Of Course, Food-as-Medicine Works

Diet in America, and much of the world, is badly broken.? How badly?? Poor overall diet quality is the single leading predictor for the risk of premature death from all causes and for the occurrence of any major chronic disease.? In other words, junk where food ought to be siphons years from lives and life from years.? Poor quality diets kill hundreds of thousands of us prematurely year in, and year out.? Welcome to 2024, when it will surely happen again.? This is the pandemic that never ends, that hides in plain sight, that invites complacency where outrage and urgency ought to be .

Fixing broken diets has powerful benefits.? Sense indicates it should be so; science affirms it.? Dedicated efforts to improve diet quality, either in general or in targeted ways, have resulted in dramatic reductions in the occurrence of diabetes ; regression of coronary atherosclerosis ; enormous reduction in the rate of heart attack in high-risk individuals; reductions in blood pressure commensurate with pharmacotherapy; reductions in LDL cholesterol commensurate with pharmacotherapy; changes in gene expression ; the lengthening of telomeres ; and so on.?

The powerful, intuitively self-evident and scientifically confirmed promise of food-as-medicine guided my efforts over a 30-year career in academic medicine .? That same promise now guides my efforts in the private sector.?

We manage what we measure, and it thus stands to reason that if we are to fix broken diets we should be taking their measure routinely.? Think of blood pressure as an analogy: we could know that “broken” blood pressure was a leading cause of heart attack and stroke at the population level, but unless we measured blood pressure in individuals, routinely, what good would that general knowledge do any of us?? Similarly, we can know that diet “on average” is badly broken, but until we take its measure in individuals- ideally, just like blood pressure, universally, as a vital sign - how can we hope to manage it in individuals?

Accordingly, when I conceived of a new way to assess dietary intake that would allow for diet quality to be treated as a vital sign, I founded a company to scale that capability.? That company was recently acquired by a major player in the food-as-medicine space , for which I now work as Chief Medical Officer.? Combined, the companies can assess dietary intake at baseline; identify a personalized goal diet; and support progress toward that goal by every means from digital coaching to meal delivery.? We do this routinely in large, generally underserved, multicultural, and often food-insecure populations.

The results are consistently gratifying.? Diet quality, measured objectively with the Healthy Eating Index 2020 , improves meaningfully from baseline.? Weight, generally excessive at baseline, declines meaningfully.? When biomarkers -blood pressure, lipids, glucose, glycohemoglobin, insulin- are measured, these reliably improve in tandem with diet quality.? Our real-world experiences show decisively that diet quality can be fixed when broken, and that fixing it confers all of the expected benefits.

How is it, then, that a recent study in JAMA Internal Medicine has invited the expression of doubt in some quarters about the value proposition of food-as-medicine?? The answer, as so often with contrarian research results played up by the media, devolves to devilry in the statistical and methodologic details .

The study in question describes itself as an “intensive” food-as-medicine program for adults with type 2 diabetes, but that declaration is suspect.? Participants were given groceries, along with recipes and coaching.? To those of us involved in the delivery of prepared, health-goal-oriented, personalized meals- this seems rather far from “intensive.”? We can agree that preparing healthful meals for one’s family is a great idea, yet still acknowledge doubts that it happened reliably.? Of note, members of the intervention group actually gained weight- suggesting that study-provided groceries were consumed, to one degree or another, in addition to, rather than instead of, prior food choices.? That was unintended.

The study’s signature strength- randomization- was also a critical weakness .? Participants were randomly assigned to receive the free grocery and coaching intervention right away, or after a delay.? Those assigned to “wait” were, of course, exposed to the basic intentions and methods of the study- a prerequisite to informed consent.? This level of exposure to a research agenda is an intervention in its own right, as made famous by a large, federal study called MRFIT .? The “doubts” about food-as-medicine attributed to this study are not because the intervention group failed to improve their glycohemoglobin (the primary outcome measure), but rather because the control group improved theirs so much, the between-group difference (the essential measure in a randomized trial) was not significant.

To be clear, the intervention group did improve more- across an array of measures, from glycemic control to fruit and vegetable intake- than the wait-listed control group.? But both improved quite meaningfully, shrinking the between-group difference.

So, perhaps you are thinking: well, then, if mere awareness of study objectives and methods does so much, why not just talk to people about food-as-medicine, and let them do the rest?? For one reason, the actual intervention produced greater results; they failed to reach significance for want of statistical power, not for want of effect.? For another, as noted above, this was not a very intensive intervention, and the magnitude of effects varies with the true intensity of causes.? But third, and probably most important- this was a very select group of participants, unlikely to represent readiness-to-change in the general population.? The researchers assessed nearly 4000 candidates for the study, of whom just over 1000 (i.e., only about 25%) were invited to participate.? Of these, just 500 (50% of the 25%) consented, and of these just 349- less than 10% of the candidate pool- generated data at the 6-month mark.

This is not a criticism; when I have run randomized controlled trials, our recruitment and enrollment numbers followed similar patterns.? But it is an important and universal consideration: steps taken to improve the controls in a study (i.e., enhancements of internal validity) are often at the expense of generalizability (a.k.a, external validity).? One may not confidently presume that real-world populations in general will behave like this highly-filtered sub-group.

There is more: participants increased their use of medications for diabetes during the study as well.? The drugs included metformin, a time-honored mainstay, as well as the powerful new drugs we all hear so much about these days, the GLP1s- of which Ozempic is an example.? To the degree that medication use improved the primary outcome measure, glycohemoglobin, diet could not improve it- because you can’t fix the same broken thing twice over.? Study eligibility criteria included a poorly-controlled glycohemoglobin at baseline, and this information alone would be a potent prompt to all study participants to address that matter.? Clinicians confronted with those numbers would be obligated to intervene.? Improvement in the primary outcome due to pharmacotherapy would reduce the statistical power of the study to attribute benefits to diet.?

Despite all of the above, the intervention did deliver improvements in accord with expectations.? Just about every measure of interest (weight, as noted above, an important exception) got better in both groups, and got better by more in the intervention group.? That’s hardly cause to start doubting the promise of food-as-medicine .? Rather, it’s reason to scrutinize the inevitable caveats and provisos that encumber a wait-list-as-control-group randomized trial.? With the great strengths of RCTs comes our collective responsibility to acknowledge their at-times quite important liabilities .

When the fine print is parsed, the new study appends to the already extensive and compelling evidence in support of food-as-medicine.? It further invites us to establish what “intensive” should mean, and warns us that better food “in addition to” is not reliably the same as better food “instead of.”? If the latter is our goal, the intervention must account for it.?

We have new, powerful means to both measure, and manage, overall diet quality in a personalized manner and at scale.? Detailed knowledge of where diet is, relative to where it ought to be in the service of specific health objectives, allows for customized goal diets, across an expanse of motivations, nutritional inclinations (e.g., vegan; low-carb; gluten-free; etc.), and culture-based cuisines (e.g., Southern, Latin, Asian, South Asian, Mediterranean, etc.).? Guiding meal delivery accordingly, we have whole new ways to empower at-risk populations to eat what they enjoy even as it helps them enjoy better health; to love the food that loves them back.? As these powerful assets are deployed- both in the service of public health, and in research- the true promise of food-as-medicine will be ever less subject to doubt.

?

-fin

David L. Katz, MD, MPH is the founder of Diet ID, Inc and Chief Medical Officer for Tangelo .? He is past president of the American College of Lifestyle Medicine .