Topical antibiotics for treating bacterial keratitis: a network meta-analysis.

backgroundInfectious keratitis, commonly known as corneal infection, is a major cause of blindness, affecting approximately six million people globally and resulting in around two million cases of monocular blindness annually. The incidence varies widely worldwide, with higher rates in low- and middle-income countries due to various risk factors, including agricultural injuries and other accidental trauma, limited access to health care, and low levels of health literacy. Bacterial keratitis (BK) is the most prevalent form in higher-income regions, contributing to significant morbidity and healthcare burden. If not diagnosed and treated promptly, BK can damage the cornea and result in corneal scarring, visual impairment and/or blindness. Broad-spectrum topical antibiotics remain the primary treatment, with regional microbiological profiles and antimicrobial resistance patterns influencing therapeutic choices. However, in view of the substantial heterogeneity in clinical practice, the optimal choice of topical antibiotics for BK remains uncertain. Addressing this unanswered question may help inform current practice and improve the clinical outcomes of BK.

objectivesTo compare the benefits and harms of topical antibiotics for treating BK and to rank interventions by performing a systematic review and network meta-analysis (NMA). SEARCH

methodsWe searched CENTRAL, MEDLINE, Embase, two other databases, and two trials registries together with reference checking and contact with study authors (where necessary). The latest search date was 8 August 2024. There were no restrictions on language or year of publication. SELECTION CRITERIA: We included randomized controlled trials (RCTs) in which different types of topical antibiotics (e.g. ciprofloxacin, moxifloxacin, vancomycin, etc.) and/or placebo were compared in participants with BK (diagnosed clinically or microbiologically, or both). DATA COLLECTION AND ANALYSIS: We used standard Cochrane methodology. Our outcomes were mean days to healing, mean size of epithelial defect, mean size of infiltrate, mean corrected and uncorrected distance visual acuity, and adverse effects. We assessed risk of bias using the RoB 2 tool and the certainty of evidence using the CINeMA framework for the primary NMA results of our critical outcome. MAIN

resultsWe included 23 parallel-group RCTs that enrolled 2692 participants diagnosed with BK. The studies were conducted in Australia, Canada, India, Iran, Israel, Japan, the Philippines, Serbia, Thailand, the UK, and the USA. The majority of participants were of working age, with a mean age ranging from 26 to 66 years, and 58% were male. We classified six types of interventions: fluoroquinolone monotherapy, cephalosporin monotherapy, penicillin monotherapy, dual therapy, triple therapy, and other monotherapy (povidone-iodine, honey, placebo), yielding 10 pair-wise comparisons. We judged 12 studies (54.5%) to be at high risk of bias and 10 studies (45.5%) to raise some concerns for bias. Based on the critical outcome (mean days to healing) analyzed by surface under the cumulative ranking curve (SUCRA), vancomycin + ceftazidime (SUCRA of 83.8), moxifloxacin (SUCRA of 83.1), and cefazolin + tobramycin (SUCRA of 71.3) were shown to be the most effective treatments for BK. When compared with ciprofloxacin monotherapy (the comparison group), the following showed evidence of faster healing time (by more than two to seven days): moxifloxacin (mean difference [MD] -6.81, 95% confidence interval [CI] -13.83 to 0.20; moderate-certainty evidence), vancomycin + ceftazidime (MD -6.18, 95% CI -10.24 to -2.12; low-certainty evidence), cefazolin + tobramycin (MD -5.57, 95% CI -12.87 to 1.74; moderate-certainty evidence), gatifloxacin (MD -3.84, 95% CI -9.12 to 1.43; low-certainty evidence), cefazolin + gentamicin (MD -2.58, 95% CI -6.45 to 1.30; low-certainty evidence), and honey (MD -2.44, 95% CI -4.42 to -0.46; low-certainty evidence). Conversely, lomefloxacin (MD -0.94, 95% CI -3.88 to 2.00; moderate-certainty evidence) and ofloxacin (MD -0.70, 95% CI -0.90 to -0.50; high-certainty evidence) showed similar healing time to ciprofloxacin with less than one-day difference. Compared with vancomycin + ceftazidime, ofloxacin (MD 5.48, 95% CI 1.41 to 9.55; low-certainty evidence), lomefloxacin (MD 5.24, 95% CI 1.50 to 8.98; low-certainty evidence), and cefazolin + gentamicin (MD 3.60, 95% CI 2.38 to 4.82; low-certainty evidence) showed evidence of longer time to heal (by three to six days). Of the important outcomes, including mean size of epithelial defect, mean size of infiltrate, mean corrected and uncorrected distance visual acuity, and adverse effects, only the odds of non-serious harms/non-severe harms (ranging from ocular discomfort, hyperemia, toxicity, conjunctivitis, and superficial punctate keratitis to the need for therapeutic keratoplasty) had sufficient data for analysis. The three interventions least likely to cause harm were vancomycin + ceftazidime (SUCRA of 93.1), cefazolin + gentamicin (SUCRA of 82.5), and chlorhexidine + cefazolin (SUCRA of 77.0). Regarding the odds of any non-serious or non-severe harm, vancomycin + ceftazidime was associated with fewer harms than ciprofloxacin (odds ratio [OR] 0.07, 95% CI 0.01 to 0.92), gatifloxacin (OR 0.05, 95% CI 0.00 to 0.90), and cefazolin + tobramycin (OR 0.05, 95% CI 0.00 to 0.75), whereas cefuroxime + gentamicin was found to cause more harms than ofloxacin (OR 16.13, 95% CI 1.88 to 138.47), moxifloxacin (OR 20.31, 95% CI 1.15 to 358.25), cefazolin + gentamicin (OR 96.41, 95% CI 2.52 to 3692.25), and cefazolin + chlorhexidine (OR 0.01, 95% CI 0.00 to 0.71). We did not assess the certainty of evidence for harms. AUTHORS'

conclusionsIn our NMA, mostly moderate- to very low-certainty evidence suggests that vancomycin + ceftazidime combination therapy, moxifloxacin monotherapy, and cefazolin + tobramycin combination therapy may be the most effective treatments for BK in terms of corneal healing time, whereas ciprofloxacin monotherapy is the least effective. Given that most evidence was not of high certainty, the results of this NMA should be interpreted with caution, and future research could potentially alter these findings.

fundingRQ receives grant support from the National Eye Institute (UG1EY020522). DSJT acknowledges support from the Medical Research Council/Fight for Sight Clinical Research Fellowship (MR/T001674/1) and the Birmingham Health Partners Fellowship. CH receives grant support from Glaucoma UK (183772), National Institute for Health Research Clinical Lectureship (CL-2020-18-009). The views expressed in this publication are those of the author(s) and not necessarily those of the NIHR, NHS, or the UK Department of Health and Social Care. REGISTRATION: Protocol available via doi: 10.1002/14651858.CD015350.