Five hundred outpatients undergoing endoscopy were admitted into a controlled trial comparing the incidence of thrombophlebitis following intravenous diazepam administered in the way, with the effects of either a saline flush following the diazepam or diluting the drug with the patient's own blood before injection ('barbotage'). The results were assessed using a questionnaire completed by patients two weeks after endoscopy; 80% replied. A saline flush reduced the incidence of side effects, particularly pain (P less than 0.05). 'Barbotage' gave the highest incidence of side effects. Saline flush is therefore recommended as a means of reducing the thrombophlebitis which may follow intravenous diazepam.