To compare the distribution and prognostic effect of the breast cancer molecular subtypes in young and elderly breast cancer patients.Patients and methods:
Our study population (n = 822) consisted of all early breast cancer patients primarily treated with surgery in our center between 1985 and 1996. A total of 142/822 fresh frozen tissues were available with good quality RNA and analyzed by gene expression microarray. Gene expression molecular subtypes were determined by correlation to the expression centroids of 534 “intrinsic” genes. Sections of a tissue micro array containing formalin-fixed paraffin-embedded tumor tissue of 714/822 patients were immunohistochemically (IHC) stained for Ki67, EGFR, CK5/6. Tumor expression of ER, PR, HER2 was previously determined. IHC molecular subtypes were defined based on expression of these markers: Luminal A: ER+ and/or PR+, HER2– and Ki67–; Luminal B: ER+ and/or PR+ and ki67+; ERBB2: ER–, PR– and HER2+; Basal-like: ER–, PR–, HER2– and EGFR+ and/or CK5/6+; Unclassified: ER–, PR–, HER2–, EGFR– and CK5/6–. IHC molecular subtypes were validated against gene expression defined molecular subtypes. Assessment of distribution and prognostic effect of molecular subtypes was stratified to age (<65 versus ≥65 years).Results:
Validation of molecular subtypes determined by IHC against gene expression revealed a substantial agreement in classification (Cohen's kappa coefficient 0.75). A statistically significant association (p = 0.02) was found between molecular subtypes and age, where Luminal tumors were more often found in elderly patients, while ERBB2, basal-like and unclassified subtypes were more often found in young patients. Molecular subtypes showed a prognostic association with outcome in young patients concerning relapse-free period (RFP) (p = 0.01) and relative survival (RS) (p < 0.001). No statistically significant prognostic effect was found for molecular subtypes in elderly patients (RFP p = 0.5; RS p = 0.1). Additional analyses showed that no molecular subtypes showed a statistically significant difference in outcome for elderly compare to young patients.Conclusion:
We have shown that molecular subtypes have a different distribution and prognostic effect in elderly compared to young breast cancer patients, emphasizing the fact that biomarkers may have different distributions and prognostic effects and therefore different implications in elderly compared to their younger counterparts. Our results support the premise that breast cancer clinical behavior is significantly affected by patient age. We suggest that competing risks of death in elderly patients, ER-driven differences and micro-environmental changes in biology are underlying these age-dependent variations in patient prognosis.