2010 FDA/Industry Statistics Workshop
- 1. Statistical Issues in Meta-Analysis of Drug Eluting Stent Data Hsini (Terry) Liao, PhD Boston Scientific Corporation Email: Terry.Liao@bsci.com The meta-analysis is a key process in most systematic reviews. In meta-analysis, the pooling of the study level outcomes from multiple studies is based on pre-specified mathematical criteria. However, many issues have been investigated and discussed in medicine research field. In drug-eluting stent industry, since the history from its first FDA approval is relatively short and the event of interest is usually rare, a couple of common concerns of meta-analysis are observed: 1) the publication bias such as insufficient number of qualified publications and 2) the study heterogeneity such as the imputation of zero events. The universal issues related to these common concerns will be discussed including measure of effect, study comparability, selection of fixed or random effects models, adjusting for imbalance baseline characteristics, multiple outcomes or time-points within a study, sensitivity analysis and posterior power analysis. Keywords: systematic review, meta-analysis