Understanding heterogeneity in meta-analysis: the role of meta-regression

Background: Meta-regression has grown in popularity in recent years, paralleling the increasing numbers of systematic reviews and meta-analysis published in the biomedical literature. However, many clinicians and decision-makers may be unfamiliar with the underlying principles and assumptions made within meta-regression leading to incorrect interpretation of their results. Aims: This paper reviews the appropriate use and interpretation of meta-regression in the medical literature, including cautions and caveats to its use. Materials & Methods: A literature search of MEDLINE (OVID) from 1966-February 2009 was conducted to identify literature relevant to the topic of heterogeneity and/or meta-regression in systematic reviews and meta-analysis. Results: Meta-analysis, a statistical method of pooling data from studies included in a systematic review, is often compromised by heterogeneity of its results. This could include clinical, methodological or statistical heterogeneity. Meta-regression, said to be a merging of meta-analytic and linear regression principles, is a more sophisticated tool for exploring heterogeneity. It aims to discern whether a linear relationship exists between an outcome measure and on or more covariates. The associations found in a meta-regression should be considered hypothesis generating and not regarded as proof of causality. Conclusions: The current review will enable clinicians and healthcare decision-makers to appropriately interpret the results of meta-regression when used within the constructs of a systematic review, and be able to extend it to their clinical practice.