The purpose of this study was to describe the clinical features of epivascular glia (EVG) using en face optical coherence tomography (OCT).
Retrospective cross-sectional study.
Single-institution en face OCT images were reviewed. Eyes displaying EVG were captured with manual internal limiting membrane (ILM) segmentation and analyzed with customized segmentation . A random age- and sex-matched control group was selected to determine relative epiretinal membrane (ERM) prevalence.
Characteristic hyper-reflective ILM plaques with dendrite-like radiations were identified using en face OCT and displayed vascular predilection. A total of 161 eyes with EVG (the EVG group) and 2,315 eyes without EVG (control group) were identified from a total cohort of 1,298 patients (or 2,476 eyes). The prevalence of EVG was 161 of 2,476 eyes (6.5%) and 119 of 1,298 patients (9.2%) in the cohort. Mean age was 79.3 ± 10.7 years old in the EVG group and 55.9 ± 24.6 years old in the control group ( P <.001). An advanced posterior vitreous detachment (PVD) stage was more common in the EVG group (grade 3: 41.7%; grade 4: 48.6%) than in the control group (grade 3: 18.5%; grade 4: 26.9%; P <.001). Contractile ERM was present in 71 of 161 eyes (44.1%) with EVG compared to 30 of 161 eyes (18.6%) in a random age- and sex-matched control cohort without EVG ( P <.001).
EVG previously described with histopathology and scanning electron microscopy can be identified using en face OCT. In this study, these lesions were associated with older age, pseudophakia, and advanced PVD, supporting the role of Müller cell activation through ILM breaks triggered by PVD, a pathogenic mechanism proposed by previous studies.
Although the pathogenesis of epiretinal membrane (ERM) formation was first identified by Iwanoff in 1865, ERMs remain poorly understood. Early histopathologic studies found that idiopathic ERMs consisted of cells that were morphologically consistent with glial cells, specifically Müller glia or astrocytes. Foos hypothesized that these glial cells migrated through microbreaks in the internal limiting membrane (ILM) caused by the development of posterior vitreous detachment (PVD). This process was noted most commonly in the regions adjacent to the large superficial retinal vessels and the lesions were referred to as epivascular glia (EVG). , It has been proposed that EVG may provide the structural scaffold for the later development of ERMs. Using scanning electron microscopy in autopsy eyes, Kishi and associates were able to identify and describe EVG, which displayed a very characteristic plaque-like appearance with a fine filamentous stellate and dendriform network of cilia-like radial extensions.
Despite the importance of EVG in the proposed mechanism of ERM pathogenesis, the identification and description of these lesions in the era of advanced retinal imaging, including optical coherence tomography (OCT), has been lacking. The advent of more sophisticated OCT systems with rapid A scan acquisition and high-density volume datasets has facilitated the development of en face OCT, which can reconstruct cross-sectional OCT volume datasets into a coronal or en face plane at specified segmentation levels. This technology has provided great insights into the pathoanatomy of the fovea and into various OCT disorders including perivenular paracentral acute middle maculopathy.
The present study used en face OCT to identify characteristic lesions at the level of the ILM that displayed features remarkably similar to those of Kishi and associates’ original electron microscopy descriptions of EVG. The prevalence and risk factors for development of EVG and the characteristic features of those lesions were studied in a large cohort of patients who underwent en face OCT.
This study was a retrospective cross-sectional analysis. Internal Review Board (IRB) approval was obtained at the University of California Los Angeles (UCLA), and all aspects of the study adhered to the tenets of the Declaration of Helsinki and Health Insurance Portability and Accountability Act (HIPAA) regulations.
Inclusion and Exclusion Criteria
The authors reviewed the entire UCLA Optovue database of patients (1,848 patients and 3,696 eyes) who underwent en face OCT angiography (OCTA) to identify en face OCT cases of EVG. In total, en face OCT scans from 1,298 patients (and 2,476 eyes), imaged using the Optovue device (Avanti RTVue XR; Optovue, Fremont, California, USA) between January 1, 2014, and August 30, 2019, as part of the standard retinal evaluation (with D.S.) at the Stein Eye Institute, were included in the initial review. Inclusion criteria were en face OCT scans of sufficient quality (quality index of 4 or greater) with signal strength index (SSI) of 50 or greater. Eyes were excluded if they possessed any retinal pathology that would predispose them to the development of secondary ERM including any history of retinal procedures (cryotherapy, laser photocoagulation, or photodynamic therapy) or retinal surgery (eg, pars plana vitrectomy) and any evidence of retinal vein occlusion, diabetic retinopathy, retinitis pigmentosa, tumor, uveitis, trauma, peripheral retinal holes, retinal tears, retinal detachments, and choroidal hemorrhages. Scans with blinking or motion artifacts limiting evaluation and analysis of EVG were excluded. A thorough chart review was performed to determine the retinal diagnosis for exclusion. Relevant demographic information, specifically age, sex, and phakic status, were collected.
Image Acquisition and Protocols
All eyes underwent imaging using spectral domain-OCT (SD-OCT) and OCTA (Avanti RTVue XR; Optovue; or Spectralis HRA + OCT; Heidelberg Engineering, Germany). Patients underwent both high-definition and non-high-definition 6 × 6-mm volume scans using the Optovue device to produce both registered cross-sectional B-scan and en face OCT. Those images were analyzed for the presence and localization of EVG and ERMs.
Presence and grading of PVD was performed on the basis of evaluation of OCT B-scans through the fovea and optic discs acquired from both the Optovue and Heidelberg systems. In some cases, additional B-scan imaging was performed using the Topcon system (Topcon 50DX; Topcon, Inc., Oakland, New Jersey, USA) to confirm PVD status.
All scans were initially evaluated for the presence of EVG, ERM, and presence and grade of PVD by the first author (C.G.). Presence of the EVG was confirmed by a second independent grader (D.W., F.G., K.T., A.A.). Presence of ERM and presence and grade of PVD were confirmed by a second independent grader (D.W., F.G., A.A.). Any discrepancies among graders were adjudicated by the senior investigator (D.S.).
Definitions and Grading Scheme of EVG, Contractile ERM, and PVD
Optimal identification of EVG was performed using customized segmentation aligned to the ILM with a 9-µm offset. The lower segmentation border was aligned 3 µm anterior to the retinal nerve fiber layer (RNFL), and the upper segmentation border was positioned 9 µm anterior to the lower segmentation line toward the vitreous side. Any abnormal segmentations were manually corrected. This customized slab segmentation of 9-µm thickness was chosen to reduce the light-reflecting signal from the RNFL on en face OCT images and to optimally highlight the EVG lesion.
The presence of a hyper-reflective band, corresponding to the EVG, at the innermost level of the retina was confirmed by using the registered cross-sectional OCT B-scan.
By contrast, a contractile ERM was characterized by the presence of epiretinal membrane causing tangential retinal traction. Identification of a contractile ERM was defined as a hyper-reflective lesion anterior to the RNFL on OCT B-scan showing hyporeflective spaces between the lesion and the inner retina and causing wrinkling or corrugation of the inner retina. With en face OCT, contractile ERMs appeared as central plaque-like hyper-reflective lesions associated with radiating hyporeflective striae corresponding to the wrinkling or irregularity of the inner retinal surface. Further confirmation was obtained by evaluating the OCTA scans, which demonstrated distortion of the retina vasculature.
The presence of PVD was assessed using the OCT B-scan through the fovea and optic disc or evaluating the OCT volume scan through the macula and optic disc. Grading was performed as previously described as follows and was defined as: grade 1: detachment of the posterior hyaloid from the macula on either side of the fovea; grade 2: detachment of the posterior hyaloid from the macula on both sides of the fovea; grade 3: detachment of the posterior hyaloid from the fovea; grade 4: detachment of the posterior hyaloid from the fovea and the optic nerve.
Image Processing and Analysis: Quantitative Analysis of the Normal ILM and EVG
The 6 × 6-mm multimodal dataset from the Optovue machine was exported to ImageJ 1.51m9 version software (National Institutes of Health, Bethesda, Maryland, USA) ( Figure 1 ). The 6 × 6-mm en face structural OCT was cropped within ImageJ, with the resulting image measured as 600 × 600 pixels. The scale was set for 100 pixels/mm for all images.
Artifacts from vessels were removed by using the cropped OCTA image. Vasculature on the default superficial capillary segmentation was isolated, selected, inverted, and overlaid on the en face OCT. This allowed for normalization and removal of the hyper-reflectivity corresponding to the retinal vasculature. Images were then binarized and converted to an 8-bit grayscale (with a total of 256 shades, where 0 = pure black and 255 = pure white for thresholding).
Threshold reflectivity was performed on the aforementioned customized segmented cropped en face OCT images that were corrected for vascular artifacts ( Figure 1 , F). This segmentation allowed for specific isolation of EVG ( Figure 1 , G). The mean EVG brightness was measured by averaging the rectangular regions of interest (0.001 mm 2 ) that were placed on 3 different regions of the EVG.
The mean EVG brightness was used as the threshold cutoff to select the EVG lesion. Using ImageJ software, selected EVG lesions were then measured for average reflectivity and total surface area (mm 2 ) of the EVG ( Figure 1 , G). For normal ILM reflectivity measurements, the inverse selection threshold was applied ( Figure 1 , H). The percentages of surface area of the EVG lesions were calculated as total surface areas of the EVG lesion divided by the total surface areas of the 6 × 6-mm scan (36 mm 2 ).
Presence of Contractile ERM in Patients With and Without EVG
To determine whether contractile ERMs occurred more frequently in patients with EVG, the presence of contractile ERMs associated with EVGs were compared to those in a randomly selected, age- and sex-matched cohort. Specifically, from the cohort of patients without EVG, 161 patients were randomly selected to match the age and sex distribution of the EVG cohort. Then we compared the percentages of contractile ERMs present in patients with EVG relative to those without EVG.
Analysis was performed using R version 3.6.3 software (R Foundation for Statistical Computing, Vienna, Austria). Mean ± standard deviation or numerical counts with percentages were presented for interval and ordinal or categorical variables, respectively. The χ 2 test was used for parametric categorical or ordinal variables while the Fisher exact test was used for nonparametric categorical or ordinal variables. For parametric interval variables, paired or nonpaired t -tests were used when samples were paired or not paired, respectively. For inter-grader analysis, Cohen’s kappa results were obtained. The generalized estimating equation (GEE) regression model was used to assess an association between the presence of EVG and age, PVD status, and phakia with correction of bilateral eye enrollment. A P value of <.05 was considered statistically significant.
Demographics of the Total Cohort and Subgroup of Patients With and Without EVG
Using the Optovue database at UCLA, the en face imaging of 1,848 patients (3,696 eyes) were evaluated to identify cases with EVG. Of these, 1,298 patients (or 2,476 eyes) met the inclusion and exclusion criteria and were analyzed ( Table 1 ). EVG was present in 161 eyes from 119 patients, and EVG was absent in 2,315 eyes of 1,179 patients. Assessment of agreement between graders for the presence of EVG showed a kappa value of 0.8 (95% confidence interval [CI]: 0.7 – 0.9; P <.01; with an overall agreement of 87.0%). EVGs were bilateral in 35% and unilateral in 65% of patients. The calculated prevalence was 161 of 2,476 eyes (6.5%) and 119 of 1,298 patients (9.2%) in this cohort. The mean age of all patients was 58.7 years ± 24.6 years old (range: 3.0 – 100.0 years). Patients with EVG (mean: 79.3 ± 10.7 years old; range: 38.0 – 100.0 years old) were significantly older than patients without EVG (mean: 55.9 ± 24.6 years old; range: 3.0 – 98.0 years old; P < .001). There were no differences pertaining to sex ( P = .10) or eye laterality ( P = .75) between patients with and without EVG. Patients with EVG were more commonly pseudophakic (98 of 161; 60.9%) compared to those without EVG (436 of 2,315; 18.8%; P <.001). GEE modeled for age and PVD status and controlled for bilateral eye involvement showed persistence of statistical significance of phakic status ( P <.001).
|Total a||Presence of EVG b||Absence of EVG c||P Value|
|Average ± SD age (range), yrs||58.7 ± 24.6 (3.0 – 100.0)||79.3 ± 10.7 (38.0 – 100.0)||55.9 ± 24.6 (3.0 – 98.0)||<.001 d|
|Sex||Females||606.0 (46.7)||64.0 (53.8)||542.0 (46.0)||.10|
|Males||692.0 (53.3)||55.0 (46.2)||637.0 (54.0)|
|Eye||OD eyes||1,246.0 (50.3)||83.0 (51.6)||1163.0 (50.2)||.75|
|OS eyes||1,230.0 (49.7)||78.0 (48.4)||1152.0 (49.8)|
|Lens status||Phakia eyes||1,541.0 (62.2)||63.0 (39.1)||1478.0 (63.8)||<.001 d|
|Pseudophakia eyes||534.0 (21.6)||98.0 (60.9)||436.0 (18.8)|
To further control for variability between age and sex, an age- and sex-matched cohort was randomly selected between patients with and without EVG. Patients with EVG were more likely to be pseudophakic (98 of 161; 60.9%) than patients without EVG (73 of 161; 45.3%; P = .005) ( Table 2 ).
|Presence of EVG a||Absence of EVG b||P Value|
|Average ± SD age (range), yrs||79.3 ± 10.7 (38.0 – 100.0)||78.9 ±10.6 (38.0 – 97.0)||.73|
|Gender||Females||64.0 (53.8)||79.0 (49.1)||.44|
|Males||55.0 (46.2)||82.0 (50.9)|
|Eye||OD eyes||83.0 (51.6)||83.0 (51.6)||1.0|
|OS eyes||78.0 (48.4)||78.0 (48.4)|
|Phakia eyes||63.0 (39.1)||88.0 (54.7)||.005 c|
|Pseudophakia eyes||98.0 (60.9)||73.0 (45.3)|
|Presence of ERM, eyes (%)||71.0 (44.1)||30.0 (18.6)||<.001 c|
|Presence eyes||137.0 (95.1)||121.0 (84.1)||.002 c|
|Absence eyes||7.0 (4.9)||23.0 (15.9)|
|Grade 1 eyes||3.0 (2.1)||1.0 (0.7)||.62|
|Grade 2 eyes||4.0 (2.7)||6.0 (4.2)||.51|
|Grade 3 eyes||60.0 (41.7)||40.0 (27.8)||<.001 c|
|Grade 4 eyes||70.0 (48.6)||74.0 (51.4)||.64|