A survey of serum testosterone levels and a range of medical and behavioral factors was conducted on 71 males aged 46 to 89 years in an extended care medical facility. Histories of myocardial infarction or heavy drinking were separately associated with diminished testosterone levels (P<.002 and P<.006, respectively). No other factor was associated with testosterone level, including age and mobility. A testosterone level of no more than 438 ng/dl best discriminated patients with myocardial infarction (P<.005, sensitivity 86.2%, and selectivity 50%). Increased incidence of myocardial infarction appeared to be influenced by testosterone in a threshold manner, since the incidence did not rise further with lesser testosterone levels once the level was below 438 ngldl. The incidence of myocardial infarction among patients with testosterone levels below this threshold was unaffected by their exposure to alcohol. Among the 21 formerly heavy drinkers, many (62%) showed testosterone levels ≤300 ngldl. Formerly heavy drinkers should be routinely considered for serum testosterone determinatio