Statistical Methods for an Individual Patient Data Meta-Analysis
Science Workshop—Spring 2019
OPEN TO THE PUBLIC | Michael D. Larsen, Professor at Saint Michael’s College Department of Mathematics & Statistics will be discussing Statistical Methods for an Individual Patient Data Meta-Analysis.
Bacterial vaginosis (BV) during pregnancy is associated with pre-term birth. Antibiotics including metronidazole and clindamycin have been studied in randomized clinical trials for treating BV. Their impact on length of pregnancy, however, has not been conclusively established. A meta-analysis is an analysis that combines results from multiple studies. To answer more focused questions about BV and groups at risk for preterm birth, however, combining published results were insufficient. Individual patient data (IPD) were gathered from several past studies, combined, and re-analyzed. This talk will describe this process, statistical models used for the analysis of the data, and a novel approach to missing data related to some studies not providing IPD. The talk will emphasize statistical issues and illustrate methods, but be mostly nontechnical in nature.