Pooling of Data from Independent Studies

Abstract

The large-scale, randomized clinical trial is generally regarded as a standard against which the credibility of other clinical investigations is evaluated. This is especially true if the effects to be studied are small or occur infrequently, and the trial has been conducted with a sample size sufficient to determine unequivocally the presence or absence of an effect. Increased sample size, or power, is therefore one immediately apparent reason for pooling data from several independent studies. This was the case in a study of several randomized trials comparing streptokinase therapy and heparin anticoagulation for the treatment of deep venous thrombosis (Goldhaber, Buring, Lipnick, & Hennikens, 1984). Pooling of results from several studies indicated a significant advantage of streptokinase treatment over heparin, but also revealed a significantly higher incidence of major bleeding complications for that treatment. Although the results of pooling could not resolve these questions of efficacy and safety, they did succeed in demonstrating a likelihood that a large-scale clinical trial would be able to do so. Pooling has also been used to extend the interpretation of data beyond the original aims of the individual studies, such as in explicating the relationship of risk factors to the incidence of coronary heart disease (Pooling Project Research Group, 1978), or to resolve ambiguous outcomes, as in evaluating the prophylactic efficacy of β-blockers after myocardial infarction (Bassan, Shalev, & Eliakim, 1984). However, because of the statistical nature of the benefits of pooling, it is not possible to use this approach to resolve truly conflicting results. That is, when several studies show statistically opposite outcomes, it is unlikely that these were caused by chance variations in the data, and the source of disagreement cannot be resolved by statistical techniques. These studies must be examined in detail to identify possible differences in experimental design, subject population, or other factors which may have affected the dynamics of the system under study. Finally, pooling can serve as an excellent vehicle for the detailed review of existing empirical or clinical procedures. The process of pooling data from independent studies can be a worthwhile and scholarly effort, likely to lead to new insights as well as the formulation of new hypotheses, provided appropriate cautions and limitations are kept in mind.