Impact of the Time to Surgery on Visual Outcomes for Rhegmatogenous Retinal Detachment Repair: A Meta-Analysis

PURPOSE
To determine the relationship between time from symptom onset or presentation to repair and visual outcomes for macula-on and macula-off rhegmatogenous retinal detachment (RRD).


DESIGN
Meta-analysis.


METHODS
We searched MEDLINE, EMBASE, and Cochrane Library for randomized controlled trials and observational studies comparing best corrected visual acuity (BCVA) based on time to RRD repair. Study identifiers, baseline characteristics, intervention characteristics, and visual outcomes were extracted. We conducted a random-effects meta-analysis. Sensitivity analyses included leave-one-out and influence analyses. Primary outcomes included mean difference (MD) in final BCVA, MD between preoperative and final BCVA (∆BCVA), and relative risk of final BCVA <0.4 logMAR for macula-off RRD repair in 0-3 vs 4-7 days and macula-on RRD repair in 0-24 vs >24 hours. Secondary outcomes assessed other time points.


RESULTS
Twenty observational studies reported on 1929 patients. Macula-off RRD repair in 0-3 days from symptom onset was superior to 4-7 days for final BCVA (MD [95% confidence interval (CI)]: -0.06[-0.09,-0.03], p<0.001), but was not different for ∆BCVA (p>0.05). Macula-on repair in 0-24 hours from presentation was superior to >24 hours for final BCVA (MD [95%CI]: -0.02[-0.03,-0.01], p<0.05), but was not different for ∆BCVA (p>0.05).


CONCLUSIONS
Macula-off RRD repair in 0-3 days from symptom onset may have better final BCVA compared to repair in 4-7 days. Macula-on RRD repair in 0-24 hours of presentation may have better final BCVA compared to repair in >24 hours. These results were supported by moderate and low-quality evidence, respectively, and may have been influenced by differences in baseline BCVA.


M
odern surgical techniques achieve anatomic reattachment after repair of vision-threatening rhegmatogenous retinal detachment (RRD) in the majority of cases. 1 , 2 However, postoperative visual outcomes can be highly variable.Functional success after RRD repair is influenced by multiple factors, including patient age, macular status, presence of comorbidities, and RRD duration. 3 , 4Of these factors, the duration from symptom onset and patient presentation to RRD repair are important predictors of future visual acuity outcomes. 4mmediate repair for every patient is ideal but may be limited by availability of operating rooms and on-site staff. 5mergent surgeries may also experience the "weekend" effect, an increase in the risk of intraoperative complications, duration of stay, and poorer outcomes that could result from limited staffing and physician coverage. 6 , 7It is therefore useful to know how long RRD repair can be delayed without increased risk of adverse visual outcomes.A thorough understanding of the relationship between visual outcomes and time to repair will help determine how to best triage RRD patients.
][10][11][12] However, disagreements in the literature have led to variability in recommendations.][16] Clarifying the association between time to repair and visual outcomes for macula-on and macula-off RRDs is cru-cial in optimizing management.A previous meta-analysis was conducted by Van Bussel and associates assessing time to RRD repair and visual outcomes following scleral buckling (SB), which found that duration of macular detachment of 0-3 days was associated with the highest RR of final BCVA < 0.4 logMAR. 17To our knowledge, no existing meta-analysis has examined the relationship between time to RRD repair and visual outcomes across multiple interventions (SB, pars plana vitrectomy [PPV], combined SB and PPV [SB + PPV], or pneumatic retinopexy [PR]).
The primary objective of this study is to determine the relationship between time to surgery and visual outcomes for macula-on and macula-off RRDs.Secondary aims include analyzing this relationship in the context of varying intervention types and patient characteristics.

• SEARCH STRATEGY AND STUDY SELECTION:
We conducted a systematic literature search using Ovid MED-LINE (2000-April 2022), EMBASE (2000-April 2022), and Cochrane CENTRAL (inception up to April 2022).Randomized controlled trials or observational studies were included if they reported on BCVA or primary retinal reattachment rates following RRD repair and analyzed these outcomes according to time from presentation or symptom (ie, central vision loss) onset to repair.Both randomized and nonrandomized trials were included as only a small number of randomized trials were likely to be available for inclusion.Studies published in languages other than English, unpublished studies, case reports, narrative reviews, editorials, and articles with repeat data from the same patient sample were excluded.
The complete search strategy can be found in Supplemental Table S1.Three independent reviewers (A.S., A.E., A.S.D.) conducted title and abstract screening and subsequent full-text screening of included abstracts.Discrepancies were resolved through consensus with the input of a fourth author (M.M.P.).Reference lists of included papers were searched to ensure that no relevant studies were missed.The study protocol was registered in the International Prospective Register of Systematic Reviews (PROS-PERO) database, CRD 42020204169.
• DATA EXTRACTION AND QUALITY ASSESSMENT: Two independent reviewers (A.S, A.E.) used standardized data collection forms to extract study identifiers (authors, journal, year of publication, study design, country of origin), baseline characteristics (number of eyes, age, gender, ethnicity, macula status, lens status, intraocular pressure, preoperative BCVA, duration of symptoms/delay to surgery), intervention characteristics (procedure type, vitrector gauge, and tamponade specifications), and outcomes (length of follow-up, BCVA at 3 months, 6 months, 12 months, and final follow-up, change in BCVA from baseline, primary reattachment rate, and adverse outcomes).
Two independent reviewers (A.S., A.E.) assessed the quality of included studies using the Risk of Bias in Nonrandomized Studies-of Interventions (ROBINS-I), which assesses bias in 9 domains: confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of reported results. 18Quality of evidence for individual outcomes was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) tool. 19All studies were also assessed for authorship conflicts of interest and industry sponsorship.
• DATA SYNTHESIS AND DATA ANALYSIS: Baseline demographics were reported as a proportion for categorical variables and means with SD for continuous variables.A random effects meta-analysis was conducted using an empirical Bayesian estimator for all outcomes.Studies were removed from meta-analysis of individual outcomes if the study did not report on that outcome, reported on the outcome but omitted a measure of dispersion (ie, SD, SE, or 95% CI), or was missing data for that outcome.Continuous outcomes were reported as a mean difference (MD) with a 95% CI, and binary outcomes as a relative risk [RR] with 95% CI.
Primary outcomes were final BCVA, change between preoperative BCVA and final BCVA ( BCVA), and relative risk (RR) of final BCVA < 0.4 logMAR (better than 20/50 Snellen) between macula-off RRD repair in 0-3 vs 4-7 days and macula-on RRD repair in 0-24 vs > 24 hours.Secondary outcomes examined other time points, the MD of operating time, the RR of primary reattachment, and the RR of complications.The most consistently reported time points in the included studies were selected for analysis.
Statistical heterogeneity was investigated using an I 2 statistic: I 2 < 0.25 was considered low heterogeneity, I 2 equal to 0.25-0.50 was considered moderate heterogeneity, and I 2 > 0.50 was considered high heterogeneity. 18,20e performed sensitivity analyses by sequentially removing each study and reanalyzing the remaining studies (ie, leave-1-out analyses), by conducting diagnostic tests to identify outliers and highly influential studies on the results (ie, influence analyses), and by generating funnel plots and assessing publication bias.We aimed to conduct subgroup analyses within each outcome according to study design, duration of follow-up, surgery performed (PPV, SB, SB + PPV, PR), lens status (phakic vs pseudophakic), and endotamponade used.Studies that did not report duration of follow-up were excluded from subgroup analysis.
P values of ≤.05 were considered statistically significant for all analyses.Data extraction and description of baseline demographics were performed on Microsoft Excel (Redmond, WA).Meta-analysis was performed on R 4.0.2(The R Project for Statistical Computing), using the metafor (version 2.4.0) and metaviz (version 0.3.1)packages.

DEMOGRAPHICS:
The search yielded 4185 articles.After removing duplicates and screening titles and abstracts, 711 articles were advanced to full-text screening.Twenty-one studies (19 case series, 2 cohort studies) met all criteria after full-text screening ( Figure 1 ).Seven studies were conducted in the United States, 11 , 14 , 16 , 21-24 followed by the United Kingdom ( k = 3), 8 , 25 , 26 Germany ( k = 2), 13 , 27 Japan ( k = 1), 28 the Netherlands ( k = 1), 29 Taiwan ( k = 1), 30 Turkey ( k = 1), 31 Canada ( k = 1), 32 Italy ( k = 1), 33 New Zealand ( k = 1), 15 Pakistan ( k = 1), 34 and Switzerland ( k = 1). 35Ten studies assessed multiple procedure types, 26 but 9 of them did not report results on time to repair stratified by procedure. 8 , 13 , 16 , 21-24 , 26 , 29 , 30 total of 1929 eyes were included at baseline.Individual study sample sizes ranged from 12 to 199 eyes.Thirty-nine percent of subjects were female (range: 18.8%-64.5%),the mean age of subjects was 58.2 (range: 39.2-64.9),and the mean postoperative follow-up time was 15.9 months (range: 3-55).Eighty percent of studies (16/20) assessed maculaoff RRDs, of which 14 studies defined duration of RRD as time from symptom onset.Time from symptom onset was defined by 10 studies as central vision loss and not specified by 4 studies.Two studies did not specify whether duration of RRD was from symptom onset or presentation.Twenty percent of studies (4/20) assessed macula-on RRDs and all defined duration of RRD as time from initial examination to repair.Given limitations of available data, our analysis assessed time from symptom onset to RRD repair for maculaoff RRDs and time from initial presentation to RRD repair for macula-on RRDs.A complete list of baseline characteristics can be found in Table 1 .
• QUALITY ASSESSMENT: Using the ROBINS-I tool (Supplemental Table S2), 55% (11/20) of observational studies had a low overall risk of bias, 40% (8/20) had a moderate risk of bias, and 5% (1/20) had a serious risk of bias.Studies received a more negative assessment for lack of statistical analysis to control for confounding (18/20), result selec-  The GRADE evaluation (Supplemental Table S3) identified that the final BCVA outcome for macula-off RRD repair in 0-3 vs 4-7 days and all outcomes for macula-off RRD repair in 0-7 vs > 7 days were supported by moderate quality evidence; RR of BCVA < 0.4 logMAR for macula-on RRD repair in 0-24 vs > 24 hours, and all outcomes for maculaoff RRD repair in 0-15 vs > 15 days were supported by very low quality evidence; and all other outcomes and endpoints were supported by low quality evidence.One study reported an author conflict of interest, and no included studies received industry sponsorship.
Baseline BCVA could not be calculated for all outcomes because of inconsistent reporting across studies.A difference in baseline BCVA was noted for the analysis of RR of final BCVA < 0.4 logMAR for macula-off RRD repair in 0-3 days vs 4-7 days (1.69 ± 0.39 logMAR vs 1.94 ± 0.32 logMAR).A similar difference in baseline BCVA was noted for the analysis of RR of final BCVA < 0.4 logMAR for macula-on RRD repair in 0-24 hours vs > 24 hours (0.12 ± 0.10 logMAR vs 0.18 ± 0.11 logMAR).Baseline BCVA for individual outcomes can be found in Table 2 .
• SENSITIVITY AND SUBGROUP ANALYSES: I 2 calculation demonstrated high heterogeneity for the final BCVA out- Mean (SDs) and proportions for each group are calculated from raw data.Mean differences and relative risk ratios were calculated from models accounting for study weighting and between-study variance.comes of 0-7 days vs > 7 days and 0-15 days vs > 15 days, and moderate heterogeneity for BCVA at 0-7 days vs > 7 days, and RR of BCVA < 0.4 logMAR at 0-3 days vs > 3 days and 0-24 hours vs > 24 hours.All other outcomes had low heterogeneity.
Additional sensitivity analyses did not identify any highly influential studies, outliers, or funnel plot asymmetry.No differences relative to the main analysis were identified in subgroup analysis for the duration of follow-up (0-3 months, 3-12 months, > 12 months) in any outcome.We were unable to conduct a subgroup analysis based on endotamponade, intervention type, or lens status because of insufficient reporting of outcomes in relevant patient cohorts across studies.

DISCUSSION
Time to repair of macula-on and macula-off RRDs has been shown to impact visual outcomes. 10 , 11][16] This meta-analysis explored the relationship between time to RRD repair and visual outcomes, analyzing 1929 eyes from 20 studies, which were collectively of low-moderate quality.Macula-on RRDs are treated urgently to prevent detachment of the fovea preoperatively, whereas macula-off RRDs are thought to have already suffered permanent foveal damage with resultant visual consequences, thus limiting the urgency of repair. 12However, this meta-analysis found that final BCVA was better when macula-off RRDs were treated in 0-3 days.This conclusion is in line with a large sample study of 847 eyes with RRD by Williamson and associates, who noted better postoperative visual acuity following surgery in 1-3 days relative to 4-6 days ( P = .013). 36 longer duration to macula-off repair was associated with worse visual outcomes when considering the 7-day and 10day thresholds; however, this was not different at the 15-day threshold.We found that macula-on RRDs were associated with better final BCVA when treated in 0-24 hours vs > 24 hours; however, the difference was small.
Heterogeneity was high for analyses of final BCVA after macula-off RRD repair in 0-7 vs > 7 days and was moderate for BCVA after macula-off RRD repair in 0-7 vs > 7 days and for RR of final BCVA < 0.4 logMAR after macula-off RRD repair in 0-3 vs 4-7 days.Given the heterogeneity, a random effects model was used, and subgroup analysis did not decrease heterogeneity in these outcomes.This was expected given our inclusion of studies with heterogeneous populations with respect to geographic region, age, ethnicity, and presence of medical and ocular comorbidities.The high heterogeneity for the analyses of final BCVA after macula-off RRD repair in 0-15 days vs > 15 days (high) and RR of final BCVA < 0.4 logMAR after macula-on RRD repair in 0-24 hours vs > 24 hours was attributable to low sample size (n = 2 studies) and a heterogeneous patient population.
Leave-1-out analyses identified several studies that had a significant influence on the observed results.Liu and associates 27 found that macula-off RRD repair in 0-3 days from symptom onset was superior to 4-7 days for final BCVA and that macula-off RRD repair in 0-7 days was superior to > 7 days for RR of final BCVA < 0.4 logMAR: this study included only patients receiving SB.Geiger and associates 22 found that macula-off RRD repair in 0-3 and 4-7 days from to assess whether the intervention type, lens status, or endotamponade used affected the observed relationship because of inconsistencies in reporting across studies.
In conclusion, modern retinal reattachment techniques are associated with excellent BCVA outcomes in most eyes.Macula-off RRD repair within 3 days of symptom onset may have a 0.06 logMAR ( ∼3 Snellen letters) superior final VA compared to repair in 4-7 days.Macula-on RRD re-paired within 24 hours of presentation may provide superior VA outcomes (0.02 logMAR, ∼1 Snellen letter) compared with repair in > 24 hours.These results were supported by evidence of moderate and low quality, respectively, and may have been influenced by differences in baseline BCVA.Future prospective studies with large sample sizes that investigate the relationship between time to RRD repair and visual outcomes are warranted.
Funding/Support: This study received no funding.Financial Disclosures: M.M.P. receives financial support (to institution) from the PSI Foundation.P.J.K. is on the advisory boards of Novartis, Alcon, Bayer, Allergan, and Novelty Nobility; receives financial support (to institution) from Bayer, Roche, and Novartis; financial support from Novartis and Bayer; and is an equity owner in ArcticDx.R.H.M. is on the advisory boards of Alcon, Bausch + Lomb, Bayer, Novartis, Allergan, and Roche; and receives financial support (to institution) from Bayer, Novartis, and Roche.All authors attest that they meet the current ICMJE criteria for authorship.

FIGURE 1 .
FIGURE 1. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram.