We conducted a single-site, randomized controlled pilot trial to determine the feasibility and estimate the effect size of an intervention which included 3 months of MTM and 6 months of MNT. This pilot study is reported according to CONSORT guidelines—pilot trial extension and is registered on ClinicalTrials.gov (NCT-NCT04034511).

We recruited English-speaking adults aged 18 and over who were enrolled in and identified by a Maryland Medicaid Insurance plan (91%) or from a clinical database (8.5%) or by direct referral from a provider (< 1%) as having T2DM with the most recent A1c level greater than 8.0%. Participants were required to have a refrigerator/freezer in the home to store meals and be staying in the area for the next 12 months. Individuals with medical conditions requiring special or additional dietary requirements (e.g., advanced chronic kidney disease with eGFR < 30 or current pregnancy), swallowing difficulties, significant food allergies and active substance use disorder were excluded.

Participants were recruited via mailed study flyer with opt-in and opt-out options. Ten days after the mailing, participants were called by Medicaid plan staff to assess interest and confirm basic eligibility. Participants were then connected directly to the study team who conducted full assessment of eligibility over the phone, followed by in-person visit. Partway through the study, we additionally took direct referrals from primary care clinicians who could contact the study team for eligibility assessment.

The study began in January 2020, with data collection completed in April 2023. Ten participants were enrolled and randomized before the onset of the COVID-19 pandemic and had both their intervention and data collection affected by temporary cessation of all study activities imposed by the IRB. Before resuming in September 2020, COVID safety mechanisms were put into place, and the MNT intervention was switched to phone call only (initially there was the potential for home visit assessments).

Participants in this parallel arm designed trial were randomly assigned 1:1 to the active intervention with usual care or usual care alone.

The active intervention was a combination of medically tailored meals (MTM), and medical nutrition therapy (MNT). MTM was managed by Moveable Feast, a Baltimore-headquartered community-based organization, and included 12 medically tailored frozen meals and a fresh produce bag (containing five additional servings) weekly for 3 months. These meals were tailored for a population with T2DM specifically based on the American Heart Association diet, providing 60 g of carbohydrates per meal. Participants were provided instructional materials on how to reheat the meals and how to add flavor to meals without added salt. The first delivery also included a bottle of house-made salt-free seasoning blend. MNT consisted of individual phone calls with a registered dietitian from Moveable Feast monthly for 6 months. These sessions were self-directed and included goal-setting and general education about nutrition, carbohydrates, physical activity, and diabetes-friendly dietary approaches. Usual care was not augmented by the study team in any way. All participants had Medicaid insurance and were encouraged to continue their routine visits with their primary care clinician and/or endocrinologist, who were free to make any changes to their patient’s medications or make other recommendations as they saw fit.

Sociodemographic information was collected at baseline. Additionally, we collected comorbidities, dietary intake, food insecurity, quality of life, and self-efficacy using standardized questionnaires at baseline, 3 months, and 6 months. Weight, blood pressure, and A1c were also measured by the study team using standardized protocols at baseline, 3 months, and 6 months (Supplemental Table 3). Data collection was overseen by trained study team members and entered into Research Electronic Data Capture (REDCap)18,19 or the online ASA24 diet recall platform. In-person data was collected at the institution’s on-site clinical research unit. Health care utilization data, including costs and the standard of care A1c at 12 months, were collected by the Medicaid insurer and transferred to the study team at the end of the study.

The primary outcome was the between-group differences in the change in A1c from baseline to 6 months. A point-of-care A1c was measured during the study visit using the Abbott Afinion AS100 Analyzer machine. We chose 6 months to assess whether the effects of the 3-month MTM were sustained and to evaluate the cumulative impact of the MTM and MNT.

Secondary outcomes included (1) between-group differences in the change in A1c from baseline to 12 months, (2) diabetes-related health care utilization and health care costs for 6 and 11 months post enrollment compared to a similar period prior to enrollment; as well as between-group differences from baseline to 3 and/or 6 months in the change in: (3) the number, type, and dosages of diabetes-related medications, (4) quality of life from the Audit of Diabetes-Dependent Quality of Life (ADDQOL), (5) self-efficacy with diabetes self-management, using the Perceived Diabetes Self-Management Scale (PDSMS), (6) blood pressure, (7) body weight, (8) changes in food insecurity, as measured by USDA’s Household Food Security Questionnaire,20 from baseline to 3 months and to 6 months, and (9) change in diet quality at baseline from baseline to 3 months and to 6 months using the ASA24 diet recall and Healthy Eating Index 2015 Scores (HEI), both total and the 13 individual components.

We summarized baseline characteristics of participants using means for continuous variables and counts/percentages for categorical variables. Analyses of outcomes were limited to those with A1c data at baseline and 6 months; no imputation was used. We used paired t-tests to conduct difference-in-differences analyses of change in the primary and secondary outcomes from baseline to 3 or 6 months between the two groups. We examined within-group changes from baseline to 3, 6, or 12 months as applicable. We used a p value of 0.05 to indicate statistical significance for all analyses. We originally proposed a sample size of 100 participants (50 per arm), which would provide 80% power to detect a difference of 1.14% in A1c. However, being a pilot study, the main goal analytically was to get an estimate of the effect size of the intervention.

Cost data were categorized into total, inpatient, emergency room, pharmacy, and remaining costs (e.g., outpatient services). We constructed time windows of 6- and 11-months duration, with enrollment as time 0, and excluding the month of enrollment. Time windows thus included (1) − 6 months to time 0, (2) − 11 months to time 0, (3) + 1 month to + 7 months, and (4) + 1 month to + 12 months.

The study was reviewed and approved by the Johns Hopkins Institutional Review Board, as well as the Office of Internal Controls and Audit Compliance (IAC) of the Maryland Department of Health.