The genetic variations underlying susceptibility to common diseases are largely unknown. Increasingly high-throughput and affordable methods of large-scale genotyping will be accompanied by discoveries of many of these susceptibility alleles. We describe a large-scale breast cancer association study with over 25,000 SNPs in 254 cases and 268 age-matched controls using allele frequency estimates derived from DNA pooling with chip-based mass spectrometry. The 74 SNPs most strongly associated with breast cancer status from the pool-based analyses were subsequently genotyped from the individual DNA sample of each study participant, identifying 52 markers of interest. Validation of these markers was accomplished by testing in two additional breast cancer collections (368 cases, 330 controls). Four common genetic variations were identified that were significantly associated with case-control status (P < 0.05), including the ICAM gene region on chromosome 19 recently published [Kammerer et al. Cancer Res. 2004 Dec. 15;64(24)]. Further association fine mapping of the regions surrounding each of these markers identified a single gene of interest in two of the four regions, and a small number of potential candidates in the remaining two. As expected, the common variations within each of these regions demonstrate modest marginal effects, with allelic odds ratio estimates between 1.3 and 2.1. However, when considered together in multi-locus models of disease risk, these markers may have substantial clinical relevance. Using a logistic regression model that included a significant two SNP interaction effect (P = 0.002), we identified one group of four-SNP genotypes having a combined relative frequency of 11% associated with a two-fold decrease in breast cancer risk, compared to the population average. Many low frequency genotypes were estimated to increase breast cancer risk more than six-fold. Although the SNPs in these four regions represent only a minor fraction of the common variations likely contributing to breast cancer risk, these results provide an example of the potential role such information may play in the development of strategies to improve patient care.

[Proc Amer Assoc Cancer Res, Volume 46, 2005]