An Empirical Study on Imbalanced Samples and Presence of Zeros in Clinical Trials

Abstract

Meta-analysis is a statistical tool in medical research to analyse exposed effect by estimating the binary outcomes from multiple clinical studies. The main objective of this paper is to study the sensitivity of parameter in a Meta-analysis from imbalanced dataset of clinical trials.  An empirical study on data from 24 clinical trials has been analysed using Bayesian random effects model. The parameter of interest is the risk difference between the exposed group and the unexposed of balanced and imbalanced groups. Presence of one or more zeros in a study could create a major impact in variability between studies. This study has focussed of extent of sensitivity in the estimation of point and confidence intervals.