Resuscitation for cardiac arrest was monitored over 4 years to examine the effect on survival of a change in the ventricular fibrillation (VF) protocol to include the routine early use of "high-dose" intravenous or transbronchial adrenaline. A significant reduction in the immediate survival of patients with VF was seen when the protocol was changed (22% after the change, 43% before). Prior predictors of poor response were similar in each group, except for the number of witnessed arrests, delay until cardiopulmonary resuscitation, and occurrence of endotracheal intubation, but multiple logistic regression showed the use of adrenaline to be an independent predictor of outcome. Early high-dose adrenaline was associated with a reduction in immediate survival in patients with persistent VF.