Editorial
The American Journal of Gastroenterology (2005) 100, 1233–1236; doi:10.1111/j.1572-0241.2005.50107.x
Metaanalyses Are Observational Studies: How Lack of Randomization Impacts Analysis
Eloise E Kaizar MS1
1Department of Statistics, Carnegie Mellon University, Pittsburgh, Pennsylvania
Correspondence: Eloise Kaizar, MS,, Department of Statistics, 132 Baker Hall, Carnegie Mellon University, Pittsburgh, PA 15213
Received 22 January 2005; Revised 0000; Accepted 7 March 2005.
Abstract
Although metaanalyses are often syntheses of many experiments, they are themselves observational studies. As such, when performing or reading metaanalyses, we must consider the sources of bias that we usually expect in an observational study. The main two sources of bias for metaanalysis are publication bias and systematic heterogeneity. I consider the nature of both of these sources, methods to detect this bias, and ways to correct for the bias.
