Choriocapillaris Impairment Is Associated With Delayed Rod-Mediated Dark Adaptation in Age-Related Macular Degeneration

Abstract

Purpose: Progress toward treatment and prevention of age-related macular degeneration (AMD) requires imaging end points that relate to vision. We investigated choriocapillaris flow signal deficits (FD%) and visual function in eyes of individuals aged ≥60 years, with and without AMD.

Methods: One eye of each participant in the baseline visit of the Alabama Study on Early Age-Related Macular Degeneration 2 (ALSTAR2; NCT04112667) was studied. AMD presence and severity was determined using the Age-Related Eye Disease Study (AREDS) grading system. FD% was quantified using macular spectral domain optical coherence tomography angiography (OCTA) scans. Vision tests included rod-mediated dark adaptation (RMDA), best-corrected visual acuity, and contrast sensitivity (photopic and mesopic), and microperimetric light sensitivity (scotopic, mesopic, and photopic). Presence of subretinal drusenoid deposits (SDD) was determined using multimodal imaging.

Results: In 410 study eyes of 410 participants (mean [SD] age = 71.7 years [5.9]), FD% was higher in early AMD (mean [SD] = 54.0% [5.5], N = 122) and intermediate AMD (59.8% [7.4], N = 92), compared to normal (52.1% [5.3], N = 196) eyes. Among visual functions evaluated, RMDA showed the strongest association with FD% (r = 0.35, P < 0.0001), followed by contrast sensitivity (r = −0.22, P < 0.0001). Eyes with SDD had worse FD% (58.3% [7.4], N = 87), compared to eyes without SDD (53.4% [6.0], N = 323, P = < 0.0001).

Conclusions: Choriocapillaris FD% were associated with AMD severity and with impaired vision, especially RMDA. Reduced metabolic transport and exchange across the choriocapillaris-Bruch's membrane retinal pigment epithelium (RPE) complex, a causal factor for high-risk soft drusen formation, also may impair photoreceptor sustenance from the circulation. This includes retinoid resupply, essential to dynamic rod function.

Age-related macular degeneration (AMD) causes legal blindness among older adults worldwide.¹ AMD is multifactorial, involving age, genetics, and lifestyle factors, such as smoking and diet. Strong evidence suggests that AMD includes an atherosclerosis-like, pro-inflammatory progression in submacular Bruch's membrane, where lipoproteins of retinal pigment epithelium (RPE) origin accumulate throughout adulthood to form high-risk drusen.^2,3 To enable treatments and preventions based on this model and others, the Alabama Study on Early Age-related Macular Degeneration 2 (ALSTAR2) is a prospective observational study seeking functional and structural biomarkers at stages before the currently approved end point (i.e. expansion of pre-existing atrophy). Lack of early biomarkers is an impediment to new treatments and prevention strategies. Optical coherence tomography (OCT)-based imaging biomarkers have the advantage of being noninvasive and rapidly acquired. However, the ideal imaging biomarker should be closely related to some aspect of visual function, which is also meaningful to patients. 

Histologic attenuation of the choriocapillaris, the innermost layer of the choroid underlying the Bruch's membrane, is a hallmark of aging⁴ and AMD.^5–7 Approximately 90% of the oxygen requirement of photoreceptors is supplied by the choriocapillaris. This microvasculature delivers glucose, retinoids, and fatty acids vital for photoreceptor function as well as removes cellular and metabolic byproducts. Choriocapillaris loss can induce RPE hypoxia and trigger upregulation of vascular endothelial growth factor, a stimulus for neovascularization. Choriocapillaris dysfunction is in the causal pathway of drusen, the largest risk factor for progression. Our model of drusen-driven AMD progression suggests that the buildup of drusen is in part due to impaired transport across aging Bruch's membrane and the choriocapillaris endothelium.^8–11 

Recent advances in optical coherence tomography angiography (OCTA) imaging allow dye-free and noninvasive visualization of the choriocapillaris meshwork. OCTA generates depth-resolved flow signal through motion contrast derived from reflectivity changes in sequential B-scans. OCTA is gaining rapid adoption in ophthalmology, and US Food and Drug Administration (FDA)-approved commercial devices are widely marketed by major manufacturers. OCTA imaging of early and intermediate AMD eyes show reduced choriocapillaris flow signal when compared to age-matched normal eyes,^12,13 and worsening with stages of non-neovascular AMD.¹⁴ Further, choriocapillaris flow deficits are associated with the presence of drusen^15,16 and subretinal drusenoid deposits (SDD).^17,18 Repeatable quantification of choriocapillaris is achievable with advancements in scan speed, signal compensation, projection artifact removal, and analytical tools.¹⁹ This enables new insights into the role of vascular dysfunction in the natural history of aging to the emergence of AMD. 

The first functional risk factor identified for AMD onset is a delayed return of light sensitivity by rod photoreceptors after exposure to a bright light flash (rod-mediated dark adaptation [RMDA]). This dynamic measure of retinoid resupply is an indicator of the metabolic exchange efficiency between rods and the systemic circulation.²⁰ Cone-mediated visual acuity is generally preserved until AMD has progressed beyond the early and intermediate stages.²¹ We recently defined a sequence of structural and functional changes preceding clinically identifiable signs of AMD.²² Retinoid resupply measured by RMDA precedes changes in steady-state scotopic sensitivity, a functional assessment of outer segment integrity. Others have shown that choriocapillaris flow signal deficits are associated with poorer steady-state scotopic sensitivity in early and intermediate AMD,²³ and may predict progression from intermediate AMD to incomplete RPE and outer retinal atrophy.²⁴ 

We hypothesize that vascular insufficiency assessed via OCTA is associated with visual impairment, especially delayed RMDA. This cross-sectional study assessed structure-function relationships between choriocapillaris using OCTA and visual functions (photopic, mesopic, and scotopic) in a large sample of older normal, early AMD, and intermediate AMD eyes. 

Methods

This cross-sectional study analyzed baseline data from the Alabama Study on Early Age-Related Macular Degeneration 2 (ALSTAR2; ClinicalTrials.gov Identifier NCT04112667). Participants ≥60 years old were recruited through the comprehensive care clinics at the Callahan Eye Hospital at the University of Alabama at Birmingham between October 2019 through September 2021 and evaluated at the Clinical Research Unit of the Department of Ophthalmology and Visual Sciences. Institutional Review Board approval was granted by the University of Alabama at Birmingham. Research adhered to the principles of the Declaration of Helsinki. All study participants provided written informed consent. 

The clinic electronic health record system was searched for patients with early or intermediate AMD using International Classification of Diseases, Tenth Revision (ICD-10) codes (H35.30*, H35.31*, and H35.36*). Records were screened for eligibility by author C.O. Exclusion criteria were (1) any eye condition or disease in either eye (other than early cataract) that can impair vision, including diabetic retinopathy, glaucoma, ocular hypertension, history of retinal diseases (e.g. retinal vein occlusion and retinal degeneration), optic neuritis, corneal disease, previous ocular trauma or surgery, and refractive error ≥ 6 diopters; (2) neurological conditions that can impair vision or judgment, including multiple sclerosis, Parkinson's disease, stroke, Alzheimer's disease, seizure disorders, brain tumor, and traumatic brain injury; (3) psychiatric disorders that could impair the ability to follow directions, answer questions about health and functioning, or to provide informed consent; (4) diabetes; and (5) any medical condition that causes significant frailty or was thought to be terminal. Persons in normal macular health from the same clinic met the same eligibility criteria except for the ICD-10 codes indicative of AMD. 

In each participant, the eye with better acuity was tested for imaging and visual function. If both eyes had the same acuity, then an eye was selected randomly. Classification into diagnostic groups was based on evaluation of three-field, color fundus photographs taken with a digital camera (450+; Carl Zeiss Meditec, Dublin, CA, USA) after dilation with 1% tropicamide and 2.5% phenylephrine hydrochloride. Presence and severity of AMD was assessed using the Age-Related Eye Disease Study (AREDS) 9-step classification system²⁵ by a trained grader (author M.E.C.) masked to other participant characteristics. Groups were defined as follows: eyes with normal macular health were classified as AREDS grade 1, early AMD as grades 2 to 4, and intermediate AMD as grades 5 to 8. We also classified eyes using the Beckman system.²⁶ Birthdate, gender, race/ethnicity, and smoking status were obtained through a self-administered questionnaire. 

Optical Coherence Tomography Angiography

Two consecutive 15 degrees × 15 degrees (approximately 4.4 × 4.4 mm) OCTA volumes from each study eye were captured using the OCT Angiography Module of the Spectralis HRA + OCT (Heidelberg Engineering, Germany; HEYEX software version 6.10.6.0).²⁷ Each fovea-centered OCTA volume comprised 384 B-scans 12 µm apart, and 3.9 µm axial and 5.7 µm transverse resolution. Projection-resolved choriocapillaris slabs were segmented using the manufacturer-recommended slab located 10 to 30 µm below Bruch's membrane.²⁸ From the two volumes captured per eye, the slab with fewer artifacts was included for analysis. Absence of OCTA artifacts like shadowing (e.g. due to floaters or dense cataract), vignetting, segmentation errors, defocus, tilt, Z-offset, or motion artifacts^29,30 were assessed by a masked, trained grader (author D.K.). OCTA slabs with scan quality ≤25 decibel signal-to-noise ratio were excluded. 

Scans were batch processed using custom software written in MATLAB 2022b (The MathWorks, Inc., Natick, MA, USA) and ImageJ (National Institutes of Health, Bethesda, MD, USA). To compensate for signal attenuation under large drusen and hyper-reflective foci, each OCTA image was multiplied by the inverse of the structural OCT image extracted from identically segmented OCTA volumes.^31,32 Compensated images were binarized using the Phansalkar local thresholding method using a 20 µm radius³³ (representative examples are shown in Supplementary Figs. S1, S2). Choriocapillaris areas directly beneath major retinal vessels were excluded to eliminate confounding sources of projection artifacts or shadowing.³⁴ Whole-image choriocapillaris flow signal deficits (FD%) within the 15 degrees × 15 degrees (approximately 4.4 × 4.4 mm) OCTA scan area were quantified. This metric represents the percentage of area lacking detectable flow signal, that is, percentage of dark pixel areas in the binary mask that are not underneath major retinal vessels.³⁵ Some OCTA studies exclude isolated FD regions with equivalent diameter smaller than 24 µm, with the presumption that these represent noise rather than true intercapillary spaces.^36–38 Our direct measurements of intercapillary distances from histology indicate that 76.8% are smaller than 24 µm (Supplementary Fig. S3).³⁹ Thus, we did not enforce a minimum cutoff. The FD% test-retest repeatability and effectiveness of choriocapillaris signal compensation methodology was performed in 2 scans consecutively captured from 30 randomly selected participants from the older normal group (AREDS1). It is possible that compensation methods promulgated for swept-source OCT may not apply well to spectral domain OCT, in which case, the area under drusen may be inadequately compensated and FD% overestimated.³² To test that possibility, we investigated whether removing drusen regions significantly affects whole-image FD% measures in a subset of study eyes. We compared FD% in choriocapillaris slabs that included regions with drusen versus excluding these regions, in 20 randomly chosen study eyes (10 early and 10 intermediate), for which drusen were identified using OCT B-scans and structural en face OCT (see Supplementary Figs. S1, S2). 

Assessment of Subretinal Drusenoid Deposits

Because SDD (also called reticular pseudodrusen), have been associated with choroidal insufficiency,⁴⁰ FD% was compared between eyes with and without SDD. As described,²² SDD were detected by a masked grader (author D.K.) using near infrared reflectance and en face OCT imaging. At least five definite lesions anywhere in the scan area, with each deposit spanning at least two cross-sectional OCT B-scans were required. This approach was adequate to capture the earliest clinically visible stages of SDD.^41,42 Agreement with a second grader (author M.E.C.) who assessed a random 14% subsample of study eyes was strong (Cohen's κ = 0.89, 95% confidence interval = 0.77–1.0). 

Visual Function Testing

A battery of vision tests was administered under photopic (cone-mediated), mesopic (rod- and cone-mediated), and scotopic conditions (rod-mediated), as described previously^22,43 (see Appendix for details). We measured best-corrected photopic and mesopic letter acuity (log minimum resolvable),^44,45 photopic and mesopic letter contrast sensitivity (log sensitivity),⁴⁶ steady-state light sensitivity under photopic, mesopic, and scotopic conditions (log sensitivity),²² and rod-mediated dark adaptation at 5 degrees (rod intercept time [RIT]).²² 

Statistical Analyses

Demographic and visual function characteristics of participants were summarized using means and standard deviations or number (percent) for continuous and categorical data, respectively. The FD% was summarized with means and standard deviations. Analysis of covariance (ANCOVA) with pairwise comparisons was used to compare visual function and FD% by AMD severity adjusting for age. As ANCOVA requires a continuous dependent variable and a categorical independent variable, AMD severity was the independent variable, and FD% and the visual function measures were the dependent variables. Age is the covariable in these models, which controls for the confounding effects of age on FD% and visual functions when comparing by AMD severity. Age-adjusted associations for FD% and visual function parameters were computed with Spearman correlations (r). In addition to age, correlation models for FD% and RIT were adjusted for SDD presence. Age-adjusted ANCOVA was used to compare FD% by SDD status and AMD status stratified by SDD status. The level of significance was P ≤ 0.05 (two-sided). All analyses were performed in SAS version 9.4. 

Results

A total of 510 eyes from 510 participants were assessed for eligibility. One hundred participants (19.6%) were excluded for significant floaters and OCTA artifacts (64 eyes), non-exudative choroidal neovascularization (18 eyes), and invalid RMDA (18 eyes). The final sample included 410 eyes from 410 participants (mean [SD] age = 71.7 [5.9] years, 245 women (59.8%), 377 non-Hispanic Whites (92.0%), and 173 self-reported current or former smokers (42.2%; Table 1). The sample included 196 older normal (47.8%), 122 early AMD (29.8%), and 92 intermediate AMD (22.4%) eyes based on the AREDS classification. Advanced age was associated with AMD severity (P < 0.0001) and higher FD% (r = 0.24, P < 0.0001). Therefore, all analyses were age-adjusted. 

Table 1.

 

Demographic Characteristics of Study Participants (410 Eyes From 410 Persons)

 

Test-retest repeatability of choriocapillaris FD% using OCTA measured in 30 older normal eyes was excellent (intraclass correlation coefficient [ICC] = 0.96, Shrout-Fleiss reliability, fixed set). These scans lacked identifiable lesions like drusen or hyper-reflective foci. Choriocapillaris FD% quantification before and after compensation did not show significant differences and showed excellent agreement in scans (ICC = 0.996, Shrout-Fleiss reliability, and fixed set, N = 30). Inspection of choriocapillaris OCTA images in 20 randomly chosen eyes with drusen (see Supplementary Figs. S1, S2) revealed that reduced and/or attenuated flow signal was not confined to areas under drusen, as is most apparent in binarized images (right columns of figures). Quantitatively, in these 20 eyes with drusen, the mean (SD) difference in FD% when including drusen versus excluding drusen was negligible, at −0.58 (0.74; Supplementary Fig. S4). Based on these findings, we did not exclude drusen areas for all study eyes, because this exclusion affects the sampling area underlying the FD% value. Eyes with large drusen have a less analyzable area within a window of constant size represented by the scan area than eyes with small drusen, thus implicitly over-weighting the contribution of individual pixels. 

Figure 1 illustrates choriocapillaris FD% across AMD severity groups. Compared to older normal eyes, FD% in early AMD, and intermediate AMD was significantly worse (age-adjusted overall P < 0.0001). Mean (SD) FD% for older normal, early AMD, and intermediate AMD groups were 52.1% (5.3), 54.0% (5.5), and 59.8% (7.4), respectively. Choriocapillaris FD% was higher in women (mean (SD) = 55.1 [6]3)) than men (53.3 [6.9]; age adjusted P = 0.0014) and was higher in current/former smokers (mean [SD] = 55.9 [7.0]) than never smokers (53.3 [6.1]; age-adjusted P = 0.0002). Representative examples of en face structural OCT and OCTA illustrating the differences in choriocapillaris FD% between older normal and intermediate AMD with drusen is shown in Figure 2. In the eye with drusen, en face OCTA showed generalized loss of choriocapillaris flow signal within the 3 mm submacular region, with deficits extending beyond the areas of drusen identified using OCT B-scans and hyper-reflective spots surrounded by a hyporeflective annuli on en face structural OCT. Supplementary Table S1 shows FD% differences between the AMD severity groups defined by the AREDS versus Beckman classification systems. By the Beckman classification, FD% in normal eyes did not differ from early AMD (P = 0.7767), but intermediate AMD eyes had higher FD% than both normal and early AMD eyes (both P < 0.0001). 

Figure 1.

 

Boxplot comparison of choriocapillaris flow signal deficits by AMD severity. Box plots indicate the median and interquartile ranges. Whiskers indicate the 5th and 95th percentiles. Notches indicate the 95% confidence interval of the median. AMD severity groups are assigned based on color fundus photography grading (AREDS 9-step).

Figure 2.

 

Representative case examples of choriocapillaris flow signal deficits in normal versus intermediate AMD. (A1) Confocal color fundus photographs showing the unremarkable macula of a 69-year-old, White woman graded as normal (AREDS1). Dotted circle shows the location of the rod-mediated dark adaptation test spot at 5 degrees superior. (A2) En face structural optical coherence tomography (OCT) slab segmented from the external limiting membrane (ELM) and interdigitation zone (IZ) boundaries. (A3) En face structural choriocapillaris slab located 10 to 30 µm below Bruch's membrane (identical slab boundaries as that of A4). (A4) En face OCT angiography (OCTA) of the choriocapillaris flow slab located 10 to 30 µm below Bruch's membrane (red lines in A5). Dark pixels denote flow signal deficits. (A5) Unremarkable foveal OCT B-scan with flow signal overlay. (B1) Confocal color fundus photographs show drusen in the left eye of a 73-year-old, White man graded as intermediate AMD (AREDS7). (B2) En face structural OCT slab show numerous drusen identified as small hyper-reflective regions each surrounded by a hyporeflective annulus. (B3) En face structural choriocapillaris slab located 10 to 30 µm below Bruch's membrane (identical slab boundaries as that of B4). (B4) En face OCTA shows generalized loss of choriocapillaris flow signal within the 3 mm submacular region, with deficits extending beyond the areas of drusen (yellow overlay in B5) labeled using OCTA B-scans (B6). Scale bars = panels A1-A3, and B1-B4 = 1 mm; panels A4, B5 = 200 µm.

Supplementary Figure S5 illustrates 15 degrees × 15 degrees choriocapillaris slabs of the entire cohort, stratified by AREDS-defined diagnostic groups. Within each group, OCTA slabs are ordered by increasing FD%. Early and intermediate AMD eyes showed greater flow signal deficits (denoted by dark pixels), compared to normal eyes. Flow signal attenuation was centered in the submacular region and appeared to be exacerbated in intermediate AMD eyes, relative to normal and early AMD eyes. 

Table 2 summarizes results of vision tests, stratified by AREDS-defined AMD status. Relative to normal and early AMD, all visual functions were significantly worse in intermediate AMD. The only function significantly worse in early AMD than in older normal eyes was RMDA (P = 0.0019). 

Table 2.

 

Summary and Age-Adjusted Comparison of Visual Functions by AMD Severity

 

Associations between choriocapillaris FD% and visual function are shown in Table 3. Greater FD% was significantly associated with worse vision in all tests (P < 0.05) except photopic and mesopic light sensitivity (P > 0.05). Among all functional tests, RMDA was most strongly associated with FD% (r = 0.35, P < 0.0001), followed by photopic contrast sensitivity (r = −0.22, P < 0.0001). Greater FD% was associated with longer RIT and lower contrast sensitivity. In comparison, photopic acuity was weakly correlated with FD% (r = 0.11, P = 0.0277). Scatter plots of functional outcomes by AMD severity group illustrate these relationships (Fig. 3). In addition to a weaker association relative to RMDA, for both acuity and contrast sensitivity, FD% largely overlapped between older normal and early AMD groups, as demonstrated by 95% confidence ellipses. The association between FD% and RIT decreased slightly when SDD presence was included in models in addition to age (r = 0.30, P < 0.0001). 

Table 3.

 

Age-Adjusted Association of Choriocapillaris Flow Signal Deficits With Rod- and Cone-Mediated Visual Function (N = 410)

 

Figure 3.

 

Association of choriocapillaris flow signal deficits with visual acuity, contrast sensitivity, and rod intercept time stratified by AMD severity. Scatterplot shows AMD severity groups in different colors. Spearman correlations (r) between choriocapillaris flow signal deficits with rod intercept time, contrast sensitivity, and best-corrected visual acuity are 0.35 (P < 0.0001), -0.22 (P < 0.0001), and 0.11 (P = 0.0277), respectively. Dotted lines represent the regression line, and shaded gray areas represent the 95% confidence intervals of the fit of the overall sample (N = 410). Ellipses show the 95% confidence level for the t-distribution of each AMD severity group.

Overall, eyes with SDD had higher FD% (58.3% [7.4], N = 87) compared to eyes without SDD (53.4% [6.0], N = 323, P < 0.0001). The proportion of eyes with SDD was highest in intermediate AMD followed by early AMD and older normal (44.6% vs. 22.1% vs. 9.7%). When stratified by SDD status, the trends in FD% change were similar between normal and early AMD (no SDD; P = 0.1473 and SDD; P = 0.0547) and when comparing normal to intermediate AMD (no SDD; P < 0.0001 and SDD; P = 0.0002; Supplementary Table S2). FD% differed between early and intermediate AMD only for eyes without SDD (P < 0.0001) and did not differ significantly among those with SDD (P = 0.0538). Supplementary Table S3 shows the within-group differences in choriocapillaris FD% by SDD status stratified by AMD status. 

Discussion

Understanding the relationship between structural biomarkers and functional measures of vision is critical for designing clinical trials that aim to arrest AMD progression. Imaging biomarkers will be most valuable as surrogate end points if they are robustly correlated with meaningful functional end points. Our findings demonstrate that choriocapillaris FD% measured using OCTA worsened with AMD severity in a well-characterized sample of adults 60 years and older. Among all visual functional parameters measured, delayed RMDA, a dynamic measure of photoreceptor sustenance, showed the strongest association with choriocapillaris impairment. In contrast, photopic visual acuity was weakly associated with FD%, and no associations were detected for steady-state photopic and mesopic light sensitivity. 

Our results extend existing literature on choriocapillaris and visual function impairment in non-neovascular AMD, primarily intermediate^24,47,48 and geographic atrophy.^49–52 Loss of parafoveal rods contributing to impaired scotopic vision is a hallmark of aging and early AMD.^53–55 Delayed RMDA is the first functional biomarker identified for incident AMD,^20,22,56 and precedes loss of cone-mediated foveal acuity in advanced disease.²⁶ In eyes with intermediate AMD, choriocapillaris signal deficits were associated with prolongation of the initial negative deflection (N1) implicit times in multifocal electroretinography.⁴⁸ In early and intermediate AMD, Nassisi and colleagues reported that reduced scotopic microperimetric sensitivity (steady-state rod function), was associated with higher FD%.²³ Findings from the current study suggest that the dynamic aspects of rod photoreceptor sustenance are also impaired. In addition, other aspects of outer retinal physiology can also potentially contribute to delayed RMDA.⁵⁷ 

Our results support the hypothesis that choriocapillaris degeneration and vascular insufficiency that initiate in aging may contribute to poor rod-mediated vision in AMD (schematized in Fig. 4). Choriocapillaris endothelium is essential to metabolic exchange between the circulation and outer retinal cells. It is thought that choriocapillaris degeneration contributes to the biogenesis of extracellular deposits (drusen and basal laminar deposit) that can further impede transport under and near the fovea. Deposits in turn portend both neovascularization and atrophy. Although drusen were not measured in this study, we observed qualitatively attenuated flow signal centered under the macula. These observations align with histological evidence of significant submacular choriocapillaris loss in clinically documented donor eyes with early AMD by Lutty and associates.^5,7 

Figure 4.

 

Vascular basis of rod- versus cone-mediated visual dysfunction in age-related macular degeneration. We hypothesize that choriocapillaris dysfunction is a causal factor for visual deficits in AMD, due to impaired retinoid and metabolic exchange to the photoreceptors. (1) In normal eyes, the choriocapillaris meshwork supplies retinoids to both rod (R) and cone (C) photoreceptors via the retinal pigment epithelium-Bruch's membrane interface. (2) Cones additionally receive retinoids from an alternate supply route from Müller glia (M) and the retinal circulation (cone-specific second visual cycle). By contrast, rods are solely dependent on the choroid for retinoid resupply. In AMD, choriocapillaris degeneration (3) and/or accumulation of drusen and other age-related extracellular deposits (4) lead to impaired retinoid exchange to photoreceptors, compromising the efficiency of classic visual cycle. (5) Thus, rod-mediated visual deficits in the parafovea occur earlier in disease because they are solely dependent on the choroidal circulation for retinoid resupply. Cone-mediated visual function in the fovea is generally preserved until later stages of disease due to the additional retinoid resupply route from Müller glia (M).

By current convention, we measured FD, the percentage of area lacking detectable OCTA flow signal. FD% is a measure of the area covered by intercapillary pillars formed by the outer collagenous layer of Bruch's membrane. It is equivalent to 100% – vascular density, a measure of the metabolic exchange interface between choriocapillaris and Bruch's membrane. In general, we report a higher FD, indicating impairment, than is commonly seen in the OCTA literature, a difference with both a biologic and technical basis. First, we did not remove small flow signal deficits from our calculations, as our histology indicates a majority of intercapillary spacings are small. Removing signal deficits <24 µm because they are less reliable³⁶ removes signal in addition to noise. Our histologic measurements of >2500 intercapillary distances from aged normal and early AMD eyes (n = 19 each) show that 76.8% are smaller than 24 µm (see Supplementary Fig. S3). Second, choriocapillaris FD% is highly variable among studies, attributable to differences in OCT technology (spectral-domain versus swept-source OCT), signal generation algorithms, image analysis methods, and study populations.^31,37,58,59 Because of these determinants, the significance of our current findings will be elucidated using the same device and methodology in the ongoing 3-year follow-up for ALSTAR2. 

In eyes with SDD, we found that FD% was worse in intermediate AMD eyes but not in early AMD eyes, when compared to normal eyes. This indicates that vascular abnormalities associated with SDD may occur later in the progression sequence than in eyes lacking SDD. Comparison between older normal and early AMD eyes failed to reach significance, perhaps due to the small number, necessitating further study. To maximize SDD detection, we used multimodal imaging, incorporating en face OCT in addition to near infrared (NIR) for high contrast over a wide area, and confirmed with OCT B-scans. RMDA was previously noted by us and others to be markedly worse in eyes with SDD,^22,60 which are proven risk factors for progression to neovascularization and atrophy. Previous research indicated that eyes with SDD have reduced visual acuity and greater choriocapillaris impairment, compared to eyes with drusen and lacking SDD.^17,18 A 5-year longitudinal study showed that both SDD-affected and unaffected areas can be associated with choriocapillaris impairment.¹⁷ 

We excluded 12.5% of participants based on the relatively stringent scan selection criteria. Exclusion criteria included floaters and common OCTA artifacts identified by a reading center study.²⁹ Frequency of OCTA artifacts is affected by participant age, severity of disease, presence of comorbidities, fixation stability, device technology, and postprocessing methods. Artifacts are inherent for any imaging modality, including OCTA. While projection artifact removal algorithms have greatly improved, other patient-level artifacts like shadowing, defocus, and movement can impede the interpretability and reproducibility of quantitative measures²⁹ like choriocapillaris FD%. We assessed the effectiveness of compensation methodology by showing that FD% before and after compensation in aged normal eyes did not differ. However, even an effective signal compensation algorithm may not adequately compensate severely attenuated or absent signal due to lesions, such as large drusen or hyper-reflective foci in intermediate to advanced AMD. In addition, the lateral resolution of commercially available devices commonly prevents visualization of the intricate choriocapillaris network seen in histology^7,61 and adaptive optics assisted OCTA.⁶² Continuing improvements in software and hardware (such as swept-source OCTA) may potentially help mitigate artifacts for better repeatability and reproducibility of OCTA based quantitative measures. 

FD% was worse in early and intermediate AMD eyes than normal eyes, when stratified by the AREDS grading system only (i.e. not by the Beckman grading system). The number of eyes considered early AMD differed between the two stratifications (N = 122 eyes for AREDS and N = 83 for Beckman). This shift is attributable to eyes with pigmentary changes only (without drusen) being considered early AMD in AREDS and intermediate AMD in Beckman. The AREDS and Beckman systems were developed for different purposes (progression versus consensus). A standardized multimodal imaging paradigm for AMD onset and progression does not yet exist. The ALSTAR2 imaging dataset may be useful in developing such a system in the future. 

Strengths of this study include a large sample, possibly the largest of OCTA imaging reported for AMD to date. Structure-function relationships were chosen to comprehensively test a wide range of light adaptation levels across all three disease severity groups. We used scan selection criteria with rigorous image quality control procedures for excluding scans with common OCTA artifacts. We also incorporated new information from histologic measurements of choriocapillaris. 

Our study also has limitations. Choriocapillaris OCTA slabs imaged using a spectral domain device are susceptible to flow signal attenuation due to overlying large drusen, hyper-reflective foci, or severe floaters. Spectral domain OCTA devices are equipped with a shorter wavelength relative to swept-source devices that have been used by other groups also studying choriocapillaris. This leads to reduced contrast and definition of this microvasculature. Given there is no consensus regarding choriocapillaris FD% quantification using different devices, and given that manufacturers use different proprietary algorithms, absolute choriocapillaris FD% cannot be compared across devices. All devices are subject to constraints of the underlying technology and are surrogates for the underlying anatomy, making longitudinal follow-up on any population essential. Finally, we did not measure axial length to correct lateral scaling before quantifying FD% in this study. Lateral scaling is important when reporting absolute measurements like foveal avascular zone diameter, vessel length, and intercapillary distances.^63,64 However, because FD% is a percentage of pixels lacking flow within the total analyzed scan area, we do not believe that our results would change substantially with image scaling. 

In conclusion, loss of choriocapillaris flow signal was associated with worse AMD disease severity and visual impairment, especially delayed RMDA in a large sample of older adults. Functionally validated structural biomarkers of early AMD, a population totaling 18 million persons in the United States,⁶⁵ can aid detection of eyes at risk for progression. FD% using OCTA is a continuous and repeatable measure of vascular supply from the choroidal circulation, using technology that is becoming widely available among retina specialists. The current findings suggest that OCTA may be able to detect changes at early stages of AMD and is thus a potential imaging biomarker for non-neovascular AMD. Measures of choriocapillaris impairment also have the advantage of being in a causal pathway for drusen biogenesis.^8,9,66 Further, choriocapillaris FD% in intermediate AMD eyes may predict 2-year disease progression to incomplete RPE and outer retinal atrophy.²⁴ Whether choriocapillaris imaging can predict AMD onset will be evaluated in the 3-year follow-up of this cohort. Topographical differences in choriocapillaris FD% may explain why RMDA is worse close to the fovea where rods and SDD are both sparse than where they are both abundant near the vascular arcades.^67–70 Our current data highlights the importance of choriocapillaris in the pathogenesis of early changes in AMD eyes, a finding with therapeutic implications,^71–73 and supports the need for standardization of OCTA analysis.⁷⁴ Future studies should probe topographical and longitudinal associations between choriocapillaris loss and vision to determine the utility of OCTA for monitoring AMD. 

Acknowledgments

Supported by R01EY029595 (C.O. and C.A.C.) and R01EY027948 (C.A.C.); P30EY03039 (C.O.); Dorsett Davis Discovery Fund (C.O.), and Alfreda J. Schueler Trust (C.O.); Unrestricted funds to the Department of Ophthalmology and Visual Sciences (UAB) from Research to Prevent Blindness, Inc., and EyeSight Foundation of Alabama. 

Disclosure: D. Kar, None; G. Corradetti, Nidek (C) (outside this project); T.A. Swain, None; M.E. Clark, None; G. McGwin Jr, None; C. Owsley, Johnson & Johnson Vision (C) (outside this project); S.R. Sadda, AbbVie/Allergan (C), Amgen (C), Apellis (C), Iveric (C), Novartis (C), Roche/Genentech (C), Regeneron (C), Janssen (C), Nanoscope (C), Biogen (C), Boehringer Ingelheim (C), Optos (C), Centervue (C), Heidelberg Engineering (C), Optos (F), Centervue (F), Heidelberg Engineering (F), Carl Zeiss Meditec (F), Nidek (F), Topcon (F); C.A. Curcio, Genentech/Hoffman LaRoche (F), Regeneron (F), Heidelberg Engineering (F), Apellis (C), Astellas (C), Boehringer Ingelheim (C), Character Biosciences (C), Osanni (C), Annexon (C) (all outside this project)