The paclitaxel (Taxol; Bristol-Myers Squibb Company, Princeton, NJ)/carboplatin combination has demonstrated promising activity in patients with incurable non-small cell lung cancer (NSCLC). Our exploratory study is designed to evaluate the efficacy of this combination as induction therapy in patients with locally advanced NSCLC, to determine the maximally tolerated doses of paclitaxel and carboplatin administered every 3 weeks during radical thoracic radiation after induction treatment, and to determine the efficacy of granulocyte colony-stimulating factor (G-CSF) priming before induction treatment, followed by conventional G-CSF, compared with conventional G-CSF alone. Eligibility stipulated Karnofsky performance status > or =70%, < or =5% weight loss, and stages IIIB or bulky IIIA NSCLC. Induction treatment consisted of two cycles of paclitaxel 175 to 225 mg/m2 infused over 3 hours combined with carboplatin (target area under the concentration-time curve of 7.5) given on days 1 and 22. On days 2 through 15 and 23 through 36, all patients received G-CSF 5 microg/kg; half were randomized to receive priming G-CSF daily for 5 days before day 1 of treatment. On day 43, thoracic radiation (60 Gy in 30 2-Gy fractions daily, 5 days a week for 6 weeks) was initiated. At dose level 1, patients received carboplatin dosed to a target area under the concentration-time curve of 3.75 and paclitaxel 67.5 mg/m2 over 3 hours on days 43 and 64. In the absence of dose-limiting toxicity, phase I escalation in three-patient cohorts proceeded to a maximum carboplatin area under the concentration-time curve of 5.0 and a paclitaxel dose of 175 mg/m2, delivered over 3 hours. To date, 35 patients (83% stage IIIB) have received induction treatment, 29 of whom are evaluable for response. Myelosuppression and neurotoxicity have been mild during induction treatment, prompting a paclitaxel dose increase to 225 mg/m2 on days 1 and 22 after the first seven patients were accrued. The phase III portion of the study evaluating G-CSF priming remains coded. Sixteen patients have received concurrent thoracic radiation and chemotherapy and are evaluable for response and toxicity. In sequential cohorts, the paclitaxel dose on days 43 and 64 has been escalated to 175 mg/m2 with only one episode each of grade 4 granulocytopenia and grade 3 anemia. In the first 13 patients evaluated, the severity of esophagitis corresponded to the length of the esophagus in the radiation treatment field: grade 1 in all six patients with esophageal exposure < or =16 cm and grade > or =2 in six of seven patients with > or =16 cm of the esophagus irradiated. Three episodes of grade > or =2 steroid-responsive pulmonary toxicity have occurred 2 to 6 months after the conclusion of concurrent thoracic radiation and chemotherapy. The major response rate is 38% to induction treatment and 59% to combined-modality treatment. Of the first 21 patients accrued, 62% survived 1 year. Induction paclitaxel/carboplatin therapy is active and well tolerated by patients with locally advanced NSCLC. The maximum tolerated doses of paclitaxel and carboplatin during concurrent thoracic radiation and the role of G-CSF priming are not yet established. Severity of esophagitis corresponds to the extent of esophagus irradiated during concurrent thoracic radiotherapy and chemotherapy.