was successfully added to your cart.

Introduction

To anyone versed in biomedicine, the so-called “hierarchy of evidence” is well-established and unquestioned. The randomized, double-blind, placebo-controlled trial (RCT) is considered the gold standard trial design, because it offers the ability to randomly allocate a treatment, minimize potential sources of bias, and compare the exposure or intervention of interest to a placebo.

There is absolutely nothing wrong with this model, or this hierarchy … if the subject of inquiry is pharmaceutical drugs or the molecular mechanisms of disease. At its core, the biomedical model is based on a presupposition that all disease can be studied with such a reductionist focus.

Only one evidential design sits atop the RCT: the systematic review and meta-analysis, which ideally includes multiple well-designed RCT’s examining the variable of interest, resulting in a greater statistical power and thus the potential for drawing more confident conclusions. There is absolutely nothing wrong with this model, or this hierarchy … if the subject of inquiry is pharmaceutical drugs or the molecular mechanisms of disease. At its core, the biomedical model is based on a presupposition that all disease can be studied with such a reductionist focus.

The evolution of nutrition as a science occurred prior to establishment of the biomedical model, in an era where diseases related to nutritional deficiency were prevalent. The characteristic feature of these disease states was short latency periods; that is, a relatively short period of time between “exposure” and the onset of symptoms (1). These short latency periods were crucial to elucidating the relationship between a given deficiency state and a particular nutrient, as rapid recovery after providing the nutrient (or more specifically the foods in which it was found) allowed for testing early hypotheses and drawing conclusions regarding diet-disease relationships (1).

These early successes for the field produced substantial advances in public health, but defined a paradigm of nutritional approaches oriented around averting deficiency states that by nature emphasized single nutrients.

For example, Thiamin [vitamin B1] was identified as the underlying cause of beriberi from observations and experiments comparing polished rice to wholegrain rice, as the polishing process leads to the loss of thiamin (2). After the 1912 discovery of certain compounds in foods that were essential to life (termed ‘vitamins’), the domino effect of vitamin identification was a renaissance for nutrition as an emerging science. Niacin [vitamin B3] deficiency was identified as a cause of pellagra, and wheat was fortified with that nutrient starting in the 1930’s. The endemic of goiter associated with iodine deficiency hypothyroidism was eradicated with the iodization of salt in 1924, while the endemic of rickets was addressed through the fortification of milk with vitamin D in 1933 (3). These early successes for the field produced substantial advances in public health, but defined a paradigm of nutritional approaches oriented around averting deficiency states that by nature emphasized single nutrients.

This single-nutrient focus would overlap with the evolution of the biomedical model. The application of the biomedical model to nutrition is based on the fundamental premise that nutrients can be isolated and formulated into a pill, and its hypothesized effects should be apparent in a randomized controlled trial. This reductionist view, coupled with residual effects of the early nutrient-focused successes of nutrition science, resulted in a paradigm of studying diet-disease relationships through a focus on isolated nutrients. However, this paradigm has persisted beyond the period of diseases associated with single-nutrient deficiencies being a public health concern, into the more recent surge of chronic lifestyle-related diseases such as diabetes and cardiovascular disease, which are characterized by longer latency periods and more complex etiology.

Instead, what has become clear is that the biomedical model has actually hindered the application of the evidence in nutrition, as it has resulted in missing the forest [the overall diet pattern] for the trees [isolated nutrients].

While the reductionist model may be useful for elucidating disease mechanisms and single-nutrient actions, it is fundamentally ill-suited to elucidate the effects of an overall dietary pattern, where numerous nutrients and bioactive food components influence complex, multifactorial disease processes (4). The evolution of nutrition science has been hampered by the expectation that the application of the biomedical model should generate a consistent evidence base. Instead, what has become clear is that the biomedical model has actually hindered the application of the evidence in nutrition, as it has resulted in missing the forest [the overall diet pattern] for the trees [isolated nutrients] (4).

High-profile biomedical purists have recently gone so far as to say that nutrition science has failed to “give reliable answers for a century”, and that no progress will be made in nutrition until epidemiology is replaced with only randomized study designs (5). In the following article series, we will explain why those contentions are misconceived.

Can We Trust Nutritional Epidemiology?

The emphasis on observational epidemiology in nutrition science has drawn substantial criticism from biomedical purists. In the traditional biomedical hierarchy of evidence, prospective cohort studies sit third on the ladder behind RCTs. However, nutrition science is faced with the logistical problem of trying to determine relationships between 1) complex diseases with long latency periods and 2) populations in which the exposure of interest – food – is a continuous, daily, polyvalent exposure.

The more pertinent question to ask is whether nutritional epidemiology is fit for the purpose of informing sensible public policies to improve population health.

From this perspective, these long-term prospective cohort studies are an under-appreciated (and frequently misunderstood) tool to examine long-term diet-disease relationships (6,7). This may be difficult to appreciate from a biomedical view, as medicine has been scarred from a history of observations generating incorrect conclusions with potentially detrimental consequences. For example, the infamous case of hormone replacement therapy [HRT] and cardiovascular disease risk is a case in point. However, it is important to avoid the assumption that nutritional epidemiology generates such spurious conclusions. The more pertinent question to ask is whether nutritional epidemiology is fit for the purpose of informing sensible public policies to improve population health.

In this regard, it is critical to separate the concepts of hierarchy of evidence (which generally describe quality of evidence) and standards of proof. For example, it is not controversial that observational epidemiology can demonstrate association, but cannot demonstrate causation. But is the clear demonstration of causation an absolute requirement for improving population health? If it were, there would be no public health policy. From the 1964 US Surgeon General’s report on smoking in which tobacco smoking was deemed a “cause” of cancers based off of observational epidemiology, to the evaluation of the public health success of vaccinations, observational research plays an integral role in the formulation and evaluation of policy. By implication, it is clear that RCTs cannot be a prerequisite to set effective policy for reducing population disease burden.

The ultimate goal for informing policy is strong external validity; that is, the generalizability of findings to the “real world” (8). As we will discuss, nutritional epidemiology provides us with a tool to achieve what other trial designs in nutrition do not.

From a practical standpoint, a very tightly controlled nutrition intervention trial must be limited in duration (8). This means that, while informative, metabolic ward studies are only useful to answer certain research questions and examine underlying mechanisms, such as the influence of diet on blood lipids as risk factors for cardiovascular disease. However, it is essentially impossible to do a long-term controlled metabolic ward study on this issue to determine cause-effect relationships, since you’d have to lock people up in a metabolic ward for decades to do so.

Similarly, performing a randomized controlled trial in nutrition involves asking subjects to change their behavior for the duration of the trial. Nutrition RCTs are therefore, by nature, behavioral trials, and therefore face the long-term issues of dietary compliance. Furthermore, we must consider that the influence of a nutrient on disease may depend on the nutrient status of the host, the dose exposure, and the stage of the natural history of the disease of interest (8,9). Nutrition RCTs going over several years therefore face significant practical limitations as well.

With this introductory understanding in mind, in the next article in this series we’ll examine the utility, limitations, and misunderstandings of the prospective cohort design for nutrition research

References

  1. Heaney R. Nutrients, Endpoints, and the Problem of Proof. The Journal of Nutrition. 2008;138(9):1591-1595.
  2. Vandenbroucke J. Adolphe Vorderman’s 1897 study on beriberi: an example of scrupulous efforts to avoid bias. Journal of the Royal Society of Medicine. 2013;106(3):108-111.
  3. Mozaffarian D. Dietary and Policy Priorities for Cardiovascular Disease, Diabetes, and Obesity. Circulation. 2016;133(2):187-225.
  4. Messina M, Lampe J, Birt D, Appel L, Pivonka E, Berry B et al. Reductionism and the Narrowing Nutrition Perspective. Journal of the American Dietetic Association. 2001;101(12):1416-1419.
  5. Trepanowski J, Ionnidis J. Perspective: Limiting Dependence on Nonrandomized Studies and Improving Randomized Trials in Human Nutrition Research: Why and How. Advances in Nutrition. 2018;Jul 1;9(4):367-377.
  6. Satija A, Stampfer M, Rimm E, Willett W, Hu F. Perspective: Are Large, Simple Trials the Solution for Nutrition Research?. Advances in Nutrition. 2018;9(4):378-387.
  7. Satija A, Yu E, Willett W, Hu F. Understanding Nutritional Epidemiology and Its Role in Policy. Advances in Nutrition. 2015;6(1):5-18.
  8. Hébert J, Frongillo E, Adams S, Turner-McGrievy G, Hurley T, Miller D et al. Perspective: Randomized Controlled Trials Are Not a Panacea for Diet-Related Research. Advances in Nutrition. 2016;7(3):423-432.
  9. Maki K, Slavin J, Rains T, Kris-Etherton P. Limitations of Observational Evidence: Implications for Evidence-Based Dietary Recommendations. Advances in Nutrition. 2014;5(1):7-15.

About the authors:

Alan Flanagan is a lawyer and nutritionist based in Dublin, Ireland. In addition to his legal practice, Alan has a Masters in Nutritional Medicine at the University of Surrey and is pursuing his PhD. Alan founded Align Health as an online coaching practice, and as a medium to communicate evidence-based nutrition and health science to a lay audience. From working professionals to professional athletes, Alan provides science-based solutions and protocols to guide motivated individuals to their goals.

Austin Baraki is a physician, strength coach, and competitive powerlifter based in Texas. His best lifts include a 615lb squat, 420lb bench press, and 675lb deadlift. In addition to being both incredibly strong and intelligent, he also has an affinity for cats. Seriously. He loves cats.

Join the discussion 2 Comments

Leave a Reply