INTRODUCTION: Xenograft models of epithelial malignancies potentially have greater correlation with clinical end points. We implanted 153 primary non-small cell lung carcinomas into non-obese diabetic-severe combined immunodeficient mice to develop primary lung cancer xenografts. Sixty-three xenografts formed. However, in 19 implantations, tumors consisted of a lymphocyte proliferation without a carcinoma component. We further characterized these lymphomas to determine clinicopathological features associated with their formation. METHODS: Lymphomas were investigated morphologically and by silver in situ hybridization to determine their species of origin. Characterization both of the xenograft lymphomas and the primary NSCLCs from which they were derived included immunohistochemistry for lymphoma markers and Epstein Barr virus Early RNA (EBER) by in situ hybridization. DNA was profiled using the MassARRAY platform; EML4-ALK translocations and lymphocyte infiltration were assessed in the primary tumor. Lymphoma formation was correlated with patient and primary tumor characteristics and survival. RESULTS: The lymphocytic tumors were EBER positive, human diffuse large B-cell lymphomas (DLBCLs). Significantly more DLBCLs that formed in mice arose in primary lung adenocarcinomas and in epithelial growth factor receptor mutant never smokers. DLBCL formation was not associated with the degree of tumor-infiltrating lymphocytes or EBER-positive lymphocytes in the primary NSCLCs. Patients whose tumors developed DLBCL had longer disease-free survival compared with patients whose tumors formed epithelial xenografts (hazard ratio: 0.44; 95% confidence interval: 0.18 -1.06, Wald p = 0.07), regardless of genotype. CONCLUSION: We hypothesize that mechanisms involved in the active suppression of viral antigens may also be involved in the suppression of tumor antigens, and may have resulted in the observed favorable clinical outcome.