Ninety-six patients with operable epidermoid cancer of the esophagus were entered into a phase III, random assignment study designed to compare the efficacy of two preoperative approaches (chemotherapy [CT] or radiation therapy [RT]). Major study end points were objective response rates, surgical outcome, and recurrence pattern. Patients were randomly assigned to receive either two cycles of cisplatin, vindesine, and bleomycin or 55 Gy of radiation before a planned surgical procedure. Postoperative crossover therapy (radiation to those receiving preoperative CT and vice versa) was given to patients with T3Nany or unresectable tumors. Objective response rates of the primary tumor to preoperative therapy were similar (RT 64%, CT 55%), as were operability rates (RT 77%, CT 75%), resection rates (RT 65%, CT 58%), and operative mortality (RT 13.5%, CT 11.1%). Significantly higher doses of CT could be administered when CT was given as initial therapy, rather than after RT/surgery. Local failure or persistence occurred in 33% of operable patients. The median survival for all patients was 11 months; 20% remain alive without disease (median follow-up, 34 months). Because of the crossover design, it was not possible to analyze survival according to the preoperative therapy arm alone. This study suggests that since CT is as effective in treating local tumor as RT, but can also potentially treat systemic disease, investigational programs using CT before surgery as part of initial treatment for localized esophageal cancer should continue. However, if a significant impact on overall survival is to be achieved, more effective chemotherapy regimens or schedules need to be identified. Outside of carefully designed clinical trials, surgery alone or radiation alone remain standard therapy.