Background Many studies indicated the human cytochrome P450 2D6 (CYP2D6) gene polymorphism was associated with acute leukemia (AL) susceptibility, however, the results were inconsistent. So we performed this meta-analysis to evaluate the relationship between CYP2D6*3 or CYP2D6*4 polymorphism and AL susceptibility. Methods We searched PubMed database up to February 20, 2013, and finally yielded 9 case-control studies including 1343 cases and 1843 controls which tested the association between CYP2D6*3 or *4 polymorphism and AL. After data extraction, we conducted a meta-analysis using the Comprehensive Meta Analysis software. Results Overall, no significant association between CYP2D6*3 or *4 polymorphism and AL risk was found in this meta-analysis (+ vs. -: OR=1.13, 95% CI=0.79-1.63; +/+ vs. -/-: OR=1.73, 95% CI=0.99-3.02; -/+ vs. -/-: OR=1.03, 95% CI=0.68-1.56; (-/+ and +/+) vs. -/-: OR=1.08, 95% CI=0.72-1.63; +/+ vs. (-/+ and -/-): OR=1.76, 95% CI=0.98-3.17). Similar results were also been found in stratified subgroup analysis. There was no publication bias. Conclusion CYP2D6*3 or *4 polymorphism might not be associated with AL susceptibility. However, the results need to be further confirmed by well-designed and high quality randomized controlled trials with larger sample sizes.