Assessing the Risk of Bias in Systematic Reviews of Health Care Interventions

Abstract

Objective Risk-of-bias assessment is a central component of systematic reviews but little conclusive empirical evidence exists on the validity of such assessments. In the context of such uncertainty, we present pragmatic recommendations that can be applied consistently across review topics, promote transparency and reproducibility in processes, and address methodological advances in the risk-of-bias assessment. Study Design Epidemiological study design principles; available empirical evidence, risk-of-bias tools, and guidance; and workgroup consensus Results We developed recommendations for assessing the risk of bias of studies of health care interventions specific to framing the focus and scope of risk-of-bias assessment; selecting risk of bias categories; choosing assessment instruments; and conducting, analyzing, and presenting results of risk-of-bias assessments. Key recommendations include transparency and reproducibility of judgments, separating risk of bias from other constructs such as applicability and precision, and evaluating risk of bias per outcome. We recommend against certain past practices, such as focusing on reporting quality, relying solely on study design, or numerical quality scores, and automatically downgrading for industry sponsorship. Conclusion Risk-of-bias assessment remains a challenging but essential step in systematic reviews. We presented standards to promote transparency of judgments.