Abstract S3-05: Integrated analysis of multidimensional genomic data on CALGB 40601 (Alliance), a randomized neoadjuvant phase III trial of weekly paclitaxel (T) and trastuzumab (H) with or without lapatinib (L) for HER2-positive breast cancer

Background: RNA profiling and mutational analyses in CALGB 40601 (NCT00770809) found significant impact on pathologic complete response (pCR) rates from tumor (intrinsic subtype, p53 mutation) and microenvironmental (immune cell) features. Integrated analysis across platforms is needed to better understand the roles of these different factors with respect to response to HER2-targeted therapies.Methods: We performed a comprehensive genomic analyses on pCR, defined as no invasive tumor in the breast, by integrating clinicopathological information with somatic mutation status, 422 segment-level DNA Copy Number Alterations (CNAs), and 510 gene expression signatures using mRNAseq and DNA exome sequencing from 213 pre-treatment tumors. Excluding 48 samples in the TL arm that was closed early due to futility, and 4 Normal-like tumors, the dataset consisted of 161 patients from TH and THL arms including 47 HER2-enriched (HER2E), 8 Basal-like, 54 Luminal A, and 52 Luminal B, all of whom received H. The main analysis was performed using the Elastic Net on multivariate logistic regression models for predicting pCR. The samples were divided into a training and a test set, then models were built to predict pCR by 10-fold cross-validation in the training set, then applying the best model onto the test set to construct ROC curves and evaluate prediction accuracy by calculating area under ROC (AUC). We also used the DawnRank, a network-based bioinformatics tool that integrates DNA and RNA data to identify driver genes, to find predictors of resistance to H-containing therapies.Results: Among clinicopathological factors, clinical estrogen/progesterone receptor (ER/PgR) status and intrinsic subtype by PAM50 were statistically associated with pCR, but treatment arm (TH vs THL) and stage were not. In the Elastic Net analysis, the models incorporating either gene signatures (AUC: 0.724) or CNAs (AUC: 0.777) were more predictive of response than mutation status model (AUC: 0.635). Gene signatures and CNAs were further combined with either mutation status (AUC: 0.773), clinical ER/PgR status (AUC: 0.787) or ER/PgR status plus intrinsic subtype (AUC: 0.784). The combination with the highest AUC comprised gene signatures, CNAs, and ER/PgR status, and demonstrated that CNAs at Chromosome (Chr.) 6p, 10q22, or 11q23, the signature of Correlation to HER2E, and a T-cell signature, positively predicted pCR and that Luminal and PgR gene signatures were negative predictors. The CN gain of Chr.6p, which contains the HLA genes, predicted for pCR and was associated with higher expression of HLA genes and B cell / IgG signatures. The CN loss of Chr.11q23 including CD3D, CD3E, and CD3G was also identified by DawnRank as a region associated with resistance.Conclusions: Tumor genetics (CNAs), tumor RNA subtype (HER2E, Luminal), and the microenvironment (immune cells) were independently predictive of response to H-containing therapies and biologically and clinically important for HER2-positive breast cancer, supporting integrated RNA- and DNA-based tumor assessments to clarify response to HER2-targeting.Support: U10CA031946/033601/180821/180882/180888.Citation Format: Tanioka M, Fan C, Carey LA, Hyslop T, Pitcher BN, Parker JA, Hoadley KA, Henry NL, Tolaney S, Dang C, Krop IE, Harris L, Berry DA, Mardis E, Perou CM, Winer EP, Hudis CA. Integrated analysis of multidimensional genomic data on CALGB 40601 (Alliance), a randomized neoadjuvant phase III trial of weekly paclitaxel (T) and trastuzumab (H) with or without lapatinib (L) for HER2-positive breast cancer [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr S3-05.

Duke Scholars

Author Terry Hyslop Biostatistics & Bioinformatics, Division of Translational Bi ...