• Holger J Schünemann, Reem A Mustafa, Jan Brozek, Nancy Santesso, Patrick M Bossuyt, Karen R Steingart, Mariska Leeflang, Stefan Lange, Tommaso Trenti, Miranda Langendam, Rob Scholten, Lotty Hooft, Mohammad Hassan Murad, Roman Jaeschke, Anne Rutjes, Jasvinder Singh, Mark Helfand, Paul Glasziou, Ingrid Arevalo-Rodriguez, Elie A Akl, Jonathan J Deeks, Gordon H Guyatt, and GRADE Working Group.
• Department of Health Research Methods, Evidence, and Impact, McMaster University, 1280 Main Street West, Hamilton, Ontario L8S4K1, Canada; Department of Medicine, McMaster University, 1280 Main Street West, Hamilton, Ontario L8S4K1, Canada; McMaster GRADE Centre, Michael DeGroote Cochrane Canada Centre, McMaster University, 1280 Main Street West, Hamilton, Ontario L8S4K1, Canada. Electronic address: schuneh@mcmaster.ca.
• J Clin Epidemiol. 2019 Jul 1; 111: 69-82.

ObjectivesThis article describes the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group's framework of moving from test accuracy to patient or population-important outcomes. We focus on the common scenario when studies directly evaluating the effect of diagnostic and other tests or strategies on health outcomes are not available or are not providing the best available evidence.Study Design And SettingUsing practical examples, we explored how guideline developers and other decision makers can use information from test accuracy to develop a recommendation by linking evidence that addresses downstream consequences. Guideline panels should develop an analytic framework that summarizes the actions that follow from applying a test and the consequences.ResultsWe describe GRADE's current thinking about the overall certainty of the evidence (also known as quality of the evidence or confidence in the estimates) arising from consideration of the often complex pathways that involve multiple tests and management options. Each link in the evidence can-and often does-lower the overall certainty of the evidence required to formulate recommendations and make decisions about tests. The frequency with which an outcome occurs and its importance will influence whether or not a particular step in the linked evidence is critical to decision-making.ConclusionsOverall certainty may be expressed by the weakest critical step in the linked evidence. The linked approach to addressing optimal testing will often require the use of decision analytic approaches. We present an example that involves decision modeling in a GRADE Evidence to Decision framework for cervical cancer screening. However, because resources and time of guideline developers may be limited, we describe alternative, pragmatic strategies for developing recommendations addressing test use.Copyright © 2019. Published by Elsevier Inc.

13 keywords...

hide…