|
Abstract:
|
Combining the results of multiple clinical trials is not a trivial process. Synthesis is the goal, but it is not appropriate to simply combine results as though they were from a single large study. In addition, it is difficult or impossible to interpret traditional confidence intervals or statistical significance levels presented for each study. Therefore, when combining information from multiple clinical trials, one must carefully and explicitly consider trial effect. We demonstrate Bayesian meta-analytic methods as alternatives to the frequentist approach, applying our methods to synthesize results from previous clinical trials of vitamin E. In particular, we develop a Bayesian hierarchal model to combine information from several trials and account for our uncertainty in the dose response relationship of vitamin E by examining several different candidate models for this relationship.
|
