Optimal use of randomized and non-randomized studies in evidence syntheses of interventions that use GRADE

Article type
Authors
Cuello-García C1, Morgan R1, Verbeek J2, Guyatt G1, Schunemann H1
1McMaster University
2Cochrane Work Review Group. Finnish Institute of Occupational Health.
Abstract
Background:
Systematic review authors, guideline developers, and other knowledge syntheses’ practitioners use randomized studies (RS) and non-randomized studies (NRS) as sources of evidence for questions about health interventions. Well-conducted RS represent the most reliable individual source of evidence for estimating relative effects, primarily because of protection against confounding. NRS, however, can provide valuable information as complementary, sequential, or replacement evidence for RS. The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach can help assessing the appropriateness of using NRS with RS in health syntheses.

Objectives:
To provide guidance on the optimal use of RS and NRS in health syntheses of interventions that use GRADE.

Methods:
This overarching work represents three consecutive mixed-methods studies on the optimal use of RS and NRS in health syntheses. Through workshops, surveys, and an iterative process of questions and feedback with face to face meetings and webinars with Cochrane authors, and guidelines and GRADE experts, we developed guidance for authors using RS and NRS in health syntheses of interventions.

Results:
We:
1) developed a framework for deciding when NRS would be helpful when conducting health syntheses;
2) present options for integrating RS and NRS when using GRADE (i.e. summary of findings tables and evidence profiles); and
3) discuss the impact of the differences in GRADE domains between RS and NRS and its effects on the overall certainty of evidence.

Conclusions:
Authors of health syntheses can use GRADE to assess the appropriateness of searching and including NRS with RS in knowledge syntheses of health interventions. Here we present several options and discuss an adequate integration of RS and NRS for increasing the certainty of a body of evidence.

Healthcare consumer involvement:
NRS are commonly considered ‘real life’ studies evaluating outcomes that sometimes are not assessed in RS. Extending the scope of the evidence in health syntheses will help consumers and healthcare providers to reach better decisions.