Critical appraisal of nonrandomized studies—A review of recommended and commonly used tools

Abstract

Rationale, aims, and objectives: When randomized controlled trial data are limited or unavailable, or to supplement randomized controlled trial evidence, health technology assessment (HTA) agencies may rely on systematic reviews of nonrandomized studies (NRSs) for evidence of the effectiveness of health care interventions. NRS designs may introduce considerable bias into systematic reviews, and several methodologies by which to evaluate this risk of bias are available. This study aimed to identify tools commonly used to assess bias in NRS and determine those recommended by HTA bodies. Methods: Appraisal tools used in NRS were identified through a targeted search of systematic reviews (January 2013-March 2017; MEDLINE and EMBASE [OVID SP]). Recommendations for the critical appraisal of NRS by expert review groups and HTA bodies were reviewed. Results: From the 686 studies included in the narrative synthesis, 48 critical appraisal tools were identified. Commonly used tools included the Newcastle-Ottawa Scale, the methodological index for NRS, and bespoke appraisal tools. Neither the Cochrane Handbook nor the Centre for Reviews and Dissemination recommends a particular instrument for the assessment of risk of bias in NRS, although Cochrane has recently developed their own NRS critical appraisal tool. Among HTA bodies, only the Canadian Agency for Drugs and Technologies in Health recommends use of a specific critical appraisal tool—SIGN 50 (for cohort or case-control studies). Several criteria including reporting, external validity, confounding, and power were examined. Conclusion: There is no consensus between HTA groups on the preferred appraisal tool. Reviewers should select from a suite of tools on the basis of the design of studies included in their review.