Between October 1981 and June 1983, the Eastern Cooperative Oncology Group (ECOG) conducted a prospectively randomized trial (EST 1581) of the four most active chemotherapy regimens for metastatic non-small-cell lung cancer (NSCLC). Four hundred eighty-six good performance status patients (PS 0 or 1; 81%) were randomized to receive cyclophosphamide, doxorubicin, methotrexate, and procarbazine (CAMP); mitomycin, vinblastine, and cisplatin (MVP); etoposide and cisplatin (VP-P); or vindesine and cisplatin (VDA-P). All regimens were administered in the doses and schedules originally reported. Complete response (CR) plus partial response (PR) rates for the four regimens were CAMP, 17%; MVP, 31%; VP-P, 20%; and VDA-P, 25%. The response rate for MVP was significantly higher in patients with squamous and adenocarcinoma histologies, but there was no impact on median survival (overall, 24.5 weeks). The duration of response did not differ by treatment as previously suggested for VDA-P. There were 15 CRs (CAMP, one; MVP, six; VP-P, two; VDA-P, six), and 12 patients have survived more than 2 years. Toxicity was significant with 20 treatment-related deaths. CAMP was significantly less toxic than the other regimens (P less than .001). VDA-P demonstrated significantly more life-threatening (seven) and lethal (three) episodes of nephrotoxicity (P less than .001) despite an aggressive hydration program that in itself caused significant morbidity. Analysis of the toxicity data showed, however, that most of the severe toxicity occurred in the 19% of patients who were initially PS 2, suggesting that they are not appropriate candidates for trials of new agents or combinations. None of these regimens can be recommended as a standard therapy for metastatic NSCLC.