The purpose of this session is to provide an overview about what methodological issues the GRADE working group has approached over the past 15 years and how it did it (including issues around qualitative research, public health), to update attendees on the latest development in GRADE, and to inform about what GRADE project groups are currently tackling and stimulate feedback and discussion about the best approaches.
NB: We are conducting a GRADE workshop to learn about GRADE on 3 October 2015. In addition, this session is followed by a GRADE working group meeting from 14:00 to 19:45 h on 5 October 2015.
Chairs of this Special Session: Elie Akl, American University of Beirut; Regina Kunz, University of Basel
1. Speaker: Signe Flottorp, Senior Scientist, Norwegian Knowledge Centre for the Health Services
Title: 15 shades of GRADE
Abstract: Fifteen years ago, GRADE was developed considering a wide range of clinical questions, but applied mainly for therapeutic interventions.
Today the GRADE approach has been applied to a range of health and health care questions and different types of research: public health, diagnosis, prognosis, environmental health, coverage decisions, network meta-analysis, qualitative research and laboratory animal research. This evolvement illustrates that GRADE provides a general framework, not only for structuring health care questions, but also to answer these questions in a systematic, transparent and explicit way.
This presentation will provide an overview of the different applications, including where adaptations to the original approach were needed and why. We will also explore to which extent wide(r) application of GRADE is sensible or desirable.
2. Speaker: Alfonso Iorio, Associate Professor, Department of Clinical Epidemiology and Biostatistics Department, McMaster University, Hamilton, Ontario, Canada
Title: GRADE application to prognostic evidence
Abstract: Observational studies addressing patients' prognosis may provide a) robust estimates of the likelihood of undesirable or desirable outcomes in both treated and untreated patients, or b) estimates of the incremental risk associated with a prognostic factor.
Patients will often find this information helpful in understanding the likely course of their disease, in planning their future and in engaging in shared decision-making with their health care providers.
We will discuss with examples the framework provided by GRADE to assess the quality of bodies of prognostic evidence. The guidance will be calibrated primarily at those conducting systematic reviews of prognostic studies, but it will be more in general useful to anyone interested in prognostic estimates and their associated confidence (e.g. a guideline panel using baseline risk estimates to estimate the absolute effect of an intervention or using evidence about a risk factor when judging the importance of a subgroup analysis).
3. Speaker: Holger Schünemann, Professor and Chair, Department of Clinical Epidemiology and Biostatistics, McMaster University
Title: GRADE for diagnosis, what is the latest?
Abstract: The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) working group published its first article about GRADE for diagnosis in 2008. Updating Cochrane authors and users about the latest development in this area is timely and needed especially with the considerable increase in diagnostic test accuracy (DTA) systematic reviews and protocols published in the Cochrane library over the last few years.
In this presentation we will update the audience about:
- Different GRADE summary of findings (SoF) tables formats for DTA systematic reviews using examples from recently published Cochrane DTA reviews.
- The latest work done by the GRADE diagnosis project group to further operationalize the GRADE domains for diagnosis by applying these concepts to Cochrane DTA systematic reviews.
After updating the audience about the latest development of GRADE for diagnosis, we will conclude by summarizing the questions that we believe are already answered. We will also highlight the areas that will benefit from empirical research.
4. Speaker: Gordon Guyatt, Professor, Department of Clinical Epidemiology and Biostatistics, McMaster University
Title: How should systematic reviews deal with loss to follow-up in primary studies?
Over the last several years our group has published four articles informing how systematic review authors can effectively address the risk of bias implications associated with loss to follow-up in primary studies.
The first two articles dealt with binary outcomes. In the first we introduced the concept of “plausible worst case” and demonstrated that up to 33% of positive trials in high impact journals lost statistical significance under a plausible worst case scenario. We then demonstrated how this method could be applied to each individual study in systematic reviews and how the resulting sensitivity analyses can inform judgments regarding risk of bias.
The second set of two articles addressed loss to follow-up in continuous outcomes. In the first, we demonstrated how making a set of increasingly stringent imputations regarding patients lost to follow-up in each primary study on the basis of results in other studies in patients for whom complete data were available could inform risk of bias decisions. The method presented in this first article was restricted to meta-analyses in which each study used the same measurement instrument. Our second article addressing continuous outcomes demonstrated how the approach could be applied when – as often happens when addressing quality of life outcomes – studies used different instruments to measure the same construct.
The suggested approaches provide rigorous and straightforward methods for making inferences about risk of bias as a result of loss to follow-up in systematic reviews and meta-analyses. The approaches can be used to inform risk of bias judgements in GRADE evidence profiles and summary of findings tables.