We performed a meta-analysis to investigate the role of XRCC1 polymorphisms Arg194Trp, Arg280His and Arg399Gln in breast cancer. The results were pooled in a manner that appropriately reflects a biological model of gene effect using a random effects logistic regression model without multiple comparisons. Forty studies from 31 reports were included with 10 465 cases and 10 888 controls at Arg194Trp, 6156 cases and 5806 controls at Arg280His, and 21 467 cases and 22 766 controls at Arg399Gln. Our analysis found a tendency towards a recessive effect of Arg280His variant in Asian population only (His/His vs. Arg/Arg+Arg/His: OR=2.27, 95% CI=0.82, 6.31). An increased breast cancer risk with a recessive effect was also suggested for Arg399Gln variant in Asian population (Gln/Gln vs. Arg/Arg+Arg/Gln: OR=1.59, 95% CI=1.22, 2.09) only. These findings suggest that polymorphisms Arg280His and Arg399Gln may modify breast cancer risk differently in Caucasian and Asian populations.