The goals of this work were to compare the relative efficacy of ondansetron, granisetron and tropisetron in a randomised double blind crossover trial, evaluating objective, subjective and pharmacoeconomic parameters. To this end, 136 patients were enrolled, 120 of whom were eligible and evaluable. Each patient received three identical chemotherapy cycles with an antiemetic protocol which consisted in dexamethasone 20 mg i.v. and a tapering dose schedule for 4 days, and a single i.v. dose of an antiserotoninergic drug in each cycle. Arm A patients received tropisetron 5 mg; arm B patients, granisetron 3 mg; and arm C patients, ondansetron 24 mg. Numbers of patients and days with emetic episodes, grade of nausea, patient preference, headaches, need for metoclopramide, nursing or medical consultation, or admission to emergency room or ward were evaluated. There was no difference in the percentage incidence of acute or delayed nausea and vomiting. Twenty-five per cent of patients preferred tropisetron, 30% preferred granisetron, and 45% preferred ondansetron (P<0.01). Toxicity was mild in less than 10% of patients. Direct and indirect costs of treatment varied from 19.74 to 28.53 euros for tropisetron, 31.07-46.51 euros for granisetron and 22.76-62.61 euros for ondansetron. There was no difference in objective activity. In the schedules studied, patients preferred ondansetron. Indirect costs amount to less than 10% of the total antiemetic cost. Direct costs varied widely and should be considered whenever an antiemetic drug is selected.