BACKGROUND: Caffeine is teratogenic in animal studies when administered at high concentrations. Previous review articles have concluded that maternal caffeine consumption does not influence the risk of congenital anomalies. These reviews were narrative rather than systematic. The objective of the current systematic review is to provide a critical appraisal of epidemiologic evidence. METHODS: A search of the MEDLINE/PUBMED database (1966-October 2004) was conducted for all published epidemiologic studies with maternal intake of caffeine as an exposure and major malformations as an outcome. Study characteristics were abstracted, internal validity evaluated, and study findings summarized. RESULTS: Twenty-five papers met the initial criteria for inclusion, of which 18 were subsequently excluded as a result of other limitations. Effect estimates for the remaining 7 studies were generally close to null. Specific subgroup analyses were summarized across studies (associations between coffee and cardiovascular malformations, coffee and oral clefts, and tea and cardiovascular malformations). Summary point estimates ranged from 1.0 to 1.2; the upper limits of all confidence intervals were less than 1.7. CONCLUSIONS: There is no evidence to support a teratogenic effect of caffeine in humans. Current epidemiologic evidence is not adequate to assess the possibility of a small change in risk of congenital anomalies resulting from maternal caffeine consumption.