OBJECTIVES: To compare the efficacy of available treatments in patients with previously treated advanced/metastatic non-small cell lung cancer (NSCLC) METHODS: Clinical trials of ceritinib, crizotinib, pemetrexed, docetaxel, erlotinib, and best supportive care (BSC) in previously treated, crizotinib-naïve NSCLC were identified in a systematic literature review (up to March 2014). Since only single-arm trials were available for ceritinib, a matching-adjusted indirect comparison between ceritinib trials and similarly-designed crizotinib trials was used to link ceritinib into the evidence network. A Bayesian-based network meta-analysis (NMA) was then conducted to compare overall survival (OS), progression-free survival (PFS) and overall response rate (ORR) among the studied treatments. Relative treatment effects and corresponding 95% credible intervals (CrIs) were estimated as well as the posterior efficacy rankings. RESULTS: Fourteen clinical trials, including two ceritinib single-arm trials (ASCEND-1 [NCT01283516], ASCEND-3 [NCT01685138]), were included. Sample sizes ranged from 49 to 488 patients per arm. Ceritinib was associated with significantly prolonged OS compared to crizotinib (hazard ratio [HR]: 0.58; 95%CrI: 0.41-0.83), pemetrexed (0.44; 0.26-0.76), docetaxel (0.47; 0.28-0.82), erlotinib (0.44; 0.26-0.80), and BSC (0.29; 0.16-0.55). Ceritinib also showed improved PFS compared to crizotinib (0.45; 0.18-1.56), pemetrexed (0.18; 0.06-0.83), docetaxel (0.17; 0.06-0.84), erlotinib (0.15; 0.05-0.90), and BSC (0.08; 0.02-0.69). Ceritinib increased the odds of achieving ORR by 38% compared to crizotinib, and was associated with ≥3 times higher ORR compared to other treatments. Ceritinib was the most efficacious treatment in terms of OS (posterior probability 99%), PFS (92%) and ORR (63%) among all treatments, followed by crizotinib. CONCLUSIONS: In this NMA of treatments for patients with previously treated advanced/metastatic NSCLC, ceritinib was associated with significantly reduced risks of death or progression, by 40-85%, and with higher ORR, compared with other active treatments. Comparisons of non-randomized treatment groups are limited by the potential for confounding due to unadjusted cross-study differences.