<p>SUPPLEMENTARY METHODS</p><p>Carolina Breast Cancer Study. The Carolina Breast Cancer Study (CBCS) is a population-based, case-control study of breast cancer conducted in 24 counties of central and eastern North Carolina between 1993 and 2001 (Newman et al., 1995;</p><p>Millikan et al., 2003). Incident cases of primary invasive and in situ breast cancer were identified using the North Carolina Central Cancer Registry, with over-sampling of</p><p>African-American and younger women. Controls were frequency-matched to cases based upon age (± 5 years) and self-reported race using randomized recruitment.</p><p>Cases were subtyped using a panel of immunohistochemistry markers that identified the</p><p>"intrinsic" breast cancer subtypes (Carey et al., 2006; Millikan et al., 2008). Definitions of subtypes were luminal A (ER+ PR+ HER2-), luminal B (ER+PR+HER2+), basal-like</p><p>(ER-PR-HER2-, ck5/6+ and/or HER1+), HER2+ (ER-PR-HER2+), and unclassified</p><p>(negative for all five markers). Procedures for recruiting and enrolling study participants and obtaining biologic samples were approved by the Institutional Review Board of the</p><p>UNC School of Medicine, and informed consent was obtained from each participant.</p><p>Genotyping of B-Myb. Genotyping for the B-Myb polymorphism at codon 427, Serine</p><p>(AGC) to Glycine (GGC) (rs2070235), was conducted using DNA extracted from peripheral blood lymphocytes. The ABI 7700 Sequence Detection System, or</p><p>"Taqman"™ assay (Applied Biosystems, ABI, Foster City, CA, USA) was employed, with polymerase chain reaction (PCR) primers and probes designed using Primer</p><p>Express TM software (ABI) and assay conditions based on the allelic discrimination protocol from Applied Biosystems. PCR primer and probe sequences are available upon request. Out of a total of 3862 samples genotyped, the failure rate was 0.4%, and complete agreement was obtained on a 10% random sample.</p><p>There were no significant departures from Hardy-Weinberg equilibrium among</p><p>African-American cases (p=0.80), African-American controls (p=0.77), Caucasian cases</p><p>(p=0.27), or Caucasian controls (p=0.51).</p><p>Cell line genotyping was performed by Polymorphic DNA Technologies, Inc</p><p>(Alameda, CA, USA). </p><p>Genotype Statistical Analysis</p><p>B-Myb genotypes were determined for 1256 cases with subtype information (500</p><p>African-American, 756 Caucasian) and 1814 controls (679 African-American, 1135</p><p>Caucasian). Odds ratios (ORs) and 95% Confidence Intervals (CIs) for each breast cancer subtype versus all controls were calculated using unconditional logistic regression. The PROC GENMOD statement in SAS (version 8.2; SAS Institute, Cary,</p><p>NC, USA) was used to incorporate offsets derived from sampling probabilities used to identify eligible study participants, race (African-American, Caucasian), and age (11- level ordinal variable representing 5-year age categories). Trend tests were conducted by calculating the p-value for B-Myb genotype coded as an ordinal variable. Likelihood ratio tests (LRTs) were used to test for modification of ORs by race.</p>