About 2000 patients have been randomised in ten trials of intravenous nitroglycerin or nitroprusside in acute myocardial infarction. Taken separately, the individual trials have all been too small to provide a reliable estimate of the effects of treatment on mortality, but collectively they provide strong evidence of benefit. In total there have been 136 nitrate and 193 control deaths, and an appropriate overview of the separate trial results indicated a "typical" reduction of 35% (SD 10) in the odds of death (2p less than 0.001, with 95% confidence limits of about one-sixth to one-half). Both nitroglycerin and nitroprusside reduced mortality, the reduction being non-significantly greater with nitroglycerin than with nitroprusside. The greatest reduction in mortality occurred during the first week or so of follow-up, with a non-significant further reduction after this early period. This suggests that the early benefit is not rapidly lost.