The combination of carboplatin and paclitaxel is an active regimen in non-small cell lung cancer (NSCLC). Historically, patients with stage III disease have manifested higher response rates than patients with metastatic disease, and patients achieving a pathologic complete response to induction chemoradiation therapy prior to surgery have shown better long-term outcome. Based upon our pilot data using high-dose carboplatin and paclitaxel, we designed a phase II trial in patients with marginally resectable stage IIIA NSCLC. Ten patients, with bulky nodal stage IIIA disease, initially received etoposide (2 g/m2) and granulocyte colony-stimulating factor (G-CSF) to mobilize peripheral blood stem cells (PBSC). Two cycles, 28 days apart, of carboplatin (AUC 12 in seven patients; AUC 16 in three patients) and paclitaxel (250 mg/m2) were administered with filgrastim (5 microg/kg) and PBSC support. After re-evaluation, patients underwent a thoracotomy followed by radiotherapy (44-60 Gy) if deemed resectable, or radiotherapy alone (60 Gy) if not resectable. The median age was 58.5 years (48-66) with a median ECOG performance status of 0 (0-1). Histology was adenocarcinoma in seven patients; the remainder had either squamous cell, large cell or bronchoalveolar carcinoma. Based on CT radiography, the overall response rate was 40%. Eight of ten patients underwent resection with four right pneumonectomies, three right upper lobectomies and one wedge resection of the right upper lobe. Six patients had a complete resection. Of eight patients resected, four were downstaged by induction therapy, three remained unchanged and one was found to have more extensive disease. The remaining two patients developed metastatic disease while receiving chemotherapy. The median dose of postoperative radiotherapy was 54 Gy (35-66 Gy). Actual median follow-up for all patients was 89 weeks (25 to 136+). The actuarial median overall survival was 124 weeks (25 to 136+) and time to progression was 57 weeks (17 to 136+). The median dose of carboplatin delivered expressed as mg/m2 was 779 (615-1540). Neutropenic fever occurred in two patients during the initial mobilization cycle only. The median number of units of RBC and/or platelets transfused was 0 (0-2 and 0-6, respectively). There were no significant non-hematologic toxicities. High-dose induction chemotherapy with stem cell rescue is feasible and safe with an acceptable response rate. Thoracotomy, including pneumonectomy and postoperative radiotherapy, were well tolerated by patients after undergoing high-dose induction chemotherapy with no apparent increase in peri-operative morbidity. The pathologic complete response rate was low--one out of ten patients. These results indicate that dose escalation of induction chemotherapy does not improve response rates even in this highly selected patient population. Accordingly, the complexity and potential toxicity of high-dose chemotherapy, as delivered in this trial as neoadjuvant treatment of non-small cell lung cancer, is not warranted.