Epiretinal Membrane: Correlations Among Clinical, Immunohistochemical, and Biochemical Features and Their Prognostic Implications

Abstract

Purpose: To investigate correlations among clinical, immunohistochemical, and biochemical characteristics of epiretinal membranes (ERMs) and their prognostic implications.

Methods: This prospective study included 120 patients with idiopathic ERM who underwent vitrectomy. Preoperative and postoperative best-corrected visual acuity (BCVA) and optical coherence tomography (OCT) were measured. Surgical ERM tissues underwent immunohistochemistry (IHC) analysis, and vitreous samples were analyzed for cytokine levels. Statistical analysis assessed associations among BCVA, OCT metrics (ERM stage, inner-retinal irregularity index [IRII], ellipsoid zone disruption [EZD]), IHC features (cell density, protein expression), and cytokines.

Results: The average age of the participants was 70.2 ± 7.6 years, and ERM stages were distributed as Stage 2 in 34 eyes (28.3%), Stage 3 in 54 eyes (45.0%), and Stage 4 in 32 eyes (26.7%). Advancing ERM stage was linked with worsening visual outcomes, higher preoperative IRII, increased EZD incidence, increased ERM cellularity, and elevated TGF-β1 level. Protein expression for myofibroblasts (αSMA), glial cells (GFAP), hyalocytes (CD45), and retinal pigment epithelium (RPE) cells (PRPH2) correlated positively with ERM stage. Multivariate analysis showed preoperative IRII significantly correlated with preoperative BCVA, GFAP, PRPH2, and type I collagen. TGF-β1 was significantly associated with preoperative IRII, αSMA, type I collagen, and N-cadherin, but not with postoperative BCVA. Postoperative BCVA was significantly correlated with preoperative BCVA, preoperative IRII, preoperative EZD, αSMA, and PRPH2.

Conclusions: Early-stage ERM is associated with better visual prognosis, fewer inner and outer retinal changes in OCT, and fewer epithelial–mesenchymal transition-related alterations in ERM tissues. Key prognostic indicators for postoperative BCVA include preoperative BCVA, IRII, EZD, myofibroblasts, and RPE cells. TGF-β1 promotes the transdifferentiation of intraocular cells, impacting OCT features directly and visual outcomes indirectly.

Idiopathic epiretinal membrane (ERM) is an age-related condition characterized by the development of a thin layer of tissue on the macula, leading to progressive vision deterioration and distortion. Its prevalence is increasing, particularly in aging populations of developed countries,^1,2 making ERM a significant public health concern. The growing number of ERM surgeries underscores the critical need to address this healthcare issue. The origin of this multifactorial disease is not yet fully understood, and current treatment primarily relies on surgical intervention.^3–6 The lack of consensus on the optimal timing for surgery, the uncertain postoperative prognosis, and the need for alternative treatment options add further complexity to managing this condition. 

Understanding the histopathological mechanisms of ERM formation, the associated retinal structural distortions, and the visual prognosis after ERM surgery remains challenging. Although optical coherence tomography (OCT) is crucial for monitoring ERM progression and evaluating surgical outcomes,^7–9 previous OCT-based biomarkers, such as various ERM staging schemes,^9–12 central subfield thickness (CST),^13,14 interdigitation zone defects,^15–18 and ellipsoid zone disruption (EZD),^15–18 have been insufficient. These metrics do not adequately capture the wrinkling of the inner retinal layer caused by the centripetal force of ERM. Moreover, measuring preoperative outer retinal variables in advanced ERM can be inaccurate due to the presence of a thick ERM and the associated thickened inner retina. New OCT biomarkers, including the recently proposed ERM staging method,⁹ which considers four distinct OCT morphologies relevant to progressive functional vision impairment, and the inner-retinal irregularity index (IRII),¹⁹ which quantifies the irregularity of the inner retinal layers, may more accurately reflect changes in the inner retina, potentially improving our understanding of the structural progression and visual prognosis of ERM. However, their effectiveness has yet to be confirmed through further study. 

The characteristic appearance of ERM on OCT, including inner retinal deformations, is believed to arise from the mechanical properties, such as tractional forces, influenced by the cellular composition, expressed proteins, cell adhesion molecules, and extracellular matrix (ECM) of ERM tissues.^20,21 Additionally, protein expression in ERM cells may be affected by inflammatory cytokines in the surrounding vitreous humor, indicating complex biochemical interactions involved in ERM. Recent studies have provided significant insights into some of these histopathological aspects related to the pathogenesis and visual prognosis of ERM.^22,23 However, these studies were retrospective, relied on categorical data for ERM cellularity, or used a single suboptimal OCT marker, such as central macular thickness (CMT) alone. More importantly, they neither considered environmental factors such as intravitreal cytokines nor employed the aforementioned new OCT-based ERM staging method or IRII.^9,12 Therefore, a comprehensive, well-designed prospective study based on a detailed clinical, histopathological, and biochemical patient database is still lacking, despite its importance for the clinical management of ERM. 

This prospective study aimed to comprehensively analyze the relationships among variables from four distinct categories, all obtained from the same individual patients: clinical visual acuity, new OCT biomarkers, histopathological features revealed by immunohistochemistry (IHC), and vitreous cytokines. We assessed how each of these variables influences ERM progression and visual prognosis, with the goal of uncovering more complex and comprehensive relationships among a greater diversity of biomarkers than previously identified in research. Additionally, we aimed to determine the optimal timing for ERM surgery and understand the impact of specific IHC markers and vitreous cytokines on ERM progression, ultimately guiding future treatment strategies. 

Methods

Patients

In this prospective study, patients between 55 and 85 years of age who were scheduled for vitrectomy due to idiopathic ERM were consecutively recruited from the outpatient clinic of the Division of Retina at Seoul Shinsegae Eye Center, Republic of Korea, from June 2023 to December 2023. The study received approval from the ethics committee of the Seoul Shinsegae Eye Center (approval no. SSGECI 2023-02-0002), and written informed consent was obtained from all participants. The research protocol complied with the tenets of the Declaration of Helsinki. 

We reviewed the medical histories of all participants. Prior to the scheduled surgery, they underwent comprehensive eye examinations, including best-corrected visual acuity (BCVA), tonometry, OCT, and ophthalmoscopy using a 90-diopter (D) lens after pupil dilation to evaluate the macular state and detect retinal abnormalities. Patients were excluded from our study if they had a history of previous vitrectomy, intraocular injection, or photocoagulation or if they had been diagnosed with secondary ERM or unstageable ERM manifestations such as pseudohole, tractional forces, and lamellar hole-associated epithelial proliferation; severe macular diseases such as intermediate or advanced age-related macular degeneration (AMD), wet AMD, macular telangiectasia, or uveitis; moderate or severe diabetic retinopathy; high myopia (axial length > 26 mm or refractive error < −6.0 D); glaucoma; or uncontrolled systemic diseases. For our final analysis, we included 120 eyes from 176 consecutive patients who underwent planned complete vitrectomy with ERM peeling (Fig. 1). Approximately 2 cc of vitreous fluid was collected with air infusion during the vitrectomy. 

Figure 1.

 

Schematic diagram illustrating the inclusion and exclusion criteria and grouping of patients with idiopathic ERM.

OCT Measurements

Spectral-domain OCT (SPECTRALIS; Heidelberg Engineering, Heidelberg, Germany) examinations were performed preoperatively and at 6 months postoperatively to measure ERM stage, CST, and IRII. OCT images were acquired using 25-line horizontal and vertical B-scans (each line spaced 250 µm apart) centered at the fovea. Each B-scan was averaged over 25 to 30 frames to enhance image quality, ensuring an automatic real-time score of 16 or above and a signal-to-noise ratio of at least 15 dB. 

Classification of OCT staging was adapted from a prior study with minimal modifications.⁹ Stage 1 involves a visible foveal pit and well-defined retinal layers, and surgery is not recommended. Stage 2 (Group 1) is marked by the disappearance of the foveal pit while retaining well-defined retinal layers. Stage 3 (Group 2) features an ectopic inner foveal layer with well-defined retinal layers, and both examiners (HSY and SC) encountered no significant issues with delineating the inner retina at this stage. Stage 4 (Group 3) is characterized by the presence of an ectopic inner foveal layer and less defined retinal layers due to inner retinal wrinkling. Enrollment in Group 3 was confined to cases where, although inner retinal delineation was difficult for at least one examiner (HSY and SC), it was still possible based on the consensus of both examiners. CST was measured over the center 1-mm circle of the Early Treatment Diabetic Retinopathy Study (ETDRS) grid on a thickness map. IRII was evaluated following a previously established protocol,¹⁹ by tracing the inner plexiform layer and retinal pigment epithelium (RPE) layers from the horizontal and vertical OCT B-scan images using the ImageJ plugin NeuronJ (National Institutes of Health, Bethesda, MD, USA) by two examiners (HSY and SC), as exemplified in Figure 2. The final IRII was the average of the horizontal and vertical IRII values. EZD, defined as a loss in the hyperreflective inner retinal bands at the fovea,^9,15,16 was assessed in both horizontal and vertical OCT sections within a 1.0-mm diameter of the central fovea. 

Figure 2.

 

OCT image of a representative Group 2 case (65-year-old female with Stage 3 ERM) showing an IRII of 1.44. The preoperative IRII is calculated as the length ratio of the inferior border of the inner plexiform layer (upper purple line) to the RPE layer (lower purple line) in the central 3-mm foveal area (yellow box) on the middle macular ring of the ETRDS grid.

Immunohistochemistry of Surgical ERM Specimens

IHC markers were initially selected through an unpublished screening study. We examined cellular components by centrifuging the vitreous of ERM patients, identifying RPE cells, hyalocytes, and glial cells. Additionally, we induced cellular changes in immortalized human Müller cells (MIO-M1) with transforming growth factor beta (TGF-β) exposure, which resulted in increased IHC expression of epithelial–mesenchymal transition (EMT) markers, such as myofibroblasts, N-cadherin, and type I collagen. These findings, which were consistent with previous reports,^24–26 guided our selection of the following markers in ERM tissues: alpha smooth muscle actin (αSMA, targeting myofibroblasts), glial fibrillary acidic protein (GFAP, targeting glial cells), cluster of differentiation 45 (CD45, targeting hyalocytes), peripherin-2 (PRPH2, targeting RPE cells), E-cadherin, N-cadherin, type I collagen (COL1), and fibronectin. 

The ERMs were carefully peeled off during vitrectomy and fixed in a 4% paraformaldehyde solution. The surgically removed ERM specimens were embedded in a maximally flattened state using Austerlitz insect pins (Fine Science Tools, Foster City, CA, USA) inside optimal cutting temperature compound (AGR1180; Agar Scientific, Essex, UK), following the method described in a previous paper.²⁰ The specimens were sectioned at a thickness of 5 µm using a cryostat (Leica CM1950; Leica Biosystems, Wetzlar, Germany). Four consecutive tissue slices were placed in a row on a microscope slide, one slide per patient. These ERM slices were isolated with a hydrophobic barrier pen, blocked for 1 hour in 5% bovine serum albumin (BSA) and 0.5% Triton X-100, and incubated overnight at 4°C with 4′,6-diamidino-2-phenylindole (DAPI) and/or primary antibody solutions. The first ERM slice was incubated with DAPI (ab104139; Abcam, Cambridge, UK). The second slice was incubated with rabbit anti-αSMA (ab150301; Abcam), mouse anti-GFAP (sc-33673; Santa Cruz Biotechnology, Dallas, TX, USA), Invitrogen rat anti-CD45 (MA5-17687; Thermo Fisher Scientific, Waltham, MA, USA), and DAPI (Fig. 3, which shows an IHC image example of Stage 3 ERM tissue). The third slice was incubated with rabbit anti-N-cadherin (ab18203; Abcam), mouse anti-PRPH2 (sc-390278; Santa Cruz Biotechnology), rat anti-E-cadherin (14-3249-82; eBioscience, San Diego, CA, USA), and DAPI. The fourth slice was incubated with rabbit anti-COL1 (ac138492; Abcam), mouse anti-fibronectin (ab281574; Abcam), and DAPI. The grouping of primary antibodies for each slice was based on the availability of host species to avoid cross-reactivity. After triple washes, the tissues were incubated for 1 hour at room temperature with secondary antibody solutions: Donkey Anti-Rabbit IgG H&L (Alexa Fluor 647) (ab150063; Abcam), Donkey Anti-Mouse IgG H&L (Alexa Fluor 555) (ab150110; Abcam), or Donkey Anti-Rat IgG H&L (Alexa Fluor 488) (ab150153, Abcam). All primary and secondary antibodies were diluted to 1:200. Following another three washes, the stained tissue slices were mounted with FluoroShield (ImmunoBioScience Corporation, Taipei City, Taiwan), covered with a microscope coverslip, and sealed with nail polish. 

Figure 3.

 

OCT images of a representative Group 2 case (73-year-old female with Stage 3 ERM), showing structural irregularities in the inner retina on both horizontal and vertical B-scans (four images on the left). The right panel shows a four-color fluorescent IHC image (63× magnification) of a surgically removed ERM tissue, sliced 5 µm thick, from the same patient.

Fluorescence Microscopy for IHC Quantification

The stained ERM tissue slices were imaged using a fluorescence microscope (THUNDER Imaging System; Leica Microsystems, Wetzlar, Germany). DAPI and Alexa Fluor 488, 555, and 647 dyes were excited by individual laser lines at wavelengths of 390, 475, 560, and 635 nm, respectively, and fluorescence signals were captured with exposure times of 100 to 500 ms, depending on signal brightness. Illumination power density and exposure time for each IHC marker were kept consistent across all tissue specimens to ensure standardized measurement results. Each tissue slice was imaged near its center with a z-stack across its thickness at 20× magnification, and a maximum intensity projection was created for each fluorescence color channel. A single z-projected IHC image was obtained per tissue slice. 

For automated quantification of histopathological characteristics of ERM, the 12-bit IHC images were analyzed using Leica Application Suite X (LAS X) Image Analysis software (version 3.8.1.26810; Leica Microsystems). In the DAPI image of the first tissue slice among the four slices per microscope slide, total cell count was quantified by automatically counting nuclei using threshold-based image analysis with automatic splitting of touching nuclei. The apparent tissue area was outlined automatically using the binary processing pre-filter mode in the software, and total cell density (counts/mm²) was calculated by dividing the total cell count by this area. For other IHC markers in the second to fourth tissue slices, protein expression levels were quantified by automatically calculating the average fluorescence intensity over the cell area using the software. 

Luminex Immunoassay

Vitreous humor samples collected during vitrectomy were stored immediately at −80°C for biochemical analysis. Cytokine concentrations were measured using an immunoassay analyzer (Luminex 200; Luminex Corporation, Austin, TX, USA) in accordance with our previous studies.^27–29 Stored vitreous samples were thawed, vortexed, and centrifuged at 10,000g for 2 minutes at 4°C to separate the supernatant, followed by activation treatment with 1-N HCl in a 1:5 acid-to-sample volume ratio and neutralization treatment with 1.2-N NaOH/0.5-M HEPES in a 1:6 base-to-sample volume ratio, each for 10 minutes. The supernatant was then diluted to 1:2.8 for TGF-β1, TGF-β2, and TGF-β3 and to 1:2 for adiponectin (APN) and vascular endothelial growth factor A (VEGFA). All cytokine concentrations were measured twice, and their average values were used for analysis. The assay sensitivity thresholds were as follows: 20.9 pg/mL for TGF-β1, 16.74 pg/mL for TGF-β2, 30.87 pg/mL TGF-β3, 850.77 pg/mL for APN, and 2.33 pg/mL for VEGFA. For measurements below these thresholds, the threshold values were recorded and used in logarithmic analyses.^27–29 

Statistical Analysis

Statistical analysis was conducted using SPSS Statistics 26 (IBM, Chicago, IL, USA). A probability value (P) of less than 0.05 was considered statistically significant. We compared baseline characteristics among groups using analysis of variance (ANOVA) or χ² test, depending on the data type. Post hoc analysis was performed using Tukey's test to compare baseline characteristics among the three groups. Preoperative versus postoperative comparisons in visual improvement and IRII within each group were analyzed using paired t-tests. Pearson and Spearman correlations were calculated to assess the relationships among the clinical, immunohistochemical, and biochemical variables. Vitreous cytokine concentrations were log-transformed to follow normal distributions for statistical analysis. In our correlation analysis, correlation coefficients (R) greater than 0.6 were considered strong correlation, 0.4 to 0.6 moderate correlation, and 0.2 to 0.4 weak correlation. A multivariate linear regression analysis was performed to identify factors predicting postoperative BCVA and visual improvement (or BCVA change) or those related to preoperative IRII and TGF-β1. 

Results

The average age of the 120 participants was 70.2 ± 7.6 years. Detailed demographics and clinical characteristics are shown in Table 1. Among these ERM patients, 34 eyes (28.3%) were categorized as Stage 2, 54 eyes (45.0%) as Stage 3, and 32 eyes (26.7%) as Stage 4. There were no significant differences in age, sex, or underlying chronic diseases such as diabetes and hypertension across the stages (all P ≥ 0.178). Preoperative BCVAs decreased with increasing ERM stage. Six months after surgery, BCVAs improved across all groups, with each group showing statistically significant visual improvement (Group 1, P < 0.001; Group 2, P < 0.001; Group 3, P < 0.001; all groups combined, P < 0.001). Similarly, IRII improved significantly in all groups except Group 1 (Group 1, P = 0.207; Group 2, P = 0.005; Group 3, P = 0.001; all groups combined, P < 0.001). Lower ERM stages demonstrated better postoperative visual outcomes (P < 0.001). The postoperative BCVA strongly correlated with preoperative BCVA (R = 0.485, P < 0.001), with a more pronounced correlation in pseudophakic patients (R = 0.619, P < 0.001) compared to phakic patients (R = 0.407, P < 0.001). Both preoperative/postoperative CST and preoperative/postoperative IRII increased with advancing ERM stage (P < 0.001 for all). The incidence of EZD, both preoperative and postoperative, also rose progressively across ERM stages (P < 0.001 and P = 0.004, respectively). 

Table 1.

 

Demographics and OCT Parameters in Idiopathic ERM Patients

 

ERM Progression Versus Immunohistochemical and Biochemical Features

Our IHC images of ERM tissues showed a progressive increase in total cell density and protein expression of αSMA in myofibroblasts, GFAP in glial cells, CD45 in hyalocytes, PRPH2 in RPE cells, and COL1 across advancing ERM stages, as determined by OCT morphology (Fig. 4). These qualitative findings were consistent with our automatic quantitative analysis (Table 2), which showed increased protein expression of cell-specific markers of myofibroblasts, glial cells, hyalocytes, and RPE cells with advancing ERM stages (P = 0.028, P = 0.022, P = 0.004, and P = 0.037, respectively). Significant differences were observed in COL1 across groups, but N-cadherin, E-cadherin, and fibronectin expression showed no significant changes. In our biochemical analysis of vitreous cytokines, TGF-β1, TGF-β2, and APN levels significantly increased across ERM stages, whereas TGF-β3 and VEGFA did not (Table 2). However, TGF-β3 and VEGFA were excluded from the further analysis due to a high prevalence of below-threshold values in 94.2% and 90.0% of the vitreous samples, respectively. 

Figure 4.

 

Representative OCT and fluorescent IHC images across different ERM stages. (A) Preoperative OCT images. (B) Postoperative OCT images taken 6 months after ERM removal. (C–F) Fluorescent IHC images (20× magnification) of four consecutive 5-µm-thick tissue slices from surgically removed ERM specimens, showing specific proteins labeled in each slice. Scale bars: 100 µm (C–F).

Table 2.

 

IHC and Biochemical (Luminex) Analysis of Eyes With Idiopathic ERM (n = 120)

 

Correlations Among BCVA, OCT Parameters, Immunohistochemical Features, and Biochemical Intravitreal Cytokines

Table 3 shows the univariate analysis between key variables. Postoperative BCVA had moderate correlations with preoperative BCVA, αSMA expression, PRPH2 expression, and TGF-β1 level (R = 0.485, R = 0.411, R = 0.444, and R = 0.475, respectively; P < 0.001 for all). Furthermore, postoperative BCVA correlated more strongly with ERM stage and preoperative IRII than with preoperative CST and preoperative EZD (R = 0.392, R = 0.536, R = 0.306, and R = 0.382, respectively; P < 0.001 for all). The ERM stage showed strong correlations with COL1 expression and TGF-β1 level (R = 0.750 and R = 0.697, respectively; P < 0.001 for both), moderate correlations with preoperative CST, preoperative EZD, preoperative IRII, and APN level (R = 0.501, R = 0.432, R = 0.559, and R = 0.455, respectively; P < 0.001 for all), and weak correlations with total cell density and the expression of αSMA, GFAP, CD45, PRPH2, and TGF-β2 level (R = 0.271, R = 0.262, R = 0.255, R = 0.303, R = 0.210, and R = 0.311; P = 0.002, P = 0.004, P = 0.005, P = 0.001, P = 0.021, and P = 0.001, respectively). Preoperative IRII exhibited a strong correlation with TGF-β1 level (R = 0.621; P < 0.001), and moderate correlations with preoperative BCVA and COL1 expression (R = 0.498 and R = 0.463, respectively; P < 0.001 for both). TGF-β1 level strongly correlated with COL1 expression (R = 0.743; P < 0.001); showed moderate correlations with EZD, αSMA, and N-cadherin expression (R = 0.422, R = 0.450, and R = 0.467, respectively; P < 0.001 for all); and had weak correlations with total cell density and the expression of GFAP, CD45, and PRPH2 (R = 0.348, R = 0.394, R = 0.369, and R = 0.397, respectively; P < 0.001 for all). 

Table 3.

 

Spearman and Pearson Correlations Among BCVA, OCT Parameters, Histopathological Features, and Intravitreal Cytokines

 

In our multivariate linear regression analysis after adjustment for age and sex, preoperative IRII showed significant positive correlations with preoperative BCVA and the expression of GFAP, PRPH2, and COL1 (P <0.001, P = 0.004, P = 0.008, and P = 0.047, respectively), but not with preoperative EZD or CD45 expression (P = 0.364 and P = 0.970, respectively). TGF-β1 level showed significant correlations with preoperative IRII; the expression of αSMA, COL1, and N-cadherin; and TGF-β2 level (P = 0.001, P <0.001, P <0.001, P <0.001, and P = 0.020, respectively), but not with preoperative CST, preoperative EZD, or the expression of GFAP, CD45, and PRPH2 level (P = 0.270, P = 0.291, P = 0.073, P = 0.544, and P = 0.334, respectively). 

Biomarker for Predicting Postoperative BCVA or Visual Improvement

Table 4 presents the multivariate analysis examining how key clinical variables, OCT parameters, histopathological features, and intravitreal cytokines influence postoperative BCVA at 6 months and visual improvement (or BCVA change). Preoperative BCVA, preoperative IRII, preoperative EZD, αSMA expression, and PRPH2 expression showed significant relationships with postoperative BCVA (P = 0.006, P = 0.002, P = 0.043, P = 0.042, and P = 0.002, respectively). For visual improvement, significant predictors were preoperative BCVA, preoperative IRII, preoperative EZD, PRPH2 expression, and N-cadherin expression (P <0.001, P = 0.021, P <0.001, P <0.001, and P = 0.038, respectively). Although APN level showed borderline significance for visual improvement (P = 0.056), it did not reach the strict threshold for statistical significance. 

Table 4.

 

Multivariate Linear Regression Analysis of Postoperative BCVA and Visual Improvement

 

Discussion

To the best of our knowledge, this is the first study to comprehensively assess the relationships among four distinct types of data—clinical visual acuity, OCT structural markers, IHC features, and vitreous cytokines—collected on a per-patient basis. Several important findings emerged from this study. 

First, early-stage ERM surgery was associated with better visual outcomes and fewer EMT-related changes in ERM tissues. Postoperative BCVA was more strongly correlated with newer OCT markers, such as ERM stage and preoperative IRII, than with conventional markers such as preoperative CST. This suggests that these new OCT parameters more accurately capture inner retinal changes that impact vision in ERM patients. Second, six of the seven IHC markers (Fig. 5) were positively associated with the new OCT biomarkers, compared to only three markers correlating with preoperative CST. This highlights the superior ability of the new OCT metrics to reflect underlying cellular changes in ERM progression. Third, TGF-β1 level in the vitreous humor was strongly correlated with both ERM stage and preoperative IRII, as well as all seven IHC markers, particularly EMT-related biomarkers such as myofibroblasts, type I collagen, and N-cadherin. However, although TGF-β1 showed a moderate correlation with postoperative BCVA in univariate analysis, this relationship did not persist in multivariate analysis. This suggests that TGF-β1 may primarily drive retinal structural changes through EMT processes in ERM tissues, as observed on OCT, without directly influencing visual acuity. Taken together, these findings reveal broader relationships among various prognostic factors (Fig. 5), suggesting that ERM pathogenesis and visual outcomes are influenced by a wider range of variables than previously recognized. 

Figure 5.

 

Schematic diagram illustrating the correlations among key variables.

Although EZD has long been recognized as a critical predictor of visual prognosis, our previous research demonstrated that retinal thickness in ectopic inner foveal layers also significantly impacts visual acuity and serves as a comparable predictor of visual outcomes.^11,12,30 Govetto et al.⁹ expanded on these findings by developing an OCT-based ERM staging system that links ERM progression with visual impairment. Similarly, Cho et al.¹⁹ demonstrated that preoperative IRII was a better predictor of postoperative BCVA than CST. Consistent with these studies, our findings confirm that both ERM stage and preoperative IRII correlate more strongly with postoperative BCVA, histopathological markers, and cytokine levels than preoperative CST. In our analysis across three groups categorized by ERM progression stages, ERM stage showed a statistically significant correlation with visual acuity, IRII, total cell density, four cell types, and levels of TGF-β1 and TGF-β2. From our multivariate analysis, we identified four key prognostic factors for postoperative BCVA: preoperative BCVA, preoperative IRII, and the protein expression of myofibroblasts and RPE cells. We also found that preoperative IRII was significantly associated with postoperative visual outcomes, specific cell types (glial cells and RPE cells), and type I collagen. These results underscore the value of integrating newer OCT metrics, such as ERM stage and IRII, with histopathological and cytokine data to provide a more comprehensive understanding of ERM progression and its impact on visual outcomes. 

The role of TGF-β in ERM progression is well established, particularly in inducing the transdifferentiation of intraocular cells—glial cells, hyalocytes, and RPE cells—into a mesenchymal phenotype, promoting myofibroblast formation and collagen production.^31,32 However, the specific roles of TGF-β subtypes in ERM progression have been underexplored. In our study, TGF-β1 showed a strong correlation with IHC markers for myofibroblasts, RPE cells, type I collagen, and newer OCT markers, such as ERM stage and preoperative IRII. TGF-β2 was moderately correlated with type I collagen and weakly associated with RPE cells and N-cadherin, whereas TGF-β3 was rarely detected. These findings suggest that TGF-β1 has a greater influence on cellular changes and retinal structure in ERM than TGF-β2 or TGF-β3. 

In line with our findings, Ngan et al.²³ reported that glial cells were the most frequent in ERM and correlated with longer symptom duration and increased CMT. Similarly, Wang et al.²² found that increased glial cell proliferation, higher total cellularity, and poor preoperative BCVA were linked to worse visual outcomes postoperatively. However, they did not find significant associations between myofibroblasts or RPE cells and final BCVA in multivariate analysis.²² In contrast, our study identified an association between the presence of myofibroblasts and RPE cells and poor visual outcomes at 6 months in multivariate analysis. This discrepancy may partly arise from differences in the ERM staging methods used. Wang et al. used the staging system of Hwang et al.,¹⁰ which focuses on retinal thickness and macular dysfunction, whereas we employed the staging scheme of Govetto et al.,⁹ which accounts for the presence and progression of ectopic inner foveal layers. In addition, Wang et al. manually measured cell densities and types as categorical variables,²² whereas we employed automated software-based IHC quantification, which treats cell densities and staining intensities as continuous variables, thereby improving measurement reproducibility and statistical power. 

In our study, both the ERM and the internal limiting membrane (ILM) were removed together during vitrectomy to minimize the risk of ERM recurrence, which made it challenging to isolate pure ERM tissue for IHC analysis. Although this approach may have introduced some ILM material into our ERM samples, the ILM is a much thinner, ECM-rich structure that is largely acellular and composed primarily of type IV collagen and laminin.³³ Additionally, N-cadherin and E-cadherin are typically more abundant in cellular environments such as the ERM, and fibronectin is likely present at much higher levels in the thicker, cell-populated ERM. Even if fibronectin content were not negligible in the ILM, we observed no statistically significant correlations involving fibronectin in our study. Given these considerations, we believe any influence from the ILM on these or other cell-specific IHC markers used in our study would be minimal, allowing our IHC results to predominantly reflect ERM characteristics and pathology. 

Despite the strengths of this study, including its prospective design and large sample size, several limitations should be mentioned. A significant proportion of patients (31.8%) were excluded due to challenges in handling small and thin ERM tissues for IHC or difficulties in segmenting inner retinal layers in Stage 4 OCT images for IRII analysis. The EZD variable may have been overestimated in Stage 4 cases with thicker ERM and retina, as OCT signal attenuation in deeper layers could reduce the accuracy of disruption identification. Additionally, our IHC analysis was conducted on sectioned portions of ERM tissue, which may not fully represent the characteristics of the entire membrane, despite its relatively small size. Future studies employing whole-mount staining and large-scale volumetric imaging techniques could help determine if local variations in IHC characteristics exist across the entire ERM, offering a more comprehensive understanding of spatial heterogeneity within the membrane. Furthermore, while IHC provided valuable insights into protein expression, we did not perform quantitative Polymerase Chain Reaction (qPCR) analysis due to the limited size of ERM tissue samples which precluded simultaneous IHC and qPCR. Future studies incorporating messenger ribonucleic acid (mRNA) analysis from different patient groups will further validate our findings. Finally, Stage 1 ERM cases were not included as they typically do not warrant surgical intervention. 

Conclusions

Our study highlights the complex relationships between intravitreal cytokines, cellular density, the protein expression of various cell types and ECM markers in ERM tissues, and how these factors correlate with OCT morphology and postoperative BCVA. We determined that the new OCT-based ERM staging scheme and preoperative IRII are predictive of visual outcomes after surgery and reflect the extent of EMT changes in ERM tissues. Key prognostic indicators for postoperative BCVA include preoperative BCVA, preoperative IRII, preoperative EZD, and the presence of myofibroblasts and RPE cells. Additionally, TGF-β1 plays a critical role in EMT-like changes in intraocular cells within ERM, influencing OCT structural features and indirectly impacting visual outcomes. These findings support early surgical intervention for ERM and suggest the potential for medical treatments aimed at preventing EMT by targeting TGF-β1, RPE proliferation, or collagen production in ERM tissues. 

Acknowledgments

Supported by a VHS Medical Center Research Grant, Republic of Korea (VHSMC24032) and Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2022R1A6A1A03063039). The sponsor or funding organization had no role in the design or conduct of this research. 

Disclosure: H.S. Yang, None; S. Choi, None; S. Kim, None; C.H. Min, None; D. Kim, None; Y. Lee, None; M. Kim, None; D. Koo, None; J. Ryu, None; J. Kim, None