• Farid Foroutan, Gordon Guyatt, Victoria Zuk, Per Olav Vandvik, Ana Carolina Alba, Reem Mustafa, Robin Vernooij, Ingrid Arevalo-Rodriguez, Zachary Munn, Pavel Roshanov, Richard Riley, Stefan Schandelmaier, Ton Kuijpers, Reed Siemieniuk, Carlos Canelo-Aybar, Holger Schunemann, and Alfonso Iorio.
• Department of Health Research Methods, Evidence, and Impact, McMaster University, Ontario, Canada; Ted Rogers Center for Heart Research, Toronto General Hospital, Ontario, Canada. Electronic address: foroutaf@mcmaster.ca.
• J Clin Epidemiol. 2020 May 1; 121: 62-70.

ObjectiveThe objective of this study was to provide guidance on the use of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to determine certainty in estimates of association between prognostic factors and future outcomes.Study Design And SettingWe developed our guidance through an iterative process that involved review of published systematic reviews and meta-analyses of prognostic factors, consultation with members, feedback, presentation, and discussion at the GRADE Working Group meetings.ResultsFor questions of prognosis, a body of observational evidence (potentially including patients enrolled in randomized controlled trials) begins as high certainty in the evidence. The five domains of GRADE for rating down certainty in the evidence, that is, risk of bias, imprecision, inconsistency, indirectness, and publication bias, as well as the domains for rating up, also apply to estimates of associations between prognostic factors and outcomes. One should determine if their ratings do not consider (noncontextualized) or consider (contextualized) the clinical context as this will may result in variable judgments on certainty of the evidence.ConclusionsThe same principles GRADE proposed for bodies of evidence addressing treatment and overall prognosis work well in assessing individual prognostic factors, both in noncontextualized and contextualized settings.Copyright © 2020 Elsevier Inc. All rights reserved.

16 keywords...

hide…