In the context of a phase III trial comparing in advanced non-small cell lung cancer (NSCLC) sequential to conventional administration of cisplatin-based chemotherapy and paclitaxel, we evaluated the activity of paclitaxel as second-line chemotherapy and investigated any relation of its efficacy with the type of failure after cisplatin. Patients received three courses of induction GIP (gemcitabine, ifosfamide, cisplatin). Non-progressing patients were randomised between three further courses of GIP or three courses of paclitaxel. Second-line paclitaxel was given to patients with primary failure (PF) to GIP and to those progressing after randomisation to further GIP (secondary failure or SF). One hundred sixty patients received second-line paclitaxel. Response rates were 7.7% for PF and 11.6% for SF (P=0.42). Median survival times (calculated from paclitaxel start) were 4.1 and 7.1 months for PF and SF (P=0.002). In multivariate analysis, three variables were independently associated with better survival: SF (hazard ratio (HR)=1.55, 95% confidence interval (CI) 1.08-2.22; P=0.02), normal haemoglobin level (HR=1.56, 95% CI 1.08-2.26; P=0.02) and minimal weight loss (HR=1.79, 95% CI 1.26-2.55; P=0.001). Paclitaxel in NSCLC patients, whether given for primary or for SF after cisplatin-based chemotherapy, demonstrates activity similar to other drugs considered active as second-line therapy.