BACKGROUND: The combination of paclitaxel with cisplatin or carboplatin has significant activity in non-small-cell lung cancer (NSCLC). This phase III study of chemotherapy-naïve advanced NSCLC patients was designed to assess whether response rate in patients receiving a paclitaxel/carboplatin combination was similar to that in patients receiving a paclitaxel/cisplatin combination. Paclitaxel was given at a dose of 200 mg/m(2) (3-h intravenous infusion) followed by either carboplatin at an AUC of 6 or cisplatin at a dose of 80 mg/m(2), all repeated every 3 weeks. Survival, toxicity and quality of life were also compared. PATIENTS AND METHODS: Patients were randomised to receive one of the two combinations, stratified according to centre, performance status, disease stage and histology. The primary analyses of response rate and survival were carried out on response-evaluable patients. Survival was also analysed for all randomised patients. Toxicity analyses were carried out on all treated patients. RESULTS: A total of 618 patients were randomised. The two treatment arms were well balanced with regard to gender (83% male), age (median 58 years), performance status (83% ECOG 0-1), stage (68% IV, 32% IIIB) and histology (38% squamous cell carcinoma). In the paclitaxel/carboplatin arm, 306 patients received a total of 1311 courses (median four courses, range 1-10 courses) while in the paclitaxel/cisplatin arm, 302 patients received a total of 1321 courses (median four courses, range 1-10 courses). In only 76% of courses, carboplatin was administered as planned at an AUC of 6, while in 96% of courses, cisplatin was given at the planned dose of 80 mg/m(2). The response rate was 25% (70 of 279) in the paclitaxel/carboplatin arm and 28% (80 of 284) in the paclitaxel/cisplatin arm (P = 0.45). Responses were reviewed by an independent radiological committee. For all randomised patients, median survival was 8.5 months in the paclitaxel/carboplatin arm and 9.8 months in the paclitaxel/cisplatin arm [hazard ratio 1.20, 90% confidence interval (CI) 1.03-1.40]; the 1-year survival rates were 33% and 38%, respectively. On the same dataset, a survival update after 22 months of additional follow-up yielded a median survival of 8.2 months in the paclitaxel/carboplatin arm and 9.8 months in the paclitaxel/cisplatin arm (hazard ratio 1.22, 90% CI 1.06-1.40; P = 0.019); the 2-year survival rates were 9% and 15%, respectively. Excluding neutropenia and thrombocytopenia, which were more frequent in the paclitaxel/carboplatin arm, and nausea/vomiting and nephrotoxicity, which were more frequent in the paclitaxel/cisplatin arm, the rate of severe toxicities was generally low and comparable between the two arms. Overall quality of life (EORTC QLQ-C30 and LC-13) was also similar between the two arms. CONCLUSIONS: This is the first trial comparing carboplatin and cisplatin in the treatment of advanced NSCLC. Although paclitaxel/carboplatin yielded a similar response rate, the significantly longer median survival obtained with paclitaxel/cisplatin indicates that cisplatin-based chemotherapy should be the first treatment option.