Introduction

Epiretinal membrane (ERM) is a relatively common retinal disorder (prevalence, 2.2–28.9%), characterized by fibrocellular proliferation on the inner retinal surface, above the internal limiting membrane (ILM) [1,2,3]. ERM involving the macular or peri-macular regions can cause visual impairment, metamorphopsia, micropsia, and occasionally monocular diplopia. The mainstay of treatment of symptomatic ERM is surgical removal of the ERM. Current modern vitreoretinal techniques involve small-gauge vitrectomy and intraocular dyes to facilitate visualization and peeling of the posterior hyaloid, ERM, and ILM [4].

In general, the ILM covers the surface of the retinal nerve fiber layer (RNFL). However, in our surgical cases, we observed several unusual forms of ERM. In these cases, after successful vitreous cortex removal, some part of the retina not covered by the ERM was not stained by indocyanine green (ICG) dye, suggesting the absence of the ILM on the RNFL surface. We called this area “bare RNFL.” We found one case series in our literature review, consisting of 23 ERM cases characterized by the presence of a localized ILM tear, among 268 patients followed up for > 6 months [5]. However, this case series was limited by relative low-resolution optical coherence tomography (OCT); moreover, preoperative OCT and intraoperative ICG staining were only performed in a few cases (14 and 13 cases, respectively). Furthermore, there were no data to associate the bare RNFL with postoperative visual outcomes and morphological restorations.

The purpose of the present study was to describe preoperative and postoperative OCT characteristics of the bare RNFL and ILM tear associated with the ERM and to assess and compare the outcomes of pars plana vitrectomy (PPV) surgery in ERM patients with and without the bare RNFL.

Methods

Patients

This study includes a retrospective consecutive chart review of patients with an ERM who underwent PPV performed by a single surgeon (J.G.K.) at Asan Medical Center, Seoul, Korea, between March 2016 and May 2019. The ERM was diagnosed based on a history of blurred or distorted vision, metamorphopsia and/or micropsia, dilated fundoscopic examination, and OCT. Data collected from patient records included age, sex, axial length, pre- and postoperative best-corrected visual acuity (BCVA), and follow-up period.

Patients with a secondary ERM (retinal detachment, diabetic retinopathy, retinal vascular occlusions, uveitis, trauma, etc.), age-related macular degeneration, glaucoma, or media opacity that could significantly interfere with OCT image acquisition, incomplete chart records, and a follow-up period of < 6 months were excluded. The study was approved by the Institutional Review Board of the Asan Medical Center (IRB No. 2019-1066) and adhered to the tenets of the Declaration of Helsinki. Due to the retrospective study design and use of de-identified patient data, the need for obtaining written informed consent was waived.

The ERM patients were divided into two groups as those with (group A) and those without (group B) bare RNFL and ILM tear. Bare RNFL and ILM tear were detected intraoperatively by the surgeon. Lesions suspected to be a torn ILM or bare RNFL were stained twice with ICG dye during surgery (Fig. 1, see Video, Online resource 1). Subsequently, ILM peeling was initiated from the adjacent stained ILM and directed toward the suspected lesion, confirming the absence or presence of the ILM. The results were re-evaluated by a masked observer (J.H.Y) who carefully studied the operation videos to confirm the surgeon’s intraoperative clinical impressions.

Fig. 1
figure 1

Representative surgical video images and preoperative optical coherence tomography (OCT) with near-infrared fundus images of patients with bare retinal nerve fiber layer (RNFL) associated with the epiretinal membrane (ERM). In eyes with ERM that received 0.25 mg/ml indocyanine green, unstained area of the internal limiting membrane (ILM), suggesting a bare RNFL (arrowhead), is located adjacent and peripheral to the ERM. In addition, OCT images show the torn ILM, RNFL schisis, and intraretinal cyst, which is located in the ganglion cell layer or inner nuclear layer

Full size image

Ocular examination

All patients underwent comprehensive ophthalmologic examinations before and 1, 3, and 6 months after surgery. Examinations included slit-lamp biomicroscopy, refraction, dilated fundoscopy, BCVA measurement, and intraocular pressure measurement. The BCVA was measured using a standard Snellen chart, and the Snellen value was converted to the logarithm of the minimum angle of resolution (logMAR) for statistical analyses. Spectral-domain (SD)-OCT with corresponding near-infrared fundus image was also acquired using Spectralis (Heidelberg Engineering Inc., Heidelberg, Germany). The color fundus photographs were reviewed to correlate with the OCT images. The distribution of the bare RNFL area was analyzed. We divided the posterior pole into five subfields, namely the superior, nasal, inferior, temporal, and central parafovea (within 3 mm) according to the Early Treatment Diabetic Retinopathy Study grid. All images were reviewed independently by one investigator (J.H.Y.) who was blinded to all clinical information.

Patients were stratified into four preoperative stages of the ERM based on the OCT B-scan images as previously described: (1) mild-ERM with negligible morphologic or anatomic disruption, (2) ERM associated with more progressive retinal distortion (absence of the foveal pit), (3) presence of an ERM with continuous ectopic inner foveal layers anomalously crossing the central foveal area (well-defined retinal layer), and (4) ERM complicated by significant retinal thickening and remarkable anatomic disruption of the macula (disrupted retinal layers) [6].

Surgical technique

All surgeries were performed by a single surgeon (J.G.K.). In both groups, hybrid 23–25-gauge PPV and ILM peeling were performed. Cataract surgeries were conducted using phacoemulsification techniques with a clear corneal approach. The vitrectomy was performed using the Alcon Constellation® platform (Alcon Laboratories, Fort Worth, Texas, USA). After performing a core vitrectomy, posterior vitreous detachment (PVD) was induced using a vitreous cutter or extensible intraocular pick. In some cases, to ensure complete separation of the vitreous cortex from the retina, 20% diluted triamcinolone acetonide was injected into vitreous cavity. Using the 23-gauge soft-tip aspiration, remnants of the posterior vitreous membrane in the macular area were removed. After staining, the posterior pole with 0.25% ICG, the ILM was grasped with an ILM forceps and peeled off in a circular manner for approximately two disc diameters around the macula. To confirm if there are any residual ILM, 0.25% ICG was injected into the vitreous again and was immediately washed out.

Statistical analysis

All statistical analyses were performed using SPSS version 20.0 (SPSS Inc., Chicago, Illinois, USA). Means and standard deviations were calculated for continuous variables. Intergroup comparisons were evaluated using the Mann-Whitney U test, Fisher’s exact test, or Chi-square test as appropriate. P < 0.05 was considered statistically significant.

Results

Among the 239 eyes that underwent PPV, a total of 158 eyes with ERM were included in the final analysis. Out of the 158 eyes, 19 (12.03%; mean age, 65.26 ± 7.47 years; age range, 54–81 years) were included in group A and 139 (87.97%; mean age, 65.09 ± 8.84 years; age range, 52–86 years) in group B. The mean axial length was 23.68 ± 1.29 mm in group A and 23.72 ± 0.99 in group B. No significant differences were found between the groups with regard to sex, age, refraction, axial length, or proportion of pseudophakic eyes. Group A had a significantly higher grade of ERM than group B (P = 0.001). Patients’ baseline characteristics are summarized in Table 1.

Table 1 Demographics and baseline clinical characteristics of the study patients
Full size table

Morphologic characteristics of ERM-associated bare RNFL and ILM tear

Table 2 summarizes clinical and OCT characteristics of patients with bare RNFL and ILM tear. Typically, the torn ILM is seen in the OCT images as a thin, mono- or multi-layered scrolled membrane. This finding was present in 18 of 19 eyes (94.74%). Additionally, RNFL schisis, which was previously described [7, 8], was present in 17 of 19 eyes (89.47%). Interestingly, RNFL schisis was observed in limited area adjacent to the torn and scrolled ILM. Figure 2 illustrates examples of bare RNFL and ILM tear with regard to the extent and location of RNFL schisis. Occasionally, an intraretinal cyst was observed at the site of schisis. Intraretinal cysts were predominantly located in the ganglion cell layer and the inner nuclear layer (INL).

Table 2 Pre- and postoperative data of patients with bare retinal nerve fiber layer and internal limiting membrane tear associated with the epiretinal membrane
Full size table
Fig. 2
figure 2

Typical spectral-domain optical coherence tomography (OCT) characteristics of the bare retinal nerve fiber layer (RNFL) and internal limiting membrane (ILM) tear associated with the epiretinal membrane (ERM). The torn and scrolled ILM and ERM are seen as a mono- or multi-layered thin scrolled membrane in OCT (white and red arrows). In addition, the RNFL schisis (yellow arrowhead) can be seen adjacent to the torn ILM. Interestingly, area of the bare RNFL where tangential traction of ERM is relieved shows no inner retinal surface and intraretinal cystic changes (asterisks)

Full size image

We found hyporeflective band on near-infrared fundus images corresponding to the edge of the ERM and torn ILM as seen on the OCT image (Fig. 3). A hyporeflective band was characterized by a convex contour toward the center of the fovea and was present in 14 of 19 eyes (73.68%). The bare RNFL was adjacent and peripheral to this hyporeflective band. Only three eyes (15.79%) presented with parafoveal involvement. In the eyes where the bare RNFL did not involve parafovea, the most frequently involved sites were superior (36.84%), inferior (21.05%), nasal (15.79%), and temporal (10.53%) quadrants (Table 3). Postoperatively, torn and scrolled ILM was removed completely on SD-OCT and the corresponding hyporeflective band was faded out. Although the surgery resolved the RNFL schisis almost completely, some irregular areas of dimpling in the RNFL were evident after 1 week, which resolved gradually within a month (Fig. 4).

Fig. 3
figure 3

Serial optical coherence tomography (OCT) images showing occurrence of bare retinal nerve fiber layer (RNFL) and internal limiting membrane (ILM) tear. a At baseline, the epiretinal membrane (ERM) with foveal pit is seen. b, c Over the next 9 months, the tangential traction secondary to ERM worsens and an ILM tear is first noted at 9 months (white arrow, c). In addition, on near-infrared fundus images, a hyporeflective band (arrowhead) corresponding to the rolled-up edge of the torn ILM, ERM, and underlying RNFL schisis is visible. d At 21 months, RNFL schisis with intraretinal cyst becomes apparent on OCT B-scan and the hyporeflective band moves toward the fovea. Bare RNFL, located adjacent to and peripheral to this hyporeflective band, shows a less-wrinkled retina as the tangential traction is relieved. e, f The ERM and torn ILM were successfully removed, and RNFL schisis with intraretinal cyst is resolved at 3 months postoperatively and observed to be stable for up to 6 months postoperatively

Full size image
Table 3 Distribution of the bare retinal nerve fiber layer
Full size table
Fig. 4
figure 4

Pre- and postoperative optical coherence tomography (OCT) with near-infrared fundus images of patients with bare retinal nerve fiber layer (RNFL) and internal limiting membrane (ILM) tear associated with the epiretinal membrane (ERM). Preoperatively, torn and scrolled ILM and ERM and RNFL schisis are seen. One week after surgery, torn and scrolled ILM are completely removed on OCT, and the hyporeflective band corresponding to the edge of torn ILM and ERM fades out. Although the area of RNFL schisis mostly resolved after surgery, irregular areas of dimpling in the RNFL were observed at 1 week after surgery, which resolved gradually over 1 month after surgery

Full size image

The preoperative near-infrared fundus image also showed discontinuity of retinal wrinkling at the junction of the bare RNFL area. In addition, the bare RNFL area appeared to be normal retinal thickness preoperatively on the macular thickness map generated using the built-in calculation system of SD-OCT.

Visual outcomes and morphological restoration after surgery

The mean preoperative logMAR value (BCVA) was slightly worse for group A than for group B (0.47 ± 0.31 and 0.32 ± 0.17, respectively; P = 0.083) but significantly improved in both the groups to 0.44 (20/55) ± 0.26 and 0.34 (20/44) ± 0.25, 0.22 (20/33) ± 0.17 and 0.21 (20/32) ± 0.19, and 0.21 (20/32) ± 0.16 and 0.15 (20/28) ± 0.17 at 1, 3, and 6 months after surgery, respectively (Table 4). No significant difference in mean BCVA change was found between the two groups.

Table 4 Changes in best-corrected visual acuity and central foveal thickness from baseline through the follow-up period in eyes with or without the internal limiting membrane tear and bare retinal nerve fiber layer
Full size table

The mean central foveal thickness (CFT) progressively improved in both groups, and no significant difference was found between them at 6 months after surgery (P = 0.140, Table 4). However, the mean change in CFT was greater in group A (134.26 ± 97.20 μm) than in group B (61.81 ± 67.81 μm; P = 0.001).

Discussion

Park et al. reported a lower recurrence rate of the ERM in the eyes that had undergone ILM peeling, and since then, several authors have advocated ILM peeling [9,10,11,12], making this maneuver popular in ERM surgeries. During these surgeries, many surgeons injected dyes to stain the ILM for its removal; occasionally, the injection was repeated to ensure adequate staining. There are several causes of negative staining of the ILM, including the ERM and remnant vitreous cortex that can be removed surgically without causing iatrogenic trauma. However, excessive attempts to peel the ILM in the bare RNFL area may damage the underlying retina while grasping or scraping the RNFL, resulting in swelling of the arcuate nerve fiber layer, dissociated optic nerve fiber layer defect, secondary paracentral macular hole, and microcysts in the INL [6, 13,14,15].

In this study, with detailed visualization of the ERM, ILM, and the retinal surface using high-resolution SD-OCT, it was possible to describe the preoperative and postoperative OCT characteristics of the bare RNFL and ILM tear associated with the ERM. Typical appearance of the bare RNFL included torn and scrolled ILM, adjacent RNFL schisis, and intraretinal cyst in the RNFL schisis area. A hyporeflective band corresponding to the rolled-up edge of the torn ILM and ERM was found in the near-infrared fundus images. Postoperatively, these appearances completely resolved on SD-OCT.

Bovey et al. demonstrated the typical appearance of the torn ILM; the fibrous preretinal band consisted of the rolled-up ILM, consistent with our finding in the near-infrared fundus image [5]. However, because of the low-resolution OCT, this study could not characterize any other appearances, such as the inner retinal surface and intraretinal cystic changes. Consequently, the pathomechanism of the bare RNFL and ILM tear associated with ERM remained speculative. They proposed the hypothesis that partial PVD exerts local traction on the ILM, resulting in the detachment of the ILM at this site. However, in our study, an evidence of partial or complete PVD was observed in 13 eyes (68.42%), whereas in 6 eyes (31.58%), the posterior vitreous remained adherent to the macula. Moreover, the fragmented ILM was not observed on the surface of the detached posterior hyaloid membrane.

Recently, Hussnain et al. reviewed their surgical cases and described similar findings of ILM tear, which they labeled as “ILM dehiscence” [7]. From a series of 41 consecutive eyes with ERM, they found RNFL schisis in over half of the eyes, and over three quarters of these eyes with RNFL schisis had a torn ILM. They also described a protrusion of the RNFL schisis through the ILM dehiscence on a horizontal B-scan and labeled it as the “spaghetti sign.” However, as shown in Fig. 2, vertical OCT B-scan taken at the same location as the horizontal OCT B-scan showing spaghetti sign demonstrated typical appearance of the torn and scrolled ILM and ERM. Therefore, we believe that the spaghetti sign on the SD-OCT indicated not only the protrusion of the RNFL schisis but also the torn and scrolled ILM and ERM. In addition, we believe that the hyporeflective band on the near-infrared fundus image in the present study was identical to what they referred to as hyperreflective areas seen on en-face imaging at the vitreoretinal interface.

It has been postulated that ERM contracture results in tangential traction on the ILM and underlying inner and outer retinal layers [5, 14, 16]. We believe that many findings from the present study supported the hypothesis that the ERM contracture contributed to the occurrence of torn ILM and bare RNFL. The ERM commonly results in centripetal forces and contraction of the posterior pole. As the ERM contracts, the maximum amplitude of the retina and tangential traction force become larger. The tangential traction on the ILM partially separates the RNFL, resulting in fibrillary surface changes (RNFL schisis). Under the tractional force of the ERM, the ILM adhering to the ERM finally detaches from the RNFL. The torn ILM, which is no longer supported by the underlying tissues, folds onto itself. Accordingly, this folding is observed on the side of the macula still covered by the ERM and results in a bare RNFL area located adjacent to and peripheral to the torn and scrolled ILM. The bare RNFL and torn ILM are believed to occur more frequently in grade 3 or 4 ERM (Tables 1 and 4), and the maximal retinal thickness was greater in group A than in group B, indicating that the ERM contraction and tangential traction were important factors in bare RNFL occurrence. Furthermore, in some cases, the excessive tangential traction was sufficient to deform the normal-layered retinal architecture, resulting in intraretinal cystic changes. Interestingly, once the ILM was torn and tangential traction was relieved, the retinal tissue was restored, resulting in the resolution of retinal wrinkling. This phenomenon was well observed in the bare RNFL areas, and RNFL schisis was not observed in the peripheral areas of the bare RNFL. Furthermore, after removing the ERM and ILM, typical appearances of the bare RNFL and torn ILM were resolved in most eyes.

Our findings raise the question whether the strong tangential traction produced by the higher grade ERMs is essential for the ILM tear. Although a grade of ERM was significantly higher in group A, two cases of grade 2 ERM also showed a bare RNFL and torn ILM in this study. A possible explanation is that the tangential tractional force caused by the ERM was asymmetric at the macula because the traction force may vary depending on the location, extent, and thickness of the ERM. In addition, multiple foci of retinal traction in the multidirectional pattern of the ERM could cause the ILM tear.

In this study, most bare RNFL areas were observed within the major vessel arcades and on the outside of the parafovea. Overall, the bare RNFL was most frequently observed in the superior and inferior quadrants (12/19 eyes, 63.16%). Interestingly, in these eyes, peripheral boundaries of the bare RNFL were located around the superior and inferior vascular arcades (Fig. 1). The ILM is known to be very thin over the major retinal vessels [17]. Therefore, we hypothesize that the traction force over major vessel arcades may cause the disruption and tearing of the ILM at this site, resulting in the distribution of the bare RNFL.

Previous studies did not evaluate the relationship between the bare RNFL/ILM tear, visual outcomes, and morphological restoration. Interestingly, although our results showed a significantly difference in terms of preoperative ERM grade, preoperative CFT, and preoperative maximum retinal thickness in the macular region between the two groups, there was no significant difference in the preoperative BCVA, which is consistent with Hirano et al.’s findings [18]. At baseline, group A tended to have a higher grade of ERM with a more disrupted retinal architecture, and this difference contribute to the greater improvement in the CFT in group A. However, despite of the greater anatomical improvement in group A, BCVA improvements were not significantly different between the two groups. We believe that the bare RNFL and torn ILM did not seem to influence the visual recovery or cause any deterioration in the preoperative visual acuity.

This study had several limitations. First, data were retrospectively collected by reviewing the medical records. Further prospective study using ultra-wide field OCT will provide a better understanding of the pathophysiologic mechanism and characteristics of bare RNFL and ILM tear. Second, a relatively small number of eyes had ILM tear and our findings should be confirmed in a larger cohort. Third, because most patients were seen at a tertiary care referral and visited our clinic with symptomatic ERM, we could not evaluate the prevalence of ILM tear in the general population. Lastly, we did not analyze the specimens histopathologically nor did we evaluate the functional visual outcomes, such as visual field, microperimetry, or severity of metamorphopsia.

In conclusion, we demonstrated typical appearances of the bare RNFL and ILM tear associated with the ERM. Contracture of the ERM may induce an ILM tear and bare RNFL, which can be seen clinically during PPV as a patch of negative ICG staining. Rolled-up edges of the torn ILM and ERM can also be observed as a hyporeflective band with a discontinuity in retinal wrinkling in the near-infrared fundus image and as a mono- or multi-layered thin scrolled membrane in the OCT images. Attempting to peel some tissues in the bare RNFL area may not be safe as it may cause unnecessary trauma and worsen the visual outcomes. Therefore, it is important that vitreoretinal surgeons meticulously analyze preoperative fundus and OCT images to recognize bare RNFL and torn ILM and minimize unnecessary surgical maneuvers in such areas. We expect that a further study with a larger patients’ cohort, could address the exact effect of bare RNFL on postoperative outcomes and improve the understanding of the underlying pathophysiologic mechanism of bare RNFL and ILM tear.