A Science of Nutrition?

John Tierney asks “Is Nutrition Science Not Really a Science?” He refers to the many nutrition related hysterias that seem to sweep the nation every few years, the most recent being the supposed discovery that trans-fats rank only slightly below arsenic in terms of their risk to human health.

I think the problem arises due to popular confusion between two different nutritional questions that science has asked. Science answered the first so successfully that lay people assumed that science could also answer the second. In truth, the second question proved much harder to answer.

The first question concerned the minimal nutrients needed by humans to ward off disease. This proved an easy question to answer because cause and effect occurred close together in time and most nutritional diseases result from the absence of a single critical nutrient. Testing a hypothesis about the cause of a nutritional disease took very little time and produced fairly unambiguous results. To prove that a Vitamin C deficiency causes scurvy a researcher need only find a test group with scurvy, give them Vitamin C and watch them recover. The same basic protocol proves that a Vitamin B1 deficiency causes beriberi or that a lack of Vitamin B3 and the amino acid tryptophan causes pellagra.

In the early 20th Century nutritional scientist quickly determined the causes and cures for dozens of nutritional diseases that had plagued humankind for all of its existence. By the 1960s, nutritional diseases had disappeared from the developed world. The public-health questions for nutritional scientists then switched to a much more difficult matter: people now ask scientists to tell them what to eat so as to optimize their long-term health and life span.

To answer this question with sure scientific rigor, scientists would have to create several test populations of several thousand individuals each, restrict each population to a specific diet and then follow their health outcomes over the course of the test subjects’ entire lives. As a practical matter, we can’t perform such experiments. We can’t ask individuals to sacrifice their entire lives for science, and no one wants to wait around for 70+ years for the answer.

Instead, scientists try to infer the correct answer from various types of proxy measurements. They change the diets of a few dozen volunteers for a few weeks or months and observe the changes that occur in their bodies. They try to statistically track large populations long term, based on study volunteers’ self-reports of their consumption. They try to compare various cultural populations that eat different diets.

While these techniques represent the best real-world tools we have, they nevertheless do not actually directly measure the phenomena we want to understand. Short-term diet changes do not necessarily reflect long-term outcomes. Statistics based on self-reports contain a vast margin for error. Cross-cultural comparison cannot control for hundreds of differences between the two groups. When it comes to the question of optimal nutrition we simply lack the practical tools to falsify any particular hypothesis. Without the ability to prove a hypothesis wrong, we never know for sure the degree to which it might be right.

Moreover, optimizing nutrition means understanding the hundreds, if not thousands, of tradeoffs between the conflicting needs of the body. Most studies focus on only one possible outcome of a diet, such as a reduction in heart disease, but do not test if reducing heart disease causes an increase in other diseases. For example, for three decades scientists warned people to limit their exposure to the sun in order to decrease the risk of skin cancer. They possessed firm inferential scientific evidence for that recommendation. Unfortunately, the studies that looked at the correlation between sun exposure and cancer did not look for the correlation between sun exposure and a decrease in other diseases. More recent research infers that reducing skin cancer by reducing sun exposure leads to a significant increase in the more lethal cancers of the intestines.

The consequences of suboptimal nutrition frighten us so that we demand that scientists give us concrete, definitive answers. If one scientist answers honestly, “I don’t know for sure,” we will keep looking until we find a scientist who will give us a definitive answer. It becomes easy for a scientist to build a political empire by emphatically pounding the table and declaring their iron certainty that eating A results in outcome B. After all, no other scientist possesses the tools to conclusively prove them wrong. Answering the scientific questions then becomes a social and political function and not a scientific one.

Making decisions based on imperfect understandings can cause more harm than good. For example, the current widespread us of trans-fats resulted from a previous hysteria in 1980s, when the same people who now war against trans-fats demanded that food providers replace the fantastically dangerous saturated fats with safe trans-fats. They based their argument on the supposedly concrete scientific understanding of the risk of saturated fats at the time. Today, we view saturated fats much more favorably than we did then, because we possess a more nuanced understanding of the role of cholesterol in the body.

We need to remember that just because scientists struggle to study a phenomenon it does not immediately follow that the current generation of scientists have the tools they need to create a testable and predictive model of the phenomenon. We need to understand when and when not a scientist or politician overstates the degree of certainty of a particular finding. We need to accept that even with all our scientific power, the true scientific answer often remains, “We do not know.”