Abstract
Meta analysis techniques, if applied appropriately, can provide a summary of the totality of evidence regarding an overall difference between a new treatment and a control group using data from multiple comparative clinical studies. The standard meta analysis procedures, however, may not give a meaningful between-group difference summary measure or identify a meaningful patient population of interest, especially when the fixed effect model assumption is not met. Moreover, a single between-group comparison measure without a reference value obtained from patients in the control arm would likely not be informative enough for clinical decision making. In this paper, we propose a simple, robust procedure based on a mixture population concept and provide a clinically meaningful group contrast summary for a well-defined target population. We use the data from a recent meta analysis for evaluating statin therapies with respect to the incidence of fatal stroke events to illustrate the issues associated with the standard meta analysis procedures as well as the advantages of our simple proposal.
Disciplines
Biostatistics | Clinical Trials | Statistical Methodology
Suggested Citation
Hasegawa, Takahiro; Claggett, Brian; Tian, Lu; Solomon, Scott D.; Pfeffer, Marc A.; and Wei, Lee-Jen, "The Myth Of Making Inferences For An Overall Treatment Efficacy With Data From Multiple Comparative Studies Via Meta-analysis" (January 2016). Harvard University Biostatistics Working Paper Series. Working Paper 207.
https://biostats.bepress.com/harvardbiostat/paper207