Macular hole (MH) is an important cause of central vision loss and distortion, with an estimated global prevalence of 0.02% to 0.8%.1 Pars plana vitrectomy (PPV) is the definite treatment, and internal limiting membrane (ILM) peeling has been the procedure of choice by most retina surgeons in this context, as it is believed the ILM-related tangential traction has an important role in the etiology and success of MH surgery.2,3
In 2010, Michalewska described the inverted ILM flap as an effective alternative in large MHs. Briefly, with this technique, the ILM is peeled off around the MH, retaining its edges attached to the edges of the hole. Then, the peeled segment is rolled/inverted over the MH, covering the defect.4 To date, there is still debate regarding the anatomic and functional results of this technique and whether it is beneficious for all types of MHs.5,6
Our aim is to compare the anatomic and functional results of the complete ILM peeling with the inverted flap technique for MH surgery. This study protocol was registered under the PROSPERO registration code CRD42018092131.
Patients and Methods
Eligible studies included patients with a MH of any size with at least 6 months of postoperative follow-up. Both prospective and retrospective designs were allowed. Only comparative studies of both techniques were considered. No restrictions were made regarding age or etiology.
An electronic database search was performed in PubMed, CENTRAL, and ClinicalTrials.gov, in addition to hand-searching through reference lists of articles. The search was last run on July 9, 2019. A detailed search strategy is provided in Table A (available at www.healio.com/OSLIRetina). Whenever necessary, we made efforts to contact the authors of published trials for further information.
Data Collection and Risk of Bias Assessment
Two investigators independently screened the studies for eligibility, evaluated the quality, and extracted the data according to PRISMA7 and MOOSE8 guidelines. Risk of bias was assessed with a modified scale based on the Newcastle-Ottawa Scale9 for comparative cohort studies, as well as on the Cochrane tool for assessing risk of bias10 for randomized controlled trials (RCTs). Particular attention was paid to imbalances regarding baseline characteristics, surgical expertise, and follow-up time. A rule of thumb was applied to define follow-up times superior to 12 months as most adequate for the purposes of study quality assessment. Disagreements were solved by consensus or a third author.
From each study, the following data were extracted if available: time of symptomatic evolution, best-corrected visual acuity (BCVA, logMAR), ILM peeling technique, tamponade agent used, surgical-related adverse events (AEs), positioning schedule, follow-up time, and MH closure.
Data Synthesis and Analysis
Considering the clinical/surgical differences inherent to different types of MH, the following groups were considered: idiopathic large MH, myopic MH without retinal detachment, and myopic MH retinal detachment (MHRD).
The primary outcome was the MH closure rate, expressed as risk ratio (RR). BCVA improvement until last observation and surgery-related AEs were defined as secondary outcomes.
Heterogeneity was assessed using the I2 statistic, respecting the following thresholds: 0% (no heterogeneity), 0% < I2 < 25% (mild heterogeneity), 25% ≤ I2 < 50% (moderate heterogeneity), and 75% ≤ I2 (great heterogeneity).10,11 A fixed-effects model was assumed unless significant heterogeneity was present (I2 > 50%), in which case a random-effects model was used instead.
Bias was evaluated using a funnel plot of the data. Sensitivity analysis was performed to confirm that our findings were not driven by any single study when I2 > 25%.11 Whenever data were presented as median and range, mean and standard deviation (SD) were appropriately pooled.12
Throughout this paper, for purposes of clarity, results of the conventional ILM peeling group will be presented first, followed by results of the inverted flap technique.
Review Manager V.5.313 (London, England) and STATA V.1514 (StataCorp, College Station, TX) were used as statistical packages. All tests were two-tailed, and a P value of less than .05 was considered significant.
An electronic database search identified 1,191 studies, with six additional relevant papers identified through other sources. After duplicates were removed, 1,149 studies were assessed through title and abstract, with 34 studies proceeding for full-text assessment.
Eighteen studies met eligibility criteria: 16 retrospective5,6,15–28 and two RCTs.4,29 Two retrospective trials (Iwasaki et al.18 and Lee et al.19) were included only in the qualitative synthesis for enrolling patients with MH closure status. The PRISMA7 study flow diagram is shown in Figure 1. Reasons for study exclusion on full-text assessment are also provided in Table B (available at www.healio.com/OSLIRetina). The characteristics of the included studies are summarized in Table 1.
Study flow diagram. ILM = internal limiting membrane; MH = macular hole
Reasons For Study Exclusion on Full-Text Assessment
Characteristics of Included Studies
Risk of Bias of Included Studies
Studies were classified as having overall moderate methodological quality. Most studies presented clear criteria for patient selection except three,5,6,19 where it is unclear if patients were consecutively enrolled to prevent selection bias. Both RCTs4,29 described simple randomization methods. Ascertainment of exposure was securely performed from surgical/clinical/imaging records across studies. All studies provided baseline demographic data for both study groups. When assessing comparability of groups, the following issues were identified: unreported duration of symptoms in 11 studies,5,6,18–21,24,26–29 unreported MH diameter in seven studies,5,15,16,20,24,25,27 MH diameter significantly larger in the inverted flap group in six studies,4,17,18,21,22,27 unreported mean axial length (AL) per study group in six studies,4,19,22,24,27,29 and unreported lens status/rates of concomitant cataract surgery in three studies.19,24,27
Indication bias was identified in two studies,23,24 as the authors reported to preferentially apply the inverted flap technique for most complicated/larger/chronic cases. Outcome assessment was performed by a blinded physician in three studies.4,17,25 Mean follow-up time was considered most adequate in five studies.4,17,19,25,28
The overall effect of the surgical learning curve could not be excluded, as nine studies suggested the inclusion of the first cases of inverted flap approach, after technique transition.6,15–17,20,21,23,26,28
One study included 620 patients,24 contributing with a study weight of 44% to this analysis. However, this paper was not included in the subgroup analyses for presenting combined results for myopic and idiopathic MH, with irretrievable data.
Summary of Results
A total of 1,469 eyes with MH were included in qualitative synthesis, with 1,403 eyes proceeding to quantitative analysis (733 complete ILM peeling, 670 inverted flap). Mean age across studies was 67.3 years ± 10.6 years (n = 1,469, 67.3 years ± 10.5 years vs. 67.3 years ± 10.6 years; P = .53; I2 = 6%). Most eyes were from female patients, with a similar proportion in both study groups (73.3% vs. 71.4%; P = .55). Overall, no differences were found between groups regarding preoperative BCVA (P = .44), proportion of phakic patients (P = .87), proportion of patients under phacoemulsification during the study period (P = .37), and mean follow-up time (P = .43). MH mean minimum diameter (531.1 μm ± 188.8 μm vs. 602.8 µm ± 223.8 µm; P < .001) and time of symptomatic evolution (10.4 months ± 20.2 months vs. 12.0 months ± 18.4 months; P < .01) were higher in the inverted flap group.
MH closure was 25% more likely in subjects submitted to inverted flap technique (76.1% vs. 93.6%; random-effects RR: 1.25; 95% confidence interval [CI], 1.14–1.38; P < .0001) (Figure 2). We re-ran the analysis without Rizzo24 for the purposes of sensitivity (largest included study), and the RR increased even further to 1.31 (P < .0001).
Macular hole closure rate (overall). ILM = internal limiting membrane; CI = confidence interval
Idiopathic Large Macular Holes
Four papers were included in this subgroup analysis4,22,27,29 comprising a total of 362 eyes (complete ILM peeling, n = 204; inverted flap, n = 158). Mean age was 66.5 years ± 8.4 years vs. 65.0 years ± 9.3 years (P = .45), and female patients were predominant in both groups (67.6% vs. 69.6%; P = .61). Mean AL was not reported in any of the studies. Mean MH minimum diameter was superior in the patients submitted to the inverted flap approach (609.8 μm ± 165.7 μm vs. 695.5 μm ± 216.5 μm; P < .001). Two papers described the mean duration of symptomatic evolution, which was well-balanced between study groups (20 months vs. 19 months and 8.7 months vs. 8.6 months).4,22
Mean baseline BCVA was 0.92 ± 0.28 logMAR vs. 0.99 ± 0.39 logMAR (P = .22), with balanced proportions of phakic patients (89% vs. 87%; P = .62) and patients submitted to phacoemulsification during study period (39% vs. 36%; P = .31). For both study groups, the most frequent tamponade was SF6 (complete ILM peeling: SF6 [59%], C3F8 [13%], air [28%]; inverted flap technique: SF6 [53%], air [36%], C3F8 [10%], and silicone oil [1%]). On average, prone positioning was performed for 3.7 days (range: 2 days to 7 days), although one study did not specify the duration of postoperative positioning.25 Mean follow-up time was 7.7 months in both study groups.
MH closure rate was 12% more likely in patients who underwent the inverted flap surgical approach (87% vs. 96%; fixed-effects RR: 1.12; 95% CI, 1.05–1.20; P < .001) (Figure 3, bottom graph).
Macular hole closure rate: subgroup analysis. CI = confidence interval
Myopic Macular Hole Without Retinal Detachment
Four studies were included in this subanalysis,6,17,21,26 enrolling 133 eyes (71 complete ILM peeling, 62 inverted flap). Groups were well-balanced in respect to the following variables: age (59.0 years ± 9.5 years vs. 60.3 years ± 10.2 years; P = .89), female gender (73% vs. 72%; P = .83), mean AL (30.3 mm ± 2.4 mm vs. 29.4 mm ± 2.4 mm; P = .40), presence of posterior staphyloma (82.2% vs. 75.7%; P = .35), presence of macular schisis (18.5% vs 18.3%; P = .96), baseline BCVA (0.79 ± 0.47 logMAR vs. 0.93 ± 0.56 logMAR; P = .31), and proportion of phakic patients (55.6% vs. 59.2%; P = .73). Mean MH minimum diameter was significantly higher in patients submitted to the inverted flap approach (347.2 μm ± 146.3 μm vs. 459.1 μm ± 216.2 μm; P < .001). One trial reported equivalent duration of symptomatic evolution (10.6 months vs. 8.2 months).17 No differences were found in the percentage of phacoemulsification during study period (31.7% vs. 26.9%; P = .58).
The following tamponade agents were used for each technique: complete ILM peeling (SF6 51%, C3F8 49%); inverted flap (SF6 55%, C3F8 45%). On average, postoperative prone positioning was maintained for 5.4 days for both groups (range: 3 days to 14 days). Mean follow-up time was 13.7 months versus 8.8 months (P = .11).
MH closure rate was 35% more likely with the inverted flap technique (70.4% vs. 95.2%; fixed-effects RR: 1.35; 95% CI, 1.14–1.59; P < .001) (Figure 3, middle graph).
Myopic Macular Hole Retinal Detachment
Seven studies were included in this subanalysis,5,15,16,17,20,25,28 assessing a total of 198 eyes (114 complete ILM peeling, 84 inverted flap technique). Mean age was balanced across groups (66.6 years ± 10.2 years vs. 65.9 years ± 11.1 years; P = .96), and both groups had a female predominance (83% vs. 80%; P = .78). Mean AL was superior in the complete ILM peeling subgroup (29.8 mm ± 4.1 mm vs. 29.1 mm ± 1.9 mm; P = .02). Of the seven studies included, three reported the percentage of posterior staphyloma in both groups (89% vs. 91%; P = .85).5,15,20 One trial compared the height of the posterior staphyloma, which was similar between groups (651 μm ± 123 μm vs. 642 μm ± 240 μm; P = .955).28
Mean duration of symptoms was higher in the inverted flap technique subgroup (5.8 months ± 8.4 vs. 9.4 months ± 21.4 months; P < .01). Mean preoperative BCVA was 1.3 ± 0.5 logMAR vs. 1.3 ± 0.6 logMAR (P = .79), with a similar proportion of phakic patients in both groups (56% vs. 60%; P = .84). Of the total sample, 40% of patients were submitted to phacoemulsification combined to vitreoretinal surgery or during follow-up (41% vs. 39%; P = .73). The following tamponade agents were used for each technique: complete ILM peeling (C3F8 68%, SF6 16%, silicone oil 12%, unspecified gas 4%), inverted flap technique (C3F8 63%, SF6 27%, silicone oil 10%). Mean duration of postoperative prone positioning was 7.5 days. Mean follow-up time was 14.0 months vs. 11.6 months (P = .36).
MH closure rate was almost doubled in patients operated with the inverted flap technique (47.4% vs. 90.5%; random-effects RR: 1.89; 95% CI, 1.31–2.73; P < .001) (Figure 3, top graph).
BCVA improved more from baseline until last observation with the inverted flap technique (logMAR mean difference [MD] −0.11; 95% CI, −0.21 to −0.01; P = .03) (Figure 4). Re-analysis without Rizzo et al.22 (sensitivity analysis) did not significantly change results (logMAR MD −0.12; 95% CI, −0.24 to 0.00; P = .05). As for subgroup analyses, no differences were found between study arms regarding BCVA improvements: idiopathic large MH (MD −0.16 logMAR; 95% CI, −0.36 to 0.05; P = .13), myopic MH without retinal detachment (MD −0.05 logMAR; 95% CI, −0.38 to 0.29; P = .78), MHRD (MD −0.18 logMAR; 95% CI, −0.41 to 0.04; P = .11) (Figure 5).
Mean difference from baseline in best-corrected visual acuity (overall). ILM = internal limiting membrane; CI = confidence interval; SD = standard deviation
Mean difference from baseline in best-corrected visual acuity: subgroup analysis. ILM = internal limiting membrane; CI = confidence interval; SD = standard deviation
Only Álvarez et al.6 reported the occurrence of postoperative complications: one case of choroidal detachment spontaneously resolved, and five cases of ocular hypertension (three ILM peeling, two ILM flap).
Other Reported Variables
Iwasaki et al.18 enrolled 24 patients with successfully closed large MHs to compare both surgical techniques regarding the postoperative recovery of external limiting membrane (ELM) and ellipsoid zone (EZ). Once again, patients in the inverted flap technique group presented with larger MH minimum diameters (551.1 μm ± 99.5 μm vs. 655.2 μm ± 112.1 µm; P = .053). This group presented with lower recovery rates for ELM (70% vs. 21.4%; P = .024) and EZ (30% vs. 0%; P = .059), as well as longer recovery periods for this layer (3.4 months ± 2.8 months vs. 11.0 months ± 1.7 months; P = .015).
Lee et al.19 analyzed the postoperative vascular displacement of 42 patients with idiopathic large MHs submitted to either one of these techniques in order to study macular topographic changes. No significant baseline discrepancies were detected. Overall, no differences were found regarding horizontal, vertical, and distance of vascular displacement.
Our analysis included 18 studies (16 retrospective, two RCT) comparing the conventional complete ILM peeling with the inverted flap technique for MH surgery, enrolling a total of 1,469 eyes. The inverted flap technique was associated to an increased chance of MH closure, as observed for all subgroups. When considering myopic MHRD, this new technique almost doubled the rate of anatomical success, similar to the results found by Yuan et al. in their meta-analysis on this subgroup.30 It is suggested the inverted flap technique, by covering the hole with an “ILM scaffold,” provides a matrix for glial and other cells proliferation, thus enhancing the rates of MH closure.30
Although subgroup analyses showed visual results similar to conventional ILM peeling (given the irreversible damage of outer retinal layers), higher closure rates provide protection against retinal redetachment, which may positively impact prognosis in the long term. Notwithstanding our visual results, it should be noted that patients submitted to the inverted flap technique presented with longer symptomatic evolution and higher MH diameters, possibly affecting the potential for visual rehabilitation and undervaluing results.
As for the safety profile of the inverted flap technique, several immediate surgical complications have been described in the literature.31 Our analysis could not compare the safety of both techniques, as surgical complications were probably significantly unreported.
The results of our study must be weighed against its limitations — most importantly, the retrospective character of most studies, large study weight being attributed to a single study, larger MH diameter and longer symptomatic evolution in the inverted flap group, and the underassessment/disclosure of several baseline characteristics. As such, we highlight the importance of full baseline characterization of both study groups regarding variables potentially impacting surgical prognosis.
Finally, few considerations are noteworthy. When choosing a surgical approach for MH patients, it is important to note that the ILM peeling (either conventional or modified technique) may affect the retinal structure and function itself; however, the role and usefulness of ILM peeling are beyond the scope of this review. Our analysis constitutes a comprehensive review assessing these three types of MH, which allows us to draw some conclusions: although the inverted flap technique definitely improves closure rates, its impact on visual function is less obvious.