[Wiley Series in Probability and Statistics] Statistical Meta-Analysis with Applications || Multivariate Meta-Analysis

✍ Scribed by Hartung, Joachim; Knapp, Guido; Sinha, Bimal K.

• Publisher: John Wiley & Sons, Inc.
• Year: 2008
• Tongue: English
• Weight: 868 KB
• Edition: 1
• Category: Article
• ISBN: 0470290897
• DOI: 10.1002/9780470386347.ch11

✦ Synopsis

In this chapter we discuss techniques for multivariate meta-analysis with some examples. It will be seen that the methods of multivariate meta-analysis can be applied quite generally, irrespective of the nature of effect of interest, be it based on means, proportions, odds ratio, or correlations. Moreover, in the context of multivariate synthesis of data, it is possible to study within-study differences between effects for different outcomes as well as different relations of potential predictors with study outcomes.

Multivariate data can arise in meta-analysis due to several reasons. First, the primary studies can be multivariate in nature because these studies may measure multiple outcomes for each subject and are typically known as multiple-endpoint studies. For example, a systematic review of prognostic marker MYCN in neuroblastoma seeks to extract two outcomes: log hazard ratio estimates for overall survival and diseasefree survival (Riley et al., 2004). Berkey et al. (1998) reported results of five studies assessing the difference in a surgical and nonsurgical procedure for treating periodontal disease with two outcomes: improvement in probing depth and improvement in attachment level. It should however be noted that not all studies in a review would have the same outcomes. For example, studies of Scholastic Aptitude Tests (SATs)