The Case Against: Nutrition Science is Not Up to the Task
Drs. Mozaffarian and Forouhi raise a crucial question in their article, “Dietary guidelines and health—is nutrition science up to the task?” That task, of course, is whether our nutrition recommendations have effectively prevented the major chronic diseases of our time. The authors, both nutritional epidemiologists, present their vigorous defense of the discipline, affirming confidently that “Modern nutrition science is up to the task.”
The reason this issue is so critically important is that we face unprecedented epidemics of obesity and diabetes, and thus, it is indeed possible that nutrition science has failed us. The case for this alternate view must be heard as well.
The success of any policy or intervention must be evaluated by relevant outcome measures. Drs. Mozaffarian and Forouhi assert that “advances in the science and application of nutrition represent expected, reassuring progress,” yet they provide no outcome measures to substantiate this progress.
If we were evaluating infectious disease prevention -- AIDS, for instance-- decreases in disease prevalence would be required to conclude that public health interventions had succeeded. The relevant outcomes for nutrition science—prevalence of obesity, type 2 diabetes, fatty liver disease, cancer, cardiovascular disease and more—cannot remotely be seen as a success. This may be why the authors scarcely mention these diseases in their article. Another measure of healthy nutrition, US life expectancy, has declined over the past several years. 
The conventional rationalization for these disconcerting trends is that people simply fail to follow guidelines. The available evidence suggests otherwise. In the U.S., national survey data confirm that trends in both macronutrient consumption  and individual food categories  faithfully follow the prescriptions of the dietary guidelines.
Ultimately Drs. Mozaffarian and Forouhi argue for a primary role of prospective cohort studies in diet guidelines, when evidence from randomized controlled trials (RCTs) is ambiguous. But these cohort studies generate associations that can reliably infer causality only in rare instances when associations are so reproducibly large (e.g., the 10- to 20-fold increased risk of lung cancer in smokers vs. non-smokers) that plausible alternative causal explanations are unimaginable. For the far smaller associations of nutritional epidemiology, the question is, how do we know that causation can be reliably inferred? The existing evidence is not promising. A 2011 analysis of 52 claims made by nutritional epidemiology tested in 12 well controlled trials found that not one of the 52 claims—0%--could be confirmed.  A 2005 analysis by Stanford epidemiologist John Ioannidis concluded that highly-cited observational findings such as those in nutrition were confirmed by RCTs in only 20 percent of cases. 
Drs. Mozaffarian and Forouhi defend their position with a revisionist history of Hormone Replacement Therapy, employing a questionable post-hoc analysis to conclude that RCTs and observational studies are concordant. This conclusion is controversial.  More importantly, the authors tacitly assume that an example from pharmacoepidemiology can fairly be extrapolated to nutritional epidemiology. They further cite “1583 meta-analyses of 228 conditions,[that] find similar close concordance between randomised trials and observational studies.” These findings, too, address observations of pharmacological or other medical interventions, not nutrition. Yet these epidemiologic disciplines cannot, de facto, be assumed to be equivalent.
Nutritional epidemiology suffers from less precise assessments of exposure – due largely to unreliable, self-reported dietary information– and far greater potential for confounding. Dietary choices are associated with subtle socio-economic, behavioral and lifestyle factors, all potentially explaining the observed associations, This is why epidemiologic associations have traditionally been considered suitable only for hypothesis generation and insufficient as rationale for establishing policy.
Despite the problematic track record of nutritional epidemiology for inferring causality, Drs. Mozaffarian and Forouhi assert that “it is appropriate to act” on ambiguous evidence, because “As in all fields, application cannot await flawless evidence or perfect understanding.” But dietary guidelines are a form of preventative medicine, not like “all fields” of science as the authors claim. If dietary guidelines are wrong, healthy people die prematurely, a point made in this very journal in 1981 by the epidemiologist Geoffrey Rose in his seminal article on the strategy of preventive medicine. Unlike medical interventions for the sick (i.e., interventional cardiology, an example elaborated upon by Drs. Mozaffarian and Forouhi), for whom risks might be worth the calculated benefits, dietary guidelines are prescribed to healthy populations with the goal of preventing disease: “This kind of population-wide preventive medicine is predicated on the assumption that it will do no harm,” wrote Rose, “and that the population following it will, at the very least on average, be healthier and live longer. It is a very different proposition than curative medicine.” It is therefore self-evident that such preventive guidelines must meet exceptional standards of evidence.
The authors’ argument closely echos the one used in 1977 to justify the U.S. Senate’s first Dietary Goals for the United States, prescribing an increase in carbohydrate consumption and decreases in total fat and cholesterol, based on much the same ambiguous trial data bolstered by epidemiologic associations. “There will undoubtedly be many people who will say we have not proven our point,” wrote Harvard nutritionist Mark Hegsted in the Senate report. Asking “what are the risks?” Hegsted answered, “There are none that can be identified and important benefits can be expected.”  A chorus of experts, including the president of the National Academies of Science and the director of the National Heart, Lung and Blood Institute, urged restraint, noting the potential for tragic unintended consequences. Indeed, the total fat and cholesterol caps have since had to be retracted, while obesity and diabetes prevalence has exploded coincident with promulgation of the guidelines.
We would like to revive that call for caution and humility. Despite methodological advances, nutritional epidemiology remains fundamentally limited by its observational nature. Guidelines relying on this circumstantial evidence can be little more than educated guesses. While public health emergencies such as the current “diabesity” epidemics compel experts to enact some, if imperfect, advice now, the risk of unintended consequences remain. Enacting prevention guidelines based on lesser evidence risks repeating what might be a tragic history of doing more harm than good.
 BMJ 2018;360:k822 doi: 10.1136/bmj.k822 (Published 16 March 2018)
 Ernst J. Schaefer et al., “Lack of Efficacy of a Food- Frequency Questionnaire in Assessing Dietary Macronutrient Intakes in Subjects Consuming Diets of Known Composition,” American Journal of Clinical Nutrition 71, no. 3 (2000): 746–751
 Alan R. Kristal, Ulrike Peters, and John D. Potter, “Is It Time to Abandon the Food Frequency Questionnaire?” Cancer Epidemiology, Biomarkers and Prevention 14, no. 12 (2005): 2826–2828
 Somdat Mahabir et al., “Calorie Intake Misreporting by Diet Record and Food Frequency Questionnaire Compared to Doubly Labeled Water among Postmenopausal Women,” European Journal of Clinical Nutrition 60, no. 4 (2005): 561–565
 Arthur Schatzkin et al., “A Comparison of a Food Frequency Questionnaire with a 24-Hour Recall for Use in an Epidemiological Cohort Study: Results from the Biomarker-Based Observing Protein and Energy Nutrition (OPEN) Study,” International Journal of Epidemiology 32, no. 6 (2003): 1054–1062.
 Rose G. Strategy of prevention. The BMJ 1981 282:1847-1851
 https://archive.org/details/CAT10527234 p.3
 Select Committee on Nutrition and Human Needs of the U.S. Senate. 1977. Cardiovascular Disease. Vol. 2, Pt 1 of Diet Related to Killer Diseases; hearings before the Select Committee on Nutrition and Human Needs of the United States Senate. Ninety-Fifth Congress. Feb 1 and 2, 1977. Washington, DC: U.S. Gov Printing Office, Washington D.C., pp. 8-33.
 Handler, P. 1980. On Bias: Does Where you Stand Really Depend on Where You Sit? In Issues and Current Studies, 1-15. Washington, DC: National Academies of Sciences.
Competing interests: No competing interests