The aim of the D:Clare trial is to evaluate the scale-up a PLA cycle intervention to prevent intermediate hyperglycaemia and T2DM and improve control of T2DM in rural Bangladesh.
We hypothesise that rapid horizontal scale-up of PLA will significantly: a) increase population-level awareness of diabetes prevention and control; b) reduce the prevalence of intermediate hyperglycaemia and diabetes by at least 30%.
Setting
The trial is set in Alfadanga upazilla, Faridpur district, in the central region of Bangladesh. Faridpur has an estimated population of 1.7 million, and is divided into nine upazillas. Alfadanga was purposefully selected as the trial location as it is less liable to flooding and land erosion than other areas of Faridpur, has not been exposed to a PLA intervention previously, and is close to an existing field office. Alfadanga has an estimated population of 120,000 people, divided into six administrative unions. The area is rural with a predominantly agricultural economy based on jute and rice; the population is mainly Bengali and Muslim, reflecting a ‘typical’ rural setting in Bangladesh.[15]
Primary healthcare is provided at union health centres, family welfare centres and community clinics. Secondary care is provided at the upazilla health complex, a charity based diabetes center in Alfadanga town, and the District Hospital in Faridpur town. There are two tertiary referral hospitals in Faridpur district, which can treat T2DM complications.
Design
This is a stepped-wedge cluster randomised controlled trial (SW-CRCT), with integrated process and economic evaluations (Fig. 1). Clusters are defined as administrative geographical areas, with a population of approximately 10,000. Each of the six unions in Alfadanga will be divided into two, forming 12 clusters. The SW-CRCT is at risk of inter-cluster contamination, where participants residing in clusters under control conditions may be exposed to the intervention in a neighbouring cluster. Therefore, we will use a ‘fried-egg’ design for our evaluation surveys, with participants residing in border areas excluded from surveys.
The trial clusters will be randomly allocated, with a 1:1 ratio, to implement the intervention at either project month 1 or 12. The intervention will be evaluated with three cross-sectional surveys, at project months 1 (baseline), 12 (midline) and 24 (endline). The trial intervention period is planned to run for 30 months, from March 2020 – September 2022 (Fig. 1).
Population
The intervention will be delivered at the community level to the whole population of Alfadanga upazilla, Faridpur district. The intervention will be available and accessible to any community member to participate, including health care providers. We will particularly encourage high-risk individuals, those aged over 30 years and those with T2DM to attend, as the groups we expect to benefit most from the intervention. The evaluation will include permanent residents – those who have lived within the study cluster for a minimum of six months, aged 30 years and older. The evaluation will exclude pregnant women.
Intervention
The intervention is community mobilisation through PLA, focused on reducing intermediate hyperglycaemia and T2DM and improving non-communicable disease behaviours and risk factors. PLA works through community groups actively engaging communities in identifying the causes of health problems, and working together to design and implement ways to address these health problems, and reflect on their progress.[16] This approach has been shown to substantially improve maternal, neonatal and child health outcomes,[17] and in the DMagic trial, reduce intermediate hyperglycaemia and T2DM.[14]
The intervention and its theory-base has been previously described.[18–20] Briefly, the intervention involves initiating separate men’s and women’s groups of 20–30 attendees. Separate groups are established to enable participation of men and women that accounts for gender norms and convenient meeting times. Equal numbers of male and female groups are planned within communities, and groups will explore working together after approximately 10 months. The groups work through four phases of PLA: 1) problem identification where participants identify and prioritise causes of diabetes and diabetes risk in their community; 2) planning together where groups and communities collectively design strategies to address the causes of diabetes that can be implemented by communities; 3) strategy implementation; 4) participatory evaluation of the strategies which they have implemented. The four-phase cycle is planned to take 18 months, with groups meeting once a month. Those attending groups are encouraged to share information in communities and households, and actively engage non-group attenders in their activities.
We plan for nine men’s and nine women’s groups in each cluster (n = 216 groups), for a coverage of 1 group per 200 adults aged 30 years and above. Groups will be led by trained and salaried facilitators, with one male and one female facilitator in each cluster; they will facilitate approximately nine meetings per month. All facilitators will have a minimum of secondary school education and will be recruited from the intervention areas. A 1-week training will cover group facilitation and basic health messages related T2DM prevention and control. The facilitators will be provided with a community action manual, picture cards and flip charts developed through formative research during the DMagic trial. These materials are aligned with standard recommendations for the prevention of T2DM provided by the Diabetic Association of Bangladesh. Male groups will be led by a male facilitator and female groups will be led by a female facilitator. Facilitators will be mentored and supported by two participatory group coordinators and a District Manager, based in Alfadanga, and a Senior Group Intervention Manager based in Dhaka. At the end of 18-months, a group volunteer will be trained in facilitation and groups will be encouraged to continue meeting.
Control
During control periods, communities will receive standard diabetes prevention and care in accordance with usual practice in the region.
Outcome
The primary outcome is the combined prevalence of intermediate hyperglycaemia (i.e. impaired fasting glucose and/or impaired glucose tolerance) and T2DM among adults aged 30 years or older, based on WHO definitions (Table 1).[21] This is the same definition used in the DMagic trial.
Secondary outcomes include: self-awareness of diabetes status; smoking prevalence; physical activity; mean population diastolic and systolic blood pressure; prevalence of hypertension (systolic blood pressure ≥ 140 mmHg or a diastolic blood pressure ≥ 90 mmHg or current treatment with antihypertensive medication); mean population body mass index; prevalence of overweight and obesity; mean population waist and hip circumference ratio; consumption of sugar; dietary diversity; knowledge of diabetes symptoms and complications; utilisation of diabetic services; psychological distress; and ability to self-manage among diabetics. The cumulative incidence of T2DM among individuals with intermediate hyperglycaemia, identified during the survey at month 1 and followed up in subsequent surveys (at 12 and 24 months), will be calculated.
Table 1
WHO definition of normoglycaemia, intermediate hyperglycaemia and T2DM [21]
Definition | Diagnostic Criteria |
Normoglycaemia | Fasting plasma glucose ≤ 6.0 mmol/l |
Intermediate Hyperglycaemia | Impaired Fasting Glucose | Fasting plasma glucose ≥ 6.1 mmol/l to < 7.0 mmol/l AND two-hour post ingestion of 75 g glucose load plasma glucose < 7.8 mmol/l |
Impaired Glucose Tolerance | Fasting plasma glucose < 7.0 mmol/l AND two-hour post ingestion of 75 g glucose load plasma glucose ≥ 7.8 mmol/l to < 11.1 mmol/l |
Type 2 Diabetes Mellitus | Fasting plasma glucose ≥ 7.0 mmol/l OR* two-hour post ingestion of 75 g glucose load plasma glucose ≥ 11.1 mmol/l |
Randomisation
Due to the nature of the intervention, communities cannot be blinded to cluster allocation. Informed community consent will be taken through discussions with local leaders. Thereafter, clusters will be randomly allocated at a public meeting in the presence of traditional leaders and representatives from communities. The name of each cluster will be written on pieces of paper and folded uniformly by study staff, placed into a container, and then drawn by a community representative. The clusters will be numbered in the order they are drawn, with clusters 1–6 implementing the intervention at month 1, and clusters 7–12 at month 12. The randomisation process will be filmed, with the consent of those present, to document the procedure.
Sampling
Participants in the impact evaluation will be selected using a three-step sampling approach. Prior to randomisation, evaluation villages used for all cross-sectional surveys will be purposively selected using the following criteria: they do not sit on a border with a neighbouring study cluster; they are not a major trading centre or administrative centre; they have a minimum of 50 households. Between two and five villages will be selected, to achieve a total of 800–1000 households per cluster based on estimated population sizes from the 2011 Bangladesh census, and confirmed through rapid study census (Table 2, Fig. 2).
A sampling frame of households with at least one eligible resident, aged over 30 years, within the evaluation villages will be generated, based on a study census completed in January 2020. A sample of 110 households will be selected using simple random sampling, with a random number generator in Stata SE 14. At the next stage, a single eligible adult will be selected from each sampled household, using simple random sampling. At each survey, a new sample of households and individuals will be generated with replacement from the same sampling frame, i.e. some individuals and households may be sampled more than once by chance. Individuals identified with intermediate hyperglycaemia in the baseline survey will be purposefully sampled for inclusion in subsequent surveys.
Table 2
Estimated population per cluster, derived from the 2011 Bangaldesh population census
Study cluster | Estimated cluster population | Estimated households in evaluation villages |
Alfadanga 1 | 9891 | 923 |
Alfadanga 2 | 8827 | 1005 |
Bana 1 | 9345 | 970 |
Bana 2 | 8421 | 867 |
Buraich 1 | 11386 | 1030 |
Buraich 2 | 13047 | 813 |
Golpalpur 1 | 10539 | 775 |
Golpalpur 2 | 9880 | 927 |
Panchuria 1 | 9257 | 706 |
Panchuria 2 | 9842 | 911 |
Tagarbanda 1 | 9628 | 812 |
Tagarbanda 2 | 10450 | 736 |
Sample size
The sample size was calculated using a presumed baseline combined prevalence of intermediate hyperglycaemia and T2DM of 30%, an assumed exchangeable correlation structure across surveys for each cluster, and an estimated intra-cluster correlation of 0.07. A sample of 110 participants per cluster, per survey will give 84% power at 95% confidence to detect a minimum of a 30% relative reduction in disease prevalence to 21% or less. An additional 10% was added to the sample size to allow for non-response, resulting in a total target sample of 3,960. We will, in addition, include individuals identified with intermediate hyperglycaemia in subsequent surveys. We estimate 360 individuals will be identified as having intermediate hyperglycaemia at baseline, and included in the month 12 and 24 surveys.
Data Collection
Data will be collected through three cross-sectional surveys at month 1, 12 and 24 (Fig. 1), by six pairs of fieldworkers (one male and one female) with three supervisors. Data collectors with completed secondary education will be recruited locally, and selected following written assessment and interview. All fieldworkers will undergo two weeks training, including: consent, survey methods and physical measurements, followed by one week supervised field practice in villages outside of the study area. During training and field practice, survey tools will be piloted and revised. Within each village, teams will be aided by a village assistant, who will receive a daily payment to coordinate communication with study participants and assist data collectors. They will inform sampled individuals about the time and location of data collection, including the need to fast overnight. A convenient location within the village will be selected for the blood and anthropometric measures and collection of questionnaire data will take place at the respondent’s home. All participants will be provided with verbal and written study information prior to taking part in anthropometric assessments and surveys, and data will only be collected following signed consent.
Individual, interviewer-administrated questionnaires will collect information on demographic and socio-economic characteristics, risk behaviours, diabetes awareness indicators, health seeking and costs of care. This questionnaire will be adapted from: the WHO Stepwise tool[22] and the 2014 Bangladesh Demographic and Health Survey,[15] and will include GAD-7 and PHQ-9 for mental health assessment,[23, 24] and an adapted Appraisal of Diabetes Scale.[25] Blood pressure will be measured using the OMRON HBP 1100 Professional Blood Pressure Monitor (Kyoto, Japan). Two measurements will be taken with a 5-minute interval, and the average taken for analysis. Height, weight, and waist and hip circumference will be measured, asking respondents to wear only light clothing and removing shoes. Blood glucose will be measured using the OneTouch Verio Flex Glucometer (Lifescan,Inc.,Milpitas,CA 95035) in mmol/L, from whole blood obtained by finger prick. Participants will be tested following an overnight fast of 8–12 hours, and then again 2-hours after receiving a 75 g glucose load dissolved in 250 ml of water. In self-reported diabetics, we will take a random blood glucose measurement only.
Data will be collected using ODK Collect on Android tablets, and downloaded via USB onto field supervisor laptops and shared via a secure cloud server. Clusters, villages, households and individuals will be linked using unique study identifiers. Personal identifiable data will be collected in order to identify sampled individuals in the community during surveys, however these will be removed following cleaning for analysis and archiving. Data quality will be promoted through: pilot testing of data collection tools; in-built range and logic checks; supervisor observation of data collection; concurrent data checking, with errors sent back to the field team for verification.
Analysis
The primary analysis will be an intention to treat analysis, at the individual level, comparing the combined prevalence of intermediate hyperglycaemia and T2DM between intervention and control periods across all clusters. Analysis will be based on regression models that include fixed effects for intervention exposure, survey timing, and sociodemographic predictors, and random effects for geographical clustering and, where appropriate individual and household clustering (e.g. if the same individual or household is randomly sampled in two surveys). Participants with missing data on the primary outcome will be excluded from analysis, as we anticipate high levels of participation. All estimates of the intervention effects will be presented with 95% confidence intervals.
Outcome data in the intervention condition are collected at 12 and 24 months after the intervention is first implemented (Fig. 1). The primary analysis will however estimate one effect of the intervention, which assumes the effect of the intervention is the same after 12 and 24 months. To explore this assumption, we a priori plan to report the estimated intervention effects at these two durations separately, fitting a model that includes two intervention duration indicators. Intuitively we can see the midline data allow the intervention effect after 12 months relative to control to be estimated, and the endline data allow the effect after 24 months relative to 12 months to be estimated. Combining these two will give the effect of the intervention after 24 months relative to control. Assuming that the intervention effect is the same at 18 months, when the intervention activities are completed, and 24 months (i.e. it does not ‘tail off’ once groups no longer have a paid facilitator), then the 24-month effect can be interpreted as the ‘full’ intervention effect.
We will also explore subgroup analyses, by socio-economic status, gender, age, and for both intermediate hyperglycaemia and T2DM separately. As we have a small number of clusters involved in the trial, we are considering the use of random effects models with a correction for the modest number of clusters. This is an active area of methodological research, and a detailed final analysis plan will be written before analysis begins.
Process Evaluation
We will conduct a mixed methods process evaluation following the UK Medical Research Council guidelines to describe intervention implementation and develop the theory of how the intervention can affect health outcomes in this context.[26] The intervention and trial design has used the extensive formative and process evaluation research from the DMagic trial.[19, 20] A process evaluation officer will collect qualitative data on how the intervention is working through: 1) group observations throughout the intervention; and 2) semi-structured and key informant interviews and group discussions in case study communities at two time points during the intervention. We will develop our hypothesis that health literacy is improved by the intervention[20] by integrating questions on health literacy in our survey tools.[27] We will also develop our hypothesis about intervention effect on diabetes stigma through qualitative research and four quantitative questions asked to self-reported diabetics in the individual questionnaires. Quantitative data to describe intervention implementation will be collected from monthly reports of group supervisors who will observe an average of 33 meetings per month.[19]
Economic Evaluation
Cost and cost-effectiveness analyses will be conducted from both the health system and societal perspectives. Costs of implementing the intervention will be collected prospectively from the project accounts and input into a customized excel-based standardised costing tool designed for this purpose. Data on utilisation of NCD services from healthcare providers and costs of seeking care will be collected from the study participants during individual surveys. Incremental cost-effectiveness ratios (ICERs) will be evaluated in terms of the cost per case of intermediate hyperglycaemia and T2DM prevented and Disability Adjusted Life Years (DALYs) averted. Robustness of the cost-effectiveness results will be assessed through a series of sensitivity analyses. In addition to cost-effectiveness analysis, an equity-impact analysis will be conducted in order to assess whether the intervention has improved the equity of health service delivery and is improving the health status of those most in need. All costs will be estimated both from a financial and economic perspective and presented in International Dollars.
Management And Oversight
The project is a collaboration between The Diabetic Association of Bangladesh and University College London. A Trial Steering Committee will be convened according to the DAMOCLES charter, with the aim to meet a minimum of once a year during the trial.[28] Their role will be to provide an independent, objective review of the study implementation and baseline data, and to advise on any extension or modification of the trial design. There are no stopping rules as we do not expect the intervention to have adverse effects at either cluster or participant level and have not planned an interim analysis to check for harm or futility.
We will also establish community advisory boards, which include community representatives from Alfadanga. This will provide advice on the research design, implementation and to support effective communication, and support feedback of adverse community events. These advisory boards will be consulted at key time points (e.g. group establishment and dissemination), and when questions or challenges arise (e.g. community conflicts).
Dissemination
We plan to disseminate the impact, economic and process evaluation findings to an academic audience, policy-makers and study communities at the end of the project. Depending on the planned audience, we will hold meetings, workings and share reports, as appropriate. We will also publish results in open-access peer-reviewed journals.