Assessment of clinical outcomes and prognostic factors following membrane peeling in idiopathic epiretinal membrane using EIFL staging system: an optical coherence tomography angiography analysis

Abstract

Background

To evaluate the associations between anatomical changes and visual outcomes after membrane peeling in eyes with different stages of idiopathic epiretinal membrane (iERM) using optical coherence tomography angiography (OCTA).

Methods

All iERM eyes were graded into four stages based on the presence of ectopic inner foveal layers (EIFL) and underwent 23-gauge vitrectomy combined with ERM and internal limiting membrane (ILM) peeling, while their fellow eyes were treated as the control group. OCTA was used to measure retinal thickness(RT), foveal avascular zone (FAZ)-related parameters and superficial and deep capillary plexus (SCP and DCP) layers using 6 × 6 mm scans before, 1 month and 3 months after surgery. In addition, best corrected visual acuity (BCVA), metamorphopsia and macular features were assessed.

Results

Forty-six subjects were included in this study. In comparison to the preoperative data, visual acuity and metamorphopsia improvement was statistically significant in four stages(P < 0.05) and the higher stage (3 and 4) achieved more pronounced improvements (P = 0.002). For higher stage, RT reduced with an increase in stage(P < 0.001), superficial and deep foveal vessel density (SFVD and DFVD) and parafoveal vessel density (PRVD) in SCP declined remarkably, FAZ area was enlarged obviously, FAZ perimeter (PERIM), foveal vessel density (FD) and PRVD in DCP increased significantly after surgery (P < 0.05). Similar to high-stage patients, those with stage 2 iERMs demonstrated a decreasing trend in central macular thickness (CMT), paraRT (parafoveal thickness), SFVD, and DFVD(P < 0.05). Nevertheless, no notable alterations were observed in other indicators. Distinct from other groups, only CMT and FD increased slightly in stage 1 iERMs (P < 0.05). Post-LogMAR BCVA and LogMAR BCVA-d (pre-LogMAR BCVA –3-month post-LogMAR BCVA) were positively correlated with preoperative stages, CMT, pre-LogMAR BCVA, SFVD, and vascular tortuosity(P < 0.05). but negatively correlated with FAZ area and DFVD (P < 0.05). Preoperative and postoperative metamorphopsia had a certain positive correlation with preoperative CMT (P < 0.05).

Conclusions

According to OCTA analysis, different EIFL stages of iERMs showed significantly functional and anatomic differences before and after membrane peeling. Low-stage patients have better post-op visual function, while high-stage patients benefit more from surgery. It also demonstrated EIFL staging system contribute doctors to manage iERMs.

Introduction

Epiretinal membrane (ERM) refers to an abnormally proliferating avascular fibrous membrane located between the posterior vitreous cortex and internal limiting membrane (ILM) in macula. It is currently believed to arise from the migration of glial cells to the retinal surface, triggered by disruptions in the ILM [1]. Depending on whether an underlying cause is present, ERM can be categorized into iERM and secondary ERM. With the widespread application of optical coherence tomography (OCT), ERM has emerged as a prevalent clinical ocular fundus disease causing visual disorders. Approximately 6.5% of the global population is affected by ERM, and aging may be an important risk factor. Vitrectomy combined with the peeling of ERM and ILM was applied widely many years to treat iERM [2]. However, given the unpredictability of surgical results, the determination of the optimal surgical timing and the question of whether to concurrently peel the internal limiting membrane have sparked debates. The grading of ERM is frequently utilized as a benchmark for selecting the appropriate surgical timing. The earliest grading criteria, proposed by Gass, were classified into grades 0–2 [3]. However, its clinical utility is limited due to the ambiguity in grading system. In 2014, Mathews introduced a new grading system for the foveal contour, which involves measuring and analyzing the thickness of the foveal center relative to the surrounding macula in radial line scans using OCT [4]. In addition, FAZ area has also been used for classification of iERM [5]. In recent years, Andrea Govetto has described a novel staging scheme that primarily relies on the presence of ectopic inner foveal layers, aiming to improve the management of iERMs [6].

OCT has proven to be the definitive diagnostic tool for iERM [7], offering unparalleled clarity in depicting ERM morphology and retinal microstructural changes. The swift advancements in OCT technology have significantly bolstered the diagnosis and therapy for retinal disorders, and the emergence of OCTA has further elevated this standard. OCTA facilitates not only three-dimensional, regional insights into retinal ultrastructure but also precise monitoring of vessel density variations within both the deep and superficial retinal layers [8].

Previous studies have primarily focused on the preoperative and postoperative changes observed through OCT in patients with ERM [9,10,11]. However, there have been relatively few studies analyzing vessel density, and most of these studies lacked stratified analysis [12]. Even fewer studies have classified patients based on the severity of inner retinal abnormalities. Therefore, combining with the new EIFL staging system, our study employed OCTA to analyze the preoperative and postoperative changes of macular microstructural features and vessel density in patients with different stages of iERM, thereby evaluating the surgical outcomes and prognostic factors of ERM and ILM peeling.

Methods

Study participants

In this study, all patients with iERM diagnosed in Wuhu Eye Hospital from January 2022 to June 2024 were collected, all of which underwent 23G vitrectomy combined with the peeling of ERM and ILM (For patients with cataracts, phacoemulsification was performed simultaneously). Postoperative follow-up was performed at one month and three months in our outpatient clinic. The study adhered to the principles of the Declaration of Helsinki, was approved by the ethics committees of the hospital and the Health Commission, and each patient signed an informed consent form before surgery.

Inclusion criteria and exclusion criteria

Inclusion Criteria: ① patients with BCVA below 0.6 or obvious visual tortuosity, blurring and other symptoms affecting daily life; ② gold leaf-like or opaque grayish-white membrane changes can be seen on fundus examination, and the peripheral retina is free of breaks, exudates, hemorrhages and other changes; and ③ OCT suggests that the macular area of the surface layer of the retina to see high-reflective ray-like changes, with or without the inner layer of the retinal abnormalities.

Exclusion Criteria: ① Fundus diseases that damage the structure and function of macula: age-related macular degeneration, diabetic retinopathy, retinal vein occlusion, macular tear, etc.; ② secondary ERM of ocular trauma, internal ocular surgery, uveitis, ischemic optic neuropathy, etc.; ③ patients with high intraocular pressure (≥ 21 mmHg) or glaucoma, especially patients with optic nerve atrophy; ④ people with obvious refractive media clouding; ⑤ people with high expectation of surgery or anxiety state.

EIFL staging system

Stage 1: thin and light ERM with the foveal pit; Stage 2: ERM with stretching of the outer nuclear layer and absence of the foveal pit; Stage 3: Ectopic inner foveal layers and absence of the foveal pit can be identified, but retinal layers remained well-defined; and Stage 4: ERM is thicker. The ectopic inner foveal layers are continuous with disrupted retinal layers [6].

Ophthalmological examination

All patients received basic ophthalmic examinations before and after surgery, including best-corrected visual acuity, intraocular pressure, and fundus examination with a slit lamp.

Amsler grid examination

During the preoperative and 3-month postoperative follow-up periods, all patients underwent the Amsler grid test with clear and evenly distributed lighting, placing the grid 30 cm from eye level after correcting for refractive errors like presbyopia or myopia. If the center area appeared blank or distorted, it was a positive result, otherwise negative.

OCTA examination

All included subjects underwent OCTA (Optovue RTVue XR; Optovue Inc, California, USA) examination before surgery, 1 month and 3 months after surgery. In order to ensure the accuracy of the examination, the pupils were dilated before the examination. High-definition OCTA images (6 × 6 mm area) centered at the fovea were acquired. All the data of retinal thickness and vessel density were collected automatically three times. Images with a signal strength index lower than 6, those of low quality were excluded. After manually inspecting to ensure reasonable segmentation of the retinal area, the quantitative vessel density analysis mode in the OCTA software was performed to localize the deep and superficial retinal layers and extract data. For the measurement of FAZ parameters, the FAZ measurement mode in the OCTA software was mainly used, and would automatically generate the FAZ area and related data. However, it was necessary to repeatedly manually check the accuracy of the automatically generated FAZ area. For images with eccentric or significantly deviated ranges, the foveal avascular area was manually drawn after fully enlarging the image and record the area, perimeter, and FD displayed by the software [13].

Cotton ball sign was identified as a hyperreflective region under the central ellipsoid zone, with poorly defined borders [6, 14].

Macular pseudo hole (MPH), a unique clinical entity, is proved to arise from the centripetal contraction of the ERM. It is important to emphasize that foveal tissue remains intact in MPH [14].

Inner retinal dimples and counting method: "Dimples," an abnormal retinal structure first observed by Tadayoni et al. in 2001 after the removal of an ERM, have been termed the dissociated optic nerve fiber layer (DONFL) [15]. Based on previous studies, the designated 6 × 6 mm area was divided into 20 × 20 squares of uniform size [16]. Within each ETDRS subfield, the count of grids containing Inner retinal dimples was manually tallied.

Data and scoring criteria

Retinal thickness (RT, μm) in 9 regions of the macula (fovea (1 mm), parafoveal temporal, superior, nasal, inferior (3 mm), perifoveal temporal, superior, nasal, inferior (6 mm)).

Vessel density (VD, %) of the SCP and DCP in the five regions encompassing the macula (temporal, superior, nasal, inferior regions (3 mm) and the fovea (1 mm)).

FAZ parameter: area (mm²) and PERIM (mm) of FAZ and foveal density (%) of the 300 μm wide area around the FAZ.

Vascular tortuosity scoring criteria (based on En-face images).

-	0	Vessels are clearly visible without significant tortuosity; retinal circulation is recognizable
+	1	Vessels are clearly visible without obvious tortuosity; retinal circulation is recognizable but smaller than normal
+ +	2	Vessels are clearly visible, small vessels are tortuosity, retinal circulation is recognizable but smaller than normal
+ + +	3	Small vessels slightly blurred and tortuosity, retinal circulation blurred and significantly reduced
+ + + +	4	Small vessels are blurred, retinal circulation was completely unrecognizable, and there is a tortuosity displacement of the upper and lower macular branching vessels

Inner retinal dimples scoring criteria (on the macular central 6 × 6 mm area of enface OCTA).

-	0	No visible dimples(Fig. 1A)
+	1	The number of inner retinal dimples < 20 (Fig. 1B)
+ +	2	20 ≤ The number of inner retinal dimples < 40 (Fig. 1C)
+ + +	3	The number of inner retinal dimples ≥ 40 (Fig. 1D)

Fig. 1

Inner retinal dimples: (A) no inner retinal dimples; (B) mild inner retinal dimples; (C) moderate inner retinal dimples; (D) severe inner retinal dimples

Surgical procedure

All patients underwent 23-G pars plana vitrectomy combined with the peeling of ERM and ILM under retrobulbar block anesthesia. The surgery was performed by the same chief surgeon with decades of experience in vitrectomy. If the patient had cataract or refractive error, cataract surgery combined with implantation of a folded monofocal aspheric IOL was performed, followed by mid-peripheral vitrectomy, and the posterior vitreous detachment and epiretinal membranes were removed by intravitreal injection of triamcinolone acetonide. Afterwards, the ILM staining was performed by using several drops of indocyanine green solution prepared with high glucose concentration. Then ILM was stripped in a circular area centered on the central concavity, with the distance between the upper and lower macular vessels as the diameter. After peeling the membrane, suction and gas–liquid exchange were carried out to decrease the residue of indocyanine green in the vitreous cavity.

Statistical analysis

All data were analyzed using SPSS 26.0 software. For data conforming to normal distribution, data were expressed as mean ± standard deviation (SD), one-way ANOVA was used for intra-group comparisons, repeated-measures ANOVA was used for inter-group comparisons. In addition, Pearson analysis was used for correlation analyses. Medians and interquartile ranges not conforming to normal distribution were expressed, and statistics were expressed by rank-sum tests, and spearman analyses were used for correlation analyses. Amsler grid results were evaluated by a chi-square test. Binary logistic regression was performed to reveal the relationships between preoperative parameter and Amsler grid results. P < 0.05 considered as statistically significant difference. Graph pad prism 8.0 was used for graphing.

Results

A total of 46 eyes in the iERM group and 14 eyes were included in the normal group. In the iERM group, there were 17 (37%) males and 29 (63%) females, with average age of 68.89 ± 1.17 years, 23 (50%) right eyes and 23 (50%) left eyes, and ocular axes (22.96 ± 0.94) mm. Specific epidemiological and basic characteristics of the iERM group were described see Table 1.

Table 1 Demographic data and baseline characteristics of patients with iERM

RT

In this study, the preoperative, 1-month and 3-month postoperative thicknesses of 9 regions of the macula were detected and analyzed (Tables 2 and 3).

• 1. Preoperative CMT and parafoveal RT were significantly higher than normal in all iERM groups (CMT_N = 242.00 ± 16.47 μm, CMT₁ = 328.27 ± 44.14 μm, CMT₂ = 383.42 ± 15.51 μm, CMT₃ = 441.00 ± 29.28 μm, and CMT₄ = 554.45 ± 47.80 μm; P < 0.0001) (parafoveal RT: P < 0.05). At the same time, the thickness increased significantly with the grading of iERM severity (P < 0.0001). Perifoveal RT was also significantly higher in the stage 3 and 4 group than in the normal group (P < 0.001), while there is no significant the difference between the stage 1 and 2 and the normal group (P > 0.05).

• 2. After performing vitrectomy combined with the peeling of ERM and ILM, all groups showed changes in retinal thickness.

Table 2 Nine regions of retinal thickness before and after surgery

Table 3 Variation of retinal thickness before and after surgery in different regions

(1) CMT and CMT-d: The CMT of the stage 2, 3 & 4 groups in the iERM group all showed different degrees of postoperative decrease (F₂ = 7.26, P₂ = 0.002; F₃ = 26.41, P₃ < 0.0001; and F₄ = 49.85, P₄ < 0.0001), yet the stage 1 group showed a trend toward a mild increase in CMT postoperatively (F₁ = 4.36 P₁ = 0.048) (Fig. 2A). In addition, there was a difference in the interaction between follow-up time and subgroup (F_stage×time = 42.80, P_stage×time < 0.0001). CMT-d = 3 months postoperative CMT-preoperative CMT. There was a significant difference in CMT-d by subgroup (F = 52.98, P < 0.0001).

Fig. 2

Changes of CMT (A), FAZ (B), SFVD (C), and DFVD (D) in eyes with iERM (preoperative, 1-month postoperative, 3-month postoperative) and in the normal eyes. * p < 0.05 vs the preoperative; pre: preoperative; post1:1-month postoperative; post3: 3-month postoperative

(2) parafoveal RT and parafoveal RT-d: Similar to the changes in CMT, the temporal (T), superior (S), nasal (N), and inferior (I) regions of parafoveal RT in the stage 2, 3, and 4 groups all decreased to different degrees in the postoperative period (stage 2: F_T2 = 21.376; P_T2 < 0.0001; F_S2 = 6.276; P_S2 = 0.005; F_N2 = 2.669; P_N2 = 0.044;F_I2 = 6.531; P_I2 = 0.004) (stage 3: F_T3 = 20.89; P_T3 < 0.0001; F_S3 = 13.26; P_S3 < 0.0001; F_N3 = 13.20; P_N3 < 0.0001;F_I3 = 14.56; P_I3 < 0.0001) (stage 4:F_T4 = 109.42; P_T4 < 0.0001; F_S4 = 56.93; P_S4 < 0.0001; F_N4 = 32.78; P_N4 < 0.0001; F_I4 = 58.59; P_I4 < 0.0001), and the degree of decline: temporal > above > below > nasal (P < 0.05). However, there was no statistically significant difference in the degree of reduction after parafovea RT in the stage 1 group. The interaction of follow-up time with subgroups likewise differed (F_stage×timeT = 46.29; P_stage×timeT < 0.0001; F_stage×timeS = 31.23; P_stage×timeS < 0.0001; F_stage×timeN = 23.27; P_stage×timeN < 0.0001; F_stage×timeI = 26.96; P_stage×timeI < 0.0001). parafoveal RT-d = 3 months postoperative parafoveal RT-preoperative parafoveal RT. There was a significant difference in parafoveal RT-d between different subgroups (T: F = 61.280, P < 0.0001; S: F = 36.289, P < 0.0001; N: F = 31.061, P < 0.0001; I: F = 35.293, P < 0.0001).

(3) perifoveal RT and perifoveal RT-d: the within-group differences between preoperative and postoperative perifoveal RT were not statistically significant between the stage 1 and 2 groups. Additionally, only the stage 3 and 4 groups showed a significant decrease in perifovea RT postoperatively (stage 3: F_T3 = 5.30; p_T3 = 0.01; F_S3 = 9.69; p_S3 < 0.0001; F_N3 = 3.55; p_N3 = 0.04; F_I3 = 7.857; p_I3 = 0.002; stage 4: F_T4 = 26.69; p_T4 < 0.0001, F_S4 = 18.38; p_S4 < 0.0001; F_N4 = 7.87; p_N4 = 0.002; F_I4 = 12.62; p_I4 < 0.0001). However, the differences between the 4 observation groups were significant, as were the interactions between follow-up time and subgroups (F_stage×timeT = 14.344; P_stage×timeT < 0.0001; F_stage×timeS = 9.92; P_stage×timeS < 0.0001; F_stage×timeN = 6.477; P_stage×timeN < 0.0001; F_stage×timeI = 5.637; P_stage×timeI = 0.001). At 3 months postoperatively, all regions of perifoveal RT in each group were not statistically different from the normal group. parafoveal RT-d = parafoveal RT at 3 months postoperatively-parafoveal RT preoperatively. there was a significant difference in parafoveal RT-d between the different subgroups (T: F = 20.71, P < 0.0001; S: F = 11.68, P < 0.0001; N: F = 8.31, P = 0.001; and I: F = 7.66, P = 0.001).

SCP

In this study, superficial vessel density in five regions of the macula were detected and analyzed (Tables 4 and 5).

Table 4 Five regions of superficial and deep vessel density before and after surgery

Table 5 Variation of superficial and deep vessel density before and after surgery in different regions

1. The preoperative SFVD was significantly higher than that of the normal group in all observation groups (SFVD_N = 15.09 ± 1.75%, SFVD₁ = 29.86 ± 6.91%, SFVD₂ = 35.89 ± 2.31%, SFVD₃ = 41.50 ± 5.14%, SFVD₄ = 46.32 ± 4.80%; P < 0.0001). Nevertheless, stage 3 and stage 4 regions had lower parafoveal VD than normal (P₃ < 0.05, P₄ < 0.0001) and no significant difference between the stage 1 and 2 groups and the normal group was observed.

(1) SFVD and SFVD-d: Compared with the control group, none of the changes in superficial foveal vessel density were statistically different in the stage 1 group in the postoperative observation group, whereas a significant decrease in SFVD was observed in the stage 2 to 4 groups after surgery (F₂ = 6.95, P₂ = 0.003; F₃ = 9.24, P₃ = 0.001; F₄ = 4.38, P₄ = 0.02, F_stage×time = 18.45, P_stage×time < 0.0001) (Fig. 2C). SFVD-d = 3 months postoperative SFVD—preoperative SFVD. SFVD-d varied significantly among different groups (K = 25.91, P < 0.0001), but did not increase with subgroups in an incremental trend.

(2) Parafoveal VD and parafoveal VD-d: The postoperative difference in parafoveal VD was statistically significant only on the temporal side and superiorly in the parafoveal VD of stage 3 vs. group 4 (stage 3: F_T = 3.74, P_T = 0.034; F_S = 5.16, P_S = 0.011; stage 4: F_T = 4.46, P_T = 0.02; F_S = 54.15, P_S = 0.026). However, the between-group differences in each index were significant (F_stage×timeT = 4.159, P_stage×timeT = 0.003, F_stage×timeT = 4.454, P_stage×timeT = 0.001, F_stage×timeT = 3.146; P_stage×timeT = 0.008a, F_stage×timeT = 3.505; P_stage×timeT = 0.006). parafoveal VD-d = 3-month postoperative parafoveal VD-preoperative parafoveal VD. The difference in parafoveal VD-d between the groups was not statistically significant.

DCP

Deep vessel density was detected and analyzed in five regions of the macula (Tables 4 and 5).

• 1. DFVD was significantly higher in the preoperative observation groups than in the normal group (DFVD_N = 28.61 ± 7.31%, DFVD₁ = 39.88 ± 5.50%, DFVD₂ = 43.97 ± 1.66%, DFVD₃ = 44.37 ± 0.99%, and DFVD₄ = 47.71 ± 1.41%; P < 0.0001). All regions of parafoveal VD were significantly lower than in the normal group (P < 0.0001), and DFVD decreased with increasing stage (P < 0.0001).

• 2. 1) DFVD and DFVD-d: DFVD in the stage 2 to 4 groups after vitrectomy all gradually decreased at each follow-up (F₂ = 4.05; P₂ = 0.027; F₃ = 11.41; P₃ = 0.001; F₄ = 8.54; P₄ = 0.008; F_stage×time = 6.909, P_stage×time < 0.0001) (Fig. 2D). There was no statistically significant difference in the changes in the stage 1 group. DFVD-d = 3-month postoperative DFVD—preoperative DFVD. The results showed that there was no significant difference in DFVD-d between the groups.

(2) Parafoveal VD and parafoveal VD-d: there was a trend of increase in all regions of parafoveal VD, and the increase was more pronounced in the stage 3 and 4 groups (stage 3: F_T = 9.79; P_T = 0.001; F_S = 25.23; P_S < 0.0001; F_N = 9.38; P_N = 0.001; F_I = 13.44; P_I < 0.0001; stage 4: F_T = 9.57; P_T < 0.001, F_S = 3.34; P_S = 0.049; F_N = 12.58; P_N < 0.0001; F_I = 4.73; P_I = 0.016), but the degree of change was not significantly different between groups of all levels and between regions. parafoveal VD-d = 3 months postoperatively parafoveal VD-preoperative parafoveal VD. The results showed that parafoveal VD-d did not show significant differences between groups.

FAZ

As shown in Table 6, the FAZ area, PERIM, and FD within 1 mm of the FAZ were observed and analyzed.

Table 6 FAZ measurement

1. Compared with the normal group (0.33 ± 0.10 mm²⁾, the preoperative FAZ area in the observation group was significantly smaller (stage 1: 0.124 ± 0.01 mm², stage 2: 0.087 ± 0.002 mm²; stage 3: 0.06 ± 0.005 mm²; stage 4: 0.026 ± 0.006 mm²; P < 0.0001). Besides, there was a tendency for an increase in the FAZ PERIM (P < 0.0001), and there was no significant difference between FD and normal group. The higher the stage, the smaller the FAZ area (P < 0.0001) and the smaller the PERIM (P < 0.0001).

2. In the 3-month postoperative follow-up, the changes in FAZ area (Fig. 2B) and PERIM were not statistically significant in the stage 1 and stage 2 groups. However, the FAZ area and PERIM increased to different degrees in the stage 3 and stage 4 groups (FAZ area: P₃ < 0.0001, P₄ = 0.001; PERIM: P₃ < 0.0001, P₄ = 0.001). FD was significantly increased in all groups (P₁ < 0.0001, P₂ = 0.015, P₃ = 0.03, P₄ = 0.009).

Visual acuity (LogMAR BCVA): see Table 7 and Fig. 4

Table 7 Preoperative and postoperative LogMAR BCVA

1. All observation groups complained of decreased visual acuity, blurred or tortuosity vision preoperatively, i.e., LogMAR BCVA 0.30 (0.19, 0.30) in stage 1 group, LogMAR BCVA 0.48 (0.40, 0.48) in stage 2 group, LogMAR BCVA 0.65 (0.48, 0.7) in stage 3 group, LogMAR BCVA 0.8 (0.7, 1.0). The higher the grouping level, the worse the visual acuity (r = 0.848, P < 0.0001).

2. Comparing the preoperative visual acuity, except for 2 patients in the stage 1 group who had no significant change in their postoperative visual acuity, the remaining observation groups showed different degrees of improvement in postoperative visual acuity (P₁ < 0.05, P₂ < 0.001, P₃ < 0.0001, and P₄ < 0.0001). At 3 months postoperatively, LogMAR BCVA reached 0.18 (0.10, 0.18) in stage 1 group, 0.20 (0.20, 0.30) in stage 2 group, 0.30 (0.275, 0.40) in stage 3 group, and 0.4 (0.4, 48) in stage 4 group.

3. LogMAR BCVA-d = preoperative LogMAR BCVA-3 months postoperative LogMAR BCVA. The results showed that LogMAR BCVA-d increased with the increase of stage (r = 0.5625, P < 0.0001)

Amsler grizd measurement: see Tables 8 and 9.

Table 8 Changes in metamorphopsia in iERMs before and after surgery

Table 9 Correlation analysis between metamorphopsia and LogMAR, CMT, FAZ

In the preoperative iERM group, patients with positive Amsler grid results (n_N = 20, n_P = 26, n_P%: 56.52%) showed significant differences among the four groups (P = 0.008). Pairwise comparisons revealed that the high-grade group had a significantly higher positive rate. At the 3-month postoperative follow-up, patients with metamorphopsia were still present in all groups (n_N = 31, n_P = 15, n_P%: 32.61%), but this was lower compared to the preoperative level (P = 0.021). Additionally, the high-grade group had more patients with metamorphopsia than the low-grade group (P = 0.048). The improvement rate of metamorphopsia in each stage were calculated as follows: ((number of positive cases before surgery—number of positive cases after surgery) / total number) × % = Stage 1: 0.09%, Stage 2: 0.17%, Stage 3: 0.42%, and Stage 4: 0.27%.

Logistic regression analysis between Amsler grid results and preoperative LogMAR BCVA, CMT, and FAZ area was conducted. The results showed that preoperative and postoperative metamorphopsia had a certain positive correlation with preoperative CMT (P = 0.019; P = 0.012). Nevertheless, no significant relationship was found with LogMAR BCVA or FAZ area.

Morphologic features of macula

1. Vascular tortuosity: Preoperative OCTA images: Stage 1: The mild tortuosity of vessels was observed in few small vessels. Stage 2: Multiple small vessels exhibited a wavy pattern. Stage 3: Small vessels are significantly deformed by the ERM. Stage 4: The tortuosity of the vessels and their displacement toward the foveal center was remarkable. After the operation, the tortuosity of vessels improved significantly, and the most significant improvement was observed at stage 4. See Fig. 3.

2. Postoperative inner retinal dimples: The inner retinal dimples were visible on the OCTA En-face images of the majority of patients after ERM combined with ILM peeling, and it was more obvious at the 3-month follow-up than at the 1-month follow-up, but there is no significant correlation between the distribution and number of dimples and either staging or postoperative visual acuity (P > 0.05).

4. Postoperative macular edema: A total of 9 patients in the study were found to have macular cystoid changes at the 1-month postoperative follow-up, and 4 patients still had mild macular cystoid changes at the 3-month follow-up.

Fig. 3

Morphological characteristics of preoperative and postoperative IERM with different stages: A1: Preoperative OCTA image of stage 1 iERM shows: A distinct foveal pit is present; retinal layers are well-defined; a slight increase in curvature is observed in a small number of small vessels; A2-A3: Postoperative CMT is mildly elevated, with improved tortuosity of small vessels. A4: No Inner retinal dimpling appears postoperatively. B1: Preoperative OCTA image of stage 2 iERM shows: The fovea is absent; retinal layers are well-defined; multiple small vessels exhibit a wavy pattern; B2-B3: Postoperative CMT decreases, with reappearance of the fovea and reduced tortuosity of small vessels. B4: Severely Inner retinal dimpling appeared postoperatively. C1: Preoperative OCTA image of stage 3 iERM shows: The fovea is not visible; ectopic inner foveal layers are present, with identifiable retinal layers; small vessels are significantly deformed by the ERM; C2-C3: Postoperative CMT decreases significantly, with marked improvement in vessel deformation. C4: Severely Inner retinal dimpling appeared postoperatively. D1: Preoperative OCTA image of stage 4 iERM shows: The fovea is completely indistinguishable; Continuous ectopic inner foveal layers are present, with severely disordered retinal layers; coarse vessels are deformed towards to foveal center by the epiretinal membrane; D2-D3: Postoperative retinal thickness decreases remarkably, with identifiable retinal layers and nearly normal restoration of vessel morphology. D4: moderate Inner retinal dimpling appears postoperatively. The red box marked the same position before and after the operation, showing that the preoperative vascular tortuosity was significantly improved after surgery

Correlation analysis of postoperative LogMAR BCVA and LogMAR BCVA-d

In current study, 3-month postoperative LogMAR BCVA and LogMAR BCVA-d (preoperative LogMAR BCVA—3-month postoperative LogMAR BCVA) was used to assess the significance of preoperative parameters for prognostic analysis (Fig. 4). It could be seen that LogMAR BCVA-d was positively correlated with preoperative staging (P < 0.0001), CMT (P < 0.0001), LogMAR (BCVA) (P < 0.0001), SFVD (P < 0.0001), and vessel tortuosity (P < 0.0001) and negatively correlated with FAZ area (P < 0.0001) and DFVD (p = 0.01). Similarly, LogMAR BCVA at 3 months postoperatively was positively correlated with staging (P < 0.0001), CMT (p = 0.0002), LogMAR (BCVA) (P < 0.0001), SFVD (p = 0.02), and vessel tortuosity (p = 0.0004), but negatively correlated with FAZ area (p = 0.0004) and DFVD (p = 0.02).

Fig. 4

Correlation analysis of postoperative LogMAR (BCVA) (A) and LogMAR (BCVA) -d (B) with preoperative parameters. LogMAR0: preoperative LogMAR BCVA; CMT0: preoperative CMT; FAZ0: preoperative FAZ; SFVD0: preoperative SFVD; DFVD0: preoperative DFVD; Tortuosity0: preoperative Tortuosity. LogMAR-d: Difference value of LogMAR. BCVA = (pre-op LogMAR BCVA) – (LogMAR BCVA at 3-mo post-op)

Discussion

Epiretinal membrane (ERM) is a clinically common fundus disease that significantly affects vision, yet the selection of surgical timing and the short- and long-term surgical outcomes often pose challenges for ophthalmologists. This study focuses on the tracking and correlation analysis of preoperative and postoperative clinical characteristics, macular thickness, morphology, and vessel density of different stages of iERM by OCTA to investigate the short-term effects of surgery and how to effectively predict postoperative effects before surgery, which will be more conducive to surgeons to grasp the timing of the surgery, to facilitate the preoperative communication between doctors and patients as well as provide new ideas for the improvement of surgical methods.

In clinical practice, the selection of diagnostic and therapeutic methods for epiretinal membrane (ERM) is primarily guided by the grading criteria established by OCT. However, previous grading systems were characterized by relatively ambiguous and imprecise boundaries or involved cumbersome measurement procedures, thereby limiting their effectiveness in guiding clinical diagnosis and treatment. In recent years, with the improvement of OCT image resolution and data acquisition, Andrea Govetto and other scholars have proposed a novel staging system. This system classifies iERMs into four stages based primarily on the presence of continuous ectopic inner foveal layers. This method allows for quicker grading judgments, better explanations of patients' visual loss, and more accurate assessments of patient prognosis [6]. Therefore, this method was adopted for grading for our study. In addition to possessing all the advantages of OCT, OCTA can also analyze macular thickness and vessel density by region and layer, providing more evidence for our clinical observations and prognosis analysis. Consequently, we opted to utilize OCTA for preoperative evaluations and postoperative follow-ups, specifically adopting the high-definition 6 × 6 mm scanning mode.

Reduction of retinal thickness has been found to be the most obvious change after surgery for ERM, as confirmed by many previous studies [9, 17, 18]. Here, it could be seen that retinal thickness in all regions of the macula gradually decreased over time after surgery in the stage 2 to stage 4 groups. Moreover, the degree of decrease was also related to the staging, with the higher the stage, the more pronounced the degree of change. However, the CMT in the stage 1 group showed a mild elevation at follow-up within 3 months after surgery. Therefore, the central foveal thickness (CFT) manually was re-measured and found no increase in the CFT, Instead, the increase in CMT was due to an increase in retinal thickness around the fovea. However, the OCTA measurement of the CMT was in a 1-mm-diameter circular area centered on the fovea, which led to the increased result. The elevation in retinal thickness surrounding the fovea in the stage 1 group can be ascribed to the continued presence of the foveal pit. In this group, there is a lack of notable traction on the retina, and the inner retinal layers remain largely unchanged. The traction force, optic nerve edema, or reactive hyperplasia that arise following the peeling of the ILM and ERM may play a role in this increase [19]. It was found that postoperative changes in parafoveal RT were also significantly different among the four groups, with the degree of change still related to the grouping. The degree of decrease was as follows: temporal > superior = inferior > nasal. The significant thickness reduction on the temporal region may be related to its particular position or the direction the surgeon used to peel off the membrane [20]. Other scholars have also confirmed that the RT of nasal region is not significantly reduced after surgery [21]. In addition, we measured perifoveal RT and found that there was no statistically significant difference in perifoveal RT between the stage 1 and stage 2 groups before and after surgery, while perifoveal RT decreased significantly after surgery in the stage 3 and stage 4 groups, with the degree of decrease correlated with the grading. We consider that this is because the degree of macular retinal deformation is small in the stage 1 and stage 2 groups, and thus the impact of membrane peeling surgery on perifoveal retinal thickness is relatively minor. Three months after surgery, parafoveal RT and perifoveal RT in all groups were close to normal values, and the retina tended to stabilize.

Due to the traction of the epiretinal membrane, the macula becomes contracted, vessels are distorted, and FAZ decreases in size significantly. Here, as the grade increased, the FAZ area reduced. It is almost impossible to measure in the stage 3 to 4 group, which is similar to the findings of previous scholars [6]. However, regarding the changes in FAZ after surgery, many scholars hold different views. Mario R. Romano compared iERM with those secondary to diabetic retinopathy and found that changes in FAZ were not significant in iERMs [22]. In contrast, KITAGAWA concluded that the FAZ was significantly enlarged after surgery compared to before [23]. In our postoperative follow-up, we found that the change in FAZ area was not significant in stage 1 and stage 2 groups, while the FAZ area increased in the stage 3 and stage 4 groups but remained significantly smaller than normal. This may stem from the traction effect of the ERM on the retina is minor in the lower-stage groups, resulting in a mild decrease in preoperative FAZ area compared to the normal group. In higher-stage groups, the traction causes severe deformation of the fovea. After membrane peeling to relieve macular folds, the FAZ is re-exposed, but due to severe damage to the inner retina, the FAZ cannot return to normal. Therefore, although membrane peeling surgery can improve the FAZ, its effectiveness is limited and only significant for ERM of stage 3 or above.

One of the greatest advantages of OCTA is to detect retinal vessel density in the macula. Numerous previous studies have shown that the superficial capillary plexus (SCP) is significantly higher in patients with iERM prior to surgery compared to normal controls [12, 13]. Preoperative SFVD was significantly higher in patients with ERM, but the higher the staging, the more pronounced the increase. This is associated with the centripetal tortuosity and crumpling of vessels due to ERM and the narrowing of FAZ, resulting in increased vessel density per unit area. After membrane peeling surgery, except for the stage 1 group which showed no significant difference, there was a notable decrease in vessel density (VD) in all other groups. Nevertheless, the extent of the decrease was not related to the subgroups. Regarding the reasons for the decrease in vessel density other researchers have previously suggested that the stripping of the internal limiting membrane during membrane peeling surgery may damage retinal vessels, leading to a decrease in VD. We believe that, besides retinal damage, the relief of macular centripetal shrinkage, enlargement of the FAZ, and the toxic effects of the dye indocyanine green may also contribute to this decrease. Unlike SFVD, preoperative superficial parafoveal VD showed a declining trend compared to the normal group only in the stage 3–4 group. Besides, the parafoveal VD in the stage 4 group rose at 1 month postoperatively and then fell again at 3 months postoperatively. By comparing enface images, we found that this may be related to measurement errors resulting from the lower vessel density in areas with significant thickening and shrinkage of the MEM in the stage 3–4 group. Although the surgery of the epiretinal membrane was limited to the superficial layer only, the deep vessel density of the patient was also significantly increased. The DFVD decreased to different degrees while the parafoveal VD tended to increase at the three-month postoperative follow-up without significant difference. There is no accurate reason for the decrease in the DFVD. speculate that the traction of the ERM may lead to an increase in deep vessel density, and after the traction is relieved, the FAZ improves, resulting in a decrease in VD per unit area. However, the significant increase in deep parafoveal VD may be due to indirect stimulation from the membrane peeling surgery, and the exact reason remains to be further elucidated.

Many morphological changes have been observed in the macula associated with macular epiretinal membranes (ERM). Vascular tortuosity is a common morphological feature of ERM, primarily caused by irregular traction forces exerted by the membrane on various directions of the retina. Currently, there are no established criteria for assessing this tortuosity. Some scholars calculated the tortuosity of retinal vessels by using two parameters: the actual vessel length in the vessel section and the direct vessel branching point distance [24]. In current study, after conducting a simple scoring system, we found significant differences in the degrees of tortuosity among different stages, with all groups showing improvement post-surgery, and the most significant changes observed in the stage 4 group. For the majority of patients, except for a smaller FAZ, the vascular morphology at 3 months post-surgery showed no significant difference from normal.

The wide-angle (6 × 6 mm) En-face scanning images allowed for direct observation of the entire area between the superior and inferior macular vessels. We found that 95% of patients exhibited a significant dimple, characterized by a depression of the inner retinal layers, postoperatively. In a few patients, this sign was confined to the foveal area, while in others, it was distributed throughout the area where ILM was peeled, but did not exceed the peeled area and followed the distribution of arcuate nerve fibers. The range of the dimples observed was consistent with previous studies [25]. and some studies have also found that the inferior temporal quadrant is the most susceptible area [20, 26, 27]. It is worth mentioning that in our study, two patients in the stage 1 group did not exhibit the dimple, and paradoxically showed an increase in parafoveal VD. However, due to the small number of cases, no statistical analysis was performed. Previous related studies have suggested that the pathogenesis of the dimple sign may be due to damage to the arcuate nerve fiber layer caused by ILM peeling [28], or the toxic effects of the staining agent indocyanine green [29, 30]. However, we believe that the ILM serves as the basal layer of Müller cells, covering the surface of the nerve fiber layer to adhere and stabilize it. When the ILM is peeled, Müller cells are damaged, and the nerve fiber layer loses partial support, leading to cracks. This is also a possible cause. During follow-up, some patients were found to have scotomas, which may be related to the dimples. More cases and more detailed studies, such as those on microperimetry and the correlation with retinal vessel density, are needed to further elucidate this issue. However, it still serves as a reminder for ophthalmologists to protect the retina during surgical procedures, minimize the area of ILM peeling, and consider simple ERM peeling for loosely adhered membranes. If recurrence of the ERM is a concern, or if a small area of ILM peeling is performed after ERM peeling, efforts should be made to minimize damage. In addition, preoperative attention should be paid to the patient's intraocular pressure and retinal nerve fiber layer thickness. For patients with high intraocular pressure or mild optic nerve atrophy, peeling surgery may exacerbate the reduction in nerve fiber layer thickness, thereby increasing the risk of visual field defects or even vision loss.

Many previous studies have demonstrated significant improvements in visual acuity after ERM peeling surgery [9, 11, 31]. In our study, except for two patients in the stage 1 group who showed no change in visual acuity after surgery, the visual acuity of the remaining patients improved to varying degrees, with a more pronounced improvement observed in higher stages. Correlation analysis between LogMAR BCVA and LogMAR BCVA-d with various preoperative indicators revealed that patients exhibiting better preoperative visual acuity, lower stages, decreased SFVD, lesser vascular tortuosity, larger FAZ areas, and increased DFVD were likely to achieve better postoperative visual acuity, albeit with lesser improvements compared to their preoperative state.

Metamorphopsia is a frequent manifestation of iERMs. Previous studies have demonstrated that iERMs had an improvement in metamorphopsia. This is associated with the thickness of CFT [32] and the inner retinal layers [33]. Consistent with these findings, this study suggested that patients experiencing preoperative and postoperative metamorphopsia were more prevalent in the high-grade group, which correlated with preoperative CMT but not with visual acuity or FAZ area. It is hypothesized that ERM with higher CMT exert more pronounced traction on the macula, leading to more severe metamorphopsia. However, further research using quantitative methods for assessing visual tortuosity in conjunction with OCTA is needed to elucidate the specific causes. Additionally, the improvement was most obvious in the stage 3 group, suggesting that stage 3 might represent the optimal surgical timing for iERM with metamorphopsia.

According to the visual function results, the improvement in visual acuity is limited since ERM is a chronic proliferative change and surgery causes some degree of retinal trauma, and it serves more to stop the further progression of the disease. Although the low-grade group ultimately attains superior visual acuity, the modest enhancement in vision may result in reduced patient satisfaction. For stage 3 + iERMs, postoperative vision may lower than that of lower-grade patients, but improvement is significant and acceptable. Surgery is recommended for stage 3 + , while for stage 1–2, consider symptoms, expectations, mental state, and surgeon skill. It is of great significance to carry out preoperative communication and tailored plans.

In the past, few researchers conducted subgroup analyses and discussions when studying ultrastructural and capillary plexus changes in the iERM [12, 34, 35]. Based on a new staging system, we analyzed the clinical characteristics, OCT findings, and vessel density changes both within and between groups before and after ERM surgery. This approach allows us to predict postoperative outcomes for patients before surgery and to gain a more detailed understanding of their macular ultrastructure and trends during postoperative follow-up. It assists ophthalmologists in accurately determining the optimal timing for surgery before the operation, selecting an appropriate surgical approach, estimating the risks and benefits of the surgery, and conducting a more comprehensive follow-up after surgery.

At the same time, our study has certain limitations.

• 1. The observation period is short. Our patients are still being followed up.

• 2. The number of cases is small. and more ERM cases will be included in future studies.

• 3. Superficial vessel density measurements may be affected in cases of severely stretched and distorted ERMs.

• 4. Visual field testing was not performed on patients with scotomas.

• 5. Quantitative methods for assessing metamorphopsia will be used in further research.

• 6. There is a concern about the toxicity of staining agents, so we aim to minimize the use of indocyanine green and explore alternative staining agents.

Conclusions

ERM combined with ILM peeling is the primary treatment for iERM, but considering the surgical damage, it is imperative to rigorously assess the patient's condition preoperatively and weigh the advantages and disadvantages of surgery. Our study suggested that there are differences in changes in visual function, RT, and VD before and after surgery among iERMs of different EIFL stages. This reveals that the EIFL staging system, along with OCTA, serves as a useful tool for managing iERM patients, offering clinical significance in grading iERMs, preoperative prognosis assessment, selection of surgical timing, improvement of surgical methods, and postoperative follow-up.