Swept-source optical coherence tomography angiography metrics of retinal ischaemic perivascular lesions in patients being evaluated for carotid artery stenosis and controls

Introduction

The microvasculature of the human macula is comprised of the interconnected superficial capillary plexus (SCP), intermediate capillary plexus (ICP) and deep capillary plexus (DCP).1–4 Sustained hypoxia of these plexuses can result in retinal atrophy, the severity of which depends on the magnitude and duration of the ischaemic injury.5–7 Moderate ischaemic insults may cause localised middle retinal disruption with a corresponding scotoma,5 a finding termed paracentral acute middle maculopathy (PAMM).8

Recent studies have found that PAMM could be localised to areas of the middle retina as small as 175 µm in diameter (‘micro-PAMM’)9 and have hypothesised that even in the absence of known ischaemia or scotoma, small middle retinal atrophic lesions may result from middle retinal hypoxia similarly or identically to PAMM. These studies have referred to such lesions as ‘asymptomatic [PAMM] with old focal lesions’,10 ‘resolved PAMM lesions’11 12 and ‘retinal ischaemic perivascular lesions (RIPLs)’.13 14 These studies have generally defined RIPLs, as we will refer to such lesions in this work, as focal inner nuclear layer (INL) thinning with associated outer nuclear layer (ONL) expansion and outer plexiform layer (OPL) displacement, regardless of scotoma presence (figure 1).

Request permission

RIPLs and PAMM display very similar qualitative optical coherence tomography (OCT) findings11 13 14 leading to the hypothesis that all RIPLs result specifically from PAMM. However, quantitative imaging parameters of RIPLs have not been reported, in contrast to those reported for PAMM. In this study, we sought to determine if quantitative OCT angiography (OCTA) metrics could demonstrate a local reduction in macular capillary plexus blood vessel linear density (VLD) or a change in vessel tortuosity (VT) at RIPLs. Such findings would add to current knowledge of the pathophysiology of RIPLs and their hypothesised relationship with PAMM. We performed our investigation in both patients being evaluated for carotid artery stenosis (CAS), who we expected to have a relatively large number of RIPLs, as RIPLs are prevalent in patients with cardiovascular disease, and controls.13

Materials and methods

This was a prospective cross-sectional study of consecutive patients being evaluated for CAS and age-matched and sex-matched controls. APOSTEL V.2.0 quantitative OCT reporting guidelines were used.15 The research adhered to the tenets of the Declaration of Helsinki and was conducted in accordance with the US Health Insurance Portability and Accountability Act. The funding source played no role in the design, conduct, analysis, or interpretation of the study, or the decision to publish.

We prospectively enrolled 25 consecutive CAS patients and 14 consecutive controls at 1 tertiary referral centre between September 2021 and February 2023. Written informed consent was obtained from all patients. All CAS patients presented to vascular surgery with symptoms attributable to CAS with carotid duplex ultrasound demonstrating intra-arterial plaques consistent with CAS. Control patients presented to ophthalmology for routine care. Each patient’s clinical history was reviewed, and patients with any documented ocular disease including glaucoma, macular oedema, diabetic retinopathy, intermediate or advanced dry age-related macular degeneration, wet age-related macular degeneration, central serous retinopathy, retinal vein occlusion and retinal artery occlusion were excluded from the analysis.

A colour fundus photo was obtained from each eye of each patient after pupillary dilation. Immediately afterwards, a single PLEX Elite 9000 swept-source OCTA system (Carl Zeiss Meditec, Dublin, California, USA) was used to obtain 6 mm × 6 mm macular OCTA volume scans centred at the fovea of each eye of each patient. APOSTEL V.2.0 reporting criteria for the OCTA scans are included as an online supplemental file 1.

Identification of RIPLs, defined as focal INL thinning with associated ONL expansion and displacement of the OPL without the presence of patient-reported scotoma, was performed on B scans manually by three independent graders with 82% inter-rater agreement, with an additional grader resolving intergrader queries. The OCTA manufacturer’s PLEX Elite 9000 software was used to automatically generate en face 1024-pixel × 1024-pixel images of each of the SCP and DCP, defined according to the ‘current OCTA nomenclature’ per Campbell et al,3 and to remove projection artefact from the DCP images. The foveal avascular zone (FAZ) was manually segmented in each en face reconstruction in Fiji ImageJ 2.9.0/1.53t Java 1.8.0_322 64 bit.16 The location of the centre of each identified RIPL was recorded on a copy of each reconstruction for visualisation (figure 2).

Request permission

Two methods, Fiji (‘algorithm 1’) and the manufacturer’s prototype research software (‘algorithm 2’), were used to threshold, binarize and skeletonize the en face SCP and DCP images to show the blood vessels as 1-pixel-wide lines. Fiji’s ‘Analyze Skeleton’ function was used to calculate the length of skeletonised vessels in pixels. VLD was then calculated for each en face image as ((length of vessels in pixels)/(surface area in pixels²))×1024 pixels/6 mm, giving units of per mm, as reported previously.17 18 VT was similarly examined with Fiji’s ‘Analyze Skeleton’ function by dividing the summed total length of all skeletonized blood vessels in pixels by the summed Euclidean distances between consecutive branch points and/or endpoints in pixels, giving a dimensionless ratio as previously described.19 The FAZ was excluded from all VLD and VT calculations. SCP and DCP en face images were also segmented in Fiji into grids of 32-pixel × 32-pixel squares, from which VLD and VT were calculated as described above. VLD values and the location of each square containing the centre of a RIPL (hereafter ‘each square containing a RIPL’) were plotted on a 2D grid for visualisation with Python V.3.10.5 (Python Software Foundation, Wilmington, Delaware, USA) (figure 3).

Request permission

For all analyses, RIPLs located in or immediately adjacent to squares overlapping the FAZ or in or immediately adjacent to quadrants containing vitreous floaters were excluded. Quadrants containing vitreous floaters were excluded from entire-eye VLD and VT calculations. Using Python, the VLD of each square containing an included RIPL was compared with the average VLD of the up to eight immediately adjacent squares with a paired t-test for each of the SCP and DCP (‘RIPL vs immediate surroundings’). The average VLD value of all squares containing an included RIPL in each eye with at least one included RIPL was compared with the average VLD of the entire eye with a paired t-test (‘RIPL(s) vs total macula’). For all patients with one eye with at least one included RIPL and one eye without any RIPLs, the average VLD of fellow eyes was compared with a paired t-test. The average VLD of all eyes with at least one included RIPL was compared with the average VLD of all eyes without any RIPLs with an unpaired t-test. Similar comparisons were performed for VT. All analyses were performed independently for each of the CAS and control groups, for each of the SCP and the DCP, for each of algorithms 1 and 2. A 95% CI was used for each statistical analysis.

Results

This study included 22 of 25 patients with CAS and 14 of 14 controls. Three of twenty-five patients with CAS were excluded: two for retinal vein occlusion and one for diabetic retinopathy. Two eyes were included for each patient. Clinical characteristics are summarised in table 1 and were no different between groups.

CAS patient ages ranged from 48 to 84 years (median 73.5). In total, 20 of 22 patients with CAS (91%) had at least 1 RIPL. A total of 30 eyes with RIPLs were identified in these 20 patients. Patients with CAS displayed a mean of 3.4 RIPLs total. Overall, 45% of patients with CAS displayed RIPLs in both eyes, 45% RIPLs in exactly 1 eye and 9.1% RIPLs in neither eye. Control patient ages ranged from 55 to 80 years (median 70.5). In total, 10 of 14 controls (71%) had at least 1 RIPL. A total of 15 eyes with RIPLs were identified in these 10 controls. Controls displayed a mean of 2.0 RIPLs total. Overall, 36% of controls displayed RIPLs in both eyes, 36% RIPLs in exactly 1 eye and 29% RIPLs in neither eye.

In CAS patients, total 75 RIPLs were identified and 38 were excluded (3 near FAZ; 35 in quadrant with floaters). The 37 included RIPLs were distributed among 17 eyes. Six CAS patients had one eye with at least one included RIPL and one eye without any RIPLs.

In controls, 28 total RIPLs were identified. No RIPLs were located near the FAZ. Seven RIPLs were excluded for being located or in or adjacent to quadrants with vitreous floaters. The 21 included RIPLs were distributed among 12 eyes. Four controls had one eye with at least one included RIPL and one eye without any RIPLs.

The results for VLD are presented in table 2. Of note, DCP VLD displayed a significant decrease at RIPLs in many comparisons in CAS patients and controls. Further, SCP VLD at RIPLs displayed a significant decrease in one comparison for CAS patients and a significant increase in one comparison for controls. No comparisons between eyes with RIPLs and eyes without RIPLs were significant.

No changes in VT at RIPLs were noted relative to either the immediate surroundings or total macula in the SCP or DCP in CAS patients or controls by either algorithm (all p>0.05). Similarly, no changes in VT were noted between eyes with RIPLs and eyes without RIPLs in any comparisons performed (all p>0.05).

Discussion

In our study, RIPLs were found to be prevalent in the retinas of patients with CAS (mean of 3.4 RIPLs per patient), agreeing with prior findings that RIPLs occur frequently in individuals with cardiovascular disease (2.8 RIPLs per patient).13 We observed that only 9.1% of CAS patients had no RIPLs in either eye. This is lower than the 44% of individuals with cardiovascular disease previously reported not to have RIPLs in either eye,13 which may be a consequence of CAS patients having more severe disease than the average cardiovascular disease patient. Control patients had 2.0 RIPLs per patient on average and 29% had no RIPLs in either eye, potentially reflecting lower rates of cardiovascular disease in controls than in CAS patients.

Our work also demonstrates that in both CAS patients and controls, RIPLs correlate with local DCP capillary loss on OCTA, and RIPLs are not associated with changes in retinal capillary tortuosity. This is similar to prior results for PAMM lesions, in which a statistically significant decrease in DCP VLD, but not SCP VLD, was observed in eyes with PAMM lesions relative to their healthy fellow eyes.7 Our results similarly demonstrate a statistically significant decrease in DCP VLD of RIPLs relative to the entire DCP of the same eye in CAS patients and a similar trend in controls. However, the lack of differences in VLD between eyes with RIPLs and their RIPL free fellow eyes in CAS patients and controls suggests that while retinal ischaemia occurs throughout the entire eye in eyes with classic PAMM lesions, it occurs only locally in eyes with RIPLs.

Interestingly, RIPLs tended to demonstrate a localised decrease in SCP VLD in CAS patients, but a localised increase in SCP VLD in controls. We hypothesise that this finding may represent an enhanced capacity of the SCP to respond to local DCP ischaemia in controls relative to CAS patients. That the DCP VLD demonstrated a smaller local mean decrease in controls than in CAS patients may reflect this hypothesised enhanced ability of the SCP to respond to DCP ischaemia in controls.

Our comparison of mean VLD between all eyes with RIPLs and all eyes without RIPLs did not demonstrate significant differences in SCP or DCP VLD in CAS patients or controls, suggesting that RIPLs are not associated with global retinal ischaemia, unlike classic PAMM lesions; however, this interpretation is confounded by possible interpatient differences in mean VLD. A recent report of ‘micro-PAMM’,9 in which retinal OCT demonstrated a 175 µm diameter region of increased reflectivity in the INL correlating to decreased flow on OCTA and a positive scotoma in an eye with unremarkable funduscopy, and which resolved within 2 weeks without characteristics findings of a RIPL at that time, supports the possibility that RIPLs are a localised form of PAMM lesion toward the less severe end of an ischaemic spectrum. It is possible that even in this case of ‘micro-PAMM’ a RIPL might develop, as PAMM has been noted to cause PAMM lesions even after 3–4 weeks.20

Our finding of unchanged VT in any comparison supports the idea that RIPLs reflect chronic lesions similar to PAMM lesions. Prior case reports of PAMM have demonstrated that qualitative changes in tortuosity in PAMM generally follow a pattern of occasional acute increases in tortuosity21 22 that sometimes progressively worsen23 but that eventually improve.24

Our study has several limitations that should be considered when interpreting its results. Our work does not address VLD changes at RIPLs in patients with retinal diseases, limiting the generalisability of our data. Furthermore, excluding patients with a history of retinal vein occlusion removed patients with a potentially greater number and severity of RIPLs, given an established relationship between retinal vein occlusion and development of PAMM lesions and the likelihood of a similar ischaemic pathogenesis underlying PAMM lesions and RIPLs. The low number of patients with one eye with at least one included RIPL and a paired eye without any RIPLs may mean analysis of such fellow eyes is underpowered. Despite these limitations, our study has strengths including automated calculation of OCTA parameters allowing for high reliability of results, investigation of both local and global changes in OCTA parameters allowing for determination of geographic extent of VLD changes, a relatively strict quality cut-off for OCTA images with correction for projection of SCP vessels onto the DCP to minimise potential artefact, exclusion of RIPLs near the FAZ or in or adjacent to quadrants with vitreous floaters to eliminate sources of confounding and use of two separate algorithms to assess reproducibility of the presented results.

In conclusion, OCTA image analysis demonstrated a quantitative local decrease in foveal DCP VLD at RIPLs, but no global decrease in VLD in eyes with RIPLs, in patients with CAS and controls. Our findings provide quantitative evidence that RIPLs are associated with local middle retinal microvasculature loss. This suggests that RIPLs share pathophysiology with classic PAMM lesions, supporting a position for RIPLs at the less severe end of an ischaemic continuum that includes classic PAMM lesions. Finally, the high prevalence of RIPLs here reported for patients with CAS underscores the emerging role of retinal microvascular biomarkers in the evaluation of systemic macrovascular disease.