Aggressive breast cancer : epidemiological studies addressing disease heterogeneity

Abstract

Breast cancer is either the number one or the second most common cause of cancer death in women in the world, depending on region¹. It is a cancer heterogeneous in many aspects related to the aggressiveness, such as proliferation rate, metastatic capacity and survival. This thesis is seeking to increase our understanding of aggressive breast cancer, and how risk factors may be related to it.

In study I, we compared interval cancers (IC) to cancers detected at screening (SC) and found the group of IC to be characterized by higher frequency of BRCA mutations, family history of breast cancer and use of hormonal therapy (HRT) as compared to SC. IC in non-dense breasts were enriched for aggressive tumour features, and in this group the estimates for family history and BRCA mutations were increased.

In study II, we studied if predictors intended to identify healthy women’s risk of breast cancer in future prevention efforts were skewed towards certain tumour characteristics. A 77-SNP breast cancer polygenic risk score appeared to underestimate risk in women with high grade and Oestrogen receptor (ER) negative tumours, as it was on average higher in women with low grade, ER positive tumours. The TyrerCuzic model of breast cancer risk also appeared to underestimate risk in ER negative, high grade tumours but this was restricted to early onset cases. Only mammographic density appeared to be a general risk factor/predictor for all tumours independent of prognosticators.

Study III was a case-control study where we estimated odds ratios for each of four breast cancer molecular subtypes separately in multinomial logistic regression. We found subtype heterogeneity in the odds ratios for genetic risk factors and for breastfeeding. The 77-SNP polygenic risk score was associated with all subtypes except for the basal-like subtype, which showed no association with the score. Although breastfeeding was protective for both luminal and basal-like subtypes, the magnitude and underlying mechanism appeared to differ across subtypes.

In Study IV we assessed the concordance between PAM50 gene expression-based and immunohistochemistry-based molecular subtypes. No proxy showed more than moderate concordance with PAM50, however if luminal A and B subtypes were collapsed into one category, substantial concordance was achieved. Sensitivity for HER2-enriched breast cancer as defined by PAM50 was low, at 0.36 for all proxies investigated, whereas sensitivity and specificity was high for classifying basal-like breast cancer.

¹According to Cancer facts sheets, GLOBOCAN 2012, [IARC].