Therapeutic intervention trials using randomization of cases, considered the most reliable available means of evaluating effectiveness of therapy, have serious limitations. They are subject to errors in both directions: suggesting a relation when none exist, or being unable to demonstrate existing relations. When death is used as an outcome of the study, errors are most likely to occur when populations under study contain patients with widely varying prognoses, as is the case in ischemic heart disease. Furthermore, some trials pay insufficient attention to clinical facts and clinical relevance, both in designing and interpreting results of the study. The results of the Anturane reinfarction trial may have been misinterpreted on that basis. Trials dealing with comparison of medical and surgical treatment of coronary artery disease have been disappointing in their negative or inconclusive results. Another approach is needed to reinvestigate this problem, one randomizing patients with high risk subsets of coronary disease.