Abstract 1670: Molecular subtypes from PAM50 in a breast cancer cohort: differences by patient characteristic, reproducibility

Proceedings: AACR 103rd Annual Meeting 2012‐‐ Mar 31‐Apr 4, 2012; Chicago, IL BACKGROUND Gene expression profiling has been shown to improve prognostication in clinical trials, but data are lacking on expression-based molecular subtypes in context of variables of interest for epidemiology. We provide interim data on the distribution of subtypes determined by PAM50 assay in relation to standard immunohistochemical (IHC) markers and to subject characteristics in a breast cancer cohort. METHODS The study population combines women from the Life After Cancer Epidemiology (LACE) study with invasive breast cancer stages I (>=1 cm), II, or IIIA, diagnosed in 1997-2000 and the Pathways Study with invasive breast cancer at any stage (>= 0.5 cm) diagnosed in 2006-2008. IHC results for estrogen and progesterone receptors (ER, PR) and her2-neu (HER2) were obtained from records. Race, ethnicity, and breast cancer risk factors were from self-report. For a stratified sample of cases, 1 mm punches were obtained from areas of representative tumor in formalin-fixed, paraffin-embedded tissue blocks. Expression of the PAM50 genes was determined by RT-PCR of extracted RNA. Random blind duplicate tissue punches were assayed. A published centroid-based algorithm (Parker, J Clin Oncol 2009;27:1160-7) was used to generate, for each sample, five continuous-scale normalized subtype scores and a predicted subtype classification. We estimated sample-weighted subtype distributions, Pearson correlations among subtype scores, and age-adjusted odds ratios (OR). RESULTS PAM50 subtype predictions for 702 breast tumors from a community cohort were 55% Luminal (Lum) A, 19% LumB, 11% HER2-enriched, 10% Basal-like, and 5% Normal-like. Among the 72% of tumors categorized as ER+ or PR+, HER2- by IHC, PAM50 subtypes were: 72% LumA, 20% LumB, 2% HER2-enriched, 1% Basal-like, and 5% Normal-like. Continuous-scale LumA, LumB, HER2-enriched, and Normal-like scores were mutually positively correlated. Basal-like score was inversely correlated with other subtypes. 32 of 34 (94%) blind duplicate pairs were concordant for subtype. Compared to non-Hispanic whites, African-American women were less likely to have LumB subtype, OR 0.1 (95% CI 0.03 - 0.4), and somewhat more likely to have Basal-like subtype, OR 2.6 (95% CI 0.8-8.5) and Hispanics more often had non-LumA subtypes. Women with a family history of breast cancer had a lower frequency of HER2-enriched tumors, OR 0.3 (95% CI 0.2-0.7). DISCUSSION Molecular subtyping by PAM50 shifted 23% of tumors from the low-risk IHC-based ER+ or PR+, HER2- category to predicted higher-risk subtypes. Correlations among subtype scores and (infrequent) discordance of duplicate tissue punches may be accounted for by borderline phenotypes and/or heterogeneous cancers. Interim results showing PAM50 subtype differences by race and ethnicity and by breast cancer risk factors indicate that molecular subtyping is a promising tool for describing etiologic heterogeneity. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 1670. doi:1538-7445.AM2012-1670