BACKGROUND: Our objective was to examine the effect of antenatal peer support on rates of breastfeeding initiation. METHODS: We performed a systematic review of randomized controlled trials, quasi-randomized trials and cohort studies with concurrent controls. We searched the Cochrane Library, MEDLINE, the Cumulative Index to Nursing and Allied Health Literature (CINAHL), the National Research Register and the British Nursing Index from inception or from 1980 to 2009. We carried out study selection, data abstraction and quality assessment independently and in duplicate. We defined high-quality studies as those that minimized the risk of at least three of the following types of bias: selection, performance, measurement and attrition bias. We calculated risk ratios (RRs) and 95% confidence intervals (CIs) for individual studies and undertook separate meta-analyses for high-quality trials of universal peer support and targeted peer support. We did not include low-quality studies. RESULTS: We selected 11 studies, which involved 5445 women in total. Seven of these studies (involving 4416 women in total) evaluated universal peer support, and four studies (involving 1029 women in total) targeted antenatal peer support. In the three high-quality studies of universal peer support, all involving low-income women, the relative risk for not initiating breastfeeding was 0.96 (95% CI 0.76-1.22). In the three high-quality studies of antenatal peer support that targeted women considering breastfeeding, the relative risk for not initiating breastfeeding was 0.64 (95% CI 0.41 - 0.99). INTERPRETATION: Universal antenatal peer support does not appear to improve rates of breastfeeding initiation, but targeted antenatal peer support may be beneficial. This effect may be related to context, however, so any new peer-support program should undergo concurrent high-quality evaluation.