Meta-analysis with Heterogeneity as Null-Hypothesis

Abstract

In traditional meta-analysis a single analysis-method is used for combined analysis of multiple studies with significant homogeneity of the studies as null-hypothesis. In this chapter a meta-analysis will be described for the combined analysis of open evaluation studies, with heterogeneity rather than homogeneity as null-hypothesis. An example of four studies receiving a different treatment per study, jointly including 120 patients with severe sepsis was used. Unlike traditional statistical methods, SPSS Modeler, a work bench for automatic data modeling, was able to identify the treatment 1 as the best of the four treatments with a predicted accuracy 91.8% of the decision tree output as applied.