The evidence base for the efficacy of the ketogenic diet was assessed among pediatric epileptic patients by application of a rigorous statistical meta-analysis. Nineteen studies from 392 abstracts met the inclusion criteria. The sample size was 1084 patients (mean age at initiation 5.78 +/- 3.43 years). The pooled odds ratio, using a random effects model, of treatment success (> 50% seizure reduction) among patients staying on the diet relative to those discontinuing the diet was 2.25 (95% confidence interval = 1.69-2.98). The reasons for diet discontinuation included < 50% seizure reduction (47.0%), diet restrictiveness (16.4%), and incurrent illness or diet side effects (13.2%). The results indicate that children with generalized seizures and patients who respond with > 50% seizure reduction within 3 months tend to remain on the diet longer. Although no class I or II studies have been published regarding the efficacy of the ketogenic diet, this meta-analysis shows that current observational studies reporting on the therapeutic effect of the ketogenic diet contain valuable statistical data. Future observational studies should aim for long-term follow-up, patient dropout analysis, and improved seizure type characterization.