BACKGROUND: Breast tumors from African American women have less favorable pathological characteristics and higher mortality rates than those of Caucasian women. Although socioeconomic status may influence prognosis, biological factors are also likely to contribute to tumor behavior. METHODS: Patients with invasive breast cancer were matched by age, grade, and estrogen receptor status; patients with benign disease were matched by age and diagnosis type. RNA from laser microdissected tumors and whole-sectioned nonmalignant breast tissues was hybridized to HG U133A 2.0 microarrays. Data were analyzed using Partek Genomics Suite using a cutoff of P < .001, >1.5-fold change, and results were validated by quantitative real-time polymerase chain reaction. RESULTS: Clinicopathological factors did not differ significantly between groups for age at diagnosis, tumor size or stage, lymph node or human epidermal growth receptor 2 status, intrinsic subtype, or mortality. Two-way analysis of the tumor specimens revealed 25 probes representing 23 genes differentially expressed between populations; hierarchical clustering classified 24 of 26 African American women and 25 of 26 Caucasian women correctly. In the nonmalignant specimens, 15 probes representing 13 genes were differentially expressed, including 5 genes that also differed in the tumor specimens; these genes were able to correctly classify nonmalignant breast specimens from 20 of 22 of African American women and all of the Caucasian women. CONCLUSIONS: Despite matching of tumors by pathological characteristics, molecular profiles differed between African American women and Caucasian women in both invasive tumors and benign breast tissues. These differentially expressed genes, including CRYBB2, PSPHL, and SOS1, are involved in cellular growth and differentiation, invasion, metastasis, and immune response and thus may contribute to the poor outcome in African American women. Cancer 2012;118:1334-