OBJECTIVE: The authors used a meta-analytic technique to (1) quantify the evidence of an association between duration of breastfeeding and risk of childhood acute lymphoblastic leukemia (ALL) or acute myeloblastic leukemia (AML), (2) assess the influence of socioeconomic status (SES) on any such associations, and (3) discuss the implications of these findings for the evaluation of whether breastfeeding reduces the risk of childhood leukemia. METHODS: A fixed effects model was employed to systematically combine the results of 14 case-control studies addressing the effect of short-term (< or = 6 months) and long-term (>6 months) breastfeeding on the risk of childhood ALL and/or AML. Subgroup analyses of studies that did and did not adjust for SES were also performed. RESULTS: A significant, negative association was observed between long-term breastfeeding and both ALL risk (odds ratio [OR]=0.76; 95% confidence interval [CI] 0.68, 0.84) and AML risk (OR=0.85; 95% CI 0.73, 0.98). Short-term breastfeeding was similarly protective for ALL and AML. Results for studies that adjusted and did not adjust for SES were not significantly different from the results for the 14 studies combined. CONCLUSIONS: This meta-analysis showed that both short-term and long-term breastfeeding reduced the risk of childhood ALL and AML, suggesting that the protective effect of breastfeeding might not be limited to ALL as earlier hypothesized. Potential bias introduced by different participation rates for case and control samples that differed in SES can be minimized by implementing larger case-control studies with SES-matched, population-based controls.