Risk assessment of relapse in Swiss participants of BIG 1-98 with grade 2 early breast cancer by gene expression profiling using RNA from formalin-fixed, paraffin embedded (FFPE) tissue.

Abstract

Abstract Abstract #1073 BIG 1-98 is a randomized controlled clinical trial for postmenopausal patients with early, estrogen and/or progesterone receptor (ER/PgR) positive breast cancer. It compares adjuvant therapy with 5 years of tamoxifen (T) or letrozole (L), two years of T followed by 3 years of L (T→L) or vice versa (L→T). L improved disease free survival (DFS) over T by 18% (hazard ratio 0.82, JCO 2007;25:486ff). Switzerland contributed 598/8010 patients. The patient and tumor characteristics of Swiss patients was similar to the entire BIG 1-98 population. The purpose of this study was to test a prognostic score in RNA derived from FFPE tissues.
 Scores representing ER and progesterone receptor (PgR) function, proliferation (PRO14, AURKA/STK15, BIRC5/SURV, CCNB1, CCNB2, CCNE2, CDC2, ENPF, KIF20A, MKI67, MYBL2, ORC6L, PRC1, SPAG5, TOP2A), and HER-2 were built from multiple genes for each score. Genes were selected according to an analysis of publicly available cDNA array databases. In a prior set of 82 pairs of breast cancer samples the Pearson correlation coefficients of the scores between fresh frozen and FFPE tissues were at least 0.94. Cut-offs for ER and HER-2 scores were defined as the antimodes of fitted bimodal Gaussian distributions; the ER and HER-2 scores closely agreed with immunohistochemical (IHC) data (kappa(ER)=0.868, kappa(HER-2)=0.836).
 FFPE tissues were available from 437 Swiss BIG 1-98 patients. RNA of sufficient quantity and quality was extracted from 388 blocks (89%). Forty-seven genes were quantified by QRTPCR on TaqMan Low Density Arrays (TLDAs) as reported previously (Proc AACR 2008, Abstract 974).
 The PRO14 score was positively correlated with the histological grade. PRO14 scores greater than the median and PGR scores lower than the median were associated with shorter DFS. In patients with grade 2 tumors, PRO14 scores were significant predictors of DFS in univariate Cox regression analyses. Multivariate analyses will be presented. Our findings are in agreement with an analysis of fresh frozen tissues of Belgian participants in BIG 1-98 (European Breast Cancer Conference 2008, Abstract 481).We show that RNA can be extracted from FFPE in sufficient quantity and quality to be used in QRTPCR using TLDAs. This opens the possibility of using FFPE tumor banks from controlled clinical trials for the investigation of gene expression patterns. Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 1073.