Anticoagulants and myocardial infarction. The problems of pooling, drowning, and floating.

The time and expense required for adequate-sized, randomized control trials may often tempt investigators to try to answer long-standing clinical questions by retrospective analysis of data combined from old studies. In a recent example of such "pooling," the authors concluded not only that anticoagulation is beneficial in acute myocardial infarction but also that further experimental trials would be unethical. This conclusion can be disputed because of scientific, statistical, and clinical standards that were not fulfilled in the "pooled" analysis. Scientifically, the pooling of data from different sources can be valid only if the component studies contain patients who are sufficiently similar in diagnosis, clinical severity, principal treatment, and outcome events. Mathematically, exact numbers rather than percentages should be pooled. Clinically, even when the pooling process is appropriate, recent changes in ancillary therapy may negate the applicability of the old data to current medical practice.