Reporting missing outcome data for participants in randomized trials: a methodological survey of proposed approaches

Article type
Authors
Akl EA1, Shawwa K1, Kahale L1, Agoritsas T2, Petersen RB2, Busse JW2, Carrasco A2, Ebrahim S2, Johnston BC2, Neumann I2, Sola I3, Sun X2, Vandvik P4, Alonso-Coello P3, Guyatt G2
1American University of Beirut, Lebanon
2McMaster University, Canada
3Iberoamerican Cochrane Centre, Spain
4The Norwegian Knowledge Centre for the Health Services, Norway
Abstract
Background:
When conducting a meta-analysis, authors should identify the number and characteristics of participants for whom outcome data are missing. Reports of randomized clinical trials (RCTs) do not always include this information in a consistent, detailed and clear manner.

Objectives:
Systematically, to review the literature to identify proposed approaches for how RCT authors should report missing participant data (MPD).

Methods:
We defined MPD as missing outcome data for trial participants. We considered both categorical and continuous outcome data. We searched MEDLINE and the Cochrane Methodology Register for papers devoting at least two paragraphs to describe a proposed approach for reporting MPD from RCTs. Ten pairs of reviewers selected eligible studies in duplicate and independently. One reviewer (KS) extracted data from included papers and a second reviewer (EA) verified the abstracted data. An iterative process facilitated finalizing tabular and narrative summaries of the data.

Results:
Nine of the 9138 citations retrieved by the electronic search strategy proved eligible. The Table shows the specific items that each approach recommends should be reported. These include: number of participants with MPD, reasons for MPD, flow of participants, differences in rates of MPD between trial arms, differences between participants with and those without MPD, pattern of missingness (e.g. whether at random), methods used to handle MPD in the analysis, results of any sensitivity analyses, and implication of MPD on interpreting the results.

Conclusions:
Most identified approaches invite trial authors to report the number of participants with missing outcome data and the underlying reasons. Fewer approaches focus on reporting missingness patterns, methods for handling MPD and implications of MPD on results.