Rigorous Reporting of Stepped Wedge Cluster Randomised Controlled Trials: Development of a Flow Diagram Template CURRENT

Clear, unambiguous reporting of a stepped-wedge cluster randomised controlled trials (SW-CRCTs) is a challenge, requiring attention to at least five reporting guidelines. A recent systematic review on the reporting quality of SW-CRCTs noted that only 24 out of their 39 included studies (61.5%) provided a flow diagram. The available flow diagram templates may be unable to accommodate the range of SW-CRCT designs and reporting requirements, thus hampering reporting standards. We aimed to review the degree to which current flow diagrams used in SW-CRCT reports support comprehensive reporting to standards. In the absence of a suitable model we aimed to develop a flexible SW-CRCT flow diagram template that would support reporting standards and facilitate transparency in the description of our SW-CRCT. identified We extracted from

However, it has been recognised that the reporting of cluster randomised controlled trials (CRCTs) is more complex and requires additional information to that of parallel group randomised trials to allow readers to interpret them accurately (3) (Campbell et al, 2004). The original CONSORT flow diagram is not sufficient for CRCTs, so an extended CONSORT Statement was published by Campbell et al in 2004 and revised in 2012 (4). More recently, motivated by the unique design characteristics of SW-CRCTs, a CONSORT extension for the SW-CRCT was developed and published (5).
A template flow diagram is provided in the CONSORT CRT extension, acknowledging that its exact form and content will vary in relation to the specific features of a trial and the type of analysis (4) (Campbell, 2012), and a specimen flow diagram is presented in the SW-CRCT extension (5). If the analysis is multilevel or hierarchical, the flow diagram should present the flow of both clusters and individuals through the trial, from assignment to analysis. Specifically, the flow diagram should elucidate the numbers of clusters that were randomly assigned, received intended treatment, and were analysed for the primary outcome, and include the losses and exclusions for both clusters and individual cluster members (4) (Campbell, 2012). Hemming et al (6)(2015) recommend that a flow diagram should illustrate how many participants are within each cell of the study design.
Although it appears that use of CONSORT checklists and flow diagrams has been associated with some improvement in reporting of parallel group randomised trials (7) (Plint et al, 2006), limitations in the quality of the reporting of SW-CRCTs remain (8) (Beard et al 2015). Martin et al (9)(2016) in the first systematic review of SW-CRCTs to assess reporting adherence to CONSORT items highlighted that the quality of reporting in stepped-wedge trials is suboptimal, for example, in relation to type of design and sample size calculations. We recently conducted a Phase II pilot SW-CRCT (SOCLE II), which explored the impact of a complex oral healthcare (OHC) intervention after stroke (11) (Brady et al, 2015). The clusters within this study were four stroke care wards (sites), and the intervention was delivered within each cluster with outcome data collected from both patient and staff participants. Site level cluster randomisation progressed in a stepped wedge manner, where, after a period of baseline data collection, each site converted from usual care to the enhanced OHC intervention in a randomly allocated order.
We wrestled with the challenge of ensuring our trial reported to all the various recommended criteria from current trial reporting standards within one flow diagram. Diagram templates within the reporting standards did not support reporting all of the required details. Furthermore, the limited adherence to the reporting standards identified in the aforementioned systematic review indicated that current flow diagram models may provide inadequate support in reporting to the various reporting standards.
We aimed to review the quality of flow diagrams used in previous SW-CRCTs and, where a suitable template was unavailable, we aimed to develop a flexible SW-CRCT flow diagram that could accommodate the relevant reporting standards and facilitate transparency in SW-CRCT reports.

Methods
As a basis for our SW-CRCT flow diagram evaluation we used the included trial reports in the Grayling For the purposes of our review, studies were eligible for inclusion if they were reports on a completed SW-CRCT and that included one or more flow diagrams. All 24 full text articles cited in the Grayling et al (10)(2017) review that included a flow diagram(s) were retrieved and stored. Study authors were not required to be contacted as we were only concerned with the information that was explicitly reported in the published flow diagrams.
Each of the 24 papers' flow diagrams were reviewed against 22 pre-determined criteria that we identified from across six key reporting standards that were relevant to our SOCLE II SW-CRCT. These  (Table 1).  One researcher (BB) extracted the data represented in each flow diagram against these pre-determined criteria entering all data into Microsoft® Excel. A second reviewer (HG) conducted a validation of the extraction process and uncertainties were resolved through discussion. Data were analysed using descriptive statistics and narrative summary.

Discussion
We conducted a comprehensive review of recently published SW-CRCT flow diagrams that, to our knowledge, has previously not been done before. Including a flow diagram when reporting SW-CRCTs is strongly recommended; however, when it came to reporting our SOCLE II trial we found that there were no suitable templates flexible enough to accommodate our SW-CRCT or comprehensive enough to support reporting all data items to standards. There was a wide variety in flow diagram quality, with the 'best' flow diagram only supporting the report of 13/22 (59%) of our required criteria. Additionally, due to the complexities of the diagrams they were often very difficult to interpret on their own without considering the full text.
A limitation of our review was that, although there are undoubtedly other SW-CRCT flow diagrams available, we chose to restrict our review to the 24 articles cited in Grayling et al (10) 2017 systematic review.
In the absence of a flow diagram template that considered all the items we required from the relevant reporting standards for our pilot SW-CRCT, we developed our own using Microsoft® Visio software (see Fig. 1 for the flow diagram populated with data from SOCLE II for illustrative purposes).
Developing a suitable flow diagram for our study was a challenge requiring many iterations to ensure we were able to meet the required reporting standards. The development team included medical statisticians, a service user and health services researchers to ensure that what was represented in the flow diagram correlated with the statistical analyses and facilitated transparency and reporting quality.  We believe that our flow diagram addresses the recommendations in the newly published CONSORT extension for SW-CRCTs, as it allows the reader to examine the nature of any differential inclusion and attrition by allocated sequence, treatment condition, and over time (5) (Hemming et al, 2018). An additional feature that could enhance our flow diagram would be to include time periods to allow for assessment of differential participation over time (5)    This is a list of supplementary files associated with this preprint. Click to download.