Although a number of studies have been conducted to address the relation between a gene deletion polymorphism of glutathione S-transferase M1 (GSTM1) and breast cancer, no definite conclusion has been reached and no clear risk pattern has yet to emerge for GSTM1. We first conducted case-control studies that included 1920 subjects using a genotyping method allowing the definition of GSTM1-null (-/-), homozygous wild-type (+/+), and heterozygous (+/-) genotypes. The results show that GSTM1(-/-) confers an increased risk for breast cancer development compared with that in GSTM1-present individuals (+/+ and +/-), which was subsequently confirmed by a meta-analysis of all of the 41 relevant studies (odds ratio: 1.10, P<0.001). Unexpectedly, we found that GSTM1(+/+) is also a risk genotype compared with GSTM1(+/-). Furthermore, we identified a functional polymorphism in the GSTM1 promoter region associated with breast cancer. The variant allele modifies DNA binding to the AP-2alpha transcription factor, resulting in reduced promoter activity and mRNA expression. However, this low-activity allele is associated with reduced breast cancer risk. It seems that approximately 60-70% expression from one allele of GSTM1 could suffice for protection against breast cancer; null activity and overactivity of GSTM1 are both disadvantageous. These results indicate a U-shaped association of GSTM1 with breast cancer, which challenges the linear gene-dosage effect of GSTM1 that was previously proposed. We recommend that a more complicated role for GSTM1 should be considered in breast cancer risk prediction.