Patients with recurrent and metastatic squamous cell carcinoma of the head and neck (SCCHN) were stratified by performance status, extent of disease, and prior radiotherapy and subsequently randomized to receive carboplatin (CBDCA; Bristol-Myers, Wallingford, CT) administered intravenously (IV) monthly, initially at doses of 400 mg/m2 in combination with methotrexate (MTX) given IV weekly at doses of 40 mg/m2 or MTX alone at the same dose/schedule. Significant dose-limiting myelosuppression required CBDCA dose reductions to 300 mg/m2 and, subsequently, 200 mg/m2. Nonhematological toxicities were not significant. Our study objective was to determine whether CBDCA plus MTX produce a substantial improvement in response rate over single-agent MTX. A response rate of 50% (complete [CR] plus partial response [PR]) for CBDCA plus MTX compared with 25% for MTX was specified as the difference to be detected. We employed a two-stage study design for randomized trials that allowed for early termination of studies involving relatively ineffective treatment regimens. With this design, the study could be closed after the first stage (20 patients entered onto each treatment arm) if the number of responders to CBDCA plus MTX were not superior to MTX. Five of 20 patients responded to treatment in each arm, and we were able to conclude that the addition of CBDCA to MTX is unlikely to result in a twofold increase in response rate compared with MTX alone in this group of patients. This two-stage design represents a simple and efficient method of testing the relative efficacy of new combinations containing at least one active agent against a suitable control arm in this disease. It addresses scientific and ethical issues of continuing testing with relatively ineffective treatments, and at the same time provides a reliable method for identifying very active regimens likely to represent significant therapeutic advances.