The CONSORT statement (with any subsequent updates and extensions) should be followed to ensure a complete report of each clinical trial.
This includes the reporting of a ‘participant flow’, showing for each treatment group ‘the number of participants who were randomly assigned, received intended treatment and were analysed for the primary outcome’, and ‘for each group, losses and exclusions after randomisation, together with reasons’. Others have encouraged reporting of proportions as well as just numbers.
Some additional considerations on top of the minimum standard defined by CONSORT are explained below. Some of the suggestions are derived from a report by Akl and colleagues.
If it is not possible to include this level of detail in the main trial report, further information could be made available as supplementary material.
Clear terminology
When reporting the ‘participant flow,’ use clear terminology to describe the participation changes (see principle O1 – participation can stop, reduce or change).
This could either be explicit, self-explanatory terms that are likely to mean the same thing to everyone (some suggestions are available in Resources).
If shorthand terminology like ‘withdrawn’ or ‘lost to follow-up’ are used then these need to be clearly defined and consistent with the trial protocol.
Other suggestions for best practice are available in the published literature.
Reporting detail of how participation changed
Participation in trials is often made up of various elements and can therefore change in various ways (see principle O1).
It is unlikely to be necessary to report every detail of how participation has changed, but enough detail should be reported to allow correct and complete understanding and interpretation of the trial results.
At the least, it is likely to be useful to separate the major categories of participation change, i.e. intervention stopping primarily due to clinical decisions, participation changes arising from participants’ expressed wishes, cases of loss of contact between trial staff and participants, participation changes due to loss of capacity to consent, and death.
Modifications to the CONSORT flow diagram – for example, use of dotted and solid lines – could help distinguish between participation changes that led to no further trial follow-up from those where the participant continued to provide data for the trial.
Reporting reasons for participation changes
Consider how to ensure that reported information on ‘reasons’ for ‘losses and exclusions after randomisation’ can be as useful as possible.
It may be useful to provide known information on a) whose choice the participation change was (participant or someone else), b) what the broad motivation was (e.g. primarily clinical decision, participant finding follow-up visits burdensome, or no longer able to attend due to worsening health) and c) any relevant indirect factors (e.g. could not continue in trial follow-up due to moving house).
The amount of reported detail should balance the need to be informative with the need to avoid revealing sensitive or confidential details about any participants.
Where missing data have occurred for reasons other than participation changes (e.g. outcome data not collected for technical reasons) then this should be made clear.
Present reasons for participation changes and missing data by treatment group.
Timing of participation changes
In some cases, information about the timing of participation changes might be important to include, for example if many participants in one treatment group stopped or changed their participation soon after randomisation.
Other suggestions for reporting missing data
Other suggestions have been made regarding reporting around missing data in trials (regardless of whether the missing data arises from participation changes).
Suggestions made by others include:
- Considering inclusion of a comparison of baseline characteristics for participants in each trial arm with (any) missing data and those without missing data (though some point out that this can only be indicative as trials are not likely to be powered to make formal comparisons of these groups).
- Making clear the number of participants in each trial arm with missing outcome data, and presenting this separately per outcome if need be.
Reporting in terms of estimands
Trial reporting should describe the trial results in terms of any pre-specified estimands.
Reports should make clear the methods for handling missing data in the trial analysis, any assumptions underpinning the choice of methods, and the methods and results of any missing data sensitivity analyses.
It has also been suggested by others that the possible impact of missing data on the trial’s conclusions should be included in the report’s discussion section.