The Application of Optical Coherence Tomography Angiography in Cerebral Small Vessel Disease, Ischemic Stroke, and Dementia: A Systematic Review

Objective: To investigate the application of optical coherence tomography angiography (OCTA) in cerebral small vessel disease (SVD), ischemic stroke and dementia. Methods: We conducted a systematic search in MEDLINE (from inception) and EMBASE (from 1980) to end 2019 for human studies that measured retinal parameters in cerebral SVD, ischemic stroke, and dementia using OCTA. Results: Fourteen articles (n = 989) provided relevant data. Ten studies included patients with Alzheimer disease (AD) and mild cognitive impairment (n = 679), two investigated pre-symptomatic AD participants (n = 154), and two investigated monogenic SVD patients with cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (n = 32) and Fabry disease (n = 124). Methods to reduce bias and risk factor adjustment were poorly reported. Substantial methodological variations between studies precluded a formal meta-analysis. Quantitative measurements revealed significant yet inconclusive changes in foveal avascular zone, perfusion density, and vessel density (VD) in AD, presymptomatic AD, and SVD patients. Two (n = 160) of three studies (n = 192) showed association between decreased VD and increased white matter hyperintensities. In three (n = 297) of seven studies (n = 563), better cognitive function was associated with increased VD. One study (n = 52) suggested increased VD was associated with increased ganglion cell–inner plexiform layer thickness in AD yet with no covariate adjustment. Conclusions: Changes in retinal microvasculature identified using OCTA are associated with monogenic SVD and different stages of AD, but data are limited and partly confounded by methodological differences. Larger studies with risk factors adjustment and more consistent OCTA methods are needed to fully exploit this technology. PROSPERO registration number: CRD42020166929.


INTRODUCTION
Cerebral small vessel disease (SVD) is an intrinsic disorder that affects the brain's small perforating arterioles, capillaries, and probably venules, causing various lesions seen on pathological examination or brain imaging with magnetic resonance imaging (MRI) or computed tomography (1). Cerebral SVD is a common condition in older adults and causes a wide array of clinical syndromes, including 25% of ischemic strokes and 80% of intracerebral hemorrhages, and contributes to up to 50% of dementias worldwide (1)(2)(3).
However, the changes in the cerebral microcirculation are difficult to visualize in vivo. Because the retinal microvasculature and cerebral microvasculature share similar embryologic origins as well as anatomical and physiological properties (4), investigating the network of retinal vessels may provide new insights into cerebrovascular disease and the vascular contribution to pathologic features of dementia (5). Previous literature has mainly focused on retinal fundus photography or optical coherence tomography (OCT) (5,6).
A newly presented extension of structural OCT, referred to as optical coherence tomography angiography (OCTA), represents a novel non-invasive, depth-selective modality that allows for visualization of retinal blood flow without dye injection (7). OCTA images are essentially motion-contrast images, which are based on the different backscattering of light between red blood cells and neurosensory tissue, because red blood cells are moving while neurosensory tissue is static (8). In many cases, these images are now approaching histology-level resolution (8), raising great interest in the possibility of detecting early microvascular pathological changes in people in vivo. A comprehensive review of the technology, its principles, limitations, and the clinical application in eye disease has been published (8), work that we extend here into the arena of neurological disease. OCTA may permit detection of a series of changes such as reduction in capillary vessels and perfusion density before they are visible on retinal photographs (9). Moreover, recent work suggests that highresolution OCTA images combined with machine learning may provide great potential to automate detection and quantification of microvascular changes underlying common brain diseases such as SVD, stroke, and dementia, in the future (10,11).
In this study, we aimed to conduct a systematic review of the literature to examine the application of OCTA in cerebral SVD, ischemic stroke, and dementia, including Alzheimer disease (AD), vascular dementia, frontotemporal dementia, and dementia with Lewy bodies. Specifically, our primary objective is to find out whether there are any specific changes seen in patients with cerebral SVD, ischemic stroke, and/or dementia using OCTA when compared with healthy individuals. Our secondary objective is to assess the association between OCTA changes and brain imaging findings, cognitive functions, and other retinal parameters using fundus camera imaging or standard OCT.

METHODS
This systematic review was based on a predefined protocol (PROSPERO registration no. CRD42020166929) following the Preferred Reporting Items for Systematic Reviews and Metaanalyses (12) guidelines. We refer to SVD MRI findings according to STRIVE (Standards for Reporting Vascular Changes on Neuroimaging) guidelines (13). As all analyses here are based on publicly available summary statistics and not individuallevel data, no ethical approval or informed patient consent was required.

Search Strategy
We conducted a literature search of the Medical Literature Analysis and Retrieval System Online (MEDLINE, from inception) and the Excerpta Medica Database (EMBASE, from 1980) up to December 31, 2019, using the Ovid Web Gateway (detailed search strategy in Supplementary Material). We considered only studies published in English. References of relevant articles were hand-searched, and a forward citation search was performed to identify further studies.

Inclusion and Exclusion Criteria
The studies included patients with cerebral SVD, ischemic stroke, and/or any type of dementia and compared either (1) retinal parameters using OCTA vs. control, (2) OCTA parameters and brain imaging findings, (3) OCTA parameters and cognitive function, or (4) OCTA parameters and findings on fundus camera imaging or OCT.
The following studies were excluded: (1) duplicate publications or studies not meeting the inclusion criteria, (2) review studies, (3) single-case reports, (4) nonhuman studies, (5) non-English language studies, (6) conference presentations or summaries, (7) studies without details of diagnosis criteria or definitions, and (8) studies that despite examining retinal parameters did not conduct or report results on OCTA.

Data Extraction
We initially screened all studies identified in the systematic search by abstract and title for potentially relevant articles. Duplicate articles were removed, and the remaining articles were assessed for eligibility after full-text review. Data extracted from these studies included title; authors; publication year; study aim; study type; number of patients and controls; number of male and female, mean age, participant selection criteria, and diagnostic criteria; type of OCTA device; assessment of OCTA image quality; whether OCTA associations were adjusted for demographics, other risk factors, or other covariates; and outcomes. All data were cross-checked by a second reviewer. We report covariateadjusted associations when available.

Quality Assessment
We used the STROBE checklist (www.equator-network.org) to score study methodology and assess study quality. We assigned up to 22 points using this checklist. This score was applied after study inclusion and did not influence whether a study was included in the review.
Owing to the small number of studies per category and the large methodological heterogeneity, it was not possible to combine study data by meta-analysis.

RESULTS
The search returned 2,611 articles. One additional article was found by hand-search. Among them, 621 duplicate articles were removed. The remaining 1,991 articles were screened by title and abstract only. Of those, 16 were considered potentially relevant and were assessed by full-text review. Figure 1 details numbers of articles excluded and the reasons.
Fourteen articles met the inclusion criteria. The population samples came from the United States (five articles), the Netherlands (two articles), Germany (two articles), Italy (two articles), Poland (one article), Turkey (one article), and China (one article). Multiple articles from the same study population were included only if different retinal parameters or outcomes were measured in separate articles. Table 1 summarizes the study design and patient characteristics from the articles included. The 14 articles (14-27) included comprised 13 cross-sectional studies and 1 case-control study. Although two articles (18,19) had overlapping samples, each study assessed different outcomes. Across the included studies (if two articles had an overlapping study population, the study with the larger sample was chosen), the total number of unique participants was 989. Participants in each study varied from 16 to 213 (median n = 55).

Quality Assessment
The median study quality score as per STROBE checklist (www. equator-network.org) was 16/22 (Supplementary Table 1). Scores were mainly affected by inadequate description of study design in title or abstract, not explaining rationale for study size, inadequate description of efforts made to control potential sources of bias, and inadequate discussion of generalizability of results.  Table 2). Therefore, the results from those studies are not all interchangeable (28). See Table 3 for detailed explanation of these retinal parameters.

Main Analysis
OCTA Parameters in Patients vs. Controls   (16,18); one presymptomatic AD study (24); one CADASIL study (26)] showed no difference between patient and control groups. Among these studies, one study matched for age and sex (18). Two studies (16,24) adjusted for age, one (16) of which also adjusted for spherical equivalent, and quality factor. The other study (26) did not adjust for any covariates.
Twelve studies (14-18, 20-24, 26, 27) compared the vascular density parameters (vessel density, vessel length density, perfusion density, etc.) between patients and controls. From these, eight studies [n = 569, median n = 53, seven AD and MCI studies (14,15,18,(20)(21)(22)(23); one CADASIL study (26)] suggested decreased vascular density parameters in patients compared with controls. Among them, two studies (20, 21) did not adjust for any covariates, one study (18) adjusted the results for age and sex, and in four studies (14,15,23,26), sample groups were matched for age, three of which (14,22,23) also matched for sex and one of them (23) also matched for race. However, two studies [n = 56 and n = 86, involving AD and MCI patients (16,17)] showed no difference between patient and control groups. Interestingly, one study in presymptomatic AD (n = 124) (24) found Aβ+ participants had higher vessel density than Aβ-participants after adjusting for age. Another study (n = 124) (27) found Fabry disease patients had decreased vascular density in superficial capillary plexus and increased in deep capillary plexus compared with controls after adjusting for age and sex.

OCTA Parameters in Relation to Brain Imaging Findings
Only four studies [n = 208, median n = 53, three in AD and MCI (15,16,19) and one in CADASIL (26)] investigated the relationship between OCTA findings and brain imaging findings. Two of them (n = 74 and n = 86) (15,16) involving AD and MCI participants found increased vessel density (VD) was associated with having fewer white matter hyperintensities (WMHs), as per Fazekas scores. Between these two studies, one (15) matched for age, and the other (16) adjusted for age and sex. One study (n = 16) (19) involving AD and MCI patients suggested a potential association between decreased VD and decreased perfusion density with increased volume of the inferior lateral ventricle (temporal horn), but it did not control for covariates, and patients and controls were not matched by age and/or gender. The study (n = 32) (26) in CADASIL patients found no association between OCTA parameters and the number of small infarcts or WMHs, the latter estimated using Fazekas scores.

OCTA Findings in Relation to Cognitive Function
Seven (14-16, 18, 20, 21, 23) studies investigated the relationship between OCTA parameters and cognitive function in AD and MCI participants. Three studies (n = 297) (14,18,23) suggested better cognitive function was associated with increased VD. One study (n = 52) (14) found enlarged FAZ area was associated with decreased Mini-Mental State Examination (MMSE) scores. Of these three studies, two (14, 23) were age-and sex-matched, one of which (23) was also race-matched. One study (18) did not adjust or match for covariates. However, four other studies (n = 266, median n = 64) (15,16,20,21) found no association between OCTA parameters and cognitive functions. Among them, one

OCTA retinal parameters Descriptions Study
Foveal avascular zone (FAZ) (mm 2 ) The central avascular region in human foveola is known as the foveal avascular zone (FAZ) (14, 16, 18-20, 22, 24-26) Perfusion density (PD) (%) In most Carl Zeiss devices, PD (%) was defined as the total area of perfused retinal microvasculature per unit area in a region of measurement (17)(18)(19) Vessel density or vascular density (VD) In most Optovue devices, VD (%) has the similar definition of PD in Carl Zeiss device (14,15,20,22,23,26,27) In most Carl Zeiss devices, VD (/mm) was defined as the total length of perfused retinal microvasculature per unit area in the region of measurement (16,18,19,24) VD (Dvox) is analyzed using fractal analysis with box counting (Dbox) (21) Vessel length density (VLD) VLD (%) (using Optovue device) represents the ratio of the total length occupied by all the blood vessels in a given area to the total area. All vessels were then skeletonized into 1-pixel-wide vessels The proportion of absent flow signal accounted for by signal voids >10,000 µm 2 (FV10000) and signal voids >40,000 µm 2 (FV40000) were evaluated (26) study (16) found no association, both adjusted and unadjusted for age and sex, whereas the other three studies (15, 20, 21) did not match or adjust for covariates.

OCTA Parameters and Other OCT/Fundus Imaging Parameters
Only one study (n = 52) (21) explored the relationship between OCTA features and OCT findings in patients with AD and found that increased VD was associated with increased ganglion cellinner plexiform layer (GC-IPL) thickness. However, this study did not adjust or match for any covariates. None of the studies included investigated the relationship between OCTA parameters and traditional fundus imaging.

DISCUSSION
This systematic review found 14 studies investigating OCTA in patients with cerebral SVD or dementia, published between 2017 and 2020. These studies mainly focused on AD and MCI, pre-symptomatic AD, and monogenic SVD including CADASIL and Fabry disease. No other dementias, ischemic stroke, or sporadic SVD studies met the inclusion criteria. There were no longitudinal studies. Collectively, these studies do not provide a clear conclusion of whether the retinal microvasculature is impaired in cerebral SVD, ischemic stroke, and dementia patients or not. FAZ is a capillary-free area in the foveal zone susceptible to ischemia (29). Enlargement of FAZ is a sign of ischemia (14). The analysis of the FAZ measurements yielded inconclusive results. Four studies (n = 196) (14,20,22,25) found increased FAZ in patients compared with controls, whereas four other studies (n = 455) (16,18,24,26) showed no difference between groups. It is possible that the relatively small sample size of some studies precludes detection of group differences. Another possibility is that some studies included patients in the earlier stage of the disease process, which might explain why no FAZ difference was found between groups (24). However, as ophthalmological confounders, quality factors, and other possible risk factor confounders including age were not always taken into account, studies reporting group differences may represent an overestimation of true disease effects (16). Furthermore, significant variation in FAZ area in normal eyes has been reported that may be associated with gender, central retinal thickness, etc. (30). Therefore, studies with larger sample size and good matching of potential confounders are needed to further verify these findings. Also, the lack of longitudinal studies means that it is difficult to know if the early OCTA changes could predate worsening neurological state, and longitudinal studies are urgently needed.
The decreased vascular density (including parameters such as VD, perfusion density, vessel length density, etc.) is another sign of microvasculature impairment. Eight studies (n = 569, seven with AD and MCI and one with CADASIL) (14,15,18,(20)(21)(22)(23)26) suggested decreased vascular density parameters in patients compared with controls. However, two studies (n = 56-86) (16,17) showed no difference between groups. Different sample size, disease stage, and confounder adjustment might explain the disparity in results. Interestingly, van de Kreeke et al. (n = 124) (24) found Aβ+ participants had higher VD than Aβ-participants in pre-symptomatic AD. Here, early amyloid accumulation in the retina might induce an inflammatory reaction with hypoxia, leading to increased retinal blood flow (24). Larger studies including those following patients longitudinally will help explore retinal vessel changes. Moreover, Cennamo et al. (n = 124) (27) found Fabry disease patients had decreased VD in superficial capillary plexus and increased VD in deep capillary plexus compared with the control group. The authors considered the increased VD in deep capillary plexus as a compensatory mechanism to support the reduced VD in superficial capillary plexus (27). We further explored the relationships between OCTA and brain imaging findings. Currently, cerebral small vessels are hard to visualize in vivo. Other imaging features are provided as SVD biomarkers such as WMHs, microbleeds, enlarged perivascular spaces, and lacunes (1). Because the retinal vasculature is a proxy of cerebral vessels (4), establishing the relationship between retinal microvasculature and cerebral imaging findings could aid in our understanding of disease mechanisms. This systematic review found only four studies (15,16,19,26) that investigated the relationship between OCTA parameters and brain imaging findings of SVD. Two studies (n = 74-86) (15,16) found decreased VD was associated with more WMHs, whereas another study (n = 32) (26) found no association between OCTA parameters and WMHs or the number of small infarcts, which might be due to small sample size. MRI changes, such as atrophy of the brain and consequential ventricular enlargement, can be seen in AD or SVD patients (31). A pilot study (n = 16) (19) suggested the potential association between decreased VD and perfusion density with increased volume of inferior lateral ventricle (temporal horn). These small sample studies show a promising role of OCTA to investigate SVD in vivo. A prospective cross-sectional study (32) published recently when our work was under review investigating the amyloid-positive AD-related cognitive impairment (n = 28), subcortical vascular cognitive impairment (n = 18) patients, and amyloid-negative cognitively normal subjects (n = 14) found that VD was negatively correlated with SVD scores and suggested OCTA as a potential imaging tool to screen for the degree of SVD. Further studies in larger samples comparing a range of SVD brain imaging biomarkers with OCTA parameters are needed to confirm relationships and explore the underlying pathophysiological mechanisms in SVD.
Previous studies reported that the changes in retinal vessel caliber on fundus imaging were associated with cognitive decline (5). This systematic review found seven studies investigated the relationship between OCTA findings and cognitive function. Among them, three studies (n = 297) (14,18,23) suggested better cognitive function was associated with higher VD. One study (n = 52) (14) found enlarged FAZ area was associated with decreased MMSE score. However, four other studies (n = 266) (15,16,20,21) found no association between OCTA parameters and cognitive function. A cross-sectional study (33) in 27 amnestic mild cognitive impaired patients and 29 controls published after completing this review did not find any correlation between MMSE scores and OCTA parameters. Contradictory results, relatively small samples with limited or no adjustment for covariates, and no long-term follow-up of the studies that suggest OCTA parameters as potential predictors of cognitive decline point to the necessity of larger, longitudinal studies in this area of research.
We found only one study (n = 52) (21) investigating the relationship between OCTA features and OCT findings that suggested increased VD was associated with increased GC-IPL thickness in AD patients. However, this study did not adjust for covariates. The optic nerve and retina develop as a direct extension of the diencephalon during embryonic development (4). OCT is a non-invasive method and offers high-resolution images of the retinal structure, including the neuronal layers (6). Exploring the relationship between OCT and OCTA findings especially in longitudinal studies will not only provide evidence of the temporal changes between retinal nerve layers and retinal microvasculature, but also offer potential predictors in disease progression.
We did not find any study investigating the relationship between OCTA parameters and fundus imaging. Retinal arterioles and venules typically measure 15 to 150 µm in size, whereas the capillaries measure 5 to 15 µm (34). There is an extensive literature studying the retinal vessel changes in SVD, ischemic stroke, and dementia using fundus imaging (5,35). However, recent fundus imaging can only measure relatively large retinal vessels, whereas OCTA provides high-resolution, depth-resolved information that has been available only since 2016 (8). The relationships between The methodological differences limit between-study comparisons and interpretation. The variations in the different device types, software and hardware, macular scan size, segmentation methods, and retinal parameters all add to heterogeneity. Other potentially confounding factors such as age, race, and baseline status including diagnosis of diabetes are inevitable. The asymmetry between eyes of the same individual should be considered because it might cause potential misleading results when we choose one eye as a proxy for both eyes (36). On the contrary, studies utilizing both eyes of participants should address the statistical challenges cautiously because the pair of eyes are non-independent (36). OCTA scan quality has impacts on derived parameters, and researchers should perform strict quality control to ensure good data quality, and thus results are reliable and comparable. Patients with more severe disease status may have fixation problems, which may cause poor scan quality (14). Lastly, studies reviewed here were cross-sectional, and no conclusions can be drawn on the value of OCTA measurements for evaluation of disease progression.

CONCLUSIONS
The findings of this review show promise that changes in retinal microvasculature identified using OCTA are associated with monogenic SVD and different stages of AD. Larger studies with risk factors adjustment and more consistent OCTA methods are needed to fully exploit this technology.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.