The retina from the right eye was dissected and fixed in 2% paraformaldehyde (PFA) in Tris-buffered saline (TBS) overnight at 4°C before being processed for flat-mount immunohistochemistry, as previously described.⁹ Briefly, the retina was blocked in 5% goat serum in TBS containing bovine serum albumin (BSA) and 1% Triton X-100. The retina was then incubated in a cocktail of chick anti-vimentin (Millipore, Burlington, MA, USA; AB5733) and rabbit anti-VEGF (ThermoFisher RB-9031-P0) for 72 hours and then in secondary antibody along with Ulexa Europaeus agglutinin 1 lectin (UEA 1; for labeling retinal blood vessels) in TBS Triton X-100 for 48 hours. The retinal vasculature was imaged on a Zeiss 710 confocal microscope. A large map of the entire posterior pole was created. Nine regions were randomly chosen from three distinct areas: three within the atrophic region, three in an area affected by drusen but without signs of RPE atrophy, and three in a zone free from both atrophy and drusen (as shown in Fig. 2). The percentage of the vascular area was measured using Fiji software as described previously.⁹

Figure 2.

Gross photograph and retinal flat-mount image. (A) A gross photograph of the posterior eyecup from a donor with GA, with the retina intact, shows a prominent area of RPE atrophy, appearing whitish compared to the surrounding tissue. (B) A low-magnification image of the retinal flat-mount stained with Ulex europaeus agglutinin 1 (UEA 1) reveals a substantial loss of retinal capillaries within the atrophic region. The green and red staining behind the vasculature defines the subretinal glial membrane (stained with vimentin and VEGF) which mimics the atrophic area in this and other eyes with GA. Nine distinct regions were randomly selected: three within the atrophic area (red boxes), three in a drusen-affected region without signs of atrophy (yellow boxes), and three in an area free of both atrophy and drusen (green boxes). The foveal avascular zone (FAZ) is enlarged compared with the healthy controls.

Statistical Analysis

All quantitative variables were reported as mean, median, standard deviation (SD), and interquartile range (IQR). The normality of distribution was assessed using the Shapiro-Wilk test. OCTA metrics were not normally distributed, whereas the other variables (i.e. age and visual acuity) followed a normal distribution. For within-group comparisons, paired-samples T test was used for parametric data, whereas the Wilcoxon rank test was applied for non-parametric data. For between-group comparisons, Student's T-test was used for parametric data, and the Mann-Whitney U test was applied for non-parametric data. Statistical analysis was conducted using Jamovi software (version 2.4.12.0), with the significance threshold set at 0.05. The sample size of the study was tested to be proper for a mean difference between groups of almost 10%, a power of 80% and type I error rate (α) of 5%.

Results

A total of 60 patients (60 eyes) were included in the analysis, with 20 subjects affected by GA, 20 by intermediate AMD, and 20 healthy controls. Mean ± SD age was 79.7 ± 7.3 years for patients with GA, 75.4 ± 6.9 years for patients with intermediate AMD, and 73.7 ± 6.6 years for healthy controls (P > 0.05). Eleven, 8, and 10 subjects were men in the GA, intermediate AMD, and healthy groups, respectively. Mean ± SD best-corrected visual acuity (BCVA) was significantly worse in patients with GA compared to those with intermediate AMD (0.5 ± 0.3 LogMAR versus 0.1 ± 0.1 LogMAR; P < 0.001). BCVA was 0.0 ± 0.1 LogMAR in healthy controls.

Reticular pseudodrusen (RPD; also known as subretinal drusenoid deposits) were graded as present in 17 of 20 eyes with GA and 6 of 20 eyes with intermediate AMD. The GA size was 3.43 ± 1.69 mm². Among the 20 eyes with GA, 8 had foveal involvement.

In Vivo Analysis of the Retinal Perfusion: Patients With GA Versus Patients With Intermediate AMD

The OCTA metrics results for the three groups are presented in Table 1.

Table 1.

OCTA Metrics in the Three Groups of Patients

	Groups
Metrics	Healthy Controls	Intermediate AMD Patients	GA Patients	P Value Healthy Controls Vs. Intermediate AMD Patients	P Value Healthy Controls Vs. GA Patients	P Value Intermediate AMD Patients Vs. GA Patients
Perfusion density – SCP
Mean (SD)	32.2 (1.8)	32.3 (1.8)	30.5 (2.8)	0.902	0.045	0.044
Median (IQR)	32.1 (2.3)	32.7 (1.7)	31.3 (3.2)
Vessel length density – SCP
Mean (SD)	6.0 (0.5)	6.0 (0.4)	5.7 (0.6)	0.980	0.225	0.184
Median (IQR)	6.0 (0.5)	6.0 (0.3)	5.8 (0.6)
Perfusion density – DCP
Mean (SD)	32.1 (2.6)	32.4 (2.1)	27.9 (3.9)	0.713	<0.001	<0.001
Median (IQR)	32.3 (1.6)	33.7 (3.6)	28.9 (4.1)
Vessel length density – DCP
Mean (SD)	6.0 (0.6)	6.3 (0.5)	5.2 (0.7)	0.297	0.001	<0.001
Median (IQR)	6.1 (0.6)	6.5 (0.8)	5.4 (1.0)
Perfusion density – full retina
Mean (SD)	35.7 (1.1)	35.4 (1.5)	32.4 (3.0)	0.778	<0.001	0.001
Median (IQR)	35.7 (1.5)	35.0 (2.2)	32.7 (3.1)
Vessel length density – full retina
Mean (SD)	7.1 (0.5)	7.1 (0.3)	6.3 (0.7)	0.986	<0.001	<0.001
Median (IQR)	7.2 (0.6)	7.1 (0.4)	6.5 (0.7)

AMD, age-related macular degeneration; DCP, deep capillary plexus; GA, geographic atrophy; IQR, interquartile range; OCTA, optical coherence tomography angiography; RPE, retinal pigment epithelium; SCP, superficial capillary plexus; SD, standard deviation.

Considering the SCP OCTA images, the PD was 30.5 ± 2.8% and 32.3 ± 1.8% in the GA and intermediate AMD groups, respectively (P = 0.044; Figs. 3, 4). No differences were detected in SCP PD between patients with intermediate AMD and healthy controls (32.2 ± 1.8%; P = 0.045). Similarly, VLD was lower in patients with GA (5.7 ± 0.6%) compared with those with intermediate AMD (6.0 ± 0.4%; P = 0.184), although the difference was not statistically significant.

Figure 3.

Multimodal imaging of a patient with geographic atrophy. (Top row) Fundus photography (left) reveals macular pigmentation changes, including a hypopigmented area where choroidal vessels are more visible due to geographic atrophy, as highlighted in the magnified image (middle). This region appears hypoautofluorescent on green autofluorescence imaging (right). (Second row) The en face image of the superficial capillary plexus (SCP) (left) shows reduced perfusion, particularly in the parafoveal region. The segmentation boundaries used for this analysis are shown on the right. (Third row) The en face image of the deep capillary plexus (DCP) (left) demonstrates a more pronounced reduction in perfusion, especially in the parafoveal region, compared to the SCP. The segmentation boundaries are shown on the right. (Bottom row) To assess retinal perfusion topographically, the en face sub-RPE image (left) highlights areas of hyperTDs corresponding to regions of RPE atrophy. This image was generated using segmentation boundaries positioned at 64 and 400 µm below Bruch’s membrane, as illustrated on the right. All segmentation boundaries were checked and manually adjusted, if necessary, to remove artifacts.

Figure 4.

Multimodal imaging of a patient with intermediate AMD. (Top row) Fundus photography (left) reveals macular pigmentary changes and the presence of large drusen, as highlighted in the magnified view (right). (Middle row) The en face image of the superficial capillary plexus (SCP) (left) shows preserved perfusion. The segmentation boundaries are shown on the right. (Bottom row) The en face image of the deep capillary plexus (DCP) (left) also demonstrates preserved perfusion. The segmentation boundaries are displayed on the right.

Considering the DCP OCTA images, the perfusion density was 27.9 ± 3.9% in patients with GA and 32.4 ± 2.1% in subjects with intermediate AMD (P < 0.001; see Figs. 3, 4). The VLD was 5.2 ± 0.7% and 6.3 ± 0.5% in patients with GA and patients with intermediate AMD, respectively (P < 0.001). Both DCP PD (32.1 ± 2.6%; P = 0.713) and VLD (6.9 ± 0.6%; P = 0.297) in healthy controls did not differ significantly from those in patients with intermediate AMD.

Considering the full retina OCTA images, the perfusion density was 32.4 ± 3.0% in patients with GA and 35.4 ± 1.5% in patients with intermediate AMD (P < 0.001). Accordingly, the VLD was reduced in patients with GA (6.3 ± 0.7%), as compared with subjects with intermediate AMD (7.1 ± 0.3%; P < 0.001). Even for the full retina OCTA images, no differences were detected between healthy controls and the patients with intermediate AMD.

In Vivo Topographical Analysis of the Retinal Perfusion in Patients With GA

Table 2 summarizes the topographical analysis of OCTA metrics in patients with GA. As mentioned above, in these patients, the analyzed ROI was divided based on the presence of RPE atrophy into a region with atrophy and one without (i.e. the regions without RPE atrophy).

Table 2.

OCTA Metrics in Patients With GA

	Region With RPE Atrophy (i.e. Showing HyperTDs)	Region Without RPE Atrophy	P Value
Perfusion density – SCP
Mean (SD)	29.3 (4.1)	32.3 (3.1)	0.007
Median (IQR)	29.6 (3.5)	32.7 (3.5)
Vessel length density – SCP
Mean (SD)	5.5 (0.8)	5.9 (0.6)	0.026
Median (IQR)	5.8 (0.7)	5.9 (0.4)
Perfusion density – DCP
Mean (SD)	23.8 (5.0)	32.8 (4.1)	<0.001
Median (IQR)	24.7 (4.4)	33.4 (5.2)
Vessel length density – DCP
Mean (SD)	4.5 (0.9)	6.1 (0.7)	<0.001
Median (IQR)	4.7 (0.8)	6.2 (0.9)
Perfusion density – full retina
Mean (SD)	29.7 (4.5)	35.5 (2.8)	<0.001
Median (IQR)	30.6 (4.7)	35.6 (2.8)
Vessel length density – full retina
Mean (SD)	5.7 (0.9)	6.9 (0.6)	<0.001
Median (IQR)	6.1 (0.7)	6.9 (0.7)

HyperTDs, hypertransmission defects.

Considering the SCP OCTA images, the perfusion density was 29.3 ± 4.1% and 32.3 ± 3.1% in regions with and without RPE atrophy, respectively (P = 0.007). Similarly, the VLD was reduced in regions with RPE atrophy (5.5 ± 0.8%) compared with regions without RPE atrophy (5.9 ± 0.6%; P = 0.026).

Considering the DCP OCTA images, the perfusion density was 23.8 ± 5.0% in regions with RPE atrophy and 32.8 ± 4.1% in areas without RPE atrophy (P < 0.001). The VLD was 4.5 ± 0.9% and 6.1 ± 0.7% in regions with and without atrophy, respectively (P < 0.001).

Considering the full retina OCTA images, the PD was significantly lower in regions with RPE atrophy (29.7 ± 4.5%) compared with regions without RPE atrophy (35.5 ± 2.8%) within the same patient population (P < 0.001). Accordingly, the VLD was reduced in regions with RPE atrophy (5.7 ± 0.9%) compared with regions without RPE atrophy (6.9 ± 0.6%; P < 0.001).

In Vivo Analysis of Retinal Perfusion in the Three Groups, Considering Only the Regions Without RPE Atrophy

When comparing the regions without RPE atrophy of the GA eyes to patients with intermediate AMD no differences in any OCTA metrics were detected between the two groups. In other words, the regions without RPE atrophy in GA eyes did not exhibit changes in retinal perfusion compared to the earlier stage of the disease. Table 3 summarizes the OCTA values and reports the P values for each comparison.

Table 3.

Comparisons After Excluding the Region With RPE Atrophy in Patients With GA

	Groups
Metrics	Healthy Controls	Intermediate AMD Patients	GA Patients (After Excluding RPE Atrophy Area)	P Value Healthy Controls Vs. GA Patients	P Value Intermediate AMD Vs. GA Patients
Perfusion density – SCP
Mean (SD)	32.2 (1.8)	32.3 (1.8)	32.3 (3.1)	0.878	0.960
Median (IQR)	32.1 (2.3)	32.7 (1.7)	32.7 (3.5)
Vessel length density – SCP
Mean (SD)	6.0 (0.5)	6.0 (0.4)	6.0 (0.6)	0.948	0.696
Median (IQR)	6.0 (0.5)	6.0 (0.3)	5.9 (0.5)
Perfusion density – DCP
Mean (SD)	32.1 (2.6)	32.4 (2.1)	32.8 (4.1)	0.526	0.837
Median (IQR)	32.3 (1.6)	33.7 (3.6)	33.4 (5.2)
Vessel length density – DCP
Mean (SD)	6.0 (0.6)	6.3 (0.5)	6.1 (0.7)	0.902	0.611
Median (IQR)	6.1 (0.6)	6.5 (0.8)	6.2 (0.9)
Perfusion density – full retina
Mean (SD)	35.7 (1.1)	35.4 (1.5)	35.5 (2.8)	0.975	0.960
Median (IQR)	35.7 (1.5)	35.0 (2.2)	35.6 (2.8)
Vessel length density – full retina
Mean (SD)	7.1 (0.5)	7.1 (0.3)	6.9 (0.6)	0.338	0.278
Median (IQR)	7.2 (0.6)	7.1 (0.4)	6.9 (0.7)

Interestingly, healthy controls did not show differences in perfusion or VLD compared to patients with GA when only the regions without RPE atrophy was considered in the latter (see Table 3).

Ex Vivo Analysis of the Retinal Vasculature in GA

The gross photograph demonstrates the atrophic area in the posterior pole where choroidal vessels are clearly visible and an area with drusen but no RPE loss (see Figs. 2, 5). In the retinal flat-mount image, stained with UEA lectin, there appear to be fewer branches from the primary vessels within the atrophic area as well as an overall reduction in vasculature compared with the unaffected areas (see Figs. 2, 5). The vessel density also appears reduced in the retina where drusen are observed in the gross photograph (see Figs. 2, 5). The percentage of the vascular area was significantly lower in the atrophic area (24.6 ± 2.4%) compared to the unaffected area (40.5 ± 5.3%; see Fig. 5; P < 0.001). The percentage of the vascular area was also significantly lower in the non-atrophic but affected area (i.e. the area with drusen but no atrophy; 28.4 ± 1.1%) compared to the unaffected area (P < 0.05). The unaffected area is similar to previously reported control percentage of the vascular areas, 33.9 ± 2.8%.⁹

Figure 5.

High-magnification UEA 1-stained retinal flat-mount images. High-magnification UEA 1-stained retinal flat-mount images of the nine analyzed regions (see Fig. 2) clearly demonstrate a reduced number of retinal capillaries in the atrophic area (D–F) compared with regions without atrophy or drusen (A–C).

Discussion

In our study, we identified significant alterations in the retinal vasculature of patients with AMD and GA. Notably, we conducted a topographical analysis to evaluate these vascular changes in regions with and without RPE atrophy. Our findings revealed that these changes were more pronounced in areas of RPE atrophy, whereas perfusion in regions without atrophy was comparable to that observed in patients with intermediate AMD and healthy controls. Additionally, we presented a histopathologic case illustrating retinal vessels in a patient with GA, which reinforces the close relationship between the retinal vasculature and the RPE. This case notably demonstrated a substantial reduction in retinal vessel density, primarily within areas of RPE atrophy. Collectively, our findings suggest that retinal vascular changes in GA are predominantly present in regions exhibiting RPE atrophy.

The introduction of OCTA has generated considerable interest in retinal perfusion in AMD, as understanding vascular changes in a disease that primarily involves the complex interaction among the photoreceptors, RPE cells, and the choroid could offer important insights into its pathogenesis. As previously mentioned, a prior study utilizing SD-OCTA investigated retinal perfusion in eyes with GA.²⁷ Specifically, this study analyzed 10 eyes with GA and 10 healthy control eyes, reporting reduced perfusion in both the SCP and DCP compared with healthy eyes. Notably, their topographical analysis revealed that the reduction in perfusion was more pronounced in areas with RPE atrophy. In the present study, we confirmed these findings using an SS-OCTA device, which is known to be less susceptible to shadowing artifacts, particularly when assessing deeper retinal vessels. Additionally, we included subjects with intermediate AMD to investigate whether regions without RPE atrophy in patients with GA might exhibit more extensive retinal vessel damage compared to the earlier stage of AMD (i.e. intermediate AMD). Our analysis revealed that retinal perfusion was similar between areas without RPE atrophy in GA and intermediate AMD, suggesting that vascular changes affecting retinal vessels in regions with RPE atrophy do not extend to the surrounding unaffected areas.

One of the most significant analyses in the present study was the histopathologic characterization of the retinal vessels in a patient with GA. The latter analysis revealed a substantial reduction in the percentage of the vascular area (vascular density), primarily within areas of RPE atrophy. This observation confirmed a previous report showing reduced retinal vascular density in three siblings with GA.⁹ In the previous study, the percentage of the vascular area was even lower, likely due to increased disease duration and/or severity in these donors. Although the different layers could not be conclusively analyzed with the images available, there does appear to be reduced branching of the superficial vessels and a reduction in the capillaries. This again confirms previous observations.

Although a direct one-to-one comparison between histologic and OCTA-based vascular metrics is challenging due to differences in methodology and measurement scale, the percent vascular area observed in the ex vivo analysis (24.6% in atrophic areas and 40.5% in unaffected regions) shows a trend comparable to the OCTA-derived perfusion densities measured in vivo. Specifically, OCTA metrics demonstrated lower perfusion within atrophic regions, with preserved values in non-atrophic areas. These consistent directional trends between imaging and histology provide cross-modal validation of our findings.

Regarding our in vivo analysis, our study also included age-matched healthy controls. An interesting observation was that retinal perfusion in intermediate AMD eyes and in the regions without RPE atrophy of GA eyes (i.e. regions not affected by RPE atrophy) was comparable to that in healthy individuals. This finding suggests that retinal vessels may maintain physiological perfusion until RPE atrophy occurs. It is important to note that in intermediate AMD eyes and the regions without RPE atrophy of GA eyes, we did not separately analyze retinal perfusion in regions with and without drusen. This decision was driven by the challenges of accurately distinguishing these areas in imaging assessments. Additionally, the presence of reticular pseudodrusen could have further complicated this distinction. Furthermore, because drusen are primarily located in the parafoveal region, comparing areas with and without drusen would have required accounting for regional differences in perfusion. Given these complexities, we determined that the most reliable approach was to conduct a topographical assessment while excluding only areas affected by RPE atrophy. However, advances in imaging analysis applied to OCTA will likely enable the evaluation of regional perfusion in areas with and without drusen, helping to determine whether differences exist between these regions. In this regard, the ex vivo analysis suggests that some changes may still occur in the region with drusen but without atrophy. The latter observation may also suggest that the retinal vessels are not simply reduced due to retinal degeneration and reduced perfusion needs.

These histological findings on retinal vessels align with previous reports on postmortem choroids from GA eyes, where the choriocapillaris and outer choroid were shown to be affected in GA, with these changes primarily localized to regions of RPE atrophy.^8,9,45,46 Taken together, the findings of this study, together with prior research, suggest that retinal vessels, like the choroidal vasculature, is predominantly impacted within GA-affected regions.

Several factors may explain the reduction in retinal vessels observed in areas with GA. First, regions with RPE atrophy undergo neurodegeneration that also affects the inner retina, potentially leading to a decreased demand for oxygen and nutrients, which in turn may result in reduced perfusion.⁴⁷ As mentioned above, however, the reduced vascular density in areas without RPE atrophy detected using ex vivo analysis suggests other causes for this change. It is also possible that photoreceptors within this area are affected as recent clinical imaging shows photoreceptor loss ahead of RPE loss in some cases.⁴⁸ Alternatively, these vascular changes may be linked to Müller cells. As essential glial cells, Müller cells support the retina and are closely associated with retinal blood vessels, playing a key role in regulating blood flow and maintaining the blood-retinal barrier.⁴⁹ In a recent study, Naik and colleagues⁵⁰ examined retinal flat-mounts from subretinal sodium iodate (NaIO₃) rat models and cryosections from human GA eyes, using antibodies against Müller cell proteins to assess their distribution and function. Their findings demonstrated that Müller cells in GA-affected regions exhibited functional changes, reduced density in perivascular areas, and structural remodeling, including migration into the subretinal space. Interestingly, reduced Müller cell perivascular expression Kir4.1 and aquaporin 4 were also altered on the non-atrophic aspect of the atrophic border.⁵⁰ Based on these findings, we may speculate that Müller cell dysfunction in GA regions could contribute to retinal vascular remodeling, ultimately leading to reduced perfusion in these areas.

Our study has certain limitations, including its cross-sectional design. A prospective longitudinal evaluation of retinal vessels in patients with GA could provide further insights into the vascular changes that occur following RPE atrophy. A limitation of this study is the histopathologic validation based on a single donor eye, which may not fully capture interindividual variability. Nonetheless, the observed vascular alterations were consistent with previous histological reports⁹ and supported our in vivo imaging findings. An additional limitation is the uneven distribution of RPD between the GA and intermediate AMD groups, with a higher prevalence in the GA group. This imbalance may have partially influenced the perfusion findings. Although the observed trends aligned with disease severity, future studies should control for the presence of RPD—either through matched cohorts or stratified analyses—to better isolate their potential impact. Finally, areas without RPE atrophy may still exhibit subtle changes (e.g. photoreceptor loss or Müller cell alterations), which were not assessed in the current analysis. However, our study also has notable strengths. We utilized SS-OCTA in the largest cohort analyzed to date, incorporating a detailed topographical assessment. Most importantly, we complemented our imaging findings with histopathologic data, further reinforcing our results.

In summary, this SS-OCTA study of macular retinal vessels revealed that vascular changes in eyes with GA are primarily confined to regions with RPE atrophy. Our histopathologic findings further support the idea that retinal vascular alterations predominantly occur in regions of RPE atrophy, suggesting a potential paracrine regulatory mechanism involving retinal cells or Müller cell dysfunction in these areas. Whereas histopathologic correlation provides insights in experimental models, regional OCTA analysis represents the only feasible and clinically applicable method in human subjects. Finally, given the cross-sectional design of the current study, we hope that our findings will inspire future longitudinal studies to further investigate the sequential progression of retinal vascular modifications in GA.