Meta analysis plays a key role in setting policy and in planning new research. In gerontology, meta analyses have led to treatment recommendations or new insights in such areas as memory loss with normal aging; identifying, predicting, and treating Alzheimer's disease and dementia; aspirin and the secondary prevention of stroke; drug therapies for osteoporosis; and Parkinson's Disease, among many others. While every Meta analysis is concerned both with the average treatment effect and with the variation in treatment effects from one study to the next, researchers tend to focus primarily on the former. In fact, though, the variation in the treatment effect from one study to the next may be as important, and sometimes more important, than the average treatment effect. If a treatment reduces the risk of death for younger patients while increasing the risk of death for older patients, then the average effect is, at best, misleading. If we need to choose between two potentially helpful interventions for patients, the question we need to ask is not whether or not each one works at all, but whether it works better than the other. The goal of this project is to develop software for Meta regression, the technique developed to study the variation in treatment effects and address these questions. This software will be integrated into a commercially successful computer program for Meta analysis developed under an earlier SBIR project. ? ?
| Vevea, Jack L; Woods, Carol M (2005) Publication bias in research synthesis: sensitivity analysis using a priori weight functions. Psychol Methods 10:428-43 |
