Analysis of ganglion cell-inner plexiform layer thickness in retinal vein occlusion with resolved macular edema

To analyze the retinal ganglion cell-inner plexiform layer (GCIPL) changes in retinal vein occlusion (RVO) eyes with resolved macular edema using optical coherence tomography. We compared the average and minimum GCIPL thickness in RVO eyes with fellow eyes and healthy controls including 40 unilateral RVO patients and 48 healthy subjects. The average GCIPL thickness in BRVO eyes was segmented into the affected and opposite area according to the site of lesion, comparing them with corresponding areas in fellow eyes. Furthermore, maximum central macular thickness (CMT), visual acuity (VA), and intravitreal injection times were recorded to investigate their relationship with the GCIPL thickness. Despite no significant difference in CMT (P = 0.96), the average (P = 0.02 and P < 0.001, respectively) and minimum (both P < 0.001) GCIPL thicknesses were decreased in RVO eyes with resolved macular edema after treatment in comparison to fellow eyes and healthy eyes. Maximum CMT thickness was negatively correlated with the minimum GCIPL thickness (r = − 0.47, P = 0.003). VA and average GCIPL thickness were associated (rs = − 0.49, P = 0.002). In a subgroup analysis that only included BRVO patients, the opposite area revealed no significant difference between two eyes (P = 0.91) although the affected area in BRVO eyes was decreased (P < 0.001). A decrease of GCIPL thickness in RVO was observed even after anatomic restoration and associated with VA prognosis. These GCIPL defects could be attributable to systemic risks and RVO itself, not anti-VEGF effects.


Introduction
RVO is the second most common retinal vascular disorder usually leading to severe loss of visual function with characteristic macular edema and retinal hemorrhage [1][2][3]. It can be roughly classified as branch RVO (BRVO) and central RVO (CRVO) according to anatomical location of vascular occlusion. Among several effective therapeutic measurements, anti-VEGF is considered as the firstline therapy for macular edema. However, despite the fact that many patients have resolved the macular edema after a period of follow-up, the final visual acuity could not return to normal values [4,5].

Abstract
Purpose To analyze the retinal ganglion cell-inner plexiform layer (GCIPL) changes in retinal vein occlusion (RVO) eyes with resolved macular edema using optical coherence tomography. Methods We compared the average and minimum GCIPL thickness in RVO eyes with fellow eyes and healthy controls including 40 unilateral RVO patients and 48 healthy subjects. The average GCIPL thickness in BRVO eyes was segmented into the affected and opposite area according to the site of lesion, comparing them with corresponding areas in fellow eyes. Furthermore, maximum central macular thickness (CMT), visual acuity (VA), and intravitreal injection times were recorded to investigate their relationship with the GCIPL thickness. Results Despite no significant difference in CMT (P = 0.96), the average (P = 0.02 and P < 0.001, respectively) and minimum (both P < 0.001) GCIPL thicknesses were decreased in RVO eyes with resolved macular edema after treatment in comparison to fellow eyes and healthy eyes. Maximum CMT thickness was negatively correlated with the minimum GCIPL thickness (r = − 0.47, P = 0.003). VA and average GCIPL thickness were associated Although its pathogenies are under investigation, the rising technology of OCT makes it available to quantify retinal structural changes and predict prognosis. Specific modifications, such as the photoreceptors, the external limiting membrane (ELM), or the ellipsoid zone (EZ), have been proved to have a strong correlation with the ultimate visual acuity [6,7]. Whereas fewer studies have assessed the inner retina changes, especially the ganglion cell-inner plexiform layer (GCIPL). A study [8] has shown that a significant reduction of macular GCIPL thickness of BRVO eyes in the presence of macular edema. However, another study [9] reported the average thickness of the GCIPL increased after resolution of macular edema, whereas that of the central macula and RNFL statistically decreased. The repeatability of the GCIPL thickness measurements increased after treatment of macular edema in that study. The authors explained that may result from macular contour changes, such as macular edema or atrophy, causing elevated auto-segmentation errors and unstable gazes from decreased visual acuity under the status of macular edema when taking OCT measurements.
Hence, the aim of our study was to evaluate the GCIPL thickness changes based on the "died retina" without macular edema after anti-VEGF treatment in RVO patients and investigate the influence factors causing GCIPL thinning and its relevance with final VA.

Methods
We retrospectively included 88 individuals (40 unilateral RVO patients and 48 age-matched healthy individuals) in this study, and it was carried out in accordance with the tenets of the Helsinki Declaration and approved by the institutional review board. Subjects of unilateral RVO meeting the inclusion criteria were enrolled from the ophthalmology clinic of the Affiliated Hospital of Medical College, Qingdao University from March 2019 to January 2022. Informed consent was obtained from all the participants. The resolution of macular edema in RVO eyes after treatment was defined as a CMT less than 315 μm, corresponding to normal value + 2SDs: 277 + 2 × 19 μm [10,11], associated with normal cube average thickness (CAT) values. CAT represents the average retinal thickness over the entire 6 × 6 mm square scanned area, the mean of thicknesses in nine Early Treatment Diabetic Retinopathy Study (ETDRS) subfields. Other exclusion criteria were as follows: any non-RVO-related retinopathies; optic nerve diseases, e.g., glaucoma and ischemic optic neuropathy; bilateral and ischemic RVO; previous intraocular surgery other than cataract extraction; BRVO concentrated in the macular region; a spherical refraction greater than ± 5.0 diopter (D), cylinder correction greater than ± 3.0D; unilateral RVO with retinal disease in the fellow eye; recent retinal photocoagulation (< 3 months); high IOP (≥ 21 mmHg); high cup-to-disc ratio (≥ 0.5); severe refracting media opacity; OCT imaging with macular structure disorder, cystoid macular edema and submacular fluid, and the GCIPL scanning discontinuity and misalignment which all could bring about the failure of auto-segmentation (Fig. 1); and scans with obvious blinking artifacts and a signal strength less than 6.
All the patients were administered with following treatment regimen: three monthly intravitreal injections of ranibizumab (0.5 mg/0.05 ml) after diagnosis, subsequently additional injections with pro re nata protocol according to recurred macular edema, and laser as a complementary treatment was usually implemented after the first three months if it was needed. Comprehensive ophthalmic examinations were performed monthly and first outpatient visit, including VA converted to the logarithm of the minimum angle of resolution (logMAR), intraocular pressure (IOP) measurement, slit-lamp microscope and dilated fundus examination, and OCT. Fundus angiography was repeated if necessary evaluated by the physician. Relevant parameters except the maximum CMT thickness were recorded when macular edema initially reached resolution meeting the above requirements. The value of maximum CMT could be interpreted as the most severe macular edema with maximum thickness of central macula during monthly follow-up until aforementioned edema resolution.
An age-matched control group was consecutively enrolled from subjects who visited the same hospital for various reasons: health screening checkup, treatment for unilateral ocular diseases while the contralateral eyes were eligible without systemic risks, preoperative examination of cataract with good OCT signal strength and no other disorders.
The sample size for this observational retrospective study was determined to be about 33 participants per group at 0.05 significance level and 80% power from previous similar research.
Optical coherence tomography analysis Imaging ( Fig. 2) was obtained using Cirrus highdefinition OCT (2016 Carl Zeiss Meditec, Inc) after pupil dilation. The macular cube 512 × 128 scan protocol centered on the fovea including macular thickness analysis and ganglion cell layer analysis (GCA) [12,13]. All GCIPL parameters were measured automatically with this GCA algorithm. The average, minimum (the lowest GCIPL thickness on 360 spokes extending from the inner radius to the outer radius), and sectoral thicknesses of the GCIPL are measured in an elliptical annulus (dimensions: vertical inner and outer radius of 0.5 mm and 2.0 mm, horizontal inner and outer radius of 0.6 and 2.4 mm, respectively) around the fovea. The size of the inner ring of the annulus was chosen to exclude the foveal area where the GCL is not susceptible to detect, whereas the outer ring was selected closely to the real anatomy of the macular region where the GCL is thickest in healthy eyes. All the measurement values of the device were compared to the internal normative database and generated a thickness map, a deviation map, and a color-coded significance map. Measurements were displayed in green for normal range (p = 5-95%), in yellow for borderline (1% ≤ p < 5%) and in red for outside the normal range (p < 1%).
The elliptical annulus of the GCA algorithm was segmented into six sectoral areas: the superotemporal (S1), superior (S2), superonasal (S3), inferonasal (S4), inferior (S5), and inferotemporal (S6) sectors. Considering the fact that BRVO is generally confined to one or two quadrants, rarely crossing the midline, we defined the most heavily influenced sector confirmed by fundus examination of the six sectors and Fig. 1 The cross section of scans indicated the segmentation of macular GCIPL using the GCA algorithm. The protocol automatically identifies the boundaries of the GCIPL, between the purple and yellow boundary, and a with normal segmentation centered on the fovea. A scan decentered on the fovea (b, c) and the presence of cystoid macular edema (f, g) caused the failures of auto-segmentation algorithm. Obvious segmentation errors were captured with GCIPL misalignment although central macular thickness values were within normal range with resolved macular edema (c, d). The segmentation was inaccurate due to submacular fluid collection and macular structure distortion (h, i) its adjacent sectors on both sides as the affected area. The rest of the sectors were defined as the opposite area in the subgroup analysis with BRVO (e.g., the superotemporal sector was the most heavily influenced sector, then this sector and its adjacent superior and inferotemporal sectors were collectively referred to as the affected area, and the remaining inferonasal, superonasal, and inferior sectors were the opposite area). The corresponding area in fellow eyes was defined similarly. We compared the average GCIPL thicknesses of the affected and opposite area in BRVO eyes and, respectively, collated them with the corresponding area in fellow eyes (Fig. 3).

Statistical analysis
Data analysis was carried on SPSS version 25.0 for Windows. Kolmogorov-Smirnov test was used for  The superotemporal (S1) sector was the most heavily influenced sector in the BRVO eye. Affected area (the blue area): mean value corresponding to the area covering superotemporal (S1), superior (S2) and inferotemperal (S6) sectors. Opposite area (the green area): mean of the area including superonasal (S3), inferonasal (S4), and inferior (S5) sectors determination of the distribution of the data. Categorical variables were compared using the chi-square test. Analysis of variance (ANOVA) was used in the CMT and GCIPL thickness comparisons among three groups, followed by a Bonferroni post hoc analysis to correct for multiple comparisons. Subgroup GCIPL thickness comparisons in BRVO were analyzed with the Paired t test. The Pearson test and Spearman test were, respectively, used to test the correlations between the maximum CMT and minimum GCIPL thickness and between the VA and average GCIPL thickness. P values of less than 0.05 were considered as statistically significant.

Results
This study finally enrolled 40 unilateral RVO patients with resolved macular edema and 48 healthy controls, and others were ruled out because of meeting the exclusion criteria.
The baseline characteristics of three groups are summarized in Table 1. No significant difference was obtained in terms of age (P = 0.66), gender (P = 0.56), IOP (P = 0.86), axial length (P = 0.73) that could affect the GCIPL thickness. The median of anti-VEGF injection times was 2 until the macular edema resolved.

Assessment of the GCIPL thicknesses in RVO eyes, fellow eyes and healthy controls
Although no significant difference was observed among RVO eyes, fellow eyes and healthy controls in CMT (245.73 ± 30.72, 247.28 ± 18.82 and 247.08 ± 19.63, P = 0.96), whereas statistical difference in GCIPL thickness was found. The average GCIPL thicknesses were 76.03 ± 8.37 for RVO eyes, 80.33 ± 7.99 for fellow eyes (P = 0.02 compared with the RVO eyes) and 84.27 ± 4.59 for healthy eyes (P < 0.001 compared with the RVO eyes). And the minimum GCIPL thickness was also significantly decreased in RVO eyes (58.78 ± 13.94) compared with fellow eyes (76.68 ± 9.36, P < 0.001) and healthy controls (80.90 ± 5.22, P < 0.001). Furthermore, the values of the average GCIPL thickness in fellow eyes were lower than that in healthy controls (P = 0.03). (Table 2).

Subgroup analysis based on different sectors in BRVO
In this subgroup analysis with BRVO patients (n = 32) after anti-VEGF treatment, the average GCIPL thickness of the affected area in BRVO eyes (73.50 ± 8.27) was obviously lower compared with its corresponding area in contralateral eyes (80.34 ± 8.02, P < 0.001), while there was no significant difference

Subgroup analysis according to previous laser and anti-VEGF times
Among resolved RVO eyes, no significant differences were found in average and minimum GCIPL thickness according to different injection times (average GCIPL thickness, P = 0.92; minimum GCIPL thickness, P = 0.12) and whether laser was applied (average GCIPL thickness, P = 0.35; minimum GCIPL thickness, P = 0.59) ( Table 3).

Correlations between GCIPL thickness and maximum CMT or VA
In resolved RVO eyes, the values of maximum CMT in the course of disease before our study were negatively correlated with the minimum GCIPL thickness (r = − 0.47, P = 0.003). Another correlation was between the VA (logMAR) and the average GCIPL thickness (r s = − 0.49, P = 0.002) (Fig. 4).

Discussion
This study showed that the GCIPL thickness after treatment with resolved macular edema in RVO eyes was thinner than that in fellow eyes and healthy controls, correlated with the visual acuity and the maximum CMT thickness. Although no significant difference in opposite area was detected between 2 eyes of BRVO patients, the affected-area mean GCIPL thickness was statistically decreased in occluded eyes. Reasons for GCIPL thinning may be complicated which have not yet been elucidated completely. One possibility could be the systemic risk factors, such as increasing age, hypertension, diabetes mellitus, and dyslipidemia [14,15]. Almost all RVO patients are accompanied by these risks or other hypercoagulable  [17]. Beyond specific systemic diseases, individuals with metabolic syndrome were found having thinner inner retinal layers in OCT segmentation analysis, which suggests that the inherent factors, such as dysfunctional adipose tissuederived chronic inflammation and insulin resistance, might have an effect on neurodegeneration [18,19]. Therefore, although there are certain differences in each of the risk factors, they may share the same processes that cause the subclinical inflammation with the production of reactive oxygen species and cytokines, microvascular endothelium damage as well as leukostasis, eventually leading to local tissue hypoxia and capillary non-perfusion [3,[20][21][22].
The above mechanisms help explain the existing subclinical abnormalities in the fellow eyes of RVO patients considering both eyes share the general risk factors, and this could also be reflected in our findings that the GCIPL thickness of fellow eyes was still thinner than that of healthy controls. Unlike our results presented structural changes, an adaptive optics scanning light ophthalmoscope fluorescein angiography showed decreased foveal microvascular density and increased non-perfused capillaries in fellow eyes of unilateral RVO patients compared with the healthy eyes [21]. The latter showed visible microvascular changes, and these findings from another perspective stress that general risk factors and subclinical pathological changes in RVO patients are worthy of more attention.
In addition to systemic factors, RVO itself could be a cause of GCIPL thickness defects accompanied by characteristic retinal edema and hemorrhages. Given the pivotal role of retinal cells in the transduction of visual pathway, any loss of connectivity among these cells would compromise the visual function. The bipolar cells acted as the only communication between the photoreceptors, and the ganglion cells have a degree of elasticity. The axons would be stretched owing to excessive fluid accumulation, and as a result, some would break with exceeding the elastic limits and the transmission pathway may be lost [23]. This concept underlines the correlations between increasing retinal thickness determined by OCT and GCIPL thinning and decreased visual acuity, which could be interpreted as the accumulated fluid within retina reducing input from the damaged photoreceptors to the ganglion cells, leading to visual impairments eventually. The results highlight that good vision occurs not only associated with the status of the photoreceptors and the external limiting membrane, but also, and even more importantly, the integrity of passing tissue in inner retina.
Finally, one could hold the opinion that the possible influence of anti-VEGF treatment on injured RGCs after an ischemic insult, which is currently under investigation with inconsistent data in the Fig. 4 The scatter diagrams showed the correlations between the maximum CMT during the follow-up and the minimum GCIPL thickness (a), and between the average GCIPL thickness and VA (logMAR) (b) in RVO eyes with resolved macular edema literature [24][25][26]. In our subgroup analysis based on the "dried retina" after anti-VEGF injections, the opposite area in the BRVO eyes and its corresponding area in fellow eyes had no difference. The ganglion cells in this area do not, however, undergo the characteristic ischemic injury after macular edema and hemorrhages of RVO. But at least this point suggests that anti-VEGF injections have not exerted an effect on further loss of ganglion cells in the region of BRVO eyes with subclinical lesions. Instead, the affected area in occluded eyes had a significant GCIPL thinning either comparing with the opposite area in the same eye or its corresponding area in fellow eyes. Furthermore, no significant difference was observed based on the different anti-VEGF times. Therefore, no evidence was found that anti-VEGF treatment caused the ganglion cell defects in our study. And it may be RVO itself, not anti-VEGF effect, causing the GCIPL thickness to further decrease in the occluded area in RVO eyes, which undoubtedly warrants more research.
To our knowledge, quantifying the GCIPL thickness alterations is unreliable in case of macular swelling, even providing misleading information. An increase in retinal thickness could lead to significant decrease in signal strength and the magnitude of this decrease parallel central retinal thickness and total retinal volume [27]. This is consistent with the theory "volume-scattering effect" in which light is scattered across the increased three-dimensional space of the elevated retina [28]. Moreover, recently, Lee et al. illustrated that the repeatability of GCL parameter in BRVO-affected eyes was lower in eyes before treatment than that after treatment, owing to the frequent auto-segmentation errors and unstable gaze with visual damage [9]. Hence, to improve the accuracy, this study was performed with anatomyimproved unilateral RVO cases and scans with obvious segmentation errors as well as low signal strength were ruled out using Cirrus HD-OCT proved with lowest occurrence of any artifacts [29].
Undoubtedly, several limitations exist in this study: Firstly, some data, such as the maximum macular thickness duration, the persistent time of systemic diseases, the history of oral contraceptives and smoking habits which all probably affect the GCIPL thickness changes, are missing owing to its retrospective nature. Secondly, although we exclude the correlation between anti-VEGF treatment and the decrease of GCIPL thickness, a longer follow-up with intensive intravitreal injections may be needed to testify this point in future study. Moreover, no superonasal and inferonasal BRVO patients were included in this study, and this may cause bias when comparing the affected and opposite area in the same eye. However, the primary goal of our subgroup analysis was to compare the corresponding areas of the BRVO eyes with those of the contralateral eyes to verify the effect of anti-VEGF on GCIPL degeneration.
In conclusion, despite the anatomic improvements of normal CMT values after anti-VEGF treatment in RVO, the RGC loss and visual impairment may be irreversible. Rather than anti-VEGF, systemic factors and the RVO itself should be responsible for the GCIPL thinning. Therefore, the improvement of systemic disorders is crucial for prevention and prognosis of RVO, and the integrity of the GCIPL and inner retina should be taken into account for ophthalmologists when formulating treatment plans. We believe that the findings in our study would be helpful for the management and prognosis assessment of RVO patients.