The phrase food as medicine sounds fabulous. It seems like a sure winner, especially for people with diabetes and food insecurity; plus, prevention is so much better than treatment.

One study put that notion to the test. Instead of eating high-caloric processed food, patients with diabetes and food insecurity would receive ingredients for 10 healthy meals per week from fresh-food pharmacies. The program also included dietician and nurse consultations, health coaching, and diabetes education.

So fabulous was this idea that both National Public Radio and The New Yorker covered the program in 2017. Journalist Adam Davidson dubbed the story a bipartisan way to improve medical care.

There were (some) data: An observational study published in 2019 had shown an association between medically tailored meal programs and lower healthcare use. Small randomized trials of less intense food intervention programs delivered mixed results.

But this food-as-medicine project was bigger and bolder than previous efforts. The ingredients for the 10 meals per week included whole grains, fruits and vegetables (fresh was emphasized over canned and frozen), lean proteins, and low-fat dairy products as well as staples such as salad dressing, cereal, brown rice, and bread that were tailored to patients’ needs as determined by a dietician.

Perhaps, you’ve heard this combination before in medical science: a great and plausible idea, positive observational studies, and glowing media coverage. Why not just forge ahead as we did with other well-meaning policies, such as hospital readmission penalties?

Yet, James Doyle, PhD, and colleagues from MIT, Harvard, and Cornell University knew that there was one more step. Their idea had to be tested in a a randomized controlled trial.

They carried out the trial in one urban and one rural community. The entry criteria required a diagnosis of type 2 diabetes with an A1c level ≥ 8%, self-reported food insecurity (via a two-question survey), local residence, and affiliation with the associated health system.

Did a Food-As-Medicine Strategy Improve A1c?

The investigators screened more than 3700 patients. They asked 1000 to participate; 500 consented and 349 completed the study. The control arm, which received standard care, consisted of patients put on a waiting list for the program.

On average, patients were aged 55 years with an A1c of 10.3%. Slightly more than half were female, 9% were Hispanic, and 7% were Black.

The food-as-medicine arm had many more interactions with the clinic (13 vs one) and dietician (2.7 vs 0.6) than did the control group over the 6-month study period. Patients in the active arm self-reported higher scores for healthy eating.

Yet, the primary endpoint, A1c levels, fell by similar levels in both groups. The between-group change was almost zero (0.03; 95% CI, -0.42 to 0.35; P = .86). There were also no substantial differences between the two groups in cholesterol, triglycerides, blood pressure, and fasting glucose over 6-12 months.

Remarkably, the treatment arm gained more weight than did the control arm: 1.95 kg, or just over 4 lb (95% CI, 0.07-3.83; P = .04).

Hospitalization claims were not different in the two arms, though prescriptions for metformin and glucagon-like peptide 1 medications were numerically higher in the active arm.


One word comes to mind. Wow. How could such an intensive program not work?

Food as medicine was set up to succeed. The intervention was intense: 10 meals per week for the entire household plus nurse and dietician visits. Indeed, the active arm had more than 10 times the amount of clinic visits than did the control arm.

And these were highly selected patients. They screened more than 3500 candidates to get 350 patients to finish the trial. Obviously, these were motivated patients. Plus, all patients had poorly controlled diabetes, which should accentuate any treatment differences if present.

Yet, there was no signal of any objective benefit, in either the primary outcome or other measures of cardiometabolic health vs a control arm. In fact, patients in the active arm gained weight.

Nonsignificant trials often raise questions about power, but this trial was designed to detect a 0.5-percentage point decline in A1c. The lower bound of the 95% CIs did not reach this level. And the P value was 0.86. Indeed, there seems to be evidence of absence of effect for this intervention.

Lifestyle Changes Are Hard

The first lesson highlights the challenge of lifestyle interventions. Although everyone agrees that diet, exercise, sleep, etc. play central roles in diseases such as diabetes, hypertension, and obesity, this trial shows the challenge in fighting societal norms.

This leads me to conclude that lifestyle changes must come from somewhere north of well-meaning people such as this research team. No amount of motivation seems likely to get average Americans to live like Sicilians.

As clinicians, many of us have had individual successes in transforming patients to Sicilian-type living. Most of my successes, however, have come from wealthy patients with great social support who have the luxury of being able to succeed.

Most of America does not have such luxury. Most don’t have time to shop for and cook healthy foods. Most don’t have safe ways to walk or cycle to work.

My main lesson therefore is that real health must come not from healthcare but from societal and political change. I am not optimistic.

A more hopeful lesson from this trial stems from the way these scientists did science.

Food as medicine is about as great a slogan as there is. And the concept was celebrated in major media outlets.

But they did not rest. They knew that this idea had to pass muster in a randomized controlled trial, one with a proper control arm.

That the food-as-medicine program did not improve health when properly studied does not mean the research team failed.

I see it as the opposite. Doyle and his team have shown us science done well. This is a success. There may be more to do in the use of food as medicine, but this trial shows us the importance of properly testing great ideas.

John Mandrola, MD, practices cardiac electrophysiology in Louisville Kentucky and is a writer and podcaster for Medscape. He espouses a conservative approach to medical practice. He participates in clinical research and writes often about the state of medical evidence. 


Leave a Reply

Your email address will not be published. Required fields are marked *