Introduction

An epiretinal membrane (ERM) is a vitreomacular interface disease that is characterized by pre-retinal proliferation of myofibroblastic cells associated with the extracellular matrix [1,2,3,4]. It is the most common retinal disease in adults. Usually manifesting in patients older than 50 years old, ERM shrinkage induces deformation of the underlying retina, which can cause symptoms of visual impairment and distorted vision (metamorphopsia). While a substantial number of ERMs have no identifiable cause, with idiopathic ERM known to occur from fibroglial proliferation on the inner retinal surface and secondary to a break in the internal limiting membrane during posterior vitreous detachment (PVD), some can develop secondarily in association with existing ocular pathologies, such as diabetic retinopathy, retinal vein occlusion, uveitis, cataract surgery, and previous retinal laser treatment [5,6,7]. Good functional recovery potential can be expected from the appropriate timing of intervention with vitreoretinal surgery, regardless of the idiopathic or secondary cause [8]. However, instances where ERMs separate spontaneously without the need for surgical treatment have been reported. The incidence rate of spontaneous separation ranges from 1% to 5.4% and varies considerably depending on the conditions of the study [9,10,11].

A significant association exists between PVDs and ERM formation, with PVD-induced peeling suggested as a principal mechanism influencing spontaneous membrane separation [9, 12, 13]. However, the specific significance of PVD-induced spontaneous separation of idiopathic and secondary ERM has not been emphasized, and confirming the precise mechanisms through consecutive imaging findings in large cohorts is an urgent need. Identifying the mechanisms and factors that can predict the patients who will experience ERM self-separation is crucial because this can significantly reduce unnecessary financial and psychological patient burden by avoiding surgical intervention for ERM peeling. Considering that advancing age is the most consistent risk factor for ERM [4], with the increasing age of populations worldwide, along with various eye-related diseases that can lead to secondary ERM, the importance of ERM treatment, particularly spontaneous separation, cannot be understated.

In this study, we aimed to conduct a comprehensive review of the clinical features, optical coherence tomography (OCT) imaging findings, and long-term treatment outcomes in consecutive patients who experienced spontaneous separation of ERM. We also compared these features between subgroups divided based on idiopathic and secondary ERM. Additionally, predictive factors for complete ERM separation and favorable vision outcomes following ERM separation were identified.

Methods

This retrospective study was conducted at a high-volume tertiary referral hospital—Severance Eye Hospital—affiliated with the Yonsei University College of Medicine. This study adhered to the principles outlined in the Declaration of Helsinki and was approved by the Institutional Review Board of the Severance Hospital (No. 4–2023-0964). Given the retrospective nature of the study, the requirement for informed consent was waived.

Patients diagnosed with ERM between May 2010 and January 2023 were reviewed, and those with subsequent spontaneous separation during serial follow-up examinations were included. Patients without retinal imaging documenting the presence and separation of ERM were excluded. Self-separation was defined as the disappearance and release of ERM over the foveal area. Complete separation was defined as the complete disappearance of the ERM over the entire macula and reduction of retinal folding. Electronic medical records, including imaging features on OCT and ocular biometric data, were reviewed for each patient.

Idiopathic or secondary ERM was determined by assessing the ocular histories of the patients, and additional subgroup analysis was performed using the primary ERM cause. Secondary ERM was defined as the presence of ERM in patients with various ocular histories, described in Tables 1 and 2.

Table 1 Baseline patient characteristics
Full size table
Table 2 Specific ocular history of the overall cohort
Full size table

The following clinical features and imaging findings of all patients were evaluated: (a) baseline characteristics (age at detection and self-separation, sex, laterality, presence of symptoms, time interval to spontaneous separation, and best-corrected visual acuity [BCVA] converted to the logarithm of the minimum angle of separation [logMAR]); (b) ocular histories; (c) imaging features on OCT; (d) factors associated with complete ERM separation; and (e) factors associated with favorable vision outcome after self-separation. The primary outcome measures were factors predicting complete ERM separation in the macular area and vision gain after separation.

OCT measurements

Multimodal imaging using ultrawide-field laser scanning ophthalmoscope fundus imaging (Optos PLC, Dunfermline, Scotland, United Kingdom) and spectral domain OCT (Heidelberg Engineering, Heidelberg, Germany) were used to determine the presence of ERM and PVD. OCT images were produced using a horizontal spectral domain-OCT cross-section with 25 lines evenly spaced at intervals of 240 mm. Each B-scan was averaged from 25 to 30 individual frames to enhance image quality.

The following OCT features were analyzed and compared between idiopathic and secondary ERM group: presence or absence of epiretinal proliferation, subfoveal ellipsoid defect, intraretinal fluid (IRF), including subretinal and/or IRF, foveoschisis, and dominant undulation of the retina; thickness measurements, including central retinal thickness and subfoveal choroidal thickness; and progression and recurrence of ERM, as well as complete separation over the entire macula.

Dominant undulation was defined as a minimum of two distinct raised or lowered areas measuring over 20 µm than the original contour of each segmented retinal layer at the macula. ERM progression and recurrence were characterized as the worsening of the membrane and appearance of a newly formed membrane at the location where it had previously separated on subsequent follow-up OCT examinations, respectively.

Additionally, imaging features associated with PVD were recorded and compared, including whether PVD was present before and after ERM detection and spontaneous separation in the studied eye and the fellow eye. Typical features of PVD include visible posterior hyaloid separation visible in the vitreoretinal interface on both vertical and horizontal foveomacular line, as well as the optic disc OCT scans; additionally, presence or absence of Weiss rings on ultra-widefield fundus imaging was assessed.

Figures 1 and 2 depict representative OCT images from patients diagnosed with ERM experiencing spontaneous separation through PVD.

Fig. 1
figure 1

Optical coherence tomography (OCT) serial imaging of slowly separating epiretinal membrane (ERM) over 2 years. A–F, The OCT vertical and horizontal sections, obtained at 3-month intervals, showed a hyper-reflective layer over the inner surface of the retina. Elevation of the foveal depression was observed due to traction from the ERM. G and H, After a 6-month follow-up, ERM self-separation was detected with posterior vitreous detachment (PVD). I and J, The most recent follow-up shows maintained ERM self-separation

Full size image
Fig. 2
figure 2

Two additional cases of OCT serial imaging demonstrating ERM self-separation with macular PVD. A and B, The OCT horizontal sections showed ERM with the traction of the foveal depression over a 1-year follow-up. C and D, After the 1-year follow-up, ERM self-separation was observed in conjunction with PVD. E, The hyper-reflective layer of the ERM was well-defined. F, After 3 months, vitreous detachment was in progress with ERM traction. G, Complete ERM self-separation with a clear retinal surface was observed on OCT. H, The most recent follow-up shows well-maintained ERM self-separation

Full size image

Statistical analyses

Statistical analyses were performed using SPSS 26.0 (IBM Corp., Armonk, NY, USA). Sample distribution was assessed using the Kolmogorov–Smirnov test, which revealed that the data followed a normal distribution. A chi-squared test and an independent Student's t-test were employed to compare data between the two groups (idiopathic ERM vs. secondary ERM). The determination that PVD triggered spontaneous ERM separations was made using a paired t-test. Additionally, multivariate stepwise linear regression models using binary logistic regression (backward stepwise likelihood ratio method) were used to assess the factors associated with complete ERM separation and favorable vision improvement after separation. Statistical significance was set at P < 0.05.

Results

Patient demographics and visual outcomes

Table 1 outlines the baseline demographic and clinical characteristics of the enrolled patients. This study included 50 eyes, divided into two subgroups: idiopathic ERM (28 eyes, 56%) and secondary ERM (22 eyes, 44%), all of which underwent spontaneous separation of ERMs. The overall mean age of the patients at ERM detection and self-separation was 54.5 ± 10.6 years (range, 19–72) and 56.6 ± 10.9 years (range, 24–75), respectively, with no significant difference between the two groups. Overall, ERM self-separation occurred over a mean period of 28.1 ± 25.3 months (median: 25.4 months). The idiopathic group experienced a shorter ERM separation period, with a mean duration of 23.4 ± 16.8 months, than the secondary ERM group, which showed a longer mean duration of 34.1 ± 32.6 months (Independent Student's t-test, P = 0.01).

Overall, there was a female preponderance (n = 31, 62.0%), with no significant difference between the two groups (17 eyes [61%] vs. 14 eyes [64%], respectively; P = 0.83). Additionally, no difference was observed in laterality and symptomatic status at ERM detection and separation between the two groups (Table 1).

The mean initial BCVA of ERM detection for the entire cohort was 20/28 (0.15 ± 0.20 logMAR; 0.09 ± 0.15 logMAR and 0.22 ± 0.23 logMAR in the idiopathic and secondary group, respectively). The idiopathic group showed significantly better pre-separation BCVA than the secondary group (P = 0.009). The mean BCVA after ERM self-separation was 20/25 (0.10 ± 0.15 logMAR) in the idiopathic group and 20/30 (0.19 ± 0.23 logMAR) in the secondary group, showing a significant difference (P = 0.01). Compared with 1 of 28 eyes in the idiopathic group, 7 of 22 eyes in the secondary group had gained more than one line of vision (3.6% vs. 31.8%, P = 0.005).

Specific ocular history of the overall cohort

Table 2 provides a detailed ocular history of the overall cohort, consisting of 50 eyes, and indicates which eyes were included in both groups. Among these, 21 eyes (42%) had no ocular history. The most common ocular history included laser photocoagulation (16%), followed by age-related macular degeneration (AMD), cataract surgery, and intravitreal injections (12%; mean number of injections per eye, 0.24). Other ocular histories comprised retinal vein occlusion (6%), uveitis, intraocular tumors or Coats’ disease, diabetic retinopathy (all 4%), and vitrectomy (2%). In one case of secondary ERM, the membrane appeared during follow-up after mild vitreous hemorrhage from an acute PVD but without associated retinal pathology; another case of secondary ERM was noted during the follow-up of a patient who experienced a spontaneously resolved small full-thickness macular hole that formed after release of vitreomacular adhesion. All six eyes with AMD were “dry,” and as such, five were included in the idiopathic group owing to the rationale that aging is a natural process and unlikely to have contributed to ERM formation. All 12 instances of intravitreal injections were performed in the secondary group. Figure 3 shows two cases of secondary ERM self-separation featuring representative OCT and fluorescein angiography images.

Fig. 3
figure 3

Self-separation of secondary ERM cases. Case 1 involved retinal vasculitis with ERM, showing dramatic improvement after 1 month. A–C, Fluorescein angiography of the patient revealed inflammation, indicated by vessel wall leakage. D and E, OCT detected ERM with intraretinal fluid (IRF) and the disruption of the normal retina structure. F and G, After 1 month of steroid therapy for vasculitis, the normal retinal structure recovered via self-separation of ERM, although IRF persisted. H and I, After a 1-year follow-up, self-separation of ERM was well-maintained. Case 2 is assumed to have ERM due to proliferative diabetic retinopathy. Although macular PVD was already present on OCT, complete self-separation of ERM was achieved, suggesting another mechanism for ERM self-separation in this case. J–N, ERM over the inner layer of the retina and pre-released posterior vitreous (thin white line above retinal structure) were observed on OCT. M and O, Spontaneous and complete separation of ERM was detected

Full size image

Specific tomographic features and PVD

Table 3 displays specific imaging features of the overall cohort observed in OCT imaging. Only IRF was more frequently observed in the secondary ERM group (4% vs. 23%; P = 0.04). Overall, the most common OCT findings were dominant undulation and ERM progression (32 eyes, 64%), followed by epiretinal proliferation (34%), foveoschisis (24%), ellipsoid defect (12%), IRF (12%), and subretinal fluid (2%), with no significant difference between the two groups (all P > 0.05).

Table 3 Specific tomographic features and posterior vitreous detachment (PVD)
Full size table

Overall, in both subgroups, most spontaneous ERM separation appeared to have been induced by PVD (PVD presence in pre- vs. post-separation, 18% vs. 100%, respectively; paired t-test, P < 0.001). PVD-induced separation did not significantly differ between the two groups (89% vs. 73%, P = 0.13). Overall, 13 of the 50 fellow eyes had PVD before self-separation of ERM eyes, with no differences between the two groups. Case 2, presented in Figs. 3J–O, showed ERM separation with pre-separation PVD.

Regarding the complete separation of ERM, 26 of 50 eyes (52%) experienced complete release with no sign of remaining macular membranes, and no difference was observed between the two groups (61% vs. 41%, P = 0.16). Figure 4 illustrates OCT imaging over 12 years in a patient who experienced ERM recurrence after partial self-separation.

Fig. 4
figure 4

Recurrence of ERM after partial self-separation. A–F, The OCT vertical and horizontal sections, obtained over 3 years, revealed a hyper-reflective layer over the inner surface of the retina and the elevation of the foveal depression caused by ERM. G–J, Self-separation of ERM occurred, but a small amount of ERM layer was observed in the left part of the vertical section. K and L, Partial self-separation led to worse ERM recurrence and traction of ERM, disrupting the retinal layer. M and N, The most recent follow-up showed severe retinal folding and layer disruption on OCT

Full size image

Factors associated with complete ERM separation

Multivariate logistic regression analysis demonstrated that only two factors were significantly associated with complete ERM self-separation (Table 4). These factors are self-separation interval from the time of detection (odds ratio [OR] 0.936, 95% confidence interval [CI] 0.893–0.981; P = 0.005) and the presence of IRF in OCT (OR 0.049, 95% CI 0.004–0.635; P = 0.02).

Table 4 Factors associated with complete separation of ERM
Full size table

Factors associated with favorable vision outcome after ERM self-separation

Favorable vision outcome was defined as an improvement in vision by more than one line after ERM self-separation. Multivariate logistic regression analysis revealed that two factors were significantly associated with favorable vision outcomes: secondary ERM (OR 15.224, 95% CI 1.228–188.695; P = 0.03) and lower initial BCVA (OR 267.589, 95% CI 1.942–36,874.244; P = 0.03). (Table 5).

Table 5 Factors associated with favorable vision outcome
Full size table

Discussion

We conducted a comprehensive review of 50 patients with ERM who achieved self-separation without surgical intervention. We described baseline demographic factors associated with spontaneous membrane release and sought to determine predictors for favorable anatomical and functional outcomes. Additionally, we compared clinical features and OCT findings between those with idiopathic (28 eyes) and secondary (22 eyes) causes for ERM. We discovered that both groups experienced spontaneous separation of ERM with no significant differences in age, sex, or symptomatic status. Idiopathic ERMs separated faster than secondary ERMs, and the secondary group had a longer follow-up duration and more frequent IRF findings on OCT. Favorable vision outcomes were associated with secondary ERMs and lower initial visual acuity (VA). In contrast, complete ERM separation was associated with a shorter self-separation interval and the absence of IRF in OCT imaging.

Most ERM cases appeared to separate spontaneously with PVD development. While the majority of ERMs occur in association with PVD, with the remnant posterior vitreous cortex left behind from the precortical vitreous pocket in the macular area providing a scaffold, cellular proliferation can occur on the surface of the posterior hyaloid even in the absence of PVD [14, 15]. Thus, in those ERM cases without pre-existing PVD, previous studies have suggested that the primary mechanism behind ERM self-separation involves ERM separation induced by PVD [16,17,18]. However, these studies are limited to individual case reports and small case series with short-term follow-up. Various interventions, such as laser photocoagulation, cryotherapy, surgery, and yttrium aluminum garnet laser procedures, have been used to induce PVD for ERM treatment, with varying degrees of success [19,20,21,22]. In our study, 41 of 50 eyes demonstrated self-separation induced by PVD, making up 82% of the overall study cohort. This finding strongly supports the main mechanism of spontaneous ERM separation, where PVD serves as the primary force vector on the vitreoretinal surface with vertical traction. Previous studies have explored an alternative mechanism for ERM self-separation in the presence of a pre-existing PVD. They proposed that the absence of PVD leads to tangential ERM traction, with stronger contractile forces than adhesion to the retina, resulting in spontaneous ERM separation, similar to tractional forces inducing secondary macular holes [10, 19, 23]. In our study, nine eyes spontaneously separated with pre-existing PVD, suggesting the possibility of separation through this mechanism. All but one eye showed dominant undulation of the retina on OCT, defined as a minimum of two distinct raised or lowered areas measuring over 20 µm than the original contour, which indicates strong tractional forces on the retina. Eight out of nine (89%) eyes with pre-existing PVD showed progression of retinal folding before spontaneous separation. These findings support the idea that stronger contractile forces may induce spontaneous ERM release in eyes with pre-existing PVDs.

Regarding OCT findings, idiopathic ERM typically presents as a straight, hyper-reflective line in front of the inner retinal surface, often accompanied by retinal folding [24]. Most OCT studies have focused primarily on idiopathic ERMs. Prior research has identified a significant morphological difference in OCT images between idiopathic and secondary ERM. Compared with idiopathic ERM, secondary ERM is more likely to display localized adhesions to the retina [25]. Our study revealed that only IRF was more frequently observed in the secondary ERM group among the various features analyzed. A previous study has revealed that IRF spaces are linked to disruptions in the ellipsoid zone, resulting in poorer postoperative vision for patients with ERM [26]. Our multivariate analysis suggests that IRF may be a potential risk factor for achieving complete ERM separation through self-separation, indicating that patients with idiopathic and secondary ERM displaying IRF features on OCT could be considered candidates for surgical intervention to attain a certain level of complete separation.

Furthermore, the complete separation of ERM was associated with a shorter interval for self-separation (Table 4, interval to release detection in months, OR 0.936, P = 0.005). This implies that the chance of achieving complete separation decreases as the follow-up period lengthens. Further research is required to determine the optimal follow-up duration without surgical intervention; however, our study suggests that membrane peeling could be delayed for up to 28 months, which is the overall mean self-separation interval, especially for patients without PVD. Complete separation of ERM was not associated with whether ERM etiologies were idiopathic or secondary, and secondary ERM took significantly longer to self-separate in our study. This highlights the importance of establishing an observation period that considers each patient’s medical condition and underlying eye diseases.

Patients with secondary ERM had poorer vision at the time of initial diagnosis and ERM separation. This could be attributed to the presence of various eye-related diseases that can lead to vision loss. However, favorable vision improvement, defined as a gain of more than one line on the Snellen chart, was more frequently observed in the secondary ERM group. Multivariate analysis also identified secondary ERM as a potential protective factor for favorable vision outcomes after ERM self-separation. Furthermore, surgery for secondary ERM led to a relatively favorable visual outcome and significant postoperative visual outcome improvement [27, 28]. Considering the visual prognosis of ERM progression, we cautiously suggest that careful observation may be a viable alternative to early surgical intervention, as the visual outcomes of spontaneous separation of secondary ERM are not inferior.

Similarly, lower initial BCVA at ERM detection was significantly associated with a favorable vision outcome after ERM self-separation. Patients with worse VA at the time of detection achieved more significant visual improvement after spontaneous separation. This finding aligns with those of previous research that patients with poorer VA before surgery are expected to experience a greater improvement in postoperative VA [29]. Thus, although worse VA and the presence of other eye diseases can lead surgeons to consider early surgical intervention in secondary ERM cases, the results of our study indicating favorable visual outcomes after spontaneous separation may provide reassurance to clinicians to opt for a close progress observation approach.

While our study primarily focused on the spontaneous separation of ERM without surgical intervention, surgery is a known effective and definite treatment for patients with symptomatic ERM with rapid deterioration [4, 24]. Our findings suggest that patients with ERM exhibiting IRF in OCT, those with pre-existing PVD, and individuals who have waited longer than 28 months without significant changes in OCT findings or symptoms may be appropriate candidates for surgery to achieve complete separation of ERM and symptom relief.

The retrospective nature of this study and its focus on a predominantly ethnically homogenous population introduce some limitations. Individuals with more severe ERM conditions might be disproportionately referred to our specialized medical centers, potentially introducing bias in the participant selection process. Additionally, the frequency of spontaneous ERM self-separation could not be determined, as our study exclusively included patients who had already experienced self-separation. Finally, as the study was not blinded, there may be limitations regarding the accuracy of PVD determination: while the determination of PVD presence or absence was reached through evaluation of multiple OCT imaging in conjunction with ultrawidefield fundus photography by retinal specialists, there may be a margin of error depending on the clarity of media or OCT signal strength.

However, this study possesses several strengths. These include the similarity in the initial patient profiles and characteristics, the participation of highly skilled experts from our tertiary academic hospitals, and a substantial sample size comprising individuals with extended follow-up periods and OCT features incorporated into the study. Moreover, considering that many previous studies excluded secondary ERM cases owing to concerns that the underlying disease might confound their effects on VA, we investigated the clinical and predictive factors associated with the spontaneous separation of ERM in secondary ERM cases, comparing them with those of idiopathic cases.

In conclusion, most eyes appear to naturally separate ERM owing to PVD development. Consequently, given the potential for complete spontaneous recovery, our results appear to show that there may be the option for carefully postponing surgical membrane peeling for up to 28 months in eyes without PVD and good visual acuity, whether the cause is idiopathic or secondary. Even in patients with secondary ERM, initially presenting poor vision and showing IRF on OCT, significant visual improvement may still occur after self-separation. However, considering the possibility of ERMs causing permanent damage to the outer retina, there is some room to prudently choose either close follow-up or early surgical intervention for ERM patients with decreased vision and/or visual symptoms including metamorphopsia.