ABSTRACT
The present contribution aims to provide a unifying approach to modeling treatment effect in a meta-analysis of clinical trials with binary outcome. In recent years, meta-analysis has become an essential method used to provide more reliable information on an intervention effect. Additionally, it has been demonstrated to provide a powerful statistical tool to analyze and potentially combine the results from individual studies. Numerous international publications have demonstrated the quality and the common practicability of meta-analysis (see for example Cooper and Hedges (1994), Sutton et al. (2000), DuMouchel and Normand (2000), Jones (1992), or Greenland (1994)). Important for our situation here is the availability of the number of events xTi (x
C i ) and the person-time under risk n
C i ) (total of time every person
spent under risk) in the treatment arm (control arm) for each clinical trial i involved in the meta-analysis of a total of k studies. If all persons spend identical time under risk nTi is equivalent to the sample size, and the same is true for the control arm. We call this situation of meta-analysis a meta-analysis using individually pooled data (MAIPD). Table 1.1 shows the principal layout of the required information.
