Benzene is an established cause of adult leukemia but its association with childhood leukemia remains unclear. We performed a systematic review of the epidemiologic literature on benzene and childhood leukemia, and used meta-analysis to explore causal inference, evaluate bias, and identify sources of heterogeneity. The summary relative risk estimate (RR) for childhood leukemia for all studies combined was 1.84 (95% confidence interval [CI]: 1.43, 2.37; n=23). In studies of acute myeloid leukemia (AML), the RR was 2.22 (95%CI: 1.65, 2.99; n=8). For studies of acute lymphocytic leukemia (ALL), the RR was lower (RR=1.62; 95%CI: 1.28, 2.04; n=16). Summary relative risks were higher for studies examining maternal (RR=1.96; 95%CI: 1.39, 2.78; n=13) versus paternal exposure (RR=1.23; 95%CI: 1.07, 1.41; n=14), and in studies assessing benzene exposure based on maternal occupation (RR=1.71; 95%CI: 0.91, 3.24; n=6) compared to paternal occupation (RR=1.18; 95%CI: 0.94, 1.50; n=6) or traffic density (RR=0.97; 95%CI: 0.75, 1.25; n=6). Summary relative risks were also higher for exposure during gestation (RR=1.91; 95%CI: 1.32, 2.75; n=12) compared to other periods, and remained elevated in studies adjusted for smoking or socioeconomic status. Overall, these findings suggest that benzene is associated with increased risk of childhood leukemia.