Comparison of methods of grading bodies of evidence and strength of recommendations used by clinical practice guideline groups

Article type
Authors
Norris S
Abstract
Background: Numerous clinical practice guideline groups have developed a variety of approaches for the assessment of bodies of evidence and for the formulation of clinical recommendations. Objectives: To review and compare methods for the assessment of bodies of evidence and approaches to the formulation of recommendations for clinical practice. Methods: Clinical practice guideline organizations that have developed two or more guidelines were identified using the National Guidelines Clearing House. Organization websites and MEDLINE were searched and organizational members contacted for published or unpublished information on the methods used by each organization. Data were synthesized in a qualitative manner. Results: In the initial review, 12 organizations were examined. Quality assessment of randomized controlled trials and other study designs is performed by most groups using tools developed inhouse. Most organizations have moved beyond study design as the sole criteria for assessing strength of evidence, and they separate assessment of evidence and strength of recommendations. Assessment of a body of evidence is usually performed across outcomes; only GRADE (or groups using a modified GRADE approach) assessed outcomes individually. The domains used to assess a body of evidence generally included: study design, study limitations, consistency of results, and precision. Applicability of the evidence to the users’ needs (indirectness) is explicitly addressed by GRADE (and related approaches) and by the US Preventive Services Task Force. Other aspects of indirectness, including intermediate and surrogate outcomes, are variably addressed by organizations. Of the groups that explicitly state the domains used to formulate recommendations, the basic constructs usually include the certainty and magnitude of net benefit. Values and measures of economic efficiency are included in a minority of guidelines. Conclusions: Guideline organizations vary significantly in the explicitness and transparency of their approaches. When domains for assessment of a body of evidence and/or the strength of recommendations are stated, there is considerable overlap among organizations. Variability across groups may be explained in part by the study designs, types of questions and interventions being examined, and the purpose of the recommendations. Ideally, more standardized approaches will be devised for journals and guideline organizations.