Predictive Biomarkers and Personalized Medicine A Clinical Model for Identifying Radiosensitive Tumor Genotypes in Non–Small Cell Lung Cancer

Purpose: Non–small cell lung cancer (NSCLC) includes a spectrum of radiosensitive and radioresistant tumors. However, little is known about the molecular determinants of cellular radiation responses. We examined clinical outcomes after gamma knife radiotherapy for NSCLC intracranial metastases to evaluate the use of this model for determining radiosensitive tumor genotypes. Experimental Design: Between 2005 and 2012, 239 patients withNSCLCwere enrolled in a prospective gamma knife data repository. Molecular pathology regarding EGF receptor (EGFR), ALK, and KRAS mutation status was available for 81 patients. Local and distant brain control was determined for 79 patients with 469 brainmetastases.ModifiedCox proportional hazardsmodels were established to evaluate local control for treated lesions after serial gamma knife treatments. Results: In total, 11%of patients developed in-field recurrence. No patients withmetastases from tumors with EGFR mutations (0/164 lesions) or EML4-ALK translocations (0/61 lesions) recurred in-field. In contrast, 19% of patients without these mutations and 18% of patients with KRAS mutations recurred infield (10/139 and 3/105 lesions, respectively). Rates of distant brain recurrence did not significantly differ across tumor genotypes. The predicted median in-field local control was significantly longer for EGFRmutant and ALK-translocated tumors compared with other patients with NSCLC (P < 0.001), whereas distant brain recurrence time was equivalent (P 1⁄4 0.97). On multivariate analysis, EGFR mutation, ALK translocation, andmetastasis size were independent predictors for superior local control after gamma knife