A randomly sampled population aged from 25 to 59 years and examined in 1972 was monitored for 5 years. During that period, 126 deaths from cardiovascular causes occurred among subjects with no recent history of major cardiovascular diseases. A control matched for age, sex and the conventional cardiovascular risk factors was assigned to each of the deceased subjects from the same population at risk. Rheumatoid factors or antinuclear antibodies as determined from the baseline specimens were found among 21 of the deceased (20 in the discordant pairs), but only in 7 controls (6 in the discordant pairs). This corresponds to a relative risk of death due to cardiovascular causes of 3.3 (p less than 0.01).