Impairment of Neurovascular Function in Intermediate Age-Related Macular Degeneration

Abstract

Purpose: To investigate neurovascular function in eyes with age-related macular degeneration (AMD).

Methods: Subjects with bilateral large drusen (intermediate AMD) and healthy controls ≥50 years old were recruited. The vasculature within the central 6 × 6-mm retinal area was captured using optical coherence tomography angiography (OCTA) and segmented to return superficial plexus, deep plexus, choriocapillaris, and choroid. OCTA scans were acquired without flicker light stimulation (conventional OCTA) and during flicker light stimulation to increase retinal activity and metabolic demand (functional OCTA). Vascular area density (VAD) and the vascular reactivity index (VRI; change in VAD induced by flicker stimulation) were determined and compared between control and AMD eyes.

Results: Thirty-five subjects (19 AMD cases and 16 healthy controls) participated in the study. In healthy eyes, flicker stimulation induced an increase in VAD (positive VRI, vasodilation) in the superficial plexus (P < 0.001) and deep plexus (P < 0.001). There was a trend for increased VAD in the choriocapillaris (P = 0.077), but there was no change in the choroid (P = 0.654). In AMD eyes, there was no change in VAD in response to flicker stimulation in any of the vascular layers examined (P ≥ 0.294). Linear mixed models confirmed that AMD was associated with a reduced VRI in the superficial plexus (P < 0.001) and deep plexus (P < 0.001).

Conclusions: Eyes with large drusen show a reduction in retinal vascular reactivity compared to healthy eyes, which suggests that there is impairment of retinal neurovascular function in intermediate AMD.

Translational Relevance: Functional OCTA could be used to study neurovascular function in retinal diseases.

Introduction

Age-related macular degeneration (AMD) is a complex disease, the pathogenesis of which is yet to be fully understood. Vascular abnormalities and oxidative stress have been suggested to be important factors in AMD pathogenesis due to the especially high oxygen consumption in the outer retina and the lack of operating reserves of oxygen.^1–3 In addition, there are multiple choroidal watershed zones at the submacular region that make the outer retina particularly vulnerable to ischemia.^4,5 Supporting evidence that AMD may have a vascular component includes data from epidemiological studies showing a strong association between cardiovascular risk factors (such as hypercholesterolemia, smoking, and atherosclerosis) and AMD.^6–10 Studies using optical coherence tomography (OCT) have reported thinning of the choroid and a reduced choroidal vascularity index in eyes with AMD.^11–14 More recently, using OCT angiography (OCTA), studies found that eyes with drusen are associated with choriocapillaris flow impairment,¹⁵ an increase in flow void area,^16,17 and a decrease in choroidal or choriocapillaris vascular density.^17–19 Furthermore, eyes with reticular pseudodrusen, a subphenotype of AMD, have been found to be associated with a greater impairment of choriocapillaris flow compared to eyes with conventional drusen.^20,21 Vascular changes are not restricted to the choroid and choriocapillaris, as studies have reported attenuation of retinal superficial capillary plexus density and a decrease in the number of branch points in the deep capillary plexus in eyes with early stages of AMD compared to those without drusen.^22,23 

Although studies have shown structural abnormalities and reduced circulation in both the choroidal and retinal vasculature in AMD, much less is known about the regulation of blood supply in response to a change in metabolic demand in these eyes. Choroidal blood flow is regulated by autonomic and trigeminal sensory nerve innervation.^24,25 Unlike the choroid, retinal blood flow is dependent upon a local autoregulatory mechanism.²⁴ This autoregulatory mechanism in response to a change in metabolic demand is known as neurovascular coupling,^26,27 which involves communication between elements of the neurovascular unit: pericytes, neurons, glia, and blood vessel endothelial cells. The process is vital for regulating the distribution of blood in the retina to match local changes in metabolic demand due to neuronal excitation. 

Studies of neurovascular function in AMD to date have been limited to the larger vessels in the superficial plexus due to the inability to visualize the deeper vascular layers with fundus photographs. Using the Dynamic Vessel Analyzer (DVA; Imedos Systems, Jena, Germany) in a small group of patients with neovascular AMD, Lanzl et al.²⁸ showed that flickering light-induced vasodilation of the superficial plexus was slower but reached the same mean level. Whether vascular function in retinal vessels, particularly the deep capillary plexus, which is critical for providing metabolic support to the outer retina, is altered in early stages of AMD has not been investigated. 

Recent studies have employed OCTA to quantify vascular reactivity of the superficial capillary plexus in response to metabolic challenge such as isometric exercise or flickering light in those with diabetes and in healthy controls.^29–31 In this study, we used functional OCTA to quantify vascular reactivity in retinal plexi, the choriocapillaris, and the choroid to investigate whether AMD is associated with neurovascular dysfunction. 

Methods

This prospective study was approved by the Human Ethics Committee of the Royal Victorian Eye and Ear Hospital and conducted in the Macular Research Unit at the Centre for Eye Research Australia in adherence with the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants after the study had been explained. 

Participants

Participants were recruited if they were 50 years of age or older with best-corrected visual acuity (BCVA) of 20/60 or better. Participants with AMD were required to have bilateral large drusen (>125 µm in diameter), satisfying the classification of intermediate AMD based on the Beckman classification (Figs. 1A, 1B).³² Cases with late AMD, either geographic atrophy or choroidal neovascularisation, were excluded. Control participants were required to have no drusen on any imaging modality nor any other eye conditions. Other exclusion criteria for both AMD cases and control participants were any myopic refractive error of greater than −6 diopters (D), hyperopic refractive error greater than +4 D, presence of significant ocular media opacification that obscured the fundus, or any co-existing eye conditions such as diabetic retinopathy or glaucoma. Participants were asked to abstain from alcohol and caffeine at least 3 hours before the visit to minimize potential autonomic effects on microvascular reactivity. 

Figure 1.

 

(A) Near infrared and (B) OCT images of an eye with large drusen located at the macula. (C) OCTA images of the superficial plexus of the same eye. (D) Shown are the 100 (10 × 10) defined regions of interest.

Procedures

Systemic and eye health was evaluated with general health questionnaires, BCVA testing, comprehensive eye examinations, multimodal retinal imaging, and functional OCTA. Acquisition of retinal imaging (including OCTA) was performed with a dilated pupil (at least 7 mm). BCVA was measured after subjective refraction using the Early Treatment Diabetic Retinopathy Study protocol and was recorded as the number of letters read. The eye with the better BCVA was selected as the study eye. In addition to BCVA, scotopic retinal function was also determined when possible using a scotopic perimeter (Medmont International, Victoria, Australia). Scotopic function was examined after 30 minutes of dark adaptation, as previously described, to return the mean sensitivity within the central 16° from the fovea.^33,34 Due to a long testing duration, some participants were unable to perform this test. 

Multimodal Imaging

All participants underwent near-infrared reflectance, short-wavelength fundus autofluorescence, OCT (SPECTRALIS HRA+OCT; Heidelberg Engineering, Heidelberg, Germany), and color fundus photography (CR6-45NM; Canon, Tokyo, Japan). For OCT, 97 B-scans within the central 20° × 20° of the retina were obtained, and each B-scan was the average of 15 frames. Multimodal images were graded to confirm AMD classification as per the Beckman Classification and Grading System.³² 

Functional OCTA

The superficial plexus, deep plexus, choriocapillaris, and choroidal vasculature within the central 6 × 6-mm area of the retina (Fig. 1C) was captured using a ZEISS PLEX Elite 9000 device (Carl Zeiss Meditec, Dublin, CA). In each eye, two conventional baseline OCTA scans (cOCTA1, cOCTA2) without flicker light stimulation were acquired first, followed by one functional OCTA (fOCTA) scan with flicker stimulation. Only one fOCTA was acquired because it has been shown that short-term retesting is associated with a reduction in vasomotor response to flicker stimulation, at least in the superficial plexus.³⁵ For the fOCTA scan, external flickering light-emitting diode (LED) lights (four diffuse LEDs, one in each quadrant; warm white; 10-Hz flicker; 100 cd/m²; 50% duty cycle) mounted around the lens of the OCTA system were used to induce an increase in retinal activity and metabolic demand, thus driving physiological functional hyperemia.^36–38 In normal eyes, maximal vasodilation of the superficial plexus was observed within a few seconds after flicker stimulation, and vasodilation remains relatively constant during flicker stimulation.^27,39 Acquisition of fOCTA scans was commenced approximately 10 seconds after flicker onset, and flicker stimulation remained on throughout acquisition. All OCTA acquisitions were performed under low ambient lighting (room light was turned off, ∼1 lux) for consistency and to maximize vasomotor response to flicker stimulation.⁴⁰ Scans were taken in follow-up mode such that the same retinal area was imaged for all subsequent OCTA scans. Scanning parameters were optimized to ensure that signal strength was ≥8. If the signal strength was less than 8, re-optimization was performed and the first baseline OCTA scan was repeated before acquiring the second baseline OCTA and the subsequent fOCTA scan. 

Determination of Vascular Reactivity to Flicker Stimulation

The OCTA scans were processed using the built-in ZEISS OCTA software to remove projection artifacts produced by the superficial plexus and generate en face images for the superficial plexus, deep plexus, choriocapillaris, and choroidal layer. The superficial plexus and deep plexus slabs were predefined by the OCTA software. Based on the anatomy of the choriocapillaris and the choroid and recommendations from previous studies,^41,42 the choriocapillaris (region from 4 to 14 µm below the Bruch's membrane) and the choroid (region from 70 µm below the Bruch's membrane to the choroid–scleral junction) slabs were manually defined using the global customize slab feature. Visual inspection on each B-scan did not reveal any segmentation errors in any of the slabs. OCTA en face images of the vascular layers were exported for image processing and calculation of vascular area density (defined as the percentage the pixel area encompassed by blood vessels). 

Exported en face images were processed using ImageJ (National Institutes of Health, Bethesda, MD). For each eye, the images of each layer were processed as pairs (i.e., pair of two cOCTA images to evaluate the variability of the technique or a pair of one cOCTA and one fOCTA image to assess neurovascular function). Images were binarized using the same automatic threshold applied to both images in the pair. In each OCTA en face image, we calculated the vascular area density (VAD) for the entire image, as well as for each of 100 (arbitrarily 10 × 10 grid) equally sized regions of interest (ROIs) to examine VAD topographically. To evaluate variations in VAD between OCTA images, changes in VAD between two cOCTA images (VAD_cOCTA2 – VAD_cOCTA1) were determined. The vascular response to flicker stimulation was calculated as the vascular reactivity index (VRI), or the change in VAD induced by flicker stimulation (VAD_fOCTA – VAD_cOCTA1). Positive and negative VRI values indicated flicker-induced vasodilation and vasoconstriction, respectively. 

Statistical Analysis

Student's t-test was used to compare age, body mass index (BMI), BCVA, and VAD between the AMD and healthy control groups. The distribution of gender and smoking status between the study groups was compared using a χ² test. The VAD of the entire OCTA image was used to examine the variation in VAD between OCTA scans and to determine the effect of flicker stimulation on VAD. Linear mixed models were used to examine the association between AMD and vascular reactivity in response to flicker stimulation, with consideration of repeated measures of VAD within an eye and adjusting for potential confounders such as age, gender, smoking status, and BCVA. Both the mean VRI of the entire OCTA scan (one data point per eye) and the individual VRI at each defined ROI (100 data points per eye) were used for the analysis (Fig. 1D), with the study groups as the fixed effect and sampling locations (ROIs) nested within an eye as random effects. Use of individual VRIs accounted for the hierarchical nature of the dataset and potential intra-eye correlation and is thus more sensitive to detect localized VRI changes. Furthermore, given that neurovascular coupling regulates local blood flow by coordinating the constriction and dilation of local and surrounding capillaries, with the net response typically resulting in vasodilation, the use of an individual VRI at each defined ROI and mean VRI allowed both local and net responses to be determined. Because a subset of participants also completed scotopic sensitivity testing, two linear mixed models were used for the analysis. In model 1, the association between AMD and VRI was examined in the entire study cohort without the scotopic sensitivity data. In model 2, the association between AMD and VRI was examined in a subgroup of participants with the scotopic data. Statistical analyses were performed using Stata 17 (StataCorp, College Station, TX). 

Results

Thirty-five participants (19 AMD cases and 16 healthy control subjects) were recruited. Demographic and clinical data of the participants are shown in Table 1. The two study groups were comparable in gender, age, BMI, smoking status, and BCVA. None of the participants had a history of cardiovascular diseases or diabetes. Compared to the control eyes, AMD eyes had reduced VAD in the choroid and choriocapillaris but not in the superficial or deep plexus (Table 1). 

Table 1.

 

Demographic and Clinical Data of the Participants

 

To examine variability in VAD between OCTA scans, the VADs of the two cOCTA scans (without flicker stimulation) in healthy control eyes were compared. There was no difference in VAD between two intrasession cOCTA scans in any of the layers examined, and the difference in VAD between two cOCTA scans was <2% for all layers (Fig. 2). Representative images showing retinal and choroidal vascular reactivity in response to flicker stimulation in a normal eye and in an eye with large drusen are shown in Figures 3 and 4, respectively. In the normal eye, flicker stimulation induced an overall vessel dilation in the superficial plexus (mean VRI = 10.1%), deep plexus (mean VRI = 5.0%), and choriocapillaris (mean VRI = 6.2%), with the greatest VRI detected in the superficial plexus layer. In contrast, the choroidal vessels appeared to be unchanged in most areas; however, some areas appeared to show vasoconstriction (mean VRI = −1.3%) (Fig. 3). In the AMD eye, flicker-induced vasodilation was barely detectable in any of the layers, with a mean VRI of ≤2.3% (Fig. 4). 

Figure 2.

 

(A) VAD of two cOCTA scans (without flicker stimulation). (B) Difference in VAD between the two cOCTA scans. There were no differences in VAD between the two cOCTA scans in any of the layers examined. CH, choroid; CC, choriocapillaris; DP, deep capillary plexus; SP, superficial capillary plexus. Error bars represent 95% confidence intervals.

Figure 3.

 

Representative cOCTA and fOCTA images and the VRI from a normal eye. Flicker-light stimulation resulted in an overall vasodilation (positive VRI) in the superficial plexus, deep plexus, and choriocapillaris. The choroidal vasculature appeared to be largely unchanged but showed vasoconstriction (negative VRI) in some areas.

Figure 4.

 

Representative cOCTA and fOCTA images and the VRI from an eye with large drusen. Flicker-induced vasodilation was barely detectable in any of the layers.

The VADs and VRIs of the control subjects and AMD cases are shown in Figure 5. In healthy control eyes, flicker stimulation resulted in an increase in VAD (positive VRI) in the superficial plexus (P < 0.001) and deep plexus (P < 0.001) (Figs. 5A, 5C). The VAD was marginally increased in the choriocapillaris (P = 0.077) in response to flicker stimulation, but there was no flicker-induced VAD change in the choroid (P = 0.654). In AMD eyes, there was no significant change in VAD in response to flicker stimulation detected in any of the vascular layers examined (Figs. 5B, 5D). Note that, in normal eyes, the superficial capillary plexus had the greatest vascular reactivity (greatest VRI) in response to flickering light. The VADs of the superficial plexus of each participant obtained by cOCTA and fOCTA are shown in Figure 6. 

Figure 5.

 

(A–D) VAD and VRI values for each vascular layer in normal control eyes (A, C) and AMD eyes (B, D). In normal eyes, flicker stimulation resulted in a significant increase in VAD in the superficial plexus (SP) and deep plexus (DP), a marginal increase in VAD in the choriocapillaris (CC), and no change in VAD in the choroid (CH). In AMD eyes, no significant chance in VAD was detected in any vascular layers examined. Error bars represent 95% confidence intervals.

Figure 6.

 

VAD of the superficial capillary plexus for each participant obtained by cOCTA without flicker light stimulation and by fOCTA during flicker light stimulation.

Analysis using linear mixed models confirmed that the VRI reduction in AMD eyes remained significant even after adjusting for potential confounders such as age, gender, smoking, BCVA, and scotopic sensitivity (Tables 2, 3). In model 1, using the data of the entire cohort (Table 2), compared to healthy control eyes AMD eyes were independently associated with a reduction in vascular reactivity in the superficial plexus (P < 0.001) and deep plexus (P = 0.002) but not the choriocapillaris (P = 0.090) or the choroid (P = 0.883) when using the mean VRI data. When using individual VRI data at each defined ROI, AMD eyes were independently associated with a reduction in vascular reactivity in the superficial plexus (P < 0.001), deep plexus (P < 0.001), and choriocapillaris (P < 0.001), but not the choroid (P = 0.993). The scotopic sensitivity data were available for 15 AMD cases (79%) and 10 healthy control participants (63%). Similar findings were found in model 2 when the association between AMD and VRI was examined in a subgroup of participants with scotopic data (Table 3). 

Table 2.

 

Linear Mixed Model Analysis of the Association Between AMD and VRI in the Entire Cohort (Model 1)

 

Table 3.

 

Subgroup Analysis on the Association Between AMD and VRI in Participants With Scotopic Sensitivity Data (Model 2)

 

Discussion

In this study, we used fOCTA to investigate vascular reactivity in the retina and the choroid in both healthy control eyes and AMD eyes. In healthy eyes, we confirmed that flicker-induced vasodilation using the current stimulus conditions was present in both the superficial and deep retinal capillary plexus but was absent in the choroid and choriocapillaris. Importantly, we found that eyes with large drusen were associated with impaired neurovascular coupling in the superficial plexus and deep retinal plexus. These findings support the hypothesis that retinal blood supply regulation is altered in AMD. 

VAD and VRI are different parameters, and they may be affected differently. A reduction in VAD indicates a loss of microvasculature (structural changes), whereas a reduction in VRI indicates a loss of neurovascular coupling function (functional changes). Using an external flickering light source, we were able to elicit a robust retinal neurovascular coupling (functional hyperemia) response in healthy individuals. Indeed, we found that flickering light induced a significant increase in VAD (i.e., positive VRI, vasodilation) in the superficial plexus and deep retinal plexus. We found that the greatest vessel reactivity in response to flicker stimulation was in the superficial plexus, followed by the deep plexus. This finding is consistent with increased inner retinal oxygen usage during flicker stimulation.^43,44 Our findings regarding flicker-induced inner retinal responses without a choroidal response in healthy eyes are consistent with other clinical studies that have shown that diffuse flicker stimulation has little effect on choroidal blood flow,^45–47 which can be explained by the absence of autoregulation in the choroid.^24,25 Although we did not expect to see a response to flicker stimulation from the choroid, our results are strengthened by including the choroidal VRI parameter in our analysis to serve as important negative control data. We attempted to determine whether flickering might also modulate the choriocapillaris as retinal metabolic consumption is altered. However, the current conditions did not elicit a robust choriocapillaris response to flicker stimulation. Further studies employing different stimuli may be able to elicit a relationship between outer and inner retinal blood flow. 

In this study, we found that in eyes with large drusen there was impaired neurovascular coupling in both superficial plexus and deep retinal plexus. Critically, this impairment of neurovascular function was present at a stage when the structural vessel density of those layers remained normal in this group of participants. Although several studies have found vascular abnormality beyond the choroid in AMD, including systemic vascular diseases and reduced inner retinal blood flow velocity,^48–51 we observed impaired neurovascular coupling in the retina. Such a deficit would not be detectable using a simple measure of vessel density. 

Impairment of neurovascular coupling in AMD eyes could have resulted from reduced retinal function, as impaired neuronal activity would result in a smaller flicker-induced changes in oxygen usage. We found that the association between AMD and VRI remained significant even after adjusting for scotopic sensitivity and BCVA, suggesting that the retinal neurovascular deficit is not simply due to reduced neuronal activity. However, not every participant could complete the scotopic sensitivity testing and as such is a limitation of the current study. Furthermore, this study was not designed to examine the effect of local cone function on VRI at each ROI. Future studies could consider using localized measures of retinal function (e.g., microperimetry or multifocal electroretinography)^52,53 to explore the relationship between vascular reactivity and retinal function in small regions of interest. 

Although we found that impairment of neurovascular function is associated with AMD, the role of retinal neurovascular deficits in the development and progression of AMD remains unclear. It has been proposed that chronic hypoxia and metabolic stress in the retinal pigment epithelium could lead to a defect in autophagy and mitochondrial damage and potentially be contributing factors to the onset and/or progression of AMD.^54–56 Determining whether retinal neurovascular dysfunction is a contributing factor to the pathological process of AMD or is a result of its presence will require further longitudinal investigations. Future studies should also examine retinal neurovascular function in different AMD stages, drusen loads, and phenotypes, particularly in eyes with the reticular pseudodrusen phenotype, as reticular pseudodrusen have been shown to be associated with coexistent vascular diseases⁵⁰ and choroidal vascular changes.^57,58 Due to a small sample size and a narrow range of drusen load in our AMD cohort, it was not feasible to assess the association between drusen load and VRI in this study. However, the findings from this study provide a foundation for such studies. 

Smoking has been found to have an effect on OCTA retinal microvascular parameters in diabetic patients⁵⁹ and retinal microvascular responses to variations in blood pressure during phase IV of the Valsalva maneuver in healthy subjects.⁶⁰ Although these studies showed the important effect of smoking on retinal microvascular parameters and reactivity, they did not directly examine the effect of smoking on retinal neurovascular coupling, which was the focus of this study. Because smoking is a well-recognized risk factor for AMD, we were interested in investigating whether it also influences retinal neurovascular coupling. Our data showed that smoking was not associated with the VRI, suggesting that smoking has little effect on neurovascular coupling in this study. This is likely due to the similar proportion of smokers between the study groups, with most being past smokers (70% in the AMD group and 60% in the control group) in our study. 

A study has shown that higher blood pressure (BP) is associated with sparser retinal density in the deep capillary plexus but interestingly has no effect in the superficial plexus.⁶¹ Whether BP has any influence on retinal neurovascular coupling remains unknown. In this study, we included only subjects with no history of cardiovascular disease based on medical history questionnaires. Although we could not control for BP in the analysis due to the lack of BP data, we found the greatest vascular reactivity in the superficial plexus of normal eyes, where BP does not affect microvascular density. Thus, our finding on reduced vascular reactivity in the superficial capillary plexus in eyes with large drusen is unlikely to be confounded by BP. 

The study also has some technical limitations. First, despite efforts to achieve diffuse flicker stimulation to generate a uniform and maximal vascular reactivity within the area of the OCTA scan, it is likely that there were variations in retinal illuminance within the scanned area due to the directional nature of LEDs. This might have resulted in variations in vascular reactivity response across the tested area. Therefore, the local variations in vascular response between the subjects should be interpreted with caution. A more diffuse stimulus would be useful in this respect. Nevertheless, any potential variation should be consistent between AMD and healthy control subjects. Second, it is possible that the VRIs of the deep plexus and choriocapillaris were overestimated due to imperfect projection artifact removal. However, given that superficial vessel density was the same between groups, it is unlikely that projection artifacts had a major influence on differences in the VRI between AMD and control participants, particularly in the deep capillary plexus. 

In conclusion, fOCTA can be used to assess retinal neurovascular function. Eyes with large drusen are associated with impaired neurovascular coupling in the superficial plexus and deep plexus. Further studies are needed to explore the role of neurovascular function in AMD pathogenesis. 

Acknowledgments

The authors thank Luis Alarcon-Martinez, PhD, for his insightful discussions on neurovascular communication and mechanisms. 

Supported by the Macular Disease Foundation Australia (CDL); a National Health and Medical Research Council (NHMRC) Investigator Grant (1194667 to RHG); and a NHMRC Synergy Grant (1027624 to RHG). The Centre for Eye Research Australia receives operational infrastructure support from the Victorian Government. 

Disclosure: B. Bui, None; R.H. Guymer, None; W. Heriot, None; A. Metha, None; C.D. Luu, None