Published data on the association between p21 Ser31Arg polymorphism and breast cancer risk are inconclusive. To derive a more precise estimation of the relationship, a meta-analysis was performed. Crude ORs with 95% CIs were used to assess the strength of association between the p21 Ser31Arg polymorphism and breast cancer risk. The pooled ORs were performed for co-dominant model (Ser/Arg vs. Ser/Ser, Arg/Arg vs. Ser/Ser), dominant model (Arg/Arg + Ser/Arg vs. Ser/Ser), and recessive model (Arg/Arg vs. Ser/Arg + Ser/Ser). A total of 21 studies including 22,109 cases and 29,127 controls were involved in this meta-analysis. Overall, no significant associations were found between p21 Ser31Arg polymorphism and breast cancer risk when all studies pooled into the meta-analysis. In the subgroup analysis by ethnicity, significantly increased risk was found for Caucasians (Arg/Arg vs. Ser/Ser: OR 1.496, 95% CI 1.164-1.924; and recessive model: OR 1.492, 95% CI 1.161-1.919). When stratified by study design, statistically significantly elevated risk was found for population-based studies (Ser/Arg vs. Ser/Ser: OR 1.085, 95% CI 1.019-1.156). In conclusion, this meta-analysis suggests that the p21 Ser31Arg polymorphism may be associated with breast cancer development in Caucasian. However, large sample and representative population-based studies with homogeneous breast cancer patients and well-matched controls are warranted to confirm this finding.