We have reported previously that the anti-emetic efficacy of single agent 5HT3 antagonists is not maintained when analysed with the measurement of cumulative probabilities. Presently, the most effective anti-emetic regimen is a combination of a 5HT3 antagonist plus dexamethasone. We, therefore, assessed the sustainment of efficacy of such a combination in 125 patients, scheduled to receive cisplatin > or = 70 mg m(-2) either alone or in combination with other cytotoxic drugs. Anti-emetic therapy was initiated with 10 mg of dexamethasone and 3 mg of granisetron intravenously, before cisplatin. On days 1-6, patients received 8 mg of dexamethasone and 1 mg of granisetron twice daily by oral administration. Protection was assessed during all cycles and calculated based on cumulative probability analyses using the method of Kaplan-Meier and a model for transitional probabilities. Irrespective of the type of analysis used, the anti-emetic efficacy of granisetron/dexamethasone decreased over cycles. The initial complete acute emesis protection rate of 66% decreased to 30% according to the method of Kaplan-Meier and to 39% using the model for transitional probabilities. For delayed emesis, the initial complete protection rate of 52% decreased to 21% (Kaplan-Meier) and to 43% (transitional probabilities). In addition, we observed that protection failure in the delayed emesis period adversely influenced the acute emesis protection in the next cycle. We conclude that the anti-emetic efficacy of a 5HT3 antagonist plus dexamethasone is not maintained over multiple cycles of highly emetogenic chemotherapy, and that the acute emesis protection is adversely influenced by protection failure in the delayed emesis phase.