Systematic reviews of prognosis studies IV: The use of GRADE for assessing the certainty of evidence for questions of overall prognosis and prognostic factors
Gordon Guyatt, McMaster University
Marialena Trivella, London School of Hygiene and Tropical Medicine
Nina Kreuzberger, University Hospital Cologne
Alfonso Iorio, McMaster University
Background The GRADE approach is widely used for assessing certainty in inferences drawn from evidence. The GRADE approach has been adopted to assess certainty in evidence on overall prognosis, and prognostic factors. Objective Present participants with an interactive session designed to understand and apply GRADE to evidence on overall prognosis and prognostic factors. Description We will engage participants on 10 topics related to the application of the GRADE for questions of prognosis. We will provide participants with examples from real systematic reviews. Participants will review the reviews and apply the GRADE guidance to determine the certainty . Study design: The participants will be engaged in discussions over the ideal study design when assessing evidence related to prognosis. Risk of bias: The domain will be introduced (high vs low risk of bias studies) with focus on considerations necessary for evidence related to prognostic factors (adjusted vs unadjusted associations). Inconsistency: We review the concept of inconsistency (and its relation to statistical heterogeneity). We will assess the impact of aberrant studies and their impact on the overall conclusion of the meta-analysis. Imprecision: The participants will review the different frameworks for assessment of imprecision, along with the difference between relative compared to absolute effect estimates. The workshop will provide an online calculator for determining absolute risks, necessary for the assessment of imprecision under the partially contextualized framework. Indirectness: In this segment, the attendees will compare the research question of a systematic review to the characteristics of studies addressing the research question. We will provide guidance for circumstances when it is necessary to rate down our certainty for indirectness. Publication bias: Example of a systematic review will be provided in which the assessment of publication bias based on visual inspection (funnel plot) is different compared to the statistical test. Domains for rating up and special considerations with use of GRADE: Discussion on the three domains for increasing certainty, and challenges with making judgment with domains of GRADE that are highly interrelated.
2University of Birmingham, Birmingham, UK