This phase II study evaluated efficacy of first-line erlotinib therapy for chemo-naïve patients with non-small cell lung cancer (NSCLC) by their clinicopathological and/or molecular characteristics. Eligible patients received erlotinib 150 mg daily until disease progression, followed by a gemcitabine/carboplatin doublet. By clinicopathological characteristics, the patients were categorized as squamous cell carcinoma (SQCC group), ever-smoker with adenocarcinoma (ever-smoking ADCC group), or never-smoker with adenocarcinoma (never-smoking ADCC group). Epidermal Growth Factor Receptor (EGFR) mutations were prospectively assessed by a direct sequencing method and confirmed retrospectively by the Scorpion amplified refractory mutation system (ARMS). Seventy-five patients participated in this study. The direct sequencing method detected 18 EGFR mutations while ARMS detected an additional 3 EGFR mutations and 1 second EGFR T790M mutation. The objective response rates (ORR) were 71.7% in never-smoking ADCC, 25.0% in ever-smoking ADCC, but no response in SQCC, while those of the patients with EGFR mutant and wild-type ere 85.7% and 10.0%, respectively. Even in never-smoking ADCC, the EGFR mutants responded better, with ORR of 88.9% and survived longer, with median survival time of 25.4 months, than those with wild-type EGFR with ORR of 25.0% and median survival time of 16.6 months (p<0.05). ORR for gemcitabine and carboplatin was 16.1%. The decision to administer first-line erlotinib should be decided by molecular characteristics, if known, but can be made by clinico-pathological characteristics as second best policy.