OBJECTIVE: To compare banding ligation versus beta-blockers as primary prophylaxis in patients with esophageal varices and no previous bleeding. METHODS: Randomized trials were identified through electronic databases, reference lists in relevant articles, and correspondence with experts. Three authors extracted data. Random effects meta-analysis and metaregression were performed. The reported allocation sequence generation and concealment were extracted as measures of bias control. RESULTS: The initial searches identified 1,174 references. Sixteen trials were included. In 15 trials, patients had high-risk varices. Three trials reported adequate bias control. All trials reported mortality for banding ligation (116/573 patients) and beta-blockers (115/594 patients). Mortality in the two treatment groups was not significantly different in the trials with adequate bias control (relative risk 1.22, 95% CI 0.84-1.78) or unclear bias control (RR 1.02, 95% CI 0.75-1.39). Trials with adequate bias control found no significant difference in bleeding rates (RR 0.86, 95% CI 0.55-1.35). Trials with unclear bias control found that banding ligation significantly reduced bleeding (RR 0.56, 95% CI 0.41-0.77). Both treatments were associated with adverse events. In metaregression analyses, the estimated effect of ligation was significantly more positive if trials were published as abstracts. Likewise, the shorter the follow-up, the more positive the estimated effect of ligation. CONCLUSIONS: Banding ligation and beta-blockers may be used as primary prophylaxis in high-risk esophageal varices. The estimated effect of banding ligation in some trials may be biased and was associated with the duration of follow-up. Further high-quality trials are still needed.