Data are n (%). The safety analysis included all patients who were randomly allocated to treatment groups and received at least one dose of study drug. Adverse events were prospectively defined and the Medical Dictionary for Regulatory Activities (MedDRA) version 13.0² was used for reporting adverse events. A list of events was prespecified in the trial protocol.

*One patient in erlotinib group had hepatotoxicity, two patients in the chemotherapy group had cerebrovascular accidents (one after related grade 5 infection).

Data are n (%). Adverse events were assessed according to the National Cancer Institute Common Terminology Criteria version 3.0²