Systematic reviews to evaluate causation: an overview of methods and application

Abstract

Currently systematic reviews focus on diagnosis or effectiveness of treatment. It is the understanding of disease aetiology that underpins medical education, practice and research. Whether an association meets causal criteria is usually assessed qualitatively. However, this can also be examined through evidence synthesis and systematic reviews to evaluate disease causation and mechanisms are much needed. It is important in such a review to specify the questions to be addressed regarding causal criteria such as strength, consistency, temporality, specificity, biological gradient, plausibility and experimental evidence. The next step is to conduct a thorough literature search to identify the relevant studies and to assess them for their quality, particularly in relation to the risk of bias, ascertainment of exposures and ascertainment of outcomes. Data synthesis can then examine if the observed associations in collated studies are consistent, strong and temporal using techniques such as meta-analysis, testing for heterogeneity and meta-regression. Biological plausibility and coherence with existing theories can also be examined systematically through an assessment of the basic scientific literature. Experimental evidence might also be collated and synthesised to determine if removal of a causal agent alters the outcome. Through these steps a systematic review can help to establish whether an association is causal or not.