Evaluating effects of treatment in subgroups of patients within a clinical trial: The case of non-Q-wave myocardial infarction and beta blockers