Background: Axillary lymph node involvement remains the most important prognostic factor in early-stage breast cancer. We hypothesized that molecular classification based on breast cancer biology would predict the presence of nodal involvement at diagnosis, which might aid treatment decisions regarding the axilla.
Patients and methods: From a clinically annotated tissue microarray of 4444 early-stage breast cancers, expression of estrogen receptor (ER), progesterone receptor (PgR), HER2, epidermal growth factor receptor, and cytokeratin 5/6 was determined by immunohistochemistry. Cases were classified by published criteria into molecular subtypes of luminal, luminal/HER2 positive, HER2 positive/ER negative/PgR negative, and basal. Risk of axillary nodal involvement at diagnosis was determined in 2 multivariable logistic regression models: a "core biopsy model" including molecular subtype, age, grade, and tumor size and a "lumpectomy model," which also included lymphovascular invasion. Luminal was used as the reference group. After internal validation of findings in 2 independent sets, we conducted combined analysis of both.
Results: In the core biopsy model, the molecular subtypes had a predictive effect for nodal involvement (P= .000001), with the basal subtype having an odds ratio for axillary lymph node involvement of 0.53 (95% CI, 0.41-0.69). Tumor grade (P=5.43 x 10(-12)) and size (P=8.52 x 10(-35)) were also predictive for nodal involvement. Similar results were found in the lumpectomy model, where lymphovascular invasion was also predictive (P=2.74 x 10(-115)).
Conclusion: These results indicate that the basal breast cancer molecular subtype predicts a lower incidence of axillary nodal involvement, and including biomarker profiles to predict nodal status at diagnosis could help stratification for decisions regarding axillary surgery and locoregional radiation.