Randomized controlled studies were reviewed to assess the effectiveness and safety of antiemetics used for prophylaxis in paediatric strabismus surgery. Early and late vomiting (6 and 48 h after operation, respectively), and adverse effects were evaluated using the numbers-needed-to-treat method. In 27 reports with information on 2033 children, the mean incidence of early vomiting was 54% and of late vomiting 59%, without prophylaxis. Only three drugs were studied sufficiently for firm conclusions to be drawn. In the best documented regimen (droperidol 75 micrograms kg-1), four children have to be given the drug to prevent one vomiting; of the three others, one may vomit and two would not have vomited anyway; fewer than one child in 100 may have an extrapyramidal reaction and 16 may have minor adverse effects. Metoclopramide 0.15 and 0.25 mg kg-1 was significantly better than control only for early vomiting. Propofol had a high incidence of oculocardiac reflex without conferring any significant antiemetic effect: it should not be used. The benefits of prophylactic antiemetic therapy are not proven.