Validated 3 Gene Signature Predicts Response to Neo-Adjuvant Chemotherapy in Luminal Breast Cancer – Results from GeparTrio and GeparQuattro.

Abstract

Abstract Background:Luminal breast cancers (LBC), defined as positive for estrogen receptor (ER) and negative for HER2, are of variable prognosis. We hypothesize that those showing potential sensitivity to neoadjuvant chemotherapy (NACT) are those with an unfavorable prognosis unless they achieve a pathological complete response (pCR = no invasive residual in breast and nodes). We therefore aimed to develop a Neoadjuvant Luminal Response Score (NLRS), identifying luminal tumors with a high chance for a pCR.Patients and methods:We examined mRNA expression of 57 candidate genes and 3 normalization genes by kinetic RT-PCR in FFPE tissues from pre-therapeutic core biopsies from 380 patients with LBC and cT2-T4 disease either receiving NACT with 6x docetaxel/ doxorubicin/ cyclophosphamide in GeparTrio (n=148, training cohort) or epirubicin/ cyclophosphamide followed by Docetaxel (+simultaneous or sequential capecitabine) in GeparQuattro (n=232, validation cohort). Samples were classified as LBC by mRNA expression analysis of ESR1 and Her2 using pre-defined cut-off values. The continuous NLRS was generated by linear combination of expression values of 3 out of 57 genes.The first pre-specified primary objective of training part was to demonstrate that the percentage of patients with LBC with a pCR is higher in NLRS+ tumors than in all LBC. The second pre-specified primary objective for validation part was to demonstrate that the NLRS score is able to identify the pCR cases in the LBC with a sensitivity of ≥80% (upper confidence interval [CI] limit) and a positive predictive value (PPV) of ≥20%.Results:In the GeparTrio training cohort, the NLRS was predictive of pCR by logistic regression analysis (AUC=0.88). Using a cutoff defined on the basis of the ROC curve, the NLRS was able to predict pCR with a sensitivity of 82%, a specificity of 77%. While the pCR rate in all LBC was only 8.2%, this rate increased to 24% in the group of 57 patients with a NLRS above the cutoff. The pCR rate in 91 patients with a NLRS below the cutoff was 1.1%.When patients of all 3 treatment arms of the GeparQuattro validation cohort were analyzed together, AUC was 0.69 (p=0.004). Using the pre-defined cutoff, sensitivity was 52% (95% CI: 33-71%), specificity 74% (69-79%) and PPV 16%. However, if only patients treated without capecitabine containing NACT (to be more comparable to GeparTrio) (N=68) were included, AUC was 0.72 and sensitivity increased to 63 (23-89)% and PPV to 21%, passing the pre-defined limits for positive validation.Conclusion:Our study provides proof-of-principle that it is possible to develop and validate predictive signatures for response to NACT based on pretherapeutic FFPE biopsies. The 3-gene NLRS is a first validated gene signature selecting LBC with a high chance for pCR to NACT. The results are a basis for further development of predictive signatures. A prospective evaluation will be performed in the PREDICT study, a substudy to GeparQuinto.Supported by a BMBF grant. Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 2132.