To the Editor, For most sporadic breast cancers, a substantial component of risk is due to multiple low-penetrance susceptibility genes [1]. Since estrogens and their metabolites are carcinogens in mammary glands [2], estrogen metabolismrelated genes are of particular interest in the study of breast cancer susceptibility [3–5]. Most recently, we revealed that NQO2, which catalyzes the reduction of electrophilic estrogen quinones, confers specific protection against breast cancer [6]. Two common genetic changes in NQO2 influence breast cancer risk through an altered gene expression regulation mechanism. One of these two common variants is a tri-allelic insertion/deletion polymorphism. Wang et al. [7] originally indicated that this polymorphism contains three alleles, a 29-bp insertion (I-29), a 29-bp deletion (D), and a 16-bp insertion (I-16). In their genotyping work, in which most subjects were Caucasian, the only heterozygote detected had a heterozygous I-29 and D genotype (I-29/D). Neither I-29/I-16 nor D/I-16 heterozygotes were detected. Since no heterozygote harboring the I-16 allele was observed, Wang et al. proposed that the I-16 allele is likely to be stable only when paired with another I-16 allele. However, in our case–control study, which included 878 cases and 711 controls in the first stage, we got different results [6]. Among the 1,589 individuals in the study, all of whom were Chinese women, there were 1,014 of the homozygous I-29 allele genotype (I-29/I-29), 18 of I-29/ I-16, 3 of I-16/I-16, 484 of I-29/D, 6 of I-16/D, and 64 of D/D. Most women harboring the I-16 allele were heterozygotes (I-29/I-16 and I-16/D genotypes), rather than homozygotes (I-16/I-16 genotype). The frequency of the I-16 allele in the Chinese population is 0.94%, slightly lower than in the Caucasian population (3.5%). Our findings challenge the previous assumption that the I-16 allele is stable only when paired with another I-16 allele [7]. The different pattern of the I-16 allele between the Caucasians and the Chinese could be due to either ethnicity or to the insufficient sample size in Wang’s study. Last but not least, potential genotyping errors cannot be ruled out. In our study, we used a sensitive genotyping strategy that permits identification of I-29, I-16 and D alleles. In brief, we first used polymerase chain reaction (PCR) and lengthy electrophoresis (not short-time electrophoresis) to discriminate between the I-29, I-16, and D alleles using the primers 50-CTGCCTGGAAGTCAGCAGGGTC-30 (sense) and 50CTCTTTACGCAGCGCGCCTAC-30 (antisense) [8]. In addition, the I-29 and I-16 alleles are difficult to separate through electrophoresis because the difference in length is only 13 bp, so we subsequently employed a PCR restriction fragment length polymorphism (RFLP) approach to discriminate between the I-16 and I-29 alleles. In contrast, Wang et al. used only PCR followed by electrophoresis. Figure 1a is from Wang’s study, showing PCR-based genotyping of 15 human primary fibroblast cell lines for NQO2 gene promoter allelic variants [7]. After reviewing Fig. 1a, we found that the PCR-based genotyping was not performed well and that some samples identified as I-29/ I-29 are suspected to be I-29/I-16 (e.g., sample 11 in Fig. 1a). The shape of that band is more likely to be a complex of I-29 and I-16. Our typical gel electrophoresis plots are shown in Fig. 1b and c (upper). Samples in K.-D. Yu G.-H. Di L. Fan Z. Hu A.-X. Chen Z.-M. Shao (&) Department of Breast Surgery, Breast Cancer Institute, Cancer Hospital, Institutes of Biomedical Science, Shanghai Medical College, Fudan University, 399 Ling-Ling Road, 200032 Shanghai, People’s Republic of China e-mail: zhimingshao@yahoo.com