Identifying and dealing with participants with missing outcome data in trials: Framework for systematic review authors




Poster session 2 Thursday: Evidence synthesis - methods / improving conduct and reporting


Thursday 14 September 2017 - 12:30 to 14:00


All authors in correct order:

Kahale LA1, Guyatt G2, Khamis AM1, Briel M3, Agoritsas T4, Carrasco-Labra A2, Zhang M2, Busse JW2, Akl EA1
1 American University of Beirut, Lebanon
2 McMaster University, Canada
3 University of Basel, Switzerland
4 University of Geneva, University of Geneva
Presenting author and contact person

Presenting author:

Lara Kahale

Contact person:

Abstract text
Background: For dealing with trial participants with missing outcome data (MPD) in systematic reviews (SRs), the GRADE working group recommends conducting a complete case analysis (CCA) for the primary analysis. The group also recommends conducting sensitivity analyses making assumptions about the outcomes of those participants to evaluate the robustness of results. However, it is not always clear from trial reports whether some categories of participants (e.g., non-compliers) were followed-up or not (i.e. have MPD or not). Also, it is not always clear how the trialists dealt with MPD in their own analysis. Currently, there is no guidance on how SR authors should deal with these situations.

Objectives: To provide SR authors with a framework to identify and deal with the categories of participants for whom it is not clear whether they were followed-up for the outcomes of interest and make suggestions on how to deal with these categories in their primary analysis and sensitivity analysis.

Methods: Our group conducted a number of methodological studies on the topic of interest. We then convened a group of researchers with expertise in SR methodology and MPD. The group drafted an illustration of the status of trial participants and their outcome data at the trial and SR level. We refined this illustration and developed into a framework that provides better guidance for SR authors.

Results: We are still in the process of finalising the framework. In general, when trialists are clear about whether certain categories have MPD, we recommend that authors of systematic reviews exclude these categories from the primary meta-analysis (CCA) and include them in the sensitivity analysis. When trialists made assumptions, or excluded them from the analysis, these categories should be considered as having MPD. When trialists are not clear whether certain categories have MPD, we recommend that authors of SRs consider that the categories of participants listed in table 1 have MPD.

Conclusion: Our framework will assist SR authors to identify participants with MPD and deal with them in their meta-analysis.