This multicenter study enrolled 73 patients with locally advanced or metastatic non-small-cell lung cancer (NSCLC). The study design was based on the hypothesis that the non-overlapping toxicities of a 3-drug combination of irinotecan (Camptosar, CPT-11), carboplatin (Paraplatin), and paclitaxel (Taxol) would allow them to be dosed at recommended or standard doses, respectively. Of the 32 patients in phase I portion of the trial, six received irinotecan 40 mg/m2 plus carboplatin at an area-under-curve (AUC) of 6 and paclitaxel 225 mg/m2 on an every-3-week schedule (regimen A). Due to the unexpectedly severe toxicity associated with this regimen, the carboplatin and paclitaxel doses were reduced, while the irinotecan doses were escalated from 40 to 80 to 125 mg and reduced to 100 mg/m2 (regimen B). In the phase II portion of the trial, irinotecan was dosed at 100 mg/m2 because 3 of 8 patients who received the 125 mg/m2 dose in phase I experienced first-course grade 3/4 toxicities. However, the toxicity of regimen B proved unacceptably severe in phase II; 13 of the 40 patients (35%) experienced grade 3/4 neutropenia and 30% suffered febrile or septic neutropenia. The unconfirmed objective response rate in phase I was 64.5%; median time to progression and median survival were 7.1 months and 11.6 months, respectively, and 1-year survival was 46.9%. The unconfirmed objective response rate in phase II was 60.6%, including one complete response (CR) and 19 partial responses (PRs); one CR and 16 PRs were subsequently confirmed. The efficacy and toxicity of this triple combination strongly suggest significant pharmacokinetic and/or pharmacodynamic interactions, warranting continued clinical investigation.