Postoperative vomiting is a common and unpleasant complication. Currently, however, mathematical models, such as number necessary to treat (NNT) and relative risk reduction (RRR), have been useful in the decision of which medication to use for prophylaxis. This study aimed at verifying whether dexamethasone, as compared to metoclopramide, decreases the incidence of vomiting when intravenously administered to children anesthetized with sevoflurane for ambulatory pediatric surgeries.Two hundred and thirty seven male children, aged 11 months to 12 years, physical status ASA I and II, undergoing hernia repair were included in this study. They were premedicated with oral midazolam. Anesthesia was induced and maintained with sevoflurane, nitrous oxide, and 1 microg kg(-1) fentanyl. Patients were divided in two groups: group D patients (n = 118) were given 150 microg kg(-1) dexamethasone at induction while group M (n = 119) received 150 microg kg(-1) metoclopramide at induction. The following parameters were evaluated: incidence of vomiting in the first 4 postoperative hours (PO), incidence of vomiting between 4 and 24 PO hours, NNT of both medications and RRR of dexamethasone as compared to metoclopramide.The incidence of vomiting was 9.32% for group D and 33.61% for group M during the first 4 PO hours, and 1.69% with dexamethasone and 3.36% with metoclopramide between 4 and 24 PO hours. RRR of dexamethasone related to metoclopramide in the first 4 hours was 72%. The number necessary to treat (NNT) for dexamethasone was 3.25 and for metoclopramide it was 15.66.Dexamethasone is more effective than metoclopramide in decreasing the incidence of vomiting when used during anesthetic induction with sevoflurane associated to nitrous oxide and fentanyl.