Community participatory learning and action cycle groups to reduce type 2 diabetes in Bangladesh (D:Clare): an updated study protocol for a parallel arm cluster randomised controlled trial

Abstract

The “Diabetes: Community-led Awareness, Response and Evaluation” (D:Clare) trial aims to scale up and replicate an evidence-based participatory learning and action cycle intervention in Bangladesh, to inform policy on population-level T2DM prevention and control.

The trial was originally designed as a stepped-wedge cluster randomised controlled trial, with the interventions running from March 2020 to September 2022. Twelve clusters were randomly allocated (1:1) to implement the intervention at months 1 or 12 in two steps, and evaluated through three cross-sectional surveys at months 1, 12 and 24. However, due to the COVID-19 pandemic, we suspended project activities on the 20th of March 2020. As a result of the changed risk landscape and the delays introduced by the COVID-19 pandemic, we changed from the stepped-wedge design to a wait-list parallel arm cluster RCT (cRCT) with baseline data. We had four key reasons for eventually agreeing to change designs: equipoise, temporal bias in exposure and outcomes, loss of power and time and funding considerations.

Trial registration ISRCTN42219712. Registered on 31 October 2019.

The D:Clare trial (Diabetes: Community-Led Awareness, Response and Evaluation) was designed as a cluster randomised stepped-wedge trial, in Alfadanga Upazilla, Faridpur District, Bangladesh (ISRCTN42219712) [1]. The trial aims to evaluate the impact of a scaled-up community-based participatory learning and action (PLA) cycle intervention to prevent type 2 diabetes (T2DM) in a population of 120,000 people. The study began in January 2020, with a public consent and randomisation ceremony including community and Ministry of Health and Family Welfare representatives, on the grounds that all communities in the Upazilla would eventually receive the intervention in line with the stepped-wedge approach.

Bangladesh reported its first confirmed cases of SARS-CoV-2 on the 8th of March 2020. Due to concerns about infection risk to both staff and communities, we made a decision to suspend all field-based project activities on the 20th of March 2020 (Fig. 1). Early in the pandemic, evidence emerged that uncontrolled hyperglycaemia and T2DM were risks for severe COVID-19 infections and mortality, alongside older age, obesity and heart disease [2,3,4]. Given the focus and nature of our PLA intervention, we were therefore particularly conscious that continuing the trial may have increased risks amongst vulnerable populations with non-communicable diseases. The status of the trial at the point of suspension is summarised in Table 1. Bangladesh subsequently entered into a nationwide government-declared lockdown from the 23rd of March to the 30th of May 2020, and restrictions on mass gatherings continued until the 1st of September 2020 [5]. The second serious COVID-19 wave began in March 2021, and lockdowns were again implemented between 5th April–21st April 2021 and 1st July–11th August 2021.

Table 1 Status of the D:Clare stepped-wedge trial at the point of COVID-19 field activity suspension on the 20th of March 2020

This short article summarises the changes to our original trial design, in line with the CONSERVE 2021 Statement recommendations that trials impacted by extenuating circumstances should report on modifications [6]. We detail the considerations and rationale for these changes, which may be of relevance to other randomised controlled trials underway in dynamic contexts.

Change in trial design

As a result of the changed risk landscape and the delays introduced by the COVID-19 pandemic, we decided to change from a stepped-wedge (SW-RCT) to a wait-list parallel arm cluster RCT (cRCT) with baseline data. Conceptually, our wait-list design is a parallel arm cRCT but with a commitment to implement the intervention to control clusters at the end of the trial evaluation. As detailed in Table 2, this differs from our stepped-wedge trial design in terms of the timing of roll-out of the intervention across all clusters, timing of cross-sectional data collection for evaluation, and in terms of how clusters are exposed over time, i.e. the allocated exposure (intervention or control) does not change during the trial evaluation. Our original SW design had two steps and was planned to take 30 months, with cross-sectional surveys done at months 1, 12 and 24 of intervention implementation (Fig. 1) [1]. The SW design should be resilient to temporal changes within a population, and so our original approach remained valid. However, the interruption of activities and the nature of the COVID-19 pandemic meant this design was no longer the most efficient and appropriate to meet project goals, and we presented alternative options to our Trial Steering Committee for consideration. We also engaged with community and government stakeholders to check that the proposed adaptions would be acceptable. We had four key reasons for eventually agreeing to change designs.

Table 2 Summary of key changes from original stepped-wedge design to parallel arm wait-list trial

Fig. 1

Planned D:Clare project timeline and COVID-19 interruptions

Equipoise

The D:Clare PLA intervention was shown to be effective in reducing both the 2-year cumulative incidence and prevalence of T2DM in a rural Bangladesh population during the D-Magic trial [7]. This was part of the justification for us using an SW-RCT originally, as evidence of population benefit existed, and our aim was to determine effectiveness at scale in a similar but new population. However, with the considerable change in context and the potential need to adapt the intervention components and delivery, the lack of equipoise around the PLA intervention we had previously argued was less clear. Specifically, our intervention encourages groups to meet, encourages participation from those with T2DM and NCDs and encourages collective action. In a context where COVID-19 preventive measures focused on restricting inter-household interactions, we hypothesised that PLA’s mechanism of action may be affected.

Further, if COVID-19 cases were not being diagnosed in this community setting, then group meetings had the potential to cause harm. However, by the end of 2020, there was evidence that outdoor environments posed a lower risk of transmission than indoor, crowded spaces, especially if this can be combined with the use of face masks, hand hygiene and physical distancing. Given the potential for our intervention to improve T2DM management (a key risk for poor COVID-19 outcomes), the ability to deliver in a way that would reduce transmission, and the inclusion of new stop/start rules (Fig. 2), we felt this risk could be sufficiently mitigated. We therefore decided we met the criteria for equipoise around the intervention needed to do a parallel arm cRCT.

Fig. 2

D:Clare trial stop, pause and start rules for COVID-19 adaptation

Temporal bias in exposure and outcomes

We also hypothesised that health literacy, care-seeking, dietary and physical activity behaviours, and the epidemiology of diabetes could be vastly different after lockdown restrictions were lifted — and therefore from our baseline survey. This in itself should not invalidate the SW-RCT design, but could make the interpretation and communication of the intervention impact on primary and secondary outcomes more complicated. The timing of intervention delivery relative to lockdown and social distancing measures was also likely to have an important influence on the uptake, delivery and effectiveness of the intervention. This may result in variable intervention effects between the two steps of SW implementation, which could be assessed through process evaluation, but again would complicate interpretation.

Loss of power

Our power calculation was based on achieving at least an 80% response in the first cross-sectional survey. However, we only achieved 72% recruitment at the time of interruption and saw variation in rates between clusters (49–89%). In order to then ensure the SW-RCT was sufficiently powered, we would have had to increase the sample size of all the subsequent surveys. Switching to a parallel arm trial which uses both a new baseline and endline data (assuming an autocorrelation of 0.4), we could achieve 78% power for a 30% reduction in the primary outcome and considered this a feasible alternative. The change to the number and timing of surveys and the inclusion of baseline data in outcome evaluation are notable changes to our original protocol (Table 2).

Time and funding

Finally, there was a very practical issue that we no longer had enough time to complete the SW-RCT design within the overall 36-month funded project period, using our 12-month staggered two-step design. By switching to a parallel cRCT we could complete the effectiveness evaluation within the funded project timeline, however, recognising that the parallel arm design would determine the intervention impact on a smaller population scale than we had originally intended. We then planned to source project extensions and explore the reallocation of resources to ensure that at a minimum the intervention could be delivered in control clusters as was promised to communities, but without incurring ongoing concurrent process, economic and impact evaluation costs.

Protocol updates

We made changes to three key areas of the trial protocol: study design, intervention and sample size; no amendments were made to the trial procedures for population eligibility, sampling, randomisation, blinding, data collection, or analysis of the primary or secondary outcomes. A list of registered trial protocol amendments in our ISRCTN record is summarised in Table 3.

Table 3 Summary of registered trial protocol revisions in the D:Clare ISRCTN record

We also set out COVID-19 standard operating procedures, with new stop-start rules (Fig. 2), and a COVID-19 safety protocol for staff and study participants, and consulted with a Data Monitoring and Safety Board on these infection prevention measures. For the intervention, we made the following modifications to incorporate COVID-19 measures: holding two meetings per village per month to allow for smaller groups but with the same coverage; inclusion of COVID-19 health information; re-organised meeting content to be delivered over a minimum of 13 instead of the planned 18 meetings (Table 2).

Current trial status

As of 22/04/2022: We completed a new baseline survey on 25/02/2021, with a response rate of 1,392 from 1,584 (87.9%) sampled participants, which forms the parallel arm cRCT baseline data. A total of 213 PLA groups have been formed in 6 of the 12 study clusters, and have completed 11 of a minimum of 13 planned meetings. The endline survey will be completed between August and October 2022.

Protocol version 3.0 (16/06/2021)