Su Golder and colleagues from the University of York said there was 'considerable debate' about the best way of estimating adverse event risks from particular treatments. RCTs give unbiased estimates of treatment effects, but are rarely large enough to find rare, long-term effects and exclude many types of patients, they said.
However, studies using observational data are not randomised and are open to bias through confounding different variables, they added.
The researchers found there was no difference in adverse effect estimates from meta-analyses of either RCTs or observational studies. They argue their findings have important implications for the conduct of systematic reviews of harm.
'It may be preferable for systematic reviewers of adverse effects to evaluate a broad range of studies that can help build a complete picture of any potential harm and improve the generalisability of the review without loss of validity.'