Simon R
Stat Med. 1987 Apr-May;6(3):389-96. doi: 10.1002/sim.4780060335.
The implications of pooling based overviews for the design and conduct of cancer clinical trials are discussed. Because the results of overviews may be highly publicized and disruptive to ongoing trials, the timing of overviews should be carefully considered. Pooling based overviews can be valuable supplements to major clinical trials if the trials are themselves quite similar in design and quality. Such overviews may be essential for addressing subset questions based on results from such trials. Pooling based overviews are not, however, satisfactory alternatives to major clinical trials. Pooled results of many small dissimilar studies may be misleading with regard to the efficacy of treatments contributing to the average. For toxic cancer treatments, the assumption that unexpected qualitative interactions are unlikely does not provide a satisfactory justification for pooling results of trials that are dissimilar in design or quality of conduct.