Background and rationale

Cataracts are the leading cause of global preventable and treatable blindness [1]. Cataract surgery is one of the most frequently performed surgical procedure globally with an estimated 20 million cases performed per year [2,3]. Cystoid macular edema (CME) is among the most common causes of visual impairment following cataract surgery with prevalence rates varying between 0.9–23% depending on the diagnosis method used and on the presence of patient risk factors [4,5]. The pathophysiology of CME is not fully understood but has been hypothesized to occur due to post-operative inflammation facilitating disruption of the blood-retinal barrier and leakage from the perifoveal retinal capillaries [6,7]. Patients who develop CME may experience, decreased vision, reduced contrast sensitivity, metamorphopsia and/or central scotomas [7]. While CME can be treated, patients may experience permanent vision loss after the resolution of edema due to changes in the photoreceptor architecture [8,9]. Moreover, the development of CME increases the costs associated with cataract surgery by an estimated 47% [10]. Given the frequency of cataract surgery, even small improvements in perioperative care can have substantial clinical and economic effects.

Nonsteroidal anti-inflammatory drugs (NSAIDs) and corticosteroid drugs are frequently used to prevent CME [5,11]. A 2016 Cochrane review concluded that topical NSAIDs may reduce the risk of CME but that it was unclear if there was any impact on visual function or quality of life; the value of adding NSAIDs to steroids was also deemed uncertain [12]. Since this review, multiple large randomized controlled trials (RCTs) have attempted to address the evidence gap [13–15]. Additionally, since this review, multiple RCTs have evaluated the safety and efficacy of subtenon steroids as an alternative to topical therapy [13,16]. Previous reviews evaluating prophylactic interventions for the development of CME have been narrative or focused solely on direct comparisons limiting the clinical applications that can be generated [5,12,17]. The lack of high-quality evidence is reflected by the considerable variability amongst surgeons and institutions regarding prophylactic practice patterns [7].

The importance of an evidence-based protocol will only be magnified given the projected increase in the global burden of cataracts over the next 15 years [18]. There is a critical need for an updated systematic review and meta-analysis to evaluate the impact of available prophylactic treatment options for the development of CME following cataract surgery. Given that there are multiple potentially efficacious agents, a network meta-analysis (NMA) is the ideal statistical method for investigating this topic [19–22]. Unlike a traditional pairwise meta-analysis, NMA facilitates comparisons across three or more interventions. NMA enables the inclusion of additional published trials by integrating both direct and indirect evidence into one model to provide more precise effect estimates and the relative effectiveness of competing interventions [21–23].

Objectives

The focus of this investigation will be to address the following question: In patients over the age of 18 randomized to receive prophylactic treatment to prevent CME following uncomplicated, age-related cataract surgery performed by phacoemulsification, what is the comparative effectiveness of each specific regimen of NSAIDs and/or corticosteroid therapy on the incidence of CME, retinal thickness, visual acuity, patient quality of life, intraocular pressure and adverse events at 6 weeks and 3 months following surgery? This investigation will evaluate all ocular routes and timing of NSAID and corticosteroid treatment (topical NSAIDs, topical corticosteroids, subtenon steroids, subconjunctival steroids, intravitreal steroids, and intravitreal steroid implants).

Eligibility criteria

We utilized the PICOTS framework to design the framework for the eligibility criteria [26]. Table 1 outlines these criteria:

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Eligibility criteria.

https://doi.org/10.1371/journal.pone.0314467.t001

Information sources

Searches of Ovid MEDLINE, Ovid EMBASE, and CENTRAL databases as well as the ClinicalTrials.gov and World Health Organization International Clinical Trials registries were performed from inception until February 27^th 2024, for RCTs meeting the eligibility criteria. The sensitive search strategy was devised collaboratively with a multidisciplinary research team consisting of an academic research librarian and clinicians. To identify any potentially missing studies, the electronic search will be supplemented by hand-searching the reference lists from included studies, and by contacting clinical experts in the field. Captured citations will be exported to Covidence (Veritas Health Innovation, Melbourne, Australia) for further screening.

Search strategy

The search strategy for each database and registry is displayed in S2 File.

Selection process

Two reviewers working independently and in duplicate, will review the titles and abstracts of search records and subsequently all full texts to determine eligibility. All disagreements will be resolved by discussion and, when not possible, through adjudication by a third party. Covidence (Melbourne, Australia) will be utilized to track triage decisions.

Data extraction

Following training and calibration, pairs of reviewers will work independently and in duplicate to complete the extraction. The calibration process is outlined in S1 File. The following information was extracted: the study design, the date the study was conducted, the number of participants in each group, the mean age of participants, the sex of participants, baseline characteristics of the study populations including the percentage of patients with diabetes and uveitis, the study inclusion criteria, details regarding the intervention agents used, the primary and secondary outcomes of each trial including methods in which the outcomes were assessed, the follow-up period of each trial, and whether the trial received funding. Any unclear information from the published manuscripts will be clarified with the corresponding author.

Risk of bias

Following training and calibration, pairs of reviewers, working independently and in duplicate, will assess the risk of bias for each eligible trial for each outcome of interest utilizing the Cochrane Risk of Bias Version 2 tool [36]. Disagreements will be resolved through discussion and, when not possible, through adjudication by a third party. For each eligible outcome, the risk of bias will be rated as ‘low,’ ‘some concerns’, or ‘high’ based on assessments across the following domains: the randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result.

Statistical analysis

We will present the results of dichotomous outcomes as odds ratios (ORs) and continuous outcomes as mean differences (MDs) or standardized mean differences (SMDs), with the associated 95% credible interval (CrI). For each comparison, we will perform a Bayesian network meta-analysis with random effects modelling. Estimates will be obtained using the Markov Chains Monte Carlo (MCMC) method with non-informative priors. 5,000 initial iterations will be used for adaptation, followed by 100,000 iterations for the calculation of odds ratios or MDs and their corresponding 95% CrI. Convergence will be assessed via the Brooks-Gelman-Rubin statistic [37].

Complete case analyses will be performed for all analyses. In the presence of sufficient data, analyses will be performed in three separate manners with progressively stricter criteria for defining each group. The first analysis will group patients by drug class and route of administration irrespective of the timing and dosing of the medication (e.g. patients receiving topical NSAIDs given pre-operatively and post-operatively would be combined with patients receiving topical NSAIDs solely post-operatively). The second analysis will group patients by drug and route of administration but will separate groups depending on the timing of the drug (e.g. patients receiving topical NSAIDs pre-operatively and post-operatively would be separate from patients receiving topical NSAIDs solely post-operatively). Lastly, an analysis will be performed in which drugs will be separated within each class and grouped solely by modality and peri-operative timing (e.g. topical diclofenac given post-operatively and topical ketorolac given post-operatively would inform separate nodes).

If there is insufficient evidence to perform a network meta-analysis for a particular outcome, a pairwise meta-analysis will be conducted. If there’s insufficient evidence for a pairwise analysis, then results will be synthesized narratively.

We will use node-splitting models to assess local incoherence [38]. Heterogeneity between RCTs for each direct comparison will be assessed with visual inspection of forest plots and the I² statistic. Heterogeneity of 0% to 40% will be considered as “might not be important,” 30% to 60% as “moderate heterogeneity,” 50% to 90% as “substantial heterogeneity,” and 75% to 100% as “considerable heterogeneity”.

In the event of missing measures of variability for continuous outcome measures, standard deviations will be imputed in line with the methodology from the Cochrane Handbook of Systematic Reviews of Interventions [39].

All analyses will be performed in R 4.0.3 (The R Project, Auckland, New Zealand) using the “gemtc”, “rjags”, and “dmetar”, and “meta” packages [40–43].

Subgroup analyses

We will conduct three pre-specified subgroup analyses. The groups of interest will be i) comparing patients with known diabetes at the time of randomization to those without diabetes, ii) comparing patients with a history of uveitis prior to cataract surgery to those without uveitis and iii) comparing individuals who receive femtosecond laser-assisted cataract surgery (FLACS) to those not undergoing FLACS. We hypothesize that there will be a higher rate of CME among patients with diabetes and patients with uveitis and that there will be no difference between patients undergoing FLACS [27–29]. Moreover, we hypothesize that NSAIDS either alone or in combination will be superior to steroids in patients with diabetes in preventing CME, retinal thickening and in improving visual acuity [11,44]. The method of subgroup analysis will depend on the availability of data. In the event of sufficient data, a network meta-regression will be performed evaluating the percentage of patients from each trial with the characteristic of interest. If there are insufficient data for the network meta-regression, the subgroup analysis will compare the pooled effect estimates between studies with and without the characteristic of interest, The Instrument to assess the Credibility of Effect Modification Analyses (ICEMAN tool) will be used to evaluate the credibility of subgroup analysis [45].

Confidence in cumulative evidence

We will utilize a minimally contextualized approach to assess the effect of the different interventions on the pre-specified outcomes [46]. The certainty of the evidence will be assessed following Grade of Recommendations, Assessment, Development and Evaluation (GRADE) guidelines [47,48]. The GRADE approach involves separate grading of the quality of evidence for each outcome followed by determining an overall quality of evidence across outcomes. Two reviewers will rate each domain for each comparison separately and resolve discrepancies by consensus. The certainty for each comparison and outcome will be rated as ‘high’, ‘moderate’, ‘low’ or ‘very low’, based on considerations of risk of bias, inconsistency, indirectness, publication bias, intransitivity, incoherence, and imprecision. Publication bias will be assessed using comparison-adjusted funnel plots and Egger’s test. For each outcome, we will present the most credible estimate from the NMA or direct pairwise comparison.

Sensitivity analysis

A sensitivity analysis will be performed excluding studies rated to have a high overall risk of bias. Additionally, a sensitivity analysis will be performed excluding studies for which any data was imputed to conduct the meta-analysis [39].

Treatment rankings

A strength of NMA is that the analysis can facilitate ranking the different treatments across outcomes. Traditional ranking systems such as the Surface Under the Cumulative Ranking curve (SUCRA) or P-value rankings do not account for the certainty of evidence or the precision of the ranking estimates [49]. Given these limitations, results will be interpreted using partially contextualized methods and presented across outcomes in a ranking incorporating both their relative magnitude of effect and certainty of evidence [46,50].