Follow-up of 270 subjects who declined to participate in a trial of oral anticoagulant therapy after acute myocardial infarction revealed a higher mortality (25.6%) than that for participants in the placebo group (20.2%). The excess mortality in terms of odds ratios was 1.35 (90% confidence interval 1.02-1.79). Age was the only predictor of death. The event rate from ischaemic cardiovascular disease was lower among non-consenters than among participants (49.3% vs. 74.8%), as was the rate of death during the first year of follow-up. In conclusion, the differing mortality and dissimilar patterns of specific causes of death in non-consenters and placebo-treated participants emphasize the need for caution in extrapolation of treatment effects to the non-consenting group. Thus the size of the non-consenting group has implications for generalization of the overall results.