Among the 141 eyes with PPV for ERM removal, five eyes (3.5%) developed MH postoperatively. The baseline characteristics, clinical parameters, and comparison between MH and non-MH groups are presented in Table 1. The mean age at vitrectomy for ERM was 65.70 ± 8.65 years. The mean preoperative BCVA by logMAR was 0.32 ± 0.27 (Snellen, 20 / 42). The mean preoperative CMT in OCT was 440.2 ± 94.6 μm. ILM peeling was performed during ERM surgery in 77 of 141 patients (54.6%). The mean follow-up duration was 27.28 ± 22.50 months (range, 6-112 months). The proportion of the preoperative foveal configuration of ERM was significantly different between the MH and non-MH groups (p = 0.008). The preoperative CMT was significantly thinner in the MH group (332.2 ± 77.4 μm vs. 444.6 ± 92.5 μm, p = 0.007). Age, sex, diabetes mellitus, hypertension, preoperative axial length, preoperative BCVA, preoperative EP, preoperative lens status, intraoperative ILM peeling, intraoperative air tamponade, surgeon’s experience, and follow-up duration were not significantly different between the MH and non-MH groups. In subgroup analysis, the proportion of surgeons with more than 10 years of surgical experience was significantly higher in the MH group within the ILM-peeled subgroup (100% vs. 18.7%, p = 0.041) (Supplementary Table 1). In the no-ILM peeling subgroup, the proportion of the preoperative foveal configuration of ERM was significantly different between the MH and non-MH groups (p = 0.010). The MH group showed significantly thinner preoperative CMT (283.7 ± 30.7 μm vs. 422.4 ± 91.4 μm, p = 0.003) (Supplementary Table 2).

Table 2 summarizes the clinical characteristics and surgical outcomes of the cases with MH formation in the present study. All five cases developed foveal MH. There were no eccentric MHs among the 141 eyes with vitrectomy for ERM removal. Representative cases are presented in Fig. 2A-2N (case 1) and Fig. 3A-3N (case 4). The remaining cases are shown in Supplementary Figs. 1-3. The mean age of the patients with MH formation at MH diagnosis was 67.80 ± 10.99 years. Three were female. The mean preoperative BCVA by logMAR was 0.46 ± 0.31, and the mean preoperative CMT was 332.2 ± 77.4 μm. Preoperative foveal configurations included ERM with LH in three eyes, ERM with MPH in one eye, and ERM only in one eye. EP was present in cases 1 and 2. All five patients underwent combined phacoemulsification and PPV with ERM removal; ILM peeling was performed in cases 1 and 5. In case 2, surgical induction of PVD was performed during vitrectomy, whereas all other patients had a preexisting PVD at the time of surgery. MH was diagnosed an average of 21.41 ± 23.91 months (range, 0.92-51.28 months) after the surgery for ERM removal. The average MH basal diameter was 822.40 ± 510.08 μm (range, 347-1,539 μm). In all five cases (100%), ERM was observed over time after ERM surgery and at the time of MH detection. Specifically, case 2 presented with a hazy OCT image on postoperative day 1, which made accurate evaluation difficult (Supplementary Fig. 1); however, ERM was observed 1 week after surgery and was considered residual ERM. In contrast, ERM recurrence was observed in case 1 at postoperative 3 months (Fig. 2), case 3 at postoperative 6 weeks (Supplementary Fig. 2), case 4 at postoperative 1 month (Fig. 3), and case 5 at postoperative 2 years (Supplementary Fig. 3). In four of the five patients (80%), CME was present at the MH detection. One patient (case 5) was transferred to another hospital for MH surgery at the patient’s request. Four patients underwent additional PPV and ERM/ILM peeling, and pneumatic tamponade for MH. In three out of four cases, MH was closed successfully. The patient with unclosed MH (case 2) refused to undergo reoperation. BCVA by logMAR were 0.74 ± 0.51 before MH surgery and 0.76 ± 0.49 postoperatively, with no statistical significance.

In univariate Firth bias-reduced logistic regression analysis, MH formation after PPV for ERM removal was significantly associated with the preoperative foveal configuration of ERM with MPH (odds ratio [OR], 14.07; 95% confidence interval [CI], 1.05-190.53; p = 0.046), ERM with LH (OR, 25.91; 95% CI, 3.82-288.85; p = 0.001), and preoperative CMT (OR, 0.98; 95% CI, 0.96-0.99; p = 0.002) (Table 3). The variance inflation factor (VIF) was also calculated to check the possible multicollinearity among the factors. The VIF values derived from the Firth regression analysis were 1.46 for ERM with MPH, 1.63 for ERM with LH, and 1.08 for preoperative CMT, all well below the cutoff of 10. Since no variables exceeded a VIF of 10, multicollinearity was considered absent. Consequently, ERM with MPH, ERM with LH, and preoperative CMT were included in the Firth logistic regression multivariate analysis model. In multivariate analysis, the preoperative foveal configuration of ERM with LH remained significantly associated with the MH formation (OR, 13.11; 95% CI, 1.56-160.50; p = 0.018). Preoperative CMT showed marginally significant association with the MH development (OR, 0.98; 95% CI, 0.97-1.00; p = 0.075).

In this study, the incidence of MH formation after vitrectomy for ERM removal was 3.5%. The interval from ERM surgery to detection of MH varied widely, ranging from 0.92 to 51.28 months. In the three patients who did not undergo ILM peeling, MH was found relatively early, between 0.92 and 10.82 months after surgery. In contrast, MH was found at a relatively later time in the ILM peeling group, at 42.64 and 51.28 months. The basal diameter of the MH also varied widely, ranging from 347 to 1,539 μm. In all five cases (100%) with MH formation, ERM was commonly observed at MH detection. This was due to residual or recurrent ERM following ERM surgery. In four of the five patients (80%), CME was present at the time of MH detection. Preoperative foveal configuration of ERM with LH was the significantly associated factors for MH formation after ERM surgery. Thin preoperative CMT was marginally significant factors for MH development. Among the four patients who underwent surgery for FTMH, MH was successfully closed in three cases (75%). There was no significant difference in BCVA before and after surgery among the four patients who underwent MH surgery.

Few studies have reported on the incidence of MH formation following ERM surgery. Moreover, most of these investigations analyzed the occurrence of nonfoveal, eccentric MHs, with reported incidence rates of 0.95%, 1.8%, and 5.6% [12,14,15]. In the study by Rush et al. [13], MH developed in 11 of 423 eyes (2.6%) that underwent PPV for removal of the ERM; among these, 2.1% were eccentric MHs and only 0.5% were central (foveal) MHs. In the present study, the overall incidence of MH following ERM surgery was 3.5%, which is relatively higher than the previous study (0.5%). In the study by Rush et al. [13], cases involving a combined procedure (i.e., cataract surgery performed concurrently with PPV) were excluded, and ILM peeling was performed in all cases. In contrast, the present study included cases with combined procedures and ILM peeling was performed in 55% of eyes.

In the present study, both the interval from ERM surgery to MH detection and the basal diameter of the MH varied widely. In contrast, in the study by Rush et al. [13], the two cases of foveal MH developed relatively early at 2 and 6 weeks postoperatively, and the hole diameters were comparatively small at 245 and 276 μm, respectively.

Several mechanisms have been proposed to explain the development of eccentric MH following ERM surgery: (1) iatrogenic trauma occurring when grasping the ERM or ILM with microforceps; (2) Müller cell damage caused by ILM peeling itself, which may lead to glial cell and photoreceptor apoptosis, resulting in FTMH [12,20]; (3) contraction of residual ILM or ERM after surgery; and (4) concurrent CME, whereby the opening of intraretinal cysts may lead to MH formation [16]. Rush et al. [13] suggested the presumed mechanism for foveal MH formation after ERM removal. They suggested that postoperative contracture of the residual ILM or the presence of CME could be associated with the MH formation. However, they found no association between preoperative factors and the subsequent development of MH.

In the current study, preoperative configuration of the ERM with LH was significant risk factor for the development of FTMH after ERM surgery. There have been rare reports of FTMH formation following vitrectomy for LH. Witkin et al. [21] performed vitrectomy in 4 out of 19 eyes with LH that were associated with visual decline and reported FTMH formation in one case. Parolini et al. [22] reported that in 3 out of 19 eyes with LH-associated ERM, FTMH developed after PPV with ERM and ILM peeling. Neither of these studies provided a clear discussion on the mechanism of FTMH formation. Lim et al. [23] reported a case of ERM with LH where FTMH developed after PPV and combined ERM/ILM removal, but the FTMH spontaneously closed during follow-up. They hypothesized that CME may have contributed to FTMH formation, based on the presence of small cystic changes at the time of hole development. Chehaibou et al. [24] reported an observational study of 20 eyes with LH that progressed spontaneously to FTMH. While 70% of these eyes had accompanying ERM, 30% did not, suggesting that factors other than the tangential traction by ERM may play a role in the formation of FTMH. Under normal conditions, the structural stability of the fovea is maintained by Müller cells forming the Müller cell cone and Müller cells from the foveal wall. However, disruption of the Müller cell cone occurs in LH [25-27], and degenerative changes may occur in the outer nuclear layer, which is destabilized by the loss of central Müller cells [24]. These structural alterations may lead to progressive cavitation of the LH and eventual progression to FTMH [24,25]. Based on these findings, it is plausible that structural weakening of the fovea due to Müller cell cone disruption in LH could explain why preoperative LH may serve as a risk factor for FTMH formation following ERM surgery, as observed in our study.

Preoperative thin CMT is marginally significant risk factors for the development of MH after ERM surgery. There are two main theoretical mechanisms regarding the pathogenesis of secondary MH formation: tangential traction exerted by an ERM and the development of CME [28]. In CME, cysts can directly rupture to form a FTMH, or roof of the cysts slowly dehisce leading to a FTMH by degenerative forces [28]. When the preoperative CMT is thin, it is presumed that the increasing tangential traction from residual or recurrent ERM over time, or the postoperative development of CME in the thinned vulnerable foveal tissue, may lead to the formation of a FTMH. In the study by Kang et al. [29], among 38 eyes that developed secondary MHs after initial vitrectomy for various vitreoretinal disorders, 27 eyes (71%) had evidence of ERM formation, suggesting that tangential traction from ERM may significantly contribute to the formation of secondary MH. Additionally, it was proposed that the reopening of successfully closed MH may be caused by the proliferation of a secondary ERM around the hole in most cases [30]. In the present study, among three patients who did not undergo ILM peeling during ERM surgery, secondary MHs were found at 0.92, 1.38, and 10.82 months postoperatively. These intervals were comparatively shorter than those observed in two patients who had undergone ILM peeling, in whom MHs were detected at 42.64 and 51.28 months postoperatively. This finding may be related with the fact that postoperative contracture of residual ILM could be an additional contributing factor in MH formation [13]. Furthermore, Jain et al. [11] reported that CME was present in 12 out of 29 eyes (41.3%) with secondary MHs following vitrectomy. In another recent study, 14 patients with small FTMHs (mean diameter, 166 μm) accompanied by CME were treated with topical corticosteroids, nonsteroidal anti-inflammatory drugs, and carbonic anhydrase inhibitors, resulting in successful hole closure in all cases and visual improvement from a mean BCVA of 20 / 70 to 20 / 40 [31]. In the present study, recurred or residual ERM was observed in all five patients who developed secondary MHs after ERM surgery. CME was concurrently observed at the time of MH formation in four of the five patients (80%). These findings suggest that in eyes with thinner preoperative foveal tissue, tangential traction from ERM or the development of CME may have contributed to the progression to FTMH. Another possible mechanism may involve iatrogenic traumatic injury during ERM removal. In two patients (cases 2 and 3, 40%), MHs were detected at 0.92 and 1.38 months postoperatively. Although this raises the possibility of an iatrogenic mechanism during initial vitrectomy, no evidence of a FTMH was seen intraoperatively or on OCT at 1 week postoperatively for these two patients (Supplementary Figs. 1, 2).

EP is primarily derived from Müller glial cells in the inner retina, distinguishing it from ERM, which is composed of myofibroblasts and demonstrates contractile properties. Yang et al. [32] suggested that the presence of EP may indicate chronic and severe gliosis accompanied by neurodegenerative changes. They reported patients with hole marginal ERM combined with EP are more likely to have failed closure of the MH (38%) compared with patients with no ERM (11.8%). This suggests that EP may also have a potential role in the development of MH after ERM surgery. However, in the present study, there was no statistically significant difference in the proportion of EP between the MH and non-MH groups, and the presence of preoperative EP was not a significant factor for MH formation following ERM surgery.

Among the four patients who underwent surgery for FTMH in the present study, MH was successfully closed in three cases. There was no significant difference in BCVA before and after surgery among the four patients who underwent MH surgery. However, in two cases, visual acuity measurements were incomplete: case 1 lacked postoperative BCVA, and case 2 lacked preoperative BCVA. Even when only cases 3 and 4, for which both preoperative and postoperative BCVA measurements were available, were analyzed, no significant difference was observed between preoperative and postoperative BCVA. Rush et al. [13] previously reported two cases of foveal FTMH development following ERM surgery, both of which achieved successful closure of MH. However, preoperative BCVA for MH surgery was not separately measured in that study; instead, the authors compared preoperative BCVA for ERM surgery and postoperative BCVA for MH surgery, showing a decrease from 0.5 to 0.4 and from 0.6 to 0.4, respectively, for the two cases of foveal FTMH. The hole diameters in those two cases were 245 and 276 μm. Additionally, those holes were detected at 2 and 6 weeks after ERM surgery and the postoperative BCVA were reported at 6 months postoperatively after ERM surgery, indicating a relatively short follow-up period after MH surgery. These differences in study design and timing make direct comparison with our study difficult. Further systematic studies with larger case numbers are warranted to better understand the anatomical and functional outcomes after MH formation following ERM surgery.

This study has several limitations. First, this study had a retrospective design and had an inherent risk of selection bias. Second, the size of the cases in the MH group was small. This may make the reliability of conventional logistic regression analysis questionable due to insufficient statistical power. Therefore, we analyzed the data using Firth bias-reduced logistic regression, which is more suitable for rare event data. The subgroup analysis by ILM peeling included fewer cases per subgroup, which may have limited the statistical robustness of the logistic regression model, even when applying Firth bias-reduced logistic regression. Therefore, we did not present a logistic regression model for the subgroup analysis by ILM peeling; instead, we provided only the comparison of the clinical parameters between the MH and non-MH groups for the subgroup analysis. Although Firth bias-reduced logistic regression was applied to account for rare events, the limited number of outcome cases and the heterogeneity of surgical techniques restrict the generalizability of the results. In addition, variations in instrumentation, dye usage, and peeling methods, which were not standardized across cases, may also have influenced the outcomes. Given these considerations, the current results should be interpreted with caution. However, this study included all analyzable cases of MH that developed following ERM surgery performed at our institution. The study subjects were comprehensively identified from the entire cohort of patients diagnosed with and treated surgically for ERM at our institution without data loss by extracting data by combining clinical data warehouse and electronic medical records. The study included all surgically treated ERM cases with available Heidelberg OCT data collected since the device was introduced at our hospital in January 2012. Lastly, ILM peeling was not consistently performed during ERM surgery but was selected based on the surgeon’s preference and the individual characteristics of each patient case. Despite these limitations, to our knowledge, this study is the first study to systematically report the largest cohort of foveal FTMH, rather than eccentric cases, that occurred following ERM surgery.

In conclusion, ERM with LH was identified as a significant factor associated with MH formation after ERM surgery, while thin preoperative CMT showed a marginal association. At the time of MH detection, recurred or residual ERM and postoperative CME were observed in most cases, suggesting that tangential traction caused by postoperative ERM or, postoperative CME may represent possible etiologies for MH formation. Therefore, in patients with ERM with LH or thin preoperative CMT, the possibility of MH formation after ERM surgery should be taken into account, and these patients may warrant closer postoperative monitoring for the early detection of MH development.