Guest Essay by Kip Hansen — 18 January 2024 —  400 words/5 minutes

According to the American Medical Association, in an article that is part of JAMA’s “What Doctors Wish Patients Knew™” series:

“Research has shown that diets high in ultraprocessed foods are linked to more than 30 health conditions, according to an umbrella review of meta-analyses that was published in The BMJ. Meanwhile, more exposure to ultraprocessed foods was associated with a higher risk of dying from any cause. There were also strong ties with higher consumption and cardiovascular disease-related deaths, mental health disorders and type 2 diabetes.”

This statement is based on reviews of many published nutritional epidemiology studies (hundreds of them).   There were 45 review studies or meta-analyses on UPFs published in 2024 alone.

Nutritional Epidemiology?

World-class statistician, William “Matt” Briggs,  author of the book “Uncertainty: The Soul of Modeling, Probability & Statistics”, tells us, in no uncertain terms, that:

“Epidemiology is the field which officially mistakes correlations for causations.”

This is a very serious accusation.  It is also true.  And it gets worse and worse.  It does so as a result of the very definition and practices of nutritional epidemiology:

“Nutritional Epidemiology

Nutritional epidemiology is the application of epidemiological methods to the study of how diet is related to health and disease in humans at the population level. … Epidemiologists investigate how nutrition affects the occurrence of disease by collecting data on and comparing large groups of people. Statistical methods are employed to estimate the extent to which a factor influences risk of disease in a population. This estimate is often expressed as a measure of association.”

“In epidemiological research, diet can be studied at different levels comprising intake of nutrients, foods, food groups, and/or patterns. These exposures can be measured by directly ascertaining what people eat (e.g., through the administration of questionnaires), by measuring markers of intake in biological specimens, or by estimating body size and the relative size of body compartments.”

“The exposure measure of interest in nutritional epidemiology is usually long-term diet, since the effects of intake on most health outcomes, especially those related to noncommunicable diseases, are likely to occur over extended periods.”

The abstract of that chapter from the Encyclopedia of Food and Health, 2016, reads more succinctly:

“Nutritional epidemiology is the application of epidemiological methods to the study of how diet is related to health and disease in humans at the population level. This article reviews key issues in the field of nutritional epidemiology, including a description of methods to assess dietary intake, sources of variability in diet, the relevance of total energy intake to epidemiological analyses, and errors that may arise in measuring dietary exposures.”

The problems of nutritional epidemiology appear obvious at once:

1.   “…the study of how diet is related to health and disease in humans at the population level.” 

            If these population-level relationships are attempted to be applied to individuals, we encounter the ecological fallacy: a logical error that occurs when you make assumptions about individuals based on group data.

2.   “… This estimate — extent to which a factor influences risk of disease in a population — is often expressed as a measure of association.”

            A measure of association, a correlation, is not a risk – it is a statistic.  True risks come from exposure to a causal factor and its dose, and for that, researchers need to measure the dose received by an individual who has had an exposure to an hypothesized casual factor.  Hypothesized causal factors need to be backed by biological plausibility.

For UPFs, ideas such as the purposes for certain processing steps or corporate ownership of the manufacturer of a food item cannot be,  and are not,  biologically plausible causal factors for the myriad of negative effects claimed for UPF consumption.

3.  “…These exposures can be measured by directly ascertaining what people eat (e.g., through the administration of questionnaires)…” 

            Exposure can be directly measured, but  Food Frequency Questionnaires (FFQs) and 24-hour dietary recalls (24HRs) do not directly measure exposure to or dosages of foods in a person’s diet. They are, at very best, vague estimates of dietary intake. Many nutritional epidemiological studies are based on a single “in the last year…”-type  Food Frequency Questionnaire.  Or even on a single, 24-hr Dietary recall.  These FFQs are then treated as real-world exposures and dosages.  Often participants are subsequently assumed to have the same diet over the following years or even, in some cases,  decades.

4.  “The exposure measure of interest in nutritional epidemiology is usually long-term diet,…”

          As in the previous point above, the exposure to the measure of interest, in the present case, UPFs, is almost never actually measured in any scientifically defensible way.  And, except in one or two clinical trials, dose, in practice, was never measured.

5.  “…sources of variability in diet”….  “and errors that may arise in measuring dietary exposures.”

          People’s diets do not remain the same for years and decades.  And diet is a very difficult thing to measure outside of an institutional setting.  Changes in marital status, age, becoming a parent, moving from university to professional life, changes in economic status, changes in employment, moving from one region or country to another – all these and many more factors directly change the day-to-day diets of most people.

How does this all apply to the studies on Ultraprocessed Foods?

In Modern Scientific Controversies:  The War on Food: Part 2, What are UPFs?,   we saw that the titular question was very difficult to answer definitively.  The definition is vague and based on nutritionally irrelevant factors, primarily relying on the “NOVA food groups: definition according to the extent and purpose of food processing”,  such as the purposes of processing  steps used in the food manufacturing: ensuring  product longevity, taste, palatability, enjoyable textures, etc. 

And, for that reason, exactly which diet items on diet questionnaires [FFQs/24HRs] were to be counted as UPFs in the various research efforts have varied over time and were often subjective [as in “influenced by or based on personal beliefs or feelings, rather than based on facts”] and thus different in each study.

None of the items in the lists of UPFs are distinct foods with distinct nutritional qualities – nutritional value is not considered in the definitions.    The items on the UPFs lists do not share common ingredients.  The items on the UPFs lists do not share common processes.  Many anti-UPF claims contain heavy doses of anti-transnational corporatism.

In fact, the whole subject is scientifically unsuitable for investigation.  To be scientifically sensible, the food item (or food group)  to which study participants are to be, or have been, exposed at some dosage level must be measurable and, if a number of items are being considered to be a group, the items must be commensurable—it must be a homogeneous group—a group of like things. 

It is impossible to measure the “purposes-ness” or the “ownership-by-transnational-corporation-ness” of varied items for sale in the grocery store. There is no acceptable metric for those concepts.   It is impossible to measure the relative “UPF-ness” of items sold in grocery stores, such as these ubiquitous grocery-store breads for instance:

We can only scientifically be interested in the dosage received of the particular dietary item of interest and that dosage needs to be measured. Dosage is not a binary measure – not yes/no. 

Even known poisons depend on the dosage – some have benefits from very small doses while larger doses kill – this is called hormesis. 

All the UPF prospective cohort and observational studies (cross-sectional, case-control, and cohort) and the review and meta-analyses and even the huge umbrella review of epidemiological meta-analyses (Lane et al. 2024)  suffer from the faults detailed above.   

And what is the result of all that?

A pseudo-science fad — the anti-UPF movement — that has infected nearly

every level of the health and nutrition science world.

# # # # #

Author’s Comment:

Let me speak as plainly as I dare:  the anti-UPF movement is pure “Fad Science”.  It is wasting untold research monies to bolster an ideological fight targeting the foods produced by transnational corporations on the trumped-up idea that they share something bad hidden somehow in “the extent and purpose of food processing”.

There are some voices trying to stem the rush to condemn all UPFs.

Please, it is important to understand that UPFs does NOT MEAN JUNK FOOD.  That is an entirely different concept and classification. 

Nor are UPFs the same as foods containing  “additives” or ”artificial colors” or ”preservatives”  — foods can be and are classified as UPFs even when lacking those.

UPFs are such an all-encompassing classification that it basically includes “nearly everything sold in your grocery store in a package, plastic bag, or box”.

The anti-UPF movement is a War on Food.

Thanks for reading.

# # # # #