After 15 days there were 142 deaths in the placebo group (4.9%) and 123 deaths in the metoprolol group (4.3%), a difference of 13% (p = 0.29). The 95% confidence limits for the relative effect of metoprolol ranged from an 8% excess (-8%) to a 33% reduction (+33%) in mortality. There was generally a lower mortality rate for metoprolol-treated patients in most subgroups and a consistent tendency for a more pronounced difference between the treatment groups in those subgroups with a placebo mortality rate higher than the average for all placebo patients. Most deaths were cardiac and occurred among patients who developed a definite myocardial infarction (97%) and most of these had a Q-wave infarction (83%). Using a simple model, the placebo mortality was found to increase with increasing number of 8 risk predictors defined from prestudy experience, from 0% in patients with no risk predictors to 11.6% in patients with any 5 or more of these risk factors. Similarly, there was an increase in the difference between the treatment groups in favor of metoprolol with increasing number of placebo risk factors. Metoprolol had no apparent effect in a low-mortality risk group (less than or equal to 2 risk factors), but there was a difference in mortality of 29% in favor of metoprolol in a high-risk group (greater than or equal to 3 risk factors) comprising one-third of the trial population.