Study design

We conducted a 24 week parallel-arm, randomised controlled trial comparing a whole-food, plant-based intervention with moderate exercise (PB+Ex) with SMC in adults with type 2 diabetes. The study was approved by the institutional review board at Loma Linda University (protocol no. 59105) and an ad hoc institutional review board assembled by the RMI Minister of Health.

Study population

We enrolled adults with type 2 diabetes aged 18–75 years. Participants either had an HbA1c ≥64 mmol/mol (≥8.0%) or were diagnosed with type 2 diabetes and taking glucose-lowering medication. Exclusion criteria included heart disease, changes in glucose-lowering medications in the past 3 months and physical or medical conditions that hinder participation. Enrolment was generally representative of Marshallese adults with type 2 diabetes (see the electronic supplementary material [ESM]). Participants provided written informed consent before participating, and demographic variables including biological sex were collected by self-report. Participants were enrolled in five cohorts and randomised in a 1:1 ratio generated by Microsoft Excel. Participants were randomised a few days before baseline data collection to give them time to make accommodations to attend weekly classes. Further details on recruitment, randomisation, the lifestyle intervention and the methods are provided in the protocol manuscript [32].

Standard care

The control group was treated using glucose-lowering pharmacotherapy, according to SMC in the RMI. They were instructed to maintain their current diet and physical activity levels.

PB+Ex intervention

The PB+Ex intervention is described in detail in the protocol manuscript [32]. In brief, the PB+Ex group was instructed to eat a WFPB diet permitting minimal animal products and to exercise 30–60 min/day for 24 weeks. During weeks 1–12, the PB+Ex group received prepared meals, attended group exercise sessions and received group instruction on eating healthfully, cooking, exercising and managing stress (see ESM Table 1 for a list of class topics). The intervention was culturally tailored and developed in partnership with the Marshallese government and local diabetes clinics and included Marshallese staff and popular foods (see the ESM for a detailed description of the cultural adaptation and positionality statements). The intensity of support progressively decreased, with participants attending 15–21 h/week of group classes in weeks 1–2, 8–10 h/week in weeks 3–6 and 4–5 h/week in weeks 7–12. During weeks 13–24 (the follow-up phase), participants were instructed to follow the intervention on their own.

WFPB diet

The prescribed diet was high in fibre (35 g/ 4184 kJ), low in fat (20–25% of energy; saturated fat <7% of energy), moderate in protein (10–15% of energy) and low in sodium (<2400 mg/day). During weeks 1–2 (the intensive phase), PB+Ex participants received 12 prepared meals/week and were instructed to consume no animal products and minimal ground grains and refined carbohydrates. Thereafter, participants received 2 meals/week during weeks 3–6 and 1 meal/week during weeks 7–12. During weeks 3–12, participants could consume small amounts of animal foods, oils, fat-rich foods and processed foods, following a four-tiered food classification system [33]. Specifically, they were instructed to consume 75–100% of energy from whole, unprocessed plant foods (tier 1), such as vegetables, legumes, whole grains and fruit. The remainder of their diet could include ≤25% lightly processed foods (tier 2), ≤10% moderately processed foods and moderate-fat animal products (tier 3) and ≤5% heavily processed foods and high-fat animal products (tier 4).

Exercise

The PB+Ex group was instructed to do moderate-intensity aerobic and resistance exercise 60 min/day during weeks 1–2 and 30–60 min/day during weeks 3–24. During weeks 1–2, participants attended 1 h group exercise classes 4 days/week. Thereafter, they attended group exercise classes twice a week during weeks 3–6 and once a week during weeks 7–12. Participants were also counselled to walk 10–20 min before breakfast and after lunch and dinner.

Cohort differences

To increase intervention intensity, participants in cohorts 3–5 of the PB+Ex group repeated weeks 1 and 2 during weeks 4 and 6. This included a repeat of educational sessions, meals provided and exercise classes.

Study outcomes

Outcomes were assessed at weeks 0, 2, 6, 12 and 24. The primary outcome was glycaemic control, measured by HbA1c, fasting glucose, fasting insulin, HOMA-IR and diabetes medication use. Secondary endpoints were cardiovascular risk factors, including body weight, waist circumference, lipids, systolic blood pressure (SBP), diastolic blood pressure (DBP), resting heart rate, high-sensitivity C-reactive protein (hsCRP) and cardiovascular medication use. Only lab analysts performing the serum assays were blinded.

Serum chemistry

HbA1c, glucose, insulin, total cholesterol, HDL-cholesterol, triglycerides and hsCRP were analysed blinded at the Clinical Laboratory Improvement Amendments-approved laboratory in the Ministry of Health’s Hospital, while LDL-cholesterol was calculated using the Friedewald equation. Triglyceride values >4.52 mmol/l (>400 mg/dl) were Winsorised to minimise the effect of outliers on the analyses. LDL-cholesterol values were treated as missing whenever triglyceride values exceeded 4.52 mmol/l. hsCRP values ≥95 mmol/l (≥10 mg/l) were considered indicative of acute infection and treated as missing.

Medication use

Primary care physicians and/or the Diabetes Wellness Clinic’s clinicians adjusted participants’ medications based on glucometer and/or serum glucose values. PB+Ex participants on insulin were monitored daily with glucometers and instructed to reduce insulin doses when glucose fell to <3.9 mmol/l (<70 mg/dl) or hypoglycaemic symptoms manifested. For SMC participants on insulin, their physicians were responsible for adjusting their medication doses. Diabetes medication use was quantified using the medication effect score (MES) [25], which estimates the HbA1c reduction expected from all glucose-lowering pharmacotherapy [34]. Diabetes remission was defined as achieving HbA1c <48 mmol/mol (<6.5%) after not using glucose-lowering medications for at least 3 months.

Statistical power

The required sample size was estimated using the variances observed in HbA1c and glucose in cohorts 1–2. A sample size of n=120 was needed to have 80% power to detect a 1.1 mmol/l (20 mg/dl) difference in glucose and an 11 mmol/mol (1.0%) difference in HbA1c, given α=0.05.

Statistical analyses

Analyses were performed with two-sided tests and α=0.05, primarily using SAS (version 9.4; SAS Institute; Cary, NC, USA). Baseline data were compared using independent samples t tests or the Mann–Whitney U test if neither raw nor transformed values were normal. The main analysis was intention-to-treat. Continuous data were analysed using linear mixed models, adjusting for baseline values, sex and/or cohort whenever statistically merited. Categorical data were analysed using Fisher’s exact test. Missing medication doses were singly imputed whenever a missing dose was flanked by two timepoints with identical doses and were otherwise treated as missing. When HbA1c values were missing, remission status was singly imputed by assuming HbA1c values changed by no more than 16 mmol/mol (1.5%) between weeks 0 and 2 and by no more than 33 mmol/mol (3.0%) between each subsequent pair of timepoints. Insulin and HOMA-IR were analysed only in participants not on insulin, while MES was analysed only in those taking glucose-lowering medication. The proportion of participants who decreased their medication doses was calculated in the subgroup of participants on the medication(s) at baseline. Lastly, to calculate diabetes remission and decreases in medication doses, we included everyone with sufficient data at either week 12, week 24 or both timepoints, in order to increase the sample size and improve accuracy in the estimated proportions.