A customized 1600 gene expression array for the prediction of tumor response against targeted and cytotoxic drugs

Abstract

22050 Background: In a previous study we have demonstrated a high correct prediction of xenograft testings for cytotoxic standard agents. We investigated the hypothesis that correlating drug response with gene expressions would identify gene signatures that predict the drug response of individual tumors. Methods: Cytotoxic drugs including five alkylating agents, two antimetabolites, two topoisomerase II inhibitors, two tubulin binders and the antibodies bevacizumab and cetuximab. The cytotoxic drugs were tested at their MTDs. Antitumor activity was evaluated at the minimum T/C value or after 28 days for 31–78 tumors. Tumors were considered as responsive if the different drugs effected a tumor volume inhibition to less than 11–41% (median 25%) of control tumors (T/C%). Using these criteria, on average 1/3 of the test tumors were responders and 2/3 were non-responders. Predictive gene expression signatures were determined by bioinformatic comparison of Affymetrix HG-U133 plus 2.0 profiles of responders and non-responders, verified using a leave-one-out cross- validation and for 3 drugs on an independent testing set. Results: On average, the response rate against the cytotoxic drugs for predicted responders was 83% in contrast to 34% for all tumors. Conversely, the predicted non-responders were resistant in 94% compared to 66% by random testings. For bevacizumab we obtained a 2.2-fold increase in response rate (67% vs. 30%) and for cetuximab 2,5-fold (24% vs. 61%). The correct prediction for non-response was 88% for Bevacizumab and 94% for Cetuximab. Conclusions: Prediction of tumor response to cytotoxic and antibody drugs may enhance the chance for patients to receive the best possible treatment for their cancer. Oncotest has developed a customized gene expression array (ISO17025 certified) comprising 1600 genes that predict response or resistance to 11 cytotoxic and the 2 targeted drugs. It is now evaluated in a clinical pilot study focusing on the response prediction of targeted drugs. No significant financial relationships to disclose.