Truth, & the Tribulations of Randomized Diet Trials

The volume of bad answers, bad questions, noise and nonsense conspiring to hide the simple, fundamental truths about diet and health seems to swell daily. The task of generating a signal to be heard over this din grows more challenging in tandem. Among the cries populating the cacophony of misinformation is the contention that we know nothing not directly demonstrated in a randomized controlled trial. Much as I like RCTs, having run and published the results of dozens over my career, I consider this view misguided surrender to the tyranny of trial design.

Leaving aside the fact that some extremely impressive randomized controlled trials- with interventions spanning flexitarian diets, Mediterranean diets, and more - do, in fact, demonstrate the fundamental truths about diet and health, the simple fact is that we do not always need a definitive RCT to know what we know. 

Suppose you wanted to know with something nearing certainty what specific dietary pattern was “best” for human health. How would you proceed?

Well, first, I think, you would need to define “best” in an operational (i.e., measurable) way. Does best mean lowers LDL in the short term, or does it mean raises HDL, or both? Does it mean it lowers inflammatory markers, or insulin, or blood glucose, or blood pressure? Does it mean it reduces body fat, or increases lean body mass? Does it mean all of these, or does it mean something else? Is the short term one month, or three, or a year?

I don’t think any of these, or anything like them, really satisfies what we think we mean when we say “best for health.” I think the intended meaning of that is actually rather clear: the combination of longevity, and vitality. Years in life, and life in years, if you will. I think a diet is “best for health” – and yes, I have wrestled with this very issue before- if it fuels a long, robust life free of preventable chronic diseases (e.g., heart disease, cancer, stroke, diabetes, dementia, etc.) and obesity, and endows us with the energy – both mental and physical- to do all we want and aspire to do. That, I think, is a robust definition of “best for health.”

We are obligated to wrestle comparably with the operational definition of a “specific diet.” Low fat, or low carb don’t mean much. A low fat diet could be rich in beans and lentils, or made up exclusively of lollipops. A low carb diet could cut out refined starch and added sugar, or exclude all fruits and vegetables. Let’s not belabor this, and simply concede that the relevant test to prove that one, specific dietary prescription (e.g., the Ornish diet, or the South Beach diet, or the DASH diet, etc.) is best is to establish optimized versions of the various contenders, from vegan to Paleo, and put them up against one another directly.

And now our tribulations begin. As we noted at the start, our outcome is the combination of longevity and vitality. To get at longevity, we need a very long trial; in fact, our trial needs to last a lifetime. So, just to get started, we are toying with the notion of a randomized trial running for 80-100 years.

Dietary influences begin in utero, so we should really randomize not our study subjects, but their mothers while pregnant with them. Dietary influences are salient during breast-feeding as well, and the composition of breast milk is influenced by maternal diet, so we need the mothers we enroll to agree not only to adhere to their assigned diet throughout pregnancy, but to breast feed exclusively until weaning, and adhere then as well. Only at weaning can our actual study subjects get in the game, adopting their assigned diet as babies. For our study to work, they too must adhere to the assigned diet, whatever it is, and in their case- for a lifetime.

Since we are randomizing participants, we may expect them to be alike, on average, in all ways other than their diet assignment- the very point of a randomized, controlled trial. Since we are comparing optimal versions of diets reasonably under consideration for “best diet” laurels, we may anticipate that our study participants are apt to be healthier, and longer-lived in general than the population at large, consuming the lamentable “typical” American diet. 

That’s a problem too. If our entire study sample does “well,” it raises the bar to show that one of our diets is truly, meaningfully better than another. Consider, for instance, that those assigned to an optimal vegan, or an optimal Mediterranean diet, just to name two, have remarkably low rates of chronic disease- and we are trying to show a difference between them in the rates of chronic disease. The smaller the difference we are seeking, the larger the sample size we need to find it, and assign statistical significance to it (let’s not belabor this point either; I’ve written a textbook on the topic, so trust me- it’s true). That now means we need not only a RCT unprecedented in length, but unprecedented in size, too. We need to randomize tens of thousands, if not hundreds of thousands, of pregnant women to study the effects of competing diets on the vitality and longevity of their offspring- at a cost that is staggering to contemplate, and would certainly run into the billions of dollars.

This study has not been done. This study will not be done. Whatever you do, don’t hold your breath waiting for it.

But, so what?

Let’s contrast our ostensible need for this RCT to how we know what we know about putting out house fires.

First, there has never been, to the best of my knowledge, a RCT to show that water is a better choice than gasoline. Do you think we need such a trial, to establish the legitimacy of the basic theme (i.e., use water) of the “right” approach? Would you, and your home, be willing to participate in such a trial when you call 911- knowing you might randomly be assigned to the gasoline arm of the study?

I trust we agree that observation, experience, and sense serve to establish beyond the realm of reasonable (or, even, any) doubt that water is generally good for putting out house fires, and gasoline…not so much.

But what if, as with diet, we wanted to know the “specific” fire fighting approach that was “best.” Once again, we would need to define “specific” approach, which here might mean water at different temperatures, pH, hardness versus softness, and pressure. We might compare hoses of different calibers, and such. And we would need to define “best,” which here presumably means putting out fires the fastest, with the least damage to people and property.

Consider the size, cost, and inconvenience of a randomized trial to compare water at 40°F versus 41°F; or a slight difference in water mineral content. We would again expect variations on the sensible theme of fire fighting such as these to produce very tiny differences in outcomes, meaning we would need an enormous sample, a lot of time, and a lot of money to append this bit of specificity to the fundamentals we already knew.

My friends- and everybody else- diet is the same. The want of a RCT addressing this kind of water versus that does not mire us in perpetual cluelessness about the basic approach to putting out fires. Sure, we could do RCTs to add to what we know- but the want of such studies does not expunge what we already know based on empirical evidence, long experience, observation, and sense.

If anything, the fundamentals of a health-promoting diet are better substantiated than those of fire fighting, since they are informed by long experience, the observation of large populations even of entire regions, and even over generations – as well as by a massive aggregation of research, ranging from mechanistic study in test tubes to RCTs enrolling people. We are the furthest thing from clueless about the basic care and feeding of Homo sapiens. Here, too, RCTs can append to what we know- but they are by no means the sole basis for it. 

I don’t know, frankly, whether an optimal vegan diet, or an optimal Mediterranean diet, or an optimal Asian diet, or even an optimal Paleo diet is “the best” for human health. I do know, because we all know, that a diet comprised principally of minimally processed vegetables, fruits, whole grains, beans, lentils and pulses, nuts, seeds, with plain water preferentially for thirst is the best theme for human and planetary health alike, and runs commonly through all the legitimate, specific contestants- just as water is the best theme when aiming a fire hose. 

To conclude otherwise is to misconstrue the utility of randomized trials, succumb to their tyranny, and lose our way in a bog of tribulations. To conclude otherwise is to fiddle around while the house of public health burns down to the ground.

-fin

David L. Katz

Director, Yale University Prevention Research Center; Griffin Hospital

Immediate Past-President, American College of Lifestyle Medicine

Senior Medical Advisor, Verywell.com

Founder, The True Health Initiative

Follow at: LinkedIN; Twitter; Facebook

Read at: INfluencer Blog; Huffington Post; US News & World Report; Verywell; Forbes