Several clinical and biological parameters are known to influence the efficacy of second-line erlotinib therapy for nonsmall cell lung cancer (NSCLC), but their medico-economic impact has not been evaluated. The objective of this study was to compare the incremental cost-effectiveness ratios of strategies for second-line erlotinib initiation in NSCLC: clinically guided initiation (nonsmoking females with adenocarcinoma received erlotinib; all other patients received docetaxel) and biologically guided selection (patients with epidermal growth factor receptor (EGFR) mutation received erlotinib; patients with wild-type EGFR or unknown status received docetaxel), compared with initiation with no patient selection (strategy reference). A Markov model was constructed. Outcomes (overall and progression-free survival), transition probabilities and direct medical costs (from the French third-party payer's perspective) were prospectively collected for individual patients treated with either erlotinib or docetaxel, from treatment initiation to disease progression. Published data were used to estimate utilities and post-progression costs. Sensitivity analyses were performed. The biologically and clinically guided strategies were both more efficient (incremental quality-adjusted life-yrs equal to 0.080 and 0.081, respectively) and less expensive (cost decrease equal to €5,020 and €5,815, respectively) than the no-selection strategy, and the biologically guided strategy was slightly less expensive than the clinically guided strategy. Sensitivity analyses confirmed the robustness of the results. The cost-effectiveness of second-line NSCLC treatment is improved when patients are selected on either clinical or biological grounds.