PURPOSE: Epithelial keratitis is the most common presentation of ocular infection by herpes simplex virus (HSV). Quantitative assessment of available therapy is needed to guide evidence-based ophthalmology. This study aimed to compare the efficacy of various treatments for dendritic or geographic HSV epithelial keratitis and to evaluate the role of various clinical characteristics on epithelial healing. METHODS: Following a systematic review of the literature, information from clinical trials of HSV dendritic or geographic epithelial keratitis was extracted, and the methodological quality of each study was scored. Methods of epithelial cauterization and curettage were grouped as relatively equivalent physicochemical therapy, and solution and ointment formulations of a given topical antiviral agent were combined. The proportion healed with 1 week of therapy, a scheduled follow-up day that approximated the average time of resolution with antiviral therapy, was selected as the primary outcome based on a masked evaluation of maximum treatment differences in published healing curves. The proportion healed at 14 days was recorded as supplemental information. Fixed-effects and random-effects meta-analysis models were used to obtain summary estimates by pooling results from comparative treatment trials. Hypotheses about which prognostic factors might affect epithelial healing during antiviral therapy were developed by multivariate analysis of the Herpetic Eye Disease Study dataset. RESULTS: After excluding 48 duplicate reports, 14 nonrandomized studies, 15 studies with outdated or similar treatments, and 29 trials lacking sufficient data on healing or accessibility, 76 primary reports were identified. These reports involved 4,251 patients allocated to 93 treatment comparisons of dendritic epithelial keratitis in 28 categories and 9 comparisons of geographic epithelial keratitis in 6 categories. For dendritic keratitis, idoxuridine was better than placebo at 7 days (combined odds ratio [OR], 3.59; 95% confidence interval [CI], 1.92-6.70), and at 14 days (OR, 4.17; 95% CI, 1.33-13.04), but pooling was limited by lack of homogeneity and low study quality. Direct comparisons at 1 week of treatment showed that trifluridine or acyclovir was significantly better than idoxuridine (OR, 3.12 and 4.56; 95% CI, 1.55-6.29 and 2.76-7.52, respectively), and indirect comparisons were also consistent with a clinically significant benefit. Vidarabine was not significantly better than idoxuridine in pooled treatment comparisons at 1 week (OR, 1.20; 95% CI, 0.72-2.00) but was better in 2 indirect comparisons (OR, 4.22 and 4.78; 95% CI, 1.69-10.54 and 2.15-10.65, respectively). At 14 days, trifluridine (OR, 6.05; 95% CI, 2.50-14.66), acyclovir (OR, 2.88; 95% CI, 1.39-4.78), and vidarabine (OR, 1.24; 95% CI, 0.65-2.37) were each better than idoxuridine. Trials of geographic epithelial keratitis also suggested that trifluridine, acyclovir, and vidarabine were more effective that idoxuridine. Other topical antiviral agents, such as bromovinyldeoxuridine, ganciclovir, and foscarnet, appeared equivalent to trifluridine or acyclovir. Oral acyclovir was equivalent to topical antiviral therapy and did not hasten healing when used in combination with topical treatment. Antiviral agents did not increase the speed of healing when compared to debridement but reduced the risk of recrudescent epithelial keratitis. The combination of physicochemical treatment with an antiviral agent seemed to be better than either physicochemical or antiviral treatment alone, but the heterogeneous cauterization and curettage techniques and the various treatment combinations limited valid quantitative summary effect measures. The combination of topical interferon with an antiviral agent was significantly better than antiviral therapy at 7 days (OR, 13.49; 95% CI, 7.39-24.61) but not at 14 days (OR, 2.36; 95% CI, 0.82-6.79). Finding apparent heterogeneity for some pooled estimates suggested that dissimilarities in patients, interventions, outcomes, or other logistical aspects of clinical trials occur across studies. CONCLUSIONS: The available evidence on the acute treatment of presumed HSV epithelial keratitis demonstrates the effectiveness of antiviral treatment and shows the log-logistic healing curve of treated dendritic epithelial keratitis. Topical trifluridine, acyclovir, and vidarabine were significantly more effective than idoxuridine but similar in relative effectiveness for dendritic epithelial keratitis. Physicochemical methods of removing infected corneal epithelium are effective, but adjunctive virucidal agents are needed to avert recrudescent epithelial keratitis. Whether debridement in combination with antiviral therapy is more beneficial than antiviral chemotherapy alone appears likely but remains inconclusive. The combination of topical interferon with an antiviral agent significantly speeds epithelial healing. Future trials of the acute treatment of HSV epithelial keratitis must aim to achieve adequate statistical power for assessing the primary outcome and should consider the effect of lesion size and other characteristics on treatment response.