10521 Background: Subcategories of NSCLC patients may benefit from epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKI's). Patients with cancers that harbor mutated EGFR have a higher chance of response. At present, there are no diagnostic tests to identify likely responders more reliably than by mutation testing. Previously, we have shown that a classifier based on kinase activity profiles in the presence and absence of a kinase inhibitor predicted erlotinib response of NSCLC patients in a neoadjuvant setting (Hilhorst et al. J Clin Oncol. 2010;28:7s [suppl; abstr 10566].) The aim of the current study was to evaluate this classifier using a blinded test set of patient tumor samples.
METHODS
Frozen tumor tissue was obtained from NSCLC patients (stage IA-IIIA) who had received 21 days of neo-adjuvant treatment with erlotinib prior to complete surgical resection. Tissue cryosections were lysed in M-PER buffer supplemented with phosphatase and protease inhibitors. Kinase activity profiles of the lysates were generated in the presence and absence of erlotinib on PamChip peptide microarrays .The classifier profile, as derived from the training set (n=14; Hilhorst et al. J Clin Oncol. 2010;28:7s [suppl; abstr 10566].), was applied to the blinded test set (n=13) of patients. Clinical response evaluation was based on histopathological examination of the tumor tissues. All specimens were analyzed for EGFR and KRAS mutation status.
RESULTS
Based on kinase activity profiles in the presence (on chip) and absence of erlotinib, a PLS-DA classifier was obtained that distinguished responders and non-responders in the training set (n=14, with two EGFR and 3 KRAS mutants). Leave-one-out cross validation resulted in misclassification for 3 samples, including both EGFR mutants. Application to the 13 blinded samples resulted in correct prediction of outcome for 11 samples. One of two EGFR mutants and one of three KRAS mutants were misclassified.
CONCLUSIONS
This blinded study validates the use of a classifier based on kinase activity profiles of patients' own tumor tissue to predict the response to treatment. This test that measures drug effects at the molecular level, may be an important step towards personalized medicine.