A free weekly bag of beans, whole grains and fresh produce from a Boston hospital's food pantry: the more bags people collected over almost three years, the more weight they lost. Blood sugar didn't budge

Doctors in the United States are increasingly told to treat hunger as a health problem. "Food is medicine" programmes give patients who can't afford enough food free produce, groceries or cooked meals through the health system, in the hope of better blood pressure, blood sugar and weight. The idea is popular; the results so far are not. In 2025 the American Heart Association reviewed the 14 randomised trials of such programmes in the US and found that they usually improved what people ate and how secure they felt about food, but that their effect on clinical measures was inconsistent and often not statistically significant. One reason may be dose: most programmes last weeks or months, and many hand out whatever food is available rather than the healthiest.

A short report published on 1 October 2026 in Preventing Chronic Disease, a journal of the US Centers for Disease Control and Prevention, looks at a programme that is unusual on both counts. The Massachusetts General Hospital Revere Food Pantry sits inside a community health centre near Boston and gives patients a plant-based grocery package every week: fresh fruit and vegetables, canned beans or nut butter, and whole grains, planned by a dietitian at first around the US dietary guidelines to cover three meals a day for each family member for a week. Most of the people it serves have low incomes and public health insurance, and most do not speak English as their first language. The researchers, from Mass General, Harvard Medical School, Boston Children's Hospital and the Greater Boston Food Bank, went back through the health records of everyone who had used it between December 2019 and April 2025.

Of 380 people who received packages, 68 adults met the study's conditions: at least 26 packages, and at least one measurement of blood pressure, blood sugar (HbA1c) or weight both in the year before their first package and later on. They were mostly women (47 of 68), with an average age of 57. At the start, four in five were overweight or obese, the average systolic blood pressure was 130 mm Hg, and the average HbA1c was 6.9%, in the diabetes range. On average they collected about 98 packages over 2.7 years, roughly one every week and a half. Because people came for different lengths of time, the analysis asked a dose question: did those who collected more packages see bigger changes between their before and after measurements?

For weight, the answer was yes, and it held whichever way the numbers were adjusted. Each 100 packages went with about 2.6 kg more weight lost and a BMI about 1 point lower than before. For diastolic blood pressure (the lower number) each 100 packages went with a drop of 1.5 mm Hg in the simpler models and 3.2 mm Hg in the fully adjusted one. The upper, systolic number did not change across the group, though among the 31 people who started at 130 or above, the fully adjusted model found a 5.8 mm Hg drop per 100 packages. HbA1c did not change, in the whole group or in the 16 people who started at 7% or above.

Those last blood pressure figures need a closer look, because they depend heavily on one choice. In the model adjusted for age, sex, starting value and time, the systolic change in people with high starting pressure was close to zero (−0.2 mm Hg, nowhere near significant). It became 5.8 mm Hg only when the authors added an adjustment for the stages of the COVID-19 pandemic, and the diastolic figure doubled in the same step. Adjusting for the pandemic is a reasonable idea, since clinic visits and daily life changed a great deal in those years, but when one adjustment moves a result from nothing to something, it is a sign the finding is fragile. The weight result, which barely moved between models, is the sturdier one.

The bigger limit is the one the authors name first: there was no comparison group. Everyone in the study chose to keep coming back, and the people who come back most often may differ from those who drop out in ways that also affect their health, from more stable housing to better contact with their doctors. The authors point out that the pantry may also help people stick to their medicines. Measurements came from routine clinic visits rather than research visits. And a randomised trial of a similar idea shows why a control group matters: in 2024, a programme in the mid-Atlantic US giving groceries for 10 meals a week plus dietitian and nurse support to people with diabetes and food insecurity found that HbA1c fell sharply in both the people who got the programme and those on the waiting list, leaving no real difference between them.

Our reading: this is an encouraging signal from a well-run local programme, not proof that plant-based groceries lower blood pressure. The steady link with weight over nearly three years is worth testing properly, and the authors themselves call for a randomised trial. The message for anyone else is the same as the dietary guidelines already give: a weekly shop built around beans, whole grains, fruit and vegetables is a sound base. If you are struggling to afford food, ask your clinic or doctor whether a food pantry or produce programme operates nearby. The work was funded by a Massachusetts General Hospital award, the Vitamix Foundation, the Ardmore Foundation and the US National Institutes of Health, and the authors declare no conflicts of interest.