GENOMIC CLASSIFICATION OF SECONDARY BREAST CANCER: EXPANDING APPROACHES TO DIVERSE POPULATIONS

Background: Clinical classifications of breast cancer recurrence, developed largely to exclude second primary tumors from ascertainment, vary widely between studies and lead to substantial heterogeneity in estimated incidence, prognosis, and optimal therapeutic intervention. Genomic algorithms relying on the presence of shared somatic alterations offer one alternative to clinical classifications but have not yet been evaluated in diverse populations nor compared against clinical methods. The aims of this dissertation are to 1) highlight population-level differences in the genomic features of breast cancers, particularly as a function of race and age, and 2) evaluate the accuracy of clinical recurrence classifications.Methods: Aim 1 investigates specific single nucleotide variants and copy number alterations enriched in the tumors of Black women and younger women (<50) in a pooling of two cohorts: the Carolina Breast Cancer Study (CBCS; N = 357) and the Cancer Genome Atlas (TCGA). Aim 2 uses probabilistic modelling of mutation and copy alteration prevalence to predict recurrence status in 42 CBCS tumor pairs and evaluates the sensitivity of models to assumptions about reference population.Results: In Aim 1, Black women had higher prevalence of TP53 and FAT1 mutations, MYC amplifications, and genomic instability, but lower prevalence of GATA3, CDH1, and PIK3CA mutations, than non-Black women. Subtype did not fully explain these differences, however, socioeconomic status modified associations. Age at diagnosis was not strongly related to somatic landscape. In Aim 2, estimated probabilities of recurrence were not substantially impacted by choice of reference population. Overall, pathologist classifications were 86%ivaccurate in classifying local recurrence, similar to the 88% accuracy of registry-based schema. However, pathologist determinations tended to emphasize sensitivity (more complete ascertainment of recurrence), while registries emphasized specificity (more complete ascertainment of second primary tumors).Conclusions: Breast cancer somatic biology varies by race and socioeconomic status. Black women are more likely to have tumors carrying markers of estrogen receptor (ER) negative disease, though differences were also observed among ER+ tumors, underscoring the complex interaction between social experiences and cancer biology. Despite these variations, genomic classifiers of recurrence were insensitive to choice of reference population and suggest modest misclassification of secondary tumors.