— July 21, 2026 · Grocery Deliveries Lower Blood Sugar in Type 2 Diabetes: A 460-Person Randomised Trial

Table of Contents

1. The Trial at a Glance
2. What Was Done
3. What Was Found, Number by Number
4. What an HbA1c Point Means
5. How Groceries Could Lower Blood Sugar
6. How It Fits With Earlier Research
7. What It Cannot Show
8. What the Researchers and Press Releases Said
9. What Remains Unknown
10. Sources and Primary Documents
11. Key Research Papers
12. Connections
13. Featured Videos

The Trial at a Glance

• Paper — Nau C, Wu JHY, Han B, et al. Effects of a “Food Is Medicine” Intervention on Glucose Control Among Medicaid-Insured Patients With Type 2 Diabetes: A Randomized Controlled Trial. Circulation 2026;154(7):624-635. Published online July 21, 2026; print issue August 18, 2026.
• Who — 460 Medicaid-insured adults with type 2 diabetes and at least two HbA1c results of 7.5% or higher in the past year. Mean age 59.2 years; 64.8% women; 85.2% Hispanic; 58% reported food insecurity; mean HbA1c 9.40%.
• Where and when — medical offices of one integrated health system in southern California (12 medical office buildings, according to a partner institute’s press release); enrolment November 2021 to July 2022.
• Design — randomised 1:1:1 to usual care (153), lower-dose groceries (153) or higher-dose groceries (154) for 6 months; open-label (no masking, per the trial registry). The planned main comparison pooled both grocery groups against usual care.
• What the grocery groups received — weekly home deliveries of healthy produce scaled to household size, worth $100 to $170 a month (lower dose) or $135 to $210 a month (higher dose), plus matched recipes and optional telenutrition counselling.
• Main result — HbA1c −0.66 points with groceries vs −0.25 with usual care; difference −0.40 points (95% CI −0.73 to −0.08; P = 0.016).
• Food and nutrition security — odds ratio 2.12 (95% CI 1.13 to 3.99) for food security and 3.65 (95% CI 1.84 to 7.25) for nutrition security.
• No significant change — hypertension and body mass index (BMI).
• Registration — ClinicalTrials.gov NCT05407376.

Back to Table of Contents

What Was Done

The researchers looked for adult patients who had Medicaid coverage and type 2 diabetes that had stayed poorly controlled, defined as at least two HbA1c results of 7.5% or higher in the previous 12 months. The trial registry adds that participants had to have been diagnosed with diabetes for at least 24 months, be aged 18 to 70, and be able to receive and prepare the delivered food. Pregnant women and people living in board-and-care homes where meals are provided were not eligible.

The 460 people who joined were divided by chance into three equal groups. Everyone kept their normal medical care. One group received nothing more. The other two received medically tailored groceries, delivered to the home every week for six months. The two grocery groups differed only in how much food arrived each month. The amount was scaled to the number of people in the household: $100 to $170 worth a month in the lower-dose group and $135 to $210 worth in the higher-dose group, according to the final paper. The registry gives an example: for a family of three or four, $180 a month versus $135, with more food for each additional family member up to five.

What was in the deliveries. The abstract describes the deliveries as “healthy produce”, and the registry calls the programme a “produce prescription (weekly produce delivery)”. Produce means fruits and vegetables. Neither the abstract nor the registry lists the individual foods, and the press releases speak more broadly of “healthy foods” and “healthy groceries”, so this report does not list items it could not confirm. The deliveries came with recipes matched to the food (described by a partner institute as “culturally tailored”), and participants could also use nutrition counselling delivered remotely (telenutrition; the press releases describe it as phone-based).

What was measured. The main outcome was the change in HbA1c from the start of the trial to six months. Secondary outcomes were food security and nutrition security, measured by surveys, and high blood pressure and BMI, taken from routine clinical visits. The registry names the survey for food security: the six-item US Adult Food Security Short Form, scored from 0 (high or marginal food security) to 6 (very low food security). Nutrition security asks a slightly different question: not only whether a household has enough food, but whether it can regularly get food that supports health. The researchers also compared the two grocery doses and checked whether the result held in subgroups.

The registry describes the study as a “pre-randomized” trial, a term usually used for designs in which people are assigned to a group before they are approached. How that worked in this trial is set out in the full paper, which was not available for this report.

Back to Table of Contents

What Was Found, Number by Number

Did people eat the food? Among the grocery groups, 244 participants (83.3%) said they ate most or all of the food provided. Only 63 (21.5%) used the telenutrition counselling.

Blood sugar. HbA1c went down in both arms. In the grocery groups it fell by 0.66 points over six months; in the usual-care group it fell by 0.25 points. The difference between the arms, which is the effect that can be credited to the groceries, was −0.40 points, with a 95% confidence interval of −0.73 to −0.08 (P = 0.016). The interval means the data fit a true effect anywhere from a small 0.08-point drop to a larger 0.73-point drop.

The fall in the usual-care group matters. People are often recruited into diabetes trials when their numbers are at their worst, and readings tend to drift back toward their average afterwards, with or without any programme. Only a randomised control group can show how much of an improvement is due to the programme itself. Here, about four-tenths of a point was.

Lower vs higher dose. According to the abstract, the higher-dose and lower-dose groceries “produced similar reductions”. The abstract does not give the numbers for each dose separately. In a partner institute’s press release, the two levels averaged about $135 and $175 a month.

Food and nutrition security. Compared with usual care, the grocery groups had 2.12 times the odds of being food secure (95% CI 1.13 to 3.99; P = 0.020) and 3.65 times the odds of nutrition security (95% CI 1.84 to 7.25; P < 0.001). At the start, 58% of participants had reported food insecurity.

Blood pressure and weight. “Hypertension and body mass index did not significantly change,” the abstract states. The participants’ average BMI at the start was 34.1, and 74% had raised blood pressure or hypertension, according to the 2023 interim abstract.

Consistency checks. The result was consistent in planned subgroups by sex, education, starting food security, starting nutrition security and starting HbA1c. It was also similar after statistical adjustment for demographics, food insecurity, nutrition insecurity, self-rated health, other illnesses and changes in medication, which suggests the HbA1c difference was not simply the result of one group having its diabetes drugs increased.

Earlier look at the same trial. An interim analysis presented as a conference abstract in November 2023 reported a similar effect, −0.37 points (95% CI −0.71 to −0.04), with no significant difference between the doses. At that point 87% of participants had a six-month HbA1c result.

Back to Table of Contents

What an HbA1c Point Means

Hemoglobin is the oxygen-carrying protein inside red blood cells. Sugar in the blood sticks to it slowly and permanently, a process called glycation. HbA1c is the percentage of hemoglobin that carries this sugar. Because red blood cells circulate for months, HbA1c reflects the average blood sugar over roughly the previous three months, rather than a single moment. The site’s page on A1c, fructosamine and glycation markers explains how the test is read and when it can mislead.

The ADAG study (Nathan and colleagues, 2008) compared HbA1c with about 2,700 glucose readings per person over three months in 507 people and produced a formula that converts one into the other: average glucose in mg/dL = 28.7 × HbA1c − 46.7. Applied to this trial (the arithmetic is this report’s, not the trial’s), a starting HbA1c of 9.40% corresponds to an estimated average glucose of about 223 mg/dL, and a 0.40-point difference corresponds to about 11 mg/dL lower average glucose.

Is 0.40 a meaningful amount? In the UK Prospective Diabetes Study analysis known as UKPDS 35, an observational study of 3,642 people with type 2 diabetes followed for years, every 1 percentage point lower HbA1c was associated with a 21% lower risk of any diabetes-related end point and a 37% lower risk of microvascular complications (damage to the small vessels of the eyes, kidneys and nerves), with no threshold below which the association stopped. That is an association over years. The grocery trial lasted six months and measured no complications, so it cannot say how many, if any, were prevented.

Back to Table of Contents

How Groceries Could Lower Blood Sugar

The trial tested a package, so it shows that the package worked on average, not which part of it did the work. Its design and the earlier research point to three possible routes.

More whole food in the house. Vegetables and fruits are whole foods and sources of dietary fibre; the site’s fibre page covers the research on fibre and blood sugar. The site’s page on how the body controls blood sugar describes how insulin moves sugar out of the blood after a meal. The trial registry lists a short diet-quality survey as a secondary outcome, but the abstract reports no results for diet, so it is not known from this paper how much participants’ eating actually changed. In an earlier crossover trial of delivered meals, discussed below, diet quality scores rose sharply, with more vegetables, fruits and whole grains and less added sugar.

Less food insecurity. Food insecurity, meaning not reliably having enough food because of cost, has been linked with diabetes. In a 2010 analysis of 5,094 low-income adults in the US National Health and Nutrition Examination Survey, food insecurity was associated with laboratory evidence of diabetes (adjusted relative risk 1.48, 95% CI 0.94 to 2.32), and the association was stronger and statistically significant with a stricter definition of food insecurity (2.42, 95% CI 1.44 to 4.08). The groceries clearly improved food security in this trial. Whether that improvement itself lowered HbA1c is not something the abstract tests.

Recipes and counselling. Every grocery household received recipes, but only about one participant in five used the telenutrition counselling. The abstract does not report a separate analysis of people who did and did not use it.

Back to Table of Contents

How It Fits With Earlier Research

“Food is medicine” is the name for health-care programmes that provide food as part of treatment: produce prescriptions (vouchers or cards for fruits and vegetables), medically tailored groceries (like this trial) and medically tailored meals (ready-made meals designed by dietitians). Until recently most of the evidence came from studies without a control group, and the randomised trials that did exist gave mixed results.

Uncontrolled studies looked promising. A 2021 meta-analysis by Bhat and colleagues found 13 studies of healthy food prescription programmes, 9 of them with no control group. Pooled, HbA1c fell by 0.8 points (95% CI 0.1 to 1.6), but the results varied enormously between studies (I² = 92%), and the authors rated the certainty of evidence as moderate to very low and called for large randomised trials. A 2023 evaluation by Hager and colleagues of 9 US produce prescription programmes (3,881 people at 22 sites in 12 states, median $63 a month for 4 to 10 months) reported that adults ate 0.85 more cups of fruits and vegetables a day, the odds of food insecurity dropped by about a third (odds ratio 0.63), and HbA1c fell by 0.29 points among adults with HbA1c of 6.5% or higher. That study also had no control group.

Delivered meals improved diet. In a 2019 randomised crossover trial by Berkowitz and colleagues, 44 food-insecure adults with diabetes received 10 medically tailored meals a week for 12 weeks and usual care for another 12 weeks, in random order. Diet quality on the Healthy Eating Index was 71.3 during the meal period vs 39.9 without meals, a 31.4-point difference; food insecurity (42% vs 62%) and self-reported low blood sugar episodes (47% vs 64%) were also lower. HbA1c was not the main outcome. A 2025 pilot trial by Clark and colleagues in 74 Medicaid-insured adults with an average HbA1c of 10.3% gave 12 frozen medically tailored meals and a bag of fresh produce weekly for three months, plus dietitian calls. HbA1c improved about equally in both groups (−0.7 vs −0.6 points), while food insecurity fell from 53% to 17% with the programme vs 48% to 44% without it at three months.

Two larger randomised trials found no HbA1c benefit. In a 2024 trial by Doyle and colleagues, 500 adults with type 2 diabetes, HbA1c of 8% or higher and food insecurity were randomised to start an intensive programme at once or after six months. The programme supplied healthy groceries for 10 meals a week for the whole household, plus dietitian visits, nurse checks, health coaching and diabetes education. Both groups’ HbA1c fell substantially, and the adjusted difference at six months was −0.10 points (95% CI −0.46 to 0.25; P = 0.57), not significant. The authors noted that programmes aimed at people with high readings need a control group, because readings improve without the programme too. In April 2026, Drake and colleagues reported a pragmatic trial of 2,155 people with diabetes at risk of food insecurity: a debit card loaded with $80 a month for up to 12 months, valid for fresh, frozen or canned fruits, vegetables and legumes. Only 30% of participants used 80% or more of the monthly benefit, and at 12 months HbA1c was 0.20 points higher in the card group (95% CI 0.05 to 0.35), favouring usual care, with no differences in emergency visits, blood pressure or BMI.

Where the new trial sits. Against that background, the 2026 Circulation trial is the only one of the randomised trials described here to report a statistically significant fall in HbA1c compared with a control group. Its programme differed from the null trials in several ways that can be read off the papers: food was delivered to the door each week rather than bought with a card; the monthly value ($100 to $210) was higher than the $80 card; and participants started with higher HbA1c (9.40%) than in the card trial (7.48%). None of these papers tests which of those differences matters.

Back to Table of Contents

What It Cannot Show

• Open-label. The registry records no masking. Participants and their clinicians knew who was receiving groceries, which can change behaviour and care in ways a blinded trial would rule out. HbA1c is a laboratory value, which limits this problem for the main outcome; the food-security results come from participants’ own survey answers.
• Six months only. The trial cannot show whether the HbA1c difference lasts after deliveries stop, or whether it grows or shrinks with longer delivery.
• One health system, one population. All participants were Medicaid-insured patients of one health system in southern California, 85.2% were Hispanic and none were older than 70. Results may differ elsewhere.
• No hard outcomes. The abstract reports no complications, hospital stays, emergency visits or deaths, and no cost-effectiveness analysis.
• Diet not reported. How much participants’ eating changed is not in the abstract, and “ate most or all food provided” is self-reported.
• A wide interval. The 95% confidence interval for the main result runs from 0.08 to 0.73 points, so the true effect could be small.
• Dose comparison incomplete. “Similar reductions” at the two doses is stated without per-dose numbers in the abstract. A trial of 460 split three ways has limited power to detect a difference between doses.
• Routine-care measurements. Blood pressure and BMI came from clinical visits rather than standardised trial measurements.
• Authors’ own limitations not seen. The full paper, including the authors’ discussion of their limitations, is behind a subscription and was not available for this report. The points above come from the trial design as recorded in the abstract and registry.

Back to Table of Contents

What the Researchers and Press Releases Said

The trial was a collaboration between the health system’s research department, a US university’s food-is-medicine institute and an Australian global-health research institute. Press releases appeared on July 21, 2026.

Claudia Nau, PhD, the lead author: “This trial shows that food-based support can help patients with low incomes lower their blood sugar levels above and beyond the health care they receive.”

Dariush Mozaffarian, MD, DrPH, the senior author, a cardiologist who directs the university institute: “These trial results further demonstrate that healthy food can be a powerful tool in treatment and management of chronic disease.”

Jason Wu, head of nutrition science at the Australian institute and a co-principal investigator: “For too long nutrition has been an afterthought in healthcare, despite poor diets driving immense disease burden and healthcare costs globally - not just in the United States.”

Where the press releases and the paper differ.

• A partner institute’s release, repeated by several news sites, said “the odds of food security increased by nearly 215 percent and nutrition security increased by 365 percent.” The paper reports odds ratios of 2.12 and 3.65: the odds were about 2.1 times and 3.65 times as high, which is an increase of roughly 112% and 265%, not 215% and 365%.
• The health system’s own release said blood sugar “dropped by nearly half a point over 6 months compared to people who received usual care.” The difference compared with usual care was 0.40 points; the 0.66-point figure was the fall within the grocery groups, not the difference between groups.
• The same release quoted the lead author as saying healthy food can help people lower their blood sugar “without needing more medicine.” The abstract reports that the result held after adjustment for medication changes, but it does not report medication use in each group, so that wording goes beyond what the abstract shows.

Funding and interests. According to the partner institute’s release, the trial was funded by the health system’s community-health programme. The authors’ disclosures list government and foundation research funding, and for two authors, research funding from fresh-produce industry groups or paid advisory roles with technology, nutrition and food-retail companies.

Back to Table of Contents

What Remains Unknown

• Does the effect last? No result after the six months of deliveries is reported.
• Which part matters? The food, the delivery to the door, the recipes, the counselling and the relief from food worry were given together.
• Does it prevent complications or save money? These were not outcomes of this trial, and the abstract reports no cost-effectiveness analysis.
• Would it work as a benefit at scale? A larger trial in the same health system (Parikh and colleagues, 2024) planned to randomise about 1,100 Medicaid-insured adults with HbA1c of 7.5% or higher in northern and southern California to $100 a month for healthful foods from a curated list on an online grocery and home-delivery platform, plus nutrition education, or to a control group that received free platform membership and delivery but no food money or curated list. Its design paper was published in 2024; results were not part of this July 2026 news.
• Would it work elsewhere? An Australian randomised trial, published as a protocol in 2025, plans to randomise 224 adults with type 2 diabetes and food insecurity or financial hardship to 26 weeks of weekly deliveries of fruits, vegetables, wholegrains and nuts with up to three dietitian sessions, or usual care.

Since then. The paper appeared in the August 18, 2026 print issue of Circulation (volume 154, issue 7, pages 624–635).

Back to Table of Contents

Sources and Primary Documents

Every trial figure above was taken from the PubMed abstract of the paper, the trial registry or the 2023 conference abstract, not from the press releases. The full text of the paper is paywalled and was not read.

1. Nau C, Wu JHY, Han B, Habib M, Li X, Padilla A, Nelson C, Chao C, Schwartz P, Mozaffarian D (2026). Effects of a “Food Is Medicine” Intervention on Glucose Control Among Medicaid-Insured Patients With Type 2 Diabetes: A Randomized Controlled Trial. Circulation 154(7):624-635. — doi:10.1161/CIRCULATIONAHA.125.077982 · PubMed PMID: 42478360
2. Nau C, Wu JH, Han B, Habib M, Gu Z, Li X, Chao C, Schwartz P, Mozaffarian D (2023). Abstract 14690: ProduceRX Clinical Trial. Circulation 148(Suppl_1). — doi:10.1161/circ.148.suppl_1.14690
3. ClinicalTrials.gov record NCT05407376, “Food As Medicine Study” (first posted June 7, 2022; enrolment 460; masking none). — ClinicalTrials.gov NCT05407376

Press coverage read for this report (July 21, 2026): partner institute press release; News-Medical report; and the health system’s own news release.

Back to Table of Contents

Key Research Papers

Background papers, in the order this article uses them. Each entry was checked against PubMed and Crossref before it was written down.

1. Nathan DM, Kuenen J, Borg R, Zheng H, Schoenfeld D, Heine RJ (2008). Translating the A1C assay into estimated average glucose values. Diabetes Care 31(8):1473-1478. — PubMed PMID: 18540046
2. Stratton IM, Adler AI, Neil HA, Matthews DR, Manley SE, Cull CA, et al. (2000). Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ 321(7258):405-412. — PubMed PMID: 10938048
3. Seligman HK, Laraia BA, Kushel MB (2010). Food insecurity is associated with chronic disease among low-income NHANES participants. Journal of Nutrition 140(2):304-310. — PubMed PMID: 20032485
4. Bhat S, Coyle DH, Trieu K, Neal B, Mozaffarian D, Marklund M, Wu JHY (2021). Healthy Food Prescription Programs and their Impact on Dietary Behavior and Cardiometabolic Risk Factors: A Systematic Review and Meta-Analysis. Advances in Nutrition 12(5):1944-1956. — PubMed PMID: 33999108
5. Hager K, Du M, Li Z, Mozaffarian D, Chui K, Shi P, et al. (2023). Impact of Produce Prescriptions on Diet, Food Security, and Cardiometabolic Health Outcomes: A Multisite Evaluation of 9 Produce Prescription Programs in the United States. Circulation: Cardiovascular Quality and Outcomes 16(9):e009520. — PubMed PMID: 37641928
6. Berkowitz SA, Delahanty LM, Terranova J, Steiner B, Ruazol MP, Singh R, et al. (2019). Medically Tailored Meal Delivery for Diabetes Patients with Food Insecurity: a Randomized Cross-over Trial. Journal of General Internal Medicine 34(3):396-404. — PubMed PMID: 30421335
7. Clark JM, Maw MTT, Pettway K, Chander G, Elias S, Zisow-McClean S, et al. (2025). Impact of Medically Tailored Meals on Clinical Outcomes Among Low-Income Adults with Type 2 Diabetes: A Pilot Randomized Trial. Journal of General Internal Medicine 40(8):1711-1719. — PubMed PMID: 39672984
8. Doyle J, Alsan M, Skelley N, Lu Y, Cawley J (2024). Effect of an Intensive Food-as-Medicine Program on Health and Health Care Use: A Randomized Clinical Trial. JAMA Internal Medicine 184(2):154-163. — PubMed PMID: 38147326
9. Drake C, Buckman C, Brucker A, McPeek Hinz E, Pignone M, Shaikha H, et al. (2026). Produce Prescription Subsidy for Patients With Diabetes: A Pragmatic Randomized Clinical Trial. JAMA Internal Medicine 186(4):416-424. — PubMed PMID: 41697676
10. Parikh RV, Nau CL, Tan TC, Tucher E, Vallejo JD, Jimenez JJ, et al. (2024). Rationale and design of the KP ENRICH trial: A food is medicine intervention in low-income high-risk adults with diabetes within Kaiser Permanente. Contemporary Clinical Trials 143:107601. — PubMed PMID: 38851480
11. Glover K, Gow M, Trieu K, Huang L, Law KK, Neal B, et al. (2025). Produce prescription to improve health among adults with type 2 diabetes in Australia: Protocol for a randomised controlled trial. Contemporary Clinical Trials 153:107915. — PubMed PMID: 40239800

PubMed Topic Searches

1. PubMed: Medically tailored groceries and diabetes
2. PubMed: Produce prescriptions and HbA1c
3. PubMed: Medically tailored meals, randomised trials
4. PubMed: Food insecurity and glycaemic control in type 2 diabetes

Back to Table of Contents

Connections

• News Archive
• Type 2 Diabetes — causes, complications and the research on treatment
• Diabetes — the site’s hub for every form of diabetes
• Prediabetes — the stage before type 2 diabetes
• Insulin Resistance — the problem underneath high blood sugar
• Type 2 Diabetes Remission and Reversal — what the research reports on diet and remission
• Hemoglobin A1C — the lab test this trial used as its main outcome
• A1c, Fructosamine and Glycation Markers — how long-term blood sugar tests are read
• Glycation and HbA1c — an animation of what the number measures
• How Your Body Controls Blood Sugar — insulin and glucose, animated
• Blood Sugar Remedies — foods, herbs and practices studied for blood sugar
• Vegetables — the site’s vegetable pages
• Fruits — the site’s fruit pages
• Beans for Blood Sugar and Diabetes — legumes in diabetes research
• Dietary Fiber — the part of plant food the body does not digest
• American Food and Health Policy Reset — earlier news on US food policy

Back to Table of Contents

Featured Videos