8006 Background: E and D are standard cares for previously treated patients with advanced NSCLC. Although E shows significant clinical benefits over best supportive care in the EGFR wild type tumors, it remains unknown whether E or D is more active against the disease. Methods: This is an open-label, multi-center phase III study, sponsored by the Japanese National Hospital Organization. Patients were randomized to E (150 mg, daily), or D (60 mg/m2, q3w) by the minimization method according to gender, performance status, histology, and institution. The primary endpoint was progression free survival (PFS), and secondary endpoints included overall survival (OS), response rate, safety, and analyses on EGFR wild type tumors. Eligible patients were those with pathologically proven NSCLC with stage IIIB or IV (AJCC version 6) previously treated with one or two chemotherapy regimens including at least one platinum agent, evaluable or measurable disease, and ECOG PS 0-2. Target sample size was calculated to be 280 based on the assumption that E was superior to D in PFS (3.5 months [m] v 2.5 m, α =0.05 [two sided], β =0.80). Results: From August 2009 to July 2012, 301 patients were accrued from 41 institutions. In the ITT population, 150 and 151 patients were randomly assigned to E and D, respectively, including respective 109 (73%, E) and 89 (59%, D) with EGFR wild type tumors. Median PFS and OS for E v D were 2.0 m (95% CI: 1.3-2.8) v 3.2 m (2.8-4.0, log-rank p=0.092; HR=1.22, 95% CI: 0.97-1.55), and 14.8 m (9.0-19.4) v 12.2 m (9.0-15.5, p=0.527; HR=0.91, 95% CI: 0.68-1.22), respectively. In the EGFR wild type tumors, PFS and OS for E v D were 1.3 m v 2.9 m (p=0.013; HR =1.44, 95% CI: 1.08-1.92), and 9.0 m v 9.2 m (p=0.914; HR =0.98, 95% CI: 0.69-1.39), respectively. Main grade 3/4 toxicities were rash (13.3% [E] v 0.7% [D]) and leukopenia (5.4% [E] v 62.9% [D]). Conclusions: E failed to show better PFS over D. While PFS was significantly longer in D than E in EGFR wild type tumors, the difference did not translate into OS in this pragmatic trial. Clinical trial information: 000002314.