A prospective survey was performed in 17 nuclear medicine departments during 1996 in an attempt to provide reliable data on the prevalence of adverse reactions to radiopharmaceuticals. All adverse events following radiopharmaceutical administration were recorded, irrespective of the severity or likelihood of causality, and subsequently analysed using an algorithm developed by Silberstein et al., designed to establish a cause-effect relationship. A prevalence of 11 events per 10(5) administrations was obtained (95% confidence limits 3.3-19.2). No serious of life-threatening events were reported. This rate is slightly higher than that obtained in a larger scale study in the United States (2.3 events per 10(5) administrations, 95% confidence limits 1.2-3.4). The difference may be due to the decision to include or exclude vasovagal events from the analysis, the way in which the algorithm was used and the comparative size and time scale of the two studies. The prevalence of adverse reactions is approximately 1000-fold than less that occurring with iodinated contrast media and drugs.