Stepped-wedge cluster-randomised controlled trials (SWTs) are used in a wide range of areas of public health, as well as other areas of public policy such as education and international development. SWTs can be thought of as a modified crossover design as the clusters are in both treatment arms at different times. All clusters start in the control arm and the intervention is introduced by random allocation and at regular intervals either to one cluster at a time or in small groups of clusters, until all clusters are eventually receiving the intervention.
On Sept 22nd, 2015, LSHTM hosted a Stepped-wedge Trial scientific symposium that featured leading speakers on the topic and a panel discussion.
Key Resources For Learning About Stepped-wedge Trials
We provide an overview of Stepped-Wedge Trials here, highlighting a by researchers at LSHTM that were published in Trials Journal. The paper series addressed the approaches to and challenges in the design, conduct, analysis, and reporting of Stepped-Wedge trials.
The use of SWTs has been rising since 1987, in both explanatory and . Researchers often think of conducting a SWT, as opposed to an ordinary parallel arms cluster randomised controlled trial, for one of three reasons: First, the appeal of phased implementation to avoid the challenges of introducing an intervention to a large number of clusters simultaneously. Second, ethical concerns about withholding an intervention with clear benefits from the control arm for the full duration of the trial. Finally, where a policy decision has already been made to implement the intervention and simultaneous roll out is impractical.
The final reason has three dimensions in itself. The first is political, whereby social acceptability of a phased implementation is increased by promising that the intervention will be delivered to all clusters randomly and thus fairly. The second is logistic in that assurance of fairness will also increase the likelihood of cluster participation. Finally, it allows for , whereby a phased and randomised roll-out provides an opportunity for a more rigorous evaluation.
While these reasons appear convincing, caution is advised as each poses challenges and issues that must be considered. The following paper discusses these reasons and the challenges they pose in detail:
Prost A, Binik A, Abubakar I, Roy A, De Allegri M, Mouchoux C, Dreischulte T, , , Osrin D. Logistic, ethical, and political dimensions of stepped wedge trials: critical review and case studies. Trials. 2015 Aug 17;16(1):1.
SWTs also pose several other challenges, and guidance for many aspects of their conduct are absent. Due to being a relatively new design, one challenge when dealing with SWTs is the lack of consistent terminology.
The ‘cross over point’ in a SWT is the point at which clusters change from being in the control condition to receiving the intervention. Crossovers are always in one direction, starting in control and crossing to intervention. The ‘roll out’ period is the time during which these crossovers of clusters occurs. The intervals between the crossover of clusters is called time between crossovers and sometimes also referred to as the ‘step length’.
Another challenge is the lack of standardised reporting of the design. There is a wide range of SWT designs. The three most commonly used and well defined are: Designs based on a closed cohort of participants; designs based on an open cohort; and the continuous-recruitment, short exposure designs.
In order for SWT designs to be fully reported and to enable readers to judge their strengths and weaknesses, it is important to describe how individuals participate, including the start and duration of exposure and whether they are exposed to control, intervention, or both, and how outcome measurements are obtained.
At the cluster level, clusters ‘participate’ throughout the trial and experience control and intervention conditions at different times. At the individual level however, individual participation and exposure can vary greatly between trials. Some individuals may experience both control and intervention, such as in a SWT involving patients with chronic diseases and requiring long term follow up. Others will experience either the control or the intervention, such as in a SWT involving individuals with acute conditions. Additionally, while in some trials all individuals in a cluster participate and provide outcome measurement, in others that involve large clusters (eg. big cities), only a fraction of individuals is sampled from the cluster to provide outcome measurement.
The duration of exposure to the control or the intervention, and therefore duration of follow up, will also vary between individuals and must be considered.
Individuals who participate can also contribute one or more outcome measurement, and outcome data may be collected by following individuals over time until the event of interest or a censoring event occurs, or may be collected at discrete time points over the course of follow up.
In the following paper, the authors endeavour to define the characteristics of stepped wedge trials and provide a taxonomy for clearer presentation of the three main SWT designs considering these features of individual participation and outcome measurement:
Copas AJ, , , , Baio G, . Designing a stepped wedge trial: three main designs, carry-over effects and randomisation approaches. Trials. 2015 Aug 17;16(1):1.
The paper also goes on to describe and guide the key design choices when planning a SWT: the randomisation methods, the number of steps and length of time between successive crossover points, whether the trial will be complete or incomplete, and Role of data collected before and after the rollout period in the study.
Sample size calculations is also an important aspect to consider when designing a SWT and that is considered in length in another article:
Baio G, Copas A, Ambler G, , Beard E, Omar RZ. Sample size calculation for a stepped wedge trial. Trials. 2015 Aug 17;16(1):1.
This range of study design characteristics presents challenges in the reporting and analysis of the results obtained from a SWT. Similar to reporting of the design of a SWT, there is no standard to guide reporting and analysis of results.
The total number of clusters are randomly allocated to a number of groups, each containing a small number of clusters and receiving the intervention at a different time. Additionally, data is collected from the intervention condition, on average, later than from the control condition. The resulting large number of cluster groups and the various crossovers that take place make it difficult to adapt the CONSORT diagram often used in the reporting of CRTs.
Another feature posing challenges in reporting results is the assessment of whether or not the randomisation has resulted in study conditions that are balanced at baseline in terms of important covariates.
In individually randomised trials, where there is a large number of participants, randomisation ensures that there are no imbalances between the study conditions in terms of known and unknown confounders. However, given the small number of clusters in CRTs, randomisation can no longer be relied on to ensure this balance. For this reason, a baseline table is used to present baseline characteristics of the clusters in each arm. Often in a SWT, the large number of groups may mean that presenting a table of baseline characteristics for each group is infeasible. Some researchers have moved to presenting baseline characteristics by study condition instead, e.g. the characteristics of participants as they are recruited. Clusters in the control condition will have different characteristics by the time the intervention is introduced as changes take place over time (e.g individuals changing, new policies being introduced, etc.).Therefore, a ‘baseline’ table presented by study arm includes changes over time in the outcome which undermine the purpose of a baseline table, which is reported so that the randomisation can be assessed by the reader.
The main disadvantage of the SWT design is its susceptibility to confounding by these secular trends. The presence of a secular trend in the outcome will bias the effect estimate, thus, accounting for this in the analysis is the priority of the available analysis methods.
The most commonly used method of analysis is the standard mixed model approach described in full detail in this paper:
Hussey MA, Hughes JP. Design and analysis of stepped wedge cluster randomized trials. Contemporary clinical trials. 2007 Feb 28;28(2):182-91.
The challenges faced in terms of reporting and analysing results of SWTs are discussed in the following paper, and the authors make recommendations for how they should be approached:
, , , Copas AJ, Beard E, , . Analysis and reporting of stepped wedge randomised controlled trials: synthesis and critical appraisal of published studies, 2010 to 2014. Trials. 2015 Aug 17;16(1):1.
In a paper by Hargreaves et al., the authors conclude that an ethically sound, well-designed, and well-conducted SWT with appropriate analysis can provide strong evidence of the effects of an intervention. However, researchers planning a SWT should consider all the challenges posed by SWT carefully. This paper offers a good summary of the questions researchers should answer and the issues they should consider before starting a SWT.
, Copas AJ, Beard E, Osrin D, , , , Baio G, , Prost A. Five questions to consider before conducting a stepped wedge trial. Trials. 2015 Aug 17;16(1):1.