Peer-modeled Intervention

Theoretical framework The proposed study utilises the Social Cognitive Theory (SCT). The SCT, widely utilised, shows the dynamic interaction among a person's behaviour (skills, practice and self-efficacy), environmental determinants (social norms, influence on others and access in the community), and personal factors (knowledge, expectations and attitudes), and how those factors change or affect one another. The current study targets personal factors and individual behaviours, given their importance, including their changeability and literature-based relevance.

In line with the SCT cognitive factors, the awareness component aims to increase knowledge of cardio-metabolic health and risk factors, health expectancies from behaviour change practices, and attitudes towards body size perception in relation to health.

Physical literacy comprises elements such as motivation and confidence, physical competence, and knowledge and understanding of the value of, and responsibility for, engaging in physical activities for life. Physical literacy aims to increase knowledge and understanding of the health benefits of engaging in daily physical activity as an integral part of one's life choices. This is based on the behavioural factors of skills, practice and self-efficacy of the SCT.

Food literacy is a collection of inter-related knowledge, skills and behaviour required to empower in the five core domains; i) planning for food through making feasible decisions; ii) select of quality food; iii) food preparation of tasty and hygienic meals; iv) and eat food to meet individual needs and v) ability to screen information from different platforms on healthy eating. Food literacy aims to increase knowledge and skills for consuming and preparing healthy meals that improve cardio-metabolic health. The basis to achieve food literacy is the behavioural factors of skills, practice, and self-efficacy within the SCT.

The study's primary objective is to assess the effect of a complex behavioural change intervention on the central adiposity of women of reproductive age.

The secondary objectives include:

• To assess the dietary changes of the women.
• To assess the physical activity changes of the women.
• To assess the psycho social outcomes of the women.
• To identify theory mediators for optimal cardio-metabolic health.
• To assess participants' adherence to the intervention.
• To assess participants' satisfaction towards the content and the peer modelling strategy.
• To assess the delivery accuracy of the intervention content by the peer models.

Study design

The study will be a cluster randomised 2-arm trial. The villages are clusters of distinct community units served by a specific Village Health Team (VHT).

The sample area (Kakoba Ward) comprises 12 villages, randomised into the intervention and comparison groups in a 1:1 ratio.

The study will include an active intervention phase lasting 3 months (12 weeks) and a follow-up post-intervention phase also lasting 3 months (12 weeks), totalling 6 months (24 weeks).

Study setting

The study will be conducted in Mbarara City, located in the Southwestern region of Uganda.

Mbarara City was chosen for the current study because;: i) 27% of the women are diagnosed with overweight or obesity within the region; ii) Behavioural intervention studies have been concentrated in the central region of Uganda, leaving a gap in other regions like the western region; iii) Mbarara City is among the newly made cities in Uganda, thus a given opportunity for a high urbanisation rate and food deserts; iv) Following our cross-sectional study (study reference number HS4708ES), it is imperative to conduct a behaviour change intervention that can optimise the cardio-metabolic health of the women in Mbarara City.

Study participant recruitment

With support from the village health teams, the investigators will advertise the study by distributing flyers with study details within the study area. Interested participants will be screened for eligibility within each cluster. In order to prevent cross-contamination:

• Eligible participants per cluster will not be randomised further.
• Clusters will be geographically separated to reduce interactions between participants from different study arms.
• Intervention delivery will be conducted at the cluster level through designated peer models who will be trained exclusively for their assigned study arm and instructed not to share material or information with different cluster peer models.
• Outcome assessments will be standardised and conducted independently of intervention delivery to minimise bias.

Sample size calculation

The sample size is calculated based on the Rutterford methods for sample size determination in cluster randomised trials. The parameters to consider in the sample calculation are;

• The Z value of 1.96 at a type 1 error of 5% with a power of 80%.
• The standard deviation of 15.58 for the mean waist circumference is based on our cross-sectional study (findings yet to be published).
• The effect size is 7.4 cm based on the Mediterranean diet for the central obesity study.
• The n value is 15 participants per cluster.
• The intra-cluster correlation coefficient, r = 0.001, is based on previous studies.

Thus, the sample size is calculated at 70 participants per arm. Considering a 10% attrition rate, 77 participants are required. Hence, the study aims to include 154 WRA in the trial with 77 participants per study arm.

Study delivery

The study will be delivered in 3 phases, i.e.,

1. Phase 1: The selection and training of the peer models that will deliver the intervention. With guidance from the village health teams, the investigators will select a peer model for each cluster. The following are the inclusion criteria of a peer model;

• Female
• A balance of married and unmarried
• Have at least a secondary education
• Be willing to lead a group
• Have social credibility and be accepted by the group members
• Relatable to the participants regarding their journey on optimising cardio-metabolic health
• Are currently practicing behavioural changes regarding dietary and physical activity practices. The peer models will go through a block of training, with the first session focusing on knowledge and skills in facilitating groups and health education, key cardio-metabolic concepts, measurement using a waist circumference tape measure, and sharing life experiences. In the second training block, the peer models will share their experiences, points for improvement, and support needs. There will be enhanced training in the areas they identify as needing more skills and knowledge.

1. Phase 2: Community inauguration This phase will include selecting participants in accordance with the eligibility criteria, followed by a briefing session on the intervention for the partners of married participants or key family heads. The purpose is to introduce the intervention to increase its acceptability and adherence. The goal is to enhance social support within the participant's home environment. The activities of this phase are;

• The peer models will share their experiences, including the barriers, solutions, and facilitators to changing their lifestyles for optimal cardio-metabolic health.
• The peer models will share the infographics with key messages about the study and what is required from the participants.

1. Phase 3: Intervention Participant intervention where both the intervention and comparison arm intervention components will start.

Intervention components

Intervention components will be delivered through a designed toolkit by the peer models in group sessions. The intervention components are;

1. Awareness. This will include the following activities;

• Group discussion session on perceptions of ideal and healthy body sizes, guided by body image silhouettes.
• A practical session on waist circumference measurement.
• Group discussion on the risk of cardio-metabolic health in relation to waist circumference.
• Group discussion to compare the measurements with the earlier discussed body perceptions and body silhouettes.
• Group discussion and illustrate doable actions by conducting lifestyle physical activity mapping and food sorting for optimal cardio-metabolic health.
• Peer models will share their experience and journey on what they are doing for optimal cardio-metabolic health.

1. Physical literacy This includes the following activities

• Group discussion on physical activity mapping in order to identify the different physical activities as lifestyle and recreational.
• Group discussion to differentiate between a sedentary lifestyle, moderate and vigorous physical activities.
• Group discussion on barriers to engaging in physical activity.
• Group discussions on physical activity and metabolic health based on global recommendations.
• Illustration of brisk walking as a proposed lifestyle physical activity.
• Weekly goal setting using a designed pluck card.
• Pairing of participants within the groups for social support.

1. Food literacy Food literacy will include the following activities;

• Meal planning by emphasising the availability and accessibility of fruits and vegetables in the home environment, with less fatty snacks.
• Planning for simple vegetable recipes
• Illustrations on budgeting for food.
• Group discussion on selecting the different food sources and healthier eating when away from home, especially for retail business women who eat out in restaurants.
• Conducting hands-on meal preparation sessions utilising what is accessible to the community under safety and hygiene principles, and exploring different healthy preparation and cooking methods, such as boiling. Therefore, the peer models will hold cooking demonstrations on preparing vegetables, fruits, and lower-fat foods to enhance food preparation skills.
• Using the plate model, healthy meal choices and portions will be emphasised. Each Participant will be encouraged to use it for their main meal. The investigators will create a food poster featuring the model plate and typical foods in Mbarara.
• Group discussion on screening of relevant information regarding healthy eating within the environment by mapping the information sources, the trusted sources and the relevancy of the information with the guidance of infographic facts. Intervention and follow-up phases During the active phase, the intervention components will be delivered weekly for the first 12 weeks. During the post-intervention period, there will be no contact between the peer models and the participants for 12 weeks. Data collection and measurement First, data will be collected utilising the WHO STEPwise instrument. The instrument consists of three steps. Step 1: Data on socio-demographic characteristics and lifestyle are collected.
• The socio-demographic characteristics include age, education level, marital status, employment status, child-bearing status, household size, religion, ethnic background, household expenditure and residence area.
• The lifestyle commodities collected include dietary intake and physical activity. The investigators will collect dietary intake data using a non-consecutive repeated 24-hour recall to capture the WRA's habitual food intake over the previous 24 hours on the two days of the visits. The investigators will follow the USDA automated multi-pass method in completing the 24-hour recall dietary assessment tool. The investigators will assess food literacy utilising the validated adult scale for East Africa. The investigators will assess for physical activity utilising the Global Physical Activity Questionnaire. In step 2, physical measurements are conducted to include Anthropometry measurements
• Waist circumference (cm) to the nearest 0.1 cm using a non-stretchable measuring tape.
• Height (cm) to the nearest 0.1 cm using a height board.
• Weight (kg) to the nearest 0.1 kg using a Seca digital weighing scale. Each measurement will be taken twice, and the final record will be an average of the measurements.
• Arterial stiffness The investigators will assess arterial stiffness using arteriography and record data on aortic pulse wave velocity, augmentation index, and central blood pressure. First, the investigators will measure the participants' blood pressure. After resting for 10 minutes, their pulse wave measurements will be conducted.
• Fat mass The investigators will utilise the complex body composition assessment model for total body water, fat free mass, fat mass, and muscle mass using a BodyStat 500 bioelectrical impedance analyser. In step 3, biomedical measurements
• Fasting blood glucose levels using a CardioCheck Plus.
• Total cholesterol, triglycerides, and low and high-density lipoprotein cholesterol using a cholesterol meter. The investigators will obtain blood for glucose and lipid level assessment by finger prick. These measurements will be taken based on the WHO reference for the STEPwise survey. Additional data collection Mediator measures
• Body image The investigators will assess respondents' perceptions of their current and ideal body size, along with supplementary questions on body size perceptions, using the body image silhouette. The investigators will use one dimension of body image, body size dissatisfaction, utilising the body image assessment questionnaire, which consists of 18 silhouettes of increasing size. Participants will be asked to choose their preferred and current body sizes. The difference will be used to measure dissatisfaction.
• Self-efficacy Self-efficacy for physical activity will be measured with a 10-item Likert self-efficacy for exercise behaviours scale. Likewise, the weight efficacy lifestyle questionnaire, in which higher values indicate greater belief in weight management actions such as eating.
• Self-monitoring The self-monitoring tools will be assessed on usability/adherence, completion and compilation every week.
• Social support To assess social support, the Sallis scale for diet and exercise behaviours will be utilised. Process evaluation The investigators will utilise subjective, qualitative, and quantitative methods to conduct a process evaluation of the peer models' delivery. Group discussions will be conducted to assess the delivery of the intervention by both participants and peer models. With the participants, the investigators will assess;
• The acceptability of the intervention components and peer model strategy, which of the components are most relevant and can be integrated later in their daily routines, and suggestions on how to improve the intervention contents. This evaluation is based on the UK Medical Research Council framework for process evaluation of complex interventions, focusing on implementation, intervention delivery and outcomes.
• The participants' adherence to the intervention. All participants will be required to measure and record their waist circumference before each session to ensure attendance is valid. Additionally, the investigators will record the number of sessions and the time spent in the sessions.
• To assess the participants' satisfaction towards the content and peer modelling strategy, a questionnaire will be administered post-intervention. Content satisfaction will be assessed using the following factors: time, place for peer sessions, module objectives, content, structure, and relevance. Furthermore, leadership scores on program delivery, peer interactions, peer model readiness, and peer knowledge will be incorporated. A Likert scale with scores ranging from 1 (strongly disagree) to 5 (strongly agree) will be utilised.
• Regarding the accuracy of content delivery by the peer models, the research assistants will be trained as observers during the sessions. The research assistants will take notes of the sessions and utilise a checklist to score whether the content was delivered as planned and accurately. The checklists will contain the learning objectives and activities that the peer model is expected to cover with the participant during each session.
• Following a session with the participant, the peer models will document the objectives covered. The content not covered in those specific sessions will be noted, and the reasons will be provided.
• The sessions will be randomly audio-taped and transcribed for accuracy and delivery. These will be matched to the study design's original content. Accuracy is considered a multifaceted concept and encompasses whether the objectives and activities were delivered at all, in sufficient detail, and in a technically correct manner. An omission in any of these areas is noted as inaccurate. Statistical analysis plan
• All analyses will be performed according to the intention-to-treat principle.
• RStudio or IBM SPSS (Version 23) will be used for analysis.
• Frequencies and percentages will be used to describe summary statistics for categorical data.
• Means and standard deviations (mean ± SD) will be used for continuous, normally distributed data. Skewed continuous data will be presented as median and interquartile range.
• Missing covariate values will be imputed using multiple imputation methods. -Multicollinearity between explanatory variables will be assessed using tolerance values (TOL) < 0.1 and variance inflation factors (VIF) < 10 in a linear regression step.
• The fruit and vegetable servings will be converted to grams using WHO guidelines and calculations. The calculated grams will be categorised as below 400 g or above the minimum recommended value of >400 g. The food items collected from individuals will be converted into nutrients by multiplying the grams of food portions by the nutrient content per food item using the Harvest Plus food composition database for central and eastern Uganda, the Kenyan, and the USDA food composition databases. These will then be calculated into the daily energy intake (kcal), carbohydrates, total fat, protein and dietary fibre.
• To understand the association between the dietary intake of the target population and the cardio-metabolic profile, the investigators will utilise a classificatory approach. Respondents will be classified into quintiles based on their intake of specific nutrients. The investigators will calculate their relative risks for each lower quintile, with the uppermost intake quintile as the reference.
• The investigators will obtain the total food literacy score by aggregating individual scores for each question on a four-point Likert scale (1=strongly disagree to 4=strongly agree). The scores will be recalculated as percentages, with 0 indicating a low level of food literacy and 100 indicating the highest level.
• The physical activity levels of the participants will be categorised as minimally active if they fall within the range of >600 to <3,000 metabolic equivalents to task (MET). Participants with levels below 600 MET are considered physically inactive, while those above 3,000 MET are classified as engaging in health-enhancing physical activity.
• The height and weight will be converted to body mass index (BMI), where BMI ≥25.0kg/m2 is overweight and ≥30.0kg/m2 is obesity.
• A large waist circumference (>80 cm in women) and high blood pressure (≥130/85 mm Hg) will be considered risk factors for cardio-metabolic syndrome.
• The metabolic risk factors considered for categorisation are: raised fasting blood glucose of ≥100 mg/dL, elevated triglycerides of ≥150 mg/dL, and low HDL-C <50 mg/dL.
• Cut points for fat mass in women at <35 kg will be considered.
• A pulse wave velocity of over 10 m/s is associated with an increased risk for cardiovascular disease.
• To assess the intervention effect at the end-line (3 months), the investigators will utilise a multilevel linear mixed-effects model with treatment as a fixed effect and villages/cells as random effects. The change in the outcome at the endline will be entered as the dependent variable. The investigators will model;
• The waist circumference is the primary dependent variable.
• The blood glucose, lipid, cholesterol levels, fat mass and arterial stiffness measures are the secondary dependent variables.
• The covariates are the lifestyle variables (physical activity levels and fruit and vegetable intake) and socio-demographic characteristics.
• To assess whether the intervention effects were sustained at post-intervention follow-up, a multilevel linear mixed-effects model will be built with time and the interaction of time and treatment as fixed effects. The villages and participants will be included as random effects, with the change in the outcome (from baseline to post-intervention follow-up) as the dependent variable. Statistical significance is set at 5%. Results will be reported as intervention effects and 95% confidence intervals (CI).
• The body image silhouettes will be classified as underweight (1-5), normal (6-9), overweight (10-13) and obese (14-18). The different perceptions based on the questions will be presented as frequencies or means. Paired t-tests and effect sizes will be used to assess any changes in these variables. To evaluate whether these changes are associated with the optimised cardio-metabolic health, Pearson correlation coefficients will be calculated. To understand the predictive value of treatment-related changes, multiple linear regression and analysis of covariance will be used.
• Likewise, the residual changes in self-efficacy, self-monitoring, and social support will be used as the method of change. These predictors will be adjusted for age, study arms and the baseline values of the primary outcome measures. The results of each predictor will be reported as standardised regression coefficients with p-values from the F-test. A p-value of <0.05 will be accepted as significant. The investigators will utilise the Bonferroni correction to correct for type 1 error.
• As part of the process evaluation, the investigators will thematically analyse the focus group discussions on participant acceptability and satisfaction using grounded theory.
• To evaluate for adherence, the recorded Waist circumference measures will be summed up and a total score obtained.
• The leadership scores will be analysed based on the Internal consistency of the items. Hence, determined using Cronbach's Alpha (α), where α > 0.70 denotes good internal consistency. The sum of responses for each domain (content and peer modelling) reflects participants' satisfaction, with higher scores indicating greater satisfaction. The sum will then be converted into percentages by dividing with maximum score.
• To determine how the peer models delivered the intervention content, the recorded sessions will be scored using the session content checklists. Each score will be computed as a percentage of content delivered. The scores will be averaged for each session.
• To determine how accurate the peer models were in presenting the intervention component information, the mean accuracy will be calculated by dividing the total number of items rated as accurate by the total number of items in the session checklists.
• Using the completed session content checklists, two outside raters will independently rate the comprehensiveness of the content delivered by listening to the audiotapes. Cohen's kappa statistic (k), a chance-corrected measure of agreement, will be used to assess interrater reliability among objective raters across random sessions. For the purposes of this study, if an item on a checklist is not covered in full, it is not considered delivered; it will be marked on the checklist. At a minimum, the peer models must mention the terms or keywords of a checklist item so that raters can consider whether the content of that item has been delivered. However, the raters will give peer models credit for covering the content of an item, even if the peer models provided some inaccurate information.