Arriving at a decision for early termination of a treatment group or of an entire clinical trial, due to either beneficial or adverse results, is a complex process. It may involve, among other things, the necessity to (1) determine whether the observed treatment differences are likely to represent real effects and are not due to chance; (2) weigh the importance of many different response variables, some possible trending in favor of the treatment and some against it; (3) adjust for differences in distributions of baseline characteristics among the treatment groups; (4) discern possible biases (due to the study not being double-blind) in the medical management of patients or in the diagnosis of events; and (5) evaluate treatment effects in subgroups of the study participants. Experiences in the Coronary Drug Project are described in making decisions for early termination of treatment groups. Special methods applied in conjunction with reaching these decisions included: (1) analyses designed to detect possible biases in diagnosing a major endpoint--nonfatal myocardial infarction, e.g.--biases that may have arisen because of unblinding by side effects of the estrogen treatment group; (2) a simulation study to assess the statistical significance of mortality differences between the dextrothyroxine and placebo participants in certain subgroups identified a posteriori; and (3) the development, on the basis of the accumulated data, or new hypotheses relating to treatment differences in specific subgroups and the testing of these hypotheses a few months later on the basis of new events occurring during that interval.