Sulfur Hexafluoride (SF6) versus Perfluoropropane (C3F8) in the Intraoperative Management of Macular Holes: A Systematic Review and Meta-Analysis

Purpose A systematic literature search was conducted to identify and review studies comparing SF6 to C3F8 as a tamponade agent in the intraoperative management of macular holes. Methods Publications up to October 2018 that focused on macular hole surgery in terms of primary closure, complications, and clinical outcomes were included. Forest plots were created using a weighted summary of proportion meta-analysis. Analysis was performed separately for SF6 and C3F8. A random effects model was used, and corresponding I2 heterogeneity estimates were calculated. Results Nine pertinent publications studying a total of 4,715 patients were identified in 2000 to 2017, including two randomized studies (n=206), two prospective studies (n=170), and five retrospective or registry-based studies. Similar rates of closure between SF6 and C3F8 were reported in eight out of nine studies, regardless of subgroup analyses. All studies reporting visual outcomes showed similar results when comparing SF6 to C3F8 at one to six months of follow-up. Neither agent was clearly associated with increased risk of ocular hypertension, cataract formation, or other adverse events. Meta-analytic pooling of the closure rates in the SF6 group resulted in 91.73% (95% confidence interval: 88.40 to 94.55, I2: 38.03%), and for C3F8, the closure rate was 88.36% (95% confidence interval: 85.88 to 90.63, I2: 0.0%). Conclusions Both SF6 and C3F8 appear to have achieved similar visual outcomes and primary closure rates and neither was associated with an increased risk of adverse events. Considering the more rapid visual recovery with SF6, there appears to be no evidence to support C3F8 as the tamponade agent of choice for macular hole surgery.


Introduction
Macular hole surgery was first described by Kelly and Wendel in 1991 [1]. Although their 5-step technique has remained largely unchanged, debate still exists concerning several key aspects of the procedure [2]. ese include whether to peel the internal limiting membrane (ILM), the use and type of dye, and the duration of face-down positioning following surgery [3,4].
Another heatedly debated aspect is the choice of tamponading agent [3,5,6]. e most commonly used compounds are sulfur hexafluoride (SF 6 ) and perfluoropropane (C 3 F 8 ). However, hexafluoroethane (C 2 F 6 ), room air, and silicone oil are also employed [7]. Although the procedure was initially described with use of SF 6 , subsequent clinical trials used C 3 F 8 further adding to the debate [1,[8][9][10]. To the authors' best knowledge, no systematic review or metaanalysis has yet been published, comparing the two agents. e aim of this study was to systematically review the literature for studies that compared SF 6 to C 3 F 8 as a tamponade agent in the intraoperative management of macular hole surgery, perform a pooled meta-analysis of the data, and discuss any differences found with respect to clinical outcomes and adverse effects. A flow diagram of the screening and inclusion process is illustrated in Figure 1.

Eligibility Criteria.
Full publications in English that directly compare SF 6 and C 3 F 8 in the intraoperative management of macular hole surgery, in terms of primary closure, complication rates, and clinical outcomes were included (not abstracts or letters to the editor). Studies performed on animal or cadaver eyes, as well as case reports and nonempirical opinion articles, were excluded.

Screening and Synthesis.
e review process was conducted under the guidance of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) criteria to support reporting [11]. Two reviewers (IH and MM) independently implemented the search strategy for relevant publications. Selected publications were then approved by the senior investigator who also devised the search strategy (YB).

Statistical Analyses.
To better illustrate the main anatomical outcome of primary closure, forest plots were created using a weighted summary of proportion meta-analysis. Analysis was performed separately for SF 6 and C 3 F 8 . A random effects model was used (as implemented by Der-Simonian and Laird, 1986 [12]), and corresponding I 2 heterogeneity estimates were generated. Data were tabulated and analyzed using SPSS for Windows version 22.0 software by IBM Inc. (Armonk, NY, USA). Graphs were created using MedCalc software version 16 (Mariakerke, Belgium).

Results
Data comparing the use of SF 6 and C 3 F 8 for macular hole surgery were extracted from nine studies published during 2000 to 2017 (Table 1), studying a total of 4,715 patients. Two of the studies were randomized by design and included a total of 206 patients. Two studies prospectively enrolled patients without randomization (170 patients), and the remaining five were retrospective or registry based. Figure 1 illustrates the flow of the inclusion process. e first relates to ILM removal as a routine part of the operation.
ILM peeling was routinely performed in six of the nine studies. In two studies (Essex et al. [4] and Tognetto et al. [7]), a majority of patients (97.7% and 67.7%, respectively) underwent ILM peeling while the remainder did not. e authors elected not to routinely perform ILM peeling of prospectively enrolled patients in only one study (Mulhern et al.) [16].
Tognetto et al. expanded their investigation, comparing the outcomes of patients who underwent ILM peeling with those who had not [7]. ey analyzed the retrospective data of 1,627 patients from different countries, who had been operated on for idiopathic macular hole. ey found that ILM peeling improved hole closure only for longstanding macular holes or those in advanced stages [7].
A second aspect in which studies differed was the choice of gas concentration. e SF 6 concentration was 20% in four of the nine studies. Mulhern

Postoperative Management.
Face-down positioning varied between studies. Most studies recommended two to seven days of prone positioning; however, some (e.g., Mulhern et al.) requested patients who received SF 6 maintain a prone head position for up to 4 weeks [16].
A registry-based analysis by Essex et al. sheds further light on this topic [4]. ey analyzed the outcome of 2,367 patients from Australia and New Zealand who underwent primary idiopathic macular hole surgery. Approximately 26% of these patients were not advised to remain in prone position postoperatively, and among those who were advised, most were instructed to remain prone between three and seven days. Essex et al. further compared the outcomes between patients who were not advised a postoperative facedown position to those who were and found that no prone positioning was noninferior for holes of less than 400 μm in diameter, while for larger holes noninferiority could not be concluded [4].

Primary Anatomical Closure.
Most studies used anatomical closure as the primary outcome; their results are summarized in Table 1. Closure rates ranged from 85% to 100%, with most in the 90% to 95% range.
A comparison of SF 6 and C 3 F 8 showed similar rates of closure in eight of the nine studies, including the two randomized controlled studies by Casini et al. [3,5]. One study by Essex et al. using a more stringent noninferiority analysis with a relatively narrow noninferiority margin (5%) also demonstrated noninferiority of SF 6 compared to C 3 F 8 [4]. Only one study by Tognetto et al. reported a higher success rate with SF 6 (93.6% versus 87.2%); however, the authors provide no statistical foundation for this claim [7].
Several studies further analyzed subgroups in an attempt to detect a subset of patients, which might benefit from one agent over the other. Modi  ey all found SF 6 to produce results similar to C 3 F 8 regardless of the subgroup analyzed (Table 1).

Visual Acuity.
Six studies reported visual acuity outcomes following surgery [3,5,6,[14][15][16]. All showed similar results when comparing SF 6 to C 3 F 8 at one to six months of follow-up (Table 1). Improvement following SF 6 gas ranged from 2.7 to 5.9 Early Treatment Diabetic Retinopathy Study (ETDRS) lines, while following use of C 3 F 8 ranged between 2.7 and 6.9 ETDRS lines. Five of the six studies report comparative statistics on visual improvement between groups, and all find no significant differences (Table 1). e randomized trial by Casini et al. as well as the prospective cohort study by Xirou et al. showed that visual acuity was initially better in the group treated with SF 6 but was eventually found to be similar [3,14].

Discussion
In this study, a systematic review and meta-analysis of studies comparing SF 6 and C 3 F 8 gas tamponades for macular hole surgery was performed. Macular hole closure rates were found to be similar regardless of whether SF 6 or C 3 F 8 was used and were typically in the 90-95% range. Visual acuity outcomes were also similar but tended to improve faster with SF 6 . Rates of complications (including cataract formation, ocular hypertension, and retinal tears) varied among studies and appeared to be inconsistently related to either agent.

Journal of Ophthalmology
Following macular hole repair, contact of the gas bubble with the retina causes the extrusion of subretinal fluid and maintains anatomical position, as well as possibly providing a scaffold for cellular proliferation [14]. It therefore seems reasonable that a larger, longer-acting, bubble would provide further benefit. However, some evidence suggests that hole closure occurs very early in the postoperative period, perhaps as early as the first 24 hours [17]. A recent study by Masuyama et al. showed that repair of the macular hole typically occurs between 4 and 7 days postoperatively and is presumed to be facilitated by ganglion and Muller cells [18].
Long-lasting gasses (such as C 3 F 8 ) may offer more extensive tamponade; however, they also impair vision longer. It is possible that such longer-acting agents are not required, considering the short timescales in which macular hole repair occurs. e outcomes of this review appear to support this notion, as use of SF 6 produced similar clinical outcomes to C 3 F 8 in terms of both primary closure and visual outcomes.
is study has several limitations. It included nonrandomized trials in the analysis. However, given the small number of randomized studies performing a direct comparison of SF 6 and C 3 F 8 for macular hole surgery, we elected to include relevant retrospective and registry-based studies, in order to increase the power of the meta-analysis. Given the findings of this study, the need for more prospective randomized studies evaluating this comparison may be questionable. In addition, several differences between the studies concerning the surgical technique may have introduced biases to the analyses. However, we discussed and compared variations in surgical technique found across these studies.

Conclusions
To the authors' best knowledge, this is the first systematic review and meta-analysis comparing SF 6 and C 3 F 8 as a tamponading agent for macular hole surgery. We found that SF 6 and C 3 F 8 resulted in both similar visual outcomes as well as similar primary closure rates. Neither agent was clearly associated with increased risk of ocular hypertension, cataract formation, or other adverse events. Visual recovery with SF 6 tended to occur earlier. It is probable that shorteracting tamponade agents such as SF 6 may be sufficient for macular hole surgery; we found no evidence to support C 3 F 8 as the tamponade gas of choice.

Conflicts of Interest
e authors declare that they have no conflicts of interest.
Modi et al. [6] Casini et al. [3] Briand et al. [5] Kumar et al. [13] Xirou et al. [14] Kim et al. [15] Tognetto et al. [7] Mulhern et al. [16] Total (random effects) 0.6 0.7 0.8 0.9 1.0 Figure 2: Forest plot of the proportion of patients who achieved anatomical closure following the use of SF 6 gas. Meta-analytic pooling of the proportion of patients who achieved anatomical closure following surgery in the SF 6 group. Pooled estimate was 91.73% (95% confidence interval: 88.40% to 94.55%, I 2 : 38.03%, p value for heterogeneity � 0.126). Size of the squares is proportional to the number of cases in the study. Error bars represent 95% confidence intervals. e diamond shape represents the pooled estimate. Casini et al. [3] Briand et al. [5] Kumar et al. [13] Xirou et al. [14] Kim et al. [15] Tognetto et al. [7] Mulhern et al. [16] Total (random effects) Figure 3: Forest plot of the proportion of patients who achieved anatomical closure following the use of C 3 F 8 gas. Meta-analytic pooling of the proportion of patients who achieved anatomical closure following surgery with C 3 F 8 : pooled estimate was 88.36% (95% confidence interval: 85.88% to 90.63%, I 2 : 0.0%, p value for heterogeneity � 0.864); size of the squares is proportional to the number of cases in the study. Error bars represent 95% confidence intervals. e diamond shape represents the pooled estimate. 6 Journal of Ophthalmology