Medically Tailored Groceries: Promise, Progress, and Need for Optimal Design

The concept of “food is medicine” has gained momentum as a strategy to integrate food-based nutritional interventions into healthcare as a therapy to manage and treat chronic diseases.^1,2 Yet despite its intuitive appeal and growing healthcare, public, and policy interest, the field remains limited by a paucity of randomized controlled trials that evaluate real-world implementation and clinical outcomes.^1,2 In this issue of JAMA, the GoFresh trial by Juraschek et al. represents an important contribution to the evidence base, testing whether home-delivered groceries patterned after the Dietary Approaches to Stop Hypertension (DASH) diet can improve blood pressure among Black adults living in “food desert” neighborhoods.³

In this randomized, parallel-arm trial, 180 Black adults with elevated systolic blood pressure (SBP, 120–150 mmHg) and not receiving anti-hypertensive medications were assigned to either three months of home-delivered DASH-patterned groceries with dietitian counseling or a $500 monthly stipend for self-directed grocery shopping. The DASH groceries were designed to provide the full dietary needs of each household and achieve calorie and nutrient targets including a potassium/sodium ratio above 2.0 and saturated fat intake less than 7% of energy. The primary outcome was change in office SBP at 3 months, with secondary outcomes including diastolic blood pressure (DBP), LDL cholesterol, body mass index (BMI), and dietary adherence.

At three months, the DASH group achieved greater reduction in SBP (−5.7 mmHg) compared to the self-directed group (−2.2 mmHg), with a between-group difference of −3.4 mmHg (P=0.009). DBP and LDL cholesterol also declined more in the DASH group. These effects were not sustained at six months, three months after the grocery deliveries ended. Notably, the SBP reduction was larger among adults age 65+ years (between-group difference: −12.7 mmHg), consistent with the known age-related sensitivity of BP effects of dietary sodium,⁴ but was not significantly different among those with higher or lower baseline SBP, perhaps owing to the relatively narrow range of SBP eligibility criteria.

While the GoFresh trial demonstrates that medically tailored groceries (MTG) can reduce SBP, several design limitations raise questions about internal validity, costs, and external relevance.

First, the intervention followed the framework of the low-fat DASH diet—despite randomized trials like OMNI-Heart, conducted by the original DASH investigators, showing that this is a less effective diet pattern than more Mediterranean-style DASH diets higher in unsaturated fat or plant protein.⁵ Unfortunately, the superiority of higher fat or plant protein DASH diets remains underrecognized. In GoFresh, the resulting drop in total dietary fat and unsaturated fat, and proportional increase in carbohydrates, may have been particularly problematic given that 8 in 10 participants had overweight or obesity and 1 in 3 had prediabetes. These metabolic profiles suggest that a higher-fat, lower-carbohydrate diet may have been more appropriate. The null findings for body mass index, glucose control, and other metabolic outcomes may reflect this mismatch.

Second, the intervention included substantial quantities of foods that are neither expected to improve BP or metabolic risk factors nor underconsumed by Americans. These included beef, pork, poultry, eggs, meat substitutes, and especially grains. The majority of Americans already overconsume these foods, and none have demonstrated BP-lowering effects.^6–8 Their inclusion likely diluted the intervention’s efficacy by displacing healthier foods in the index participant’s diet. This may be why no significant differences were seen in the consumption of vegetables, nuts, seeds, legumes, whole grains, or dairy—all factors that can improve BP and metabolic health.^7,8 Consistent with these self-reported dietary data, no significance change was seen in urinary potassium—a biomarker of minimally processed plant food intake—suggesting limited adherence to the healthy components of the diet. This contrasts with the DASH-Sodium trial, where potassium intake was a key mediator of BP reduction.⁹

Third, the provision of foods that neither have BP or metabolic benefits nor are underconsumed by Americans substantially increased the cost of the intervention, which averaged $1,125 per month. This far exceeds the monthly cost of effective anti-hypertensive drugs—e.g., chlorthalidone or losartan, at ~$10–15 each—and even of GLP-1 agonists, at ~$500–600 (discounted). At $1,125 per month, the BP-lowering effect observed in GoFresh is remarkably expensive, greatly exceeding thresholds for when BP-lowering therapies become cost-effective.¹⁰ On the other hand, a healthy diet can improve numerous other physiologic risk factors as well as mental health and cognitive outcomes^6–8,11,12—benefits few pill combinations can achieve that should be incorporated into cost considerations. Still, to realize greatest efficacy and cost-benefit, MTG should likely focus on the smallest set of healthful foods that are known to improve the clinical outcomes of interest and that are commonly underconsumed.

The GoFresh trial has important implications for future MTG design and policy choices. It provides insights into how MTG interventions can be structured—and the need for refinement. Target dietary patterns should reflect the best available science. The low-fat DASH model, while historically influential, is no longer the optimal choice for most adults—particularly those with metabolic risk. Mediterranean-style diets, rich in unsaturated fats and plant-based proteins and low in ultra-processed foods,¹³ should generally be prioritized.

In addition, the trial’s subgroup findings—specifically the stronger effect in older adults—suggest that, at least for BP control, MTG may be most impactful in older populations. This highlights a more general principle: like many other therapies, food is medicine interventions are most likely to provide high value, from both clinical benefit and cost perspectives, for higher risk populations. Today, a real tension exists in the design and target populations of food is medicine approaches. Many programs have origins related to anti-hunger and food security goals, creating a natural tendency to focus on primary prevention in broad populations experiencing social stressors. At the same time, payers are expecting a short-term return-on-investment or even cost-savings, which are often difficult to demonstrate in lower risk individuals. This reflects somewhat of a double standard compared to pharmacologic therapies, which are generally approved and utilized based on efficacy rather than economic return. But it is a common expectation nonetheless—one that must be thoughtfully considered in the design and evaluation of target populations for food is medicine strategies.

Finally, rather than attempting to reproduce an entire dietary pattern, MTG programs should focus on providing foods that are both underconsumed by Americans and known to improve health outcomes. This would commonly include, for example, minimally processed fruits, vegetables, nuts, seeds, legumes, whole grains, plant oils, plain yogurt, and fish.^7,8 Currently, multiple states are implementing medically tailored or medically supportive food boxes through Medicaid section 1115 demonstration waivers. In all of these, the standards or guidelines for allowable foods are quite broad, not specifying which items can and cannot be included. Furthermore, the allowable costs are relatively high, for example up to $365 per month in California and $400 per month in Massachusetts for groceries to compose 3 meals per day.^14,15

Medicare Advantage grocery benefit programs for seniors with chronic conditions have similar monthly dollar amounts and vague nutritional criteria: included foods should meet “nutritional needs” and exclude “non-healthy food,” but these terms are not defined. Testing of more highly curated interventions, providing specific healthful foods and at lower total cost, is critical to increase likelihood of efficacy and cost-effectiveness. One ongoing MTG trial among 1,100 individuals with diabetes, providing $100 per month for select items from among a curated list of healthful food groups in an online grocery platform, will provide important evidence in this regard.¹⁶

Overall, the GoFresh trial advances the field of food is medicine by demonstrating that home-delivered DASH-patterned groceries can modestly reduce SBP, DBP, and LDL cholesterol among Black adults living in food deserts. However, the intervention’s relatively low risk population, inclusion of neutral or unhealthy foods, reliance on a low-fat DASH model, and high costs limit its efficacy, scalability, and relevance to practice. Future MTG programs should focus on high risk populations, prioritize foods that are underconsumed and health-promoting, align with current evidence favoring higher fat Mediterranean-style diets, and ensure cost-effectiveness. As the field continues to evolve, additional randomized trials like GoFresh will be essential—as will thoughtful design grounded in nutritional science and public health priorities.