Maintenance Phase Reality Check: Ultra-Processed Foods

This episode is complex, so I’m breaking it into two fact-checks. This covers the first half, up until the point where Michael and Aubrey discuss the book Ultra-Processed People and the concept of food addiction. Stay tuned for part two!

Before I begin: I eat and enjoy many UPFs and don't believe they're uniformly terrible - neither do nutrition researchers. No respected scientist claims you must eliminate all UPFs to be healthy. However, this doesn't make the UPF concept "useless" (as Michael suggests). Notably, Michael wrote a piece for HuffPost in 2013 excoriating processed foods and their corporate manufacturers, describing Wonder Bread as “27 ingredients, half a teaspoon of sugar and 7 percent of your daily allowance of salt in every slice, lasts on the shelf for two weeks.” While people can change their minds with new information, Michael's current rhetoric treats anyone sharing concerns he once held as fundamentally naïve, as if holding previously championed views makes you an obvious mark for pseudoscience. It would also be interesting to know what exactly has changed his mind about corporate interests in the intervening years.

I have to admit I found this episode on ultra-processed foods a bit jarring. Michael and Aubrey have built the Maintenance Phase podcast around the concept of dismantling personal responsibility narratives - the reality is that individual food “choices” are often the result of systemic forces beyond personal control such as price, availability, convenience, and marketing (I completely agree with them on all these points). However, they sidestep the fact that food corporations deliberately engineer products to maximize consumption through calibrated combinations of salt, sugar, fat, and additives, and dismiss concerns about hyper palatability as ridiculous - despite it being a frequently used research term. They also characterize arguments about the profitability of addictive food design as ridiculous, despite this being an industry norm (also, fun fact: cigarette companies pivoted to the food market and applied their existing knowledge about creating addictive substances to drinks and foods).

MP has previously argued that pharmaceutical companies manufacture demand for weight loss drugs through manipulative marketing, yet here they suggest it's absurd to think food companies might manufacture demand through manipulative formulation. They'll acknowledge that poverty limits access to fresh produce, but won't engage with how economic constraints funnel people toward engineered foods designed to override satiety signals. In previous episodes they cite the ~$70 billion/year weight-loss industry while ignoring that snacks and confectionary products in the US generate hundreds of billions annually (note: depending on methodology, different market research firms have different figures for this, but they are all substantially greater than the $70 billion figure that MP cites - which I also do not know the methodology for).

This selective blindness to corporate influence represents more than just an intellectual inconsistency - it undermines their own central thesis. If we reject the personal responsibility framework around weight and health, we need to be willing to examine all the institutional forces at play, not just those that don't implicate our favorite snack foods. The food industry's own internal documents reveal decades of research into "bliss points" and "craveability" - but acknowledging this reality would undermine some of the podcast’s core positions.

Mostly, though, Michael and Aubrey's critique reveals a profound misunderstanding of what NOVA actually is and why it exists. The NOVA classification system is a research tool, developed because scientists hypothesized that the increasing prevalence of highly processed foods in our diets was causing excess morbidity and mortality. To test this hypothesis, researchers needed a way to systematically categorize and measure exposure to these foods across populations.

This is fundamental epidemiological research methodology. When you want to study the health effects of any exposure (e.g., smoking, air pollution, dietary patterns), you first have to define and classify that exposure in a consistent, measurable way. NOVA represents an attempt to capture the exposure of interest (industrially processed foods) so researchers can assess associations between that exposure and health outcomes.

Like any scientific tool, NOVA isn't 100% perfect. Edge cases exist, classifications can be debated, and refinements are ongoing. But defining an exposure is integral to designing studies that test hypotheses. Epidemiological research often uses proxy measures - tools that capture some underlying exposure we can't directly quantify. For example, when researchers study health outcomes associated with fruit and vegetable consumption, they're not interested in the act of eating fruits and vegetables, but rather the nutrients those foods provide (e.g., fiber, vitamins, minerals). "Fruit and vegetable consumption" serves as a proxy for a complex nutritional exposure that would be impossible to measure directly across large populations.

Similarly, NOVA classification serves as a proxy for myriad factors that accompany ultra-processing: the disruption of food matrix, the additives, and the engineering for hyper palatability.

There is an interesting discussion to be had about the physiological, social, emotional, and economic impacts of the modern day food industry, but the positions and representations made in this episode feel like they’re completely unaware of the body of existing legitimate research around this topic. This puts them in a position where they are attempting to discuss research involving complex behaviors and neurophysiological processes using what appears to be their “gut feelings” - not too different from the anti-vaccine crowd’s approach to research around vaccines.

So, here goes Part 1. For those who want to see the full spectrum of how much was wrong in this episode, a line-by-line fact check is below (note that this only includes things that are wrong or that I have comments on). I have included periodic time stamps to help listeners find each quote:

4:51 Michael: So, the first thing to know about the term “processed food” is that it's been around much longer than I knew. So, I went on that Google Ngram thing, that is how much is this term being used? The first reference that it has for processed food is from 1912.

I know this is a nitpick, but it reveals their pattern of getting details wrong: If you search “processed foods” in the Google n-gram Viewer, the first reference that comes up is in 1801, not 1912.

Michael: You can find old articles in the New York Times, decrying processed food and how it's harming people. One of the first articles I found was from 1970 called “Bread Is Fatal to Rats, But That's Not the Point.” So, this is a study where they fed rats ordinary white bread and of course, refined flour. They remove a lot of the fiber, a lot of the nutrients, etc. If you feed rats just white bread, they do die. Like, they starve to death, basically because there's not enough nutrients. However, if you feed them an enriched white bread that has these vitamins and minerals put back in, they live. Yeah, but what's interesting to me is, in this article, from the very beginning of this term, nobody could really define what processing is because one way to think about it's like, okay, you're processing the wheat, you're taking out all the nutrients. But also boiling down foods to get the nutrients out and turning them into powders and putting them into bread is also processing. It's arguably more processing.

This article is only tangentially about “processed foods” (the term is used once, in passing). It is primarily a piece covering ongoing debate amongst scientists and the FDA about enriching vs. not enriching store-bought bread.

Michael is misrepresenting both the study mentioned in the article and the argument of the article itself. In the study by Dr. Roger Williams (I can’t find the published results, so if anyone finds them, please let me know), rats were fed either: 1) commercial white bread labeled as “enriched” (contrary to Michael’s claim that it was “ordinary white bread”), or 2) “an improved bread made from flour with the following supplements: magnesium oxide, manganese sulphate, copper sulphate, calcium, phosphate, folic acid, vitamins A and E, cobalamine pantothenate, pyridoxine and the amino acid lysine, which is a vital link in the protein chain.” At this point, it seems pretty obvious to me that this study is not concerned with “processed” vs “unprocessed” - it is focused on the nutritional value of commercially available bread. The single reference to “processed foods” comes towards the end of the article, where it says, “Dr. Barnes suggested that maybe all processed foods should be studied for their nutritional value.” There is no suggestion that the “improved bread” is not “processed”. In fact, in the article about the study itself (linked above), it is very clear that Dr. Roger Williams considers his “improved bread” processed, because he says, “additives must ‘be used judiciously.’”

Michael: The term “ultra processed” is coined in 2009 by a Brazilian researcher named Carlos Monteiro, who had been doing fieldwork in Brazil in the 1970s and 1980s, originally on malnutrition. But he was noticing with these poor populations, eventually the problem of malnutrition had started to shift to what he calls overnutrition.

While a cursory internet search will suggest that “ultra processed food” was coined by Monteiro, there is a history of the term dating back to 1982 in an Esquire article titled, “The Universal Salesman.” I think this is quite interesting, because it reveals that the concept of UPFs, while not yet defined scientifically, was already circulating in culture. The text reads:

Not long ago, my wife and I entertained a few of her advertising colleagues, one of whom is a fabulous cook. This woman, whom I’ll call Beth, installed herself in the kitchen on Friday evening, and from then until Sunday she produced spectacular meals at the rate of one every' three to four hours. A stickler for fresh ingredients, Beth used only local produce, herbs from the garden, and critters snatched from the sea that day. On Monday morning, we all knew, she’d be returning to her ad-land post, helping to market one of those ultra-processed foodstuffs whose labels read like a chem text and that bear only the most tenuous resemblance to anything found naturally under God’s blue sky—the sort of product that would never find a place in Beth’s own pantry.

And a few years later, the term popped up again in a book titled, “Superimmunity for Kids” (a very dated book with a lot of bad advice and some wacky recipes). This is from page 2:

Far too many nutritional consultants are operating on the basis of theories unsupported by scientific research. Pediatricians, the main source of feeding advice to mothers, generally receive scant nutritional education themselves. And in any case, most of the principles we all learned in school are either incorrect or inappropriate to an era of ultraprocessed and manufactured foods.

Anyway, this is just to provide some of the context Mike didn’t include, because I think context is important. The idea of UPFs didn’t just come out of nowhere in 2009 - it’s been floating around for a while. For further reading, I recommend this article, which does a great job of presenting the full timeline of opposition to processed foods.

9:41 Aubrey: God, the whole time you're talking about this, I'm just thinking about other foods that are ultra processed, protein powder. Protein powder that's ultra processed.

Very unclear what Aubrey’s point is, here. Protein powder can have tons of preservatives and additives in it. It very clearly fits the definition of ultra processed. I think she’s trying to suggest that it’s “healthy” and therefore can’t be a UPF, but that mostly reveals Aubrey’s lack of understanding of nutrition and the concept of UPFs.

10:10 Aubrey (discussing content of Monteiro’s paper): “Modern diets usually do contain some unprocessed plant foods and meat and milk, but also keep several of the unhealthy features of the processed ingredients they are mostly based on, low-nutrient density, little dietary fiber, and excess simple carbohydrates, saturated fats, sodium, and trans fatty acids. What makes snacks, drinks, dishes, and meals mainly made up from the ultra processed foods different from traditional dishes and meals is that they are inalterable. They come ready to eat or heat. Diets that include a lot of ultra processed foods are intrinsically, nutritionally unbalanced and intrinsically harmful to health.”

This is paraphrasing, not a quote from the actual paper. It is important not to misrepresent primary sources. The paper says, “This explains the problem with modern diets that contain a lot of the ultra-processed foods in group 3. While these diets usually do contain some group 1 plant foods and meat and milk, they usually keep several of the unhealthy features of the group 2 ingredients they are mostly based on: low nutrient density, little dietary fibre, and excess simple carbohydrates, saturated fats, sodium, and trans fatty acids. They are also energy-dense. What makes snacks, drinks, dishes and meals mainly made up from the ultra-processed foods in group 3 different from traditional dishes and meals that also use group 2 ingredients, is that they are inalterable; they come ready-to-eat or -heat. Diets that include a lot of ultra-processed foods are intrinsically nutritionally unbalanced and intrinsically harmful to health.”

10:56 Aubrey: Like, you can say this, but we just talked about enriched flour and enriched cereals, right? That like, growing up, yes, absolutely breakfast cereal was everywhere, but all of that breakfast cereal was like very prominently labeled as being enriched with vitamin A, vitamin C, vitamin B… I don't know that you can argue necessarily that it's intrinsically, nutritionally unbalanced. Again, Athletic Greens is an ultra processed food that advertises itself as being nutritionally balanced.

Ironically, Aubrey is proving the point by attempting to disprove the point. Do you know why companies started enriching/fortifying foods with vitamins and minerals? Because those vitamins and minerals were removed during the processing of the ingredients. The goal was to improve nutrition without requiring Americans to change their diets. The National Nutrition Conference for Defense in 1941 recommended: “improving the nutritive value of certain low-cost staple food products, such as flour and bread, by enrichment with nutritive elements that have been removed from them by modern milling and refining processes. Pending further developments in the milling of grains so as to retain their full, natural, nutritive values, enrichment is an economical way to improve American dietaries almost universally, without interfering with deeply ingrained food habits. The method, however, should be used with discretion and only on the basis of findings of medical and nutritional experts.” These foods absolutely are “intrinsically nutritionally unbalanced” and that’s why they were fortified and enriched.

Also, “nutritionally balanced” is not an FDA-regulated term - literally anyone can slap that on their product. And the fact that the hosts have bought into the Athletic Greens marketing has nothing to do with UPFs.

11:31 Michael: There's also this core problem that you find in every single paper about this where it says, “Okay, it's not about the nutrients, it's not about what's inside of the food. It's about the process by which the foods are made.”

I would love to see a paper that says “it’s not about the nutrients.” What does Mike mean by “it”? What, exactly, is not about the nutrients? Mike has mentioned nutrients already multiple times while talking about this. The 2009 paper he is quoting from pretty explicitly acknowledges that nutrients are important, “How foods affect health depends on a number of factors, one of which is their relative importance within diets. One important factor is food nutrient density (nutrient per energy unit) and food energy density (energy per volume).”

Michael: And then it's like, “Okay, why are ultra processed foods bad?” It's about what's in the food. So, he's saying they're high in fat, they're high in sugar, and they're calorie dense.

Again, this is a complete misrepresentation of Monteiro’s work. If you read the 2009 paper, it’s very clear that Monteiro is outlining the way that UPFs alter eating behaviors and encourage excess caloric intake: “[UPFs] are ‘fast’ food, designed to be portable, convenient and accessible. They induce eating patterns such as ‘grazing’ and skipping main meals, eating when doing other things such as watching television, driving a car or working, and eating alone. Extremely convenient packaged products such as caloric soft drinks have created diets in which a substantial amount of energy comes in liquid form.”

12:16 We finally get to the meat of this paper and just such fucking Mike bait. The attempt to actually operationally define what ultra processed food is. So, he proposes in 2009, this classification that has three groups. Every single person who writes about ultra processed food has the same paragraph where they're like, “Well, all food is processed.” Slicing up an apple is processing it, but it has no nutritional content whatsoever. But also, Pringles, making a slurry of potato and then shaping it into chip shape is also processing. It's like, this term encompasses such a wide range of activities that it's probably better to just find a different term.

This is a terrible argument. A lot of terms have a wide range of things that fit into that. Should we use a different word for “vegetable” because a lot of foods fall into that category? No, the whole point is that it encompasses a wide range of activities and we characterize within that category to get greater nuance and detail.

This is the problem. He's saying these are forms of processing, but they don't count. It's not about whether foods are processed. It's about the intensity of the processing.

Nowhere does Monteiro say that those forms of processing don’t “count” (what does that even mean?).

13:41 Aubrey: Yeah, it just feels goofy to be like, “The dried beans are not processed, but you put them in a can and then they are.”

15:16 Aubrey (defining UPFs): These are made up of Group 2 substances to which either no or relatively small amounts of minimally processed foods are added, plus salt and other preservatives, and often also cosmetic additives such as flavors and colors. This group of foods includes breads, cookies, ice creams, chocolates, candies, breakfast cereals, cereal bars, potato chips and savory and also sweet snack products in general, and sugared and other soft drinks.

Michael: That's the kind of you know it when you see it thing.

This dismissal of this definition as a “know it when you see it thing” reveals a fundamental misunderstanding of what UPF classification is intended to capture. The difference isn't arbitrary - it's about the degree of industrial transformation and the types of ingredients involved. Let’s use beans as an example, because Michael called it “goofy” to say that dry beans are not processed but canned beans are (note: dried beans are also minimally processed, so this is a bad example, but Michael brought it up so we’ll go with it.). When you put dried beans in a can with water and salt, you preserve the original “food matrix” (the physical structure that holds nutrients together) - the beans retain their fiber structure, protein composition, and nutrient density. The canning process involves heat treatment for safety, but the end product remains recognizably beans with minimal additives.

In contrast, UPFs undergo extensive industrial processing that breaks down the original food matrix and reconstructs it using ingredients like isolated proteins, modified starches, industrial fats, and synthetic additives that don't exist in the average home kitchen. Going back to the bean example, an ultra processed bean product could be bean protein isolate reconstituted with emulsifiers, stabilizers, flavor enhancers, and preservatives.

Food processing affects how it interacts with our physiology. UPFs are often engineered to be rapidly consumed and absorbed, bypassing normal satiety mechanisms. The food matrix plays a crucial role in how our bodies process food, and ultra-processing deliberately dismantles this structure.

Calling this distinction “know it when you see it” is like saying it's arbitrary to distinguish between an orange and orange-flavored drink powder because they both contain "orange." The level of processing fundamentally changes what the food does in our bodies, which is precisely why scientists have been working to create a comprehensive classification system.

16:10 Aubrey (paraphrasing the definition of UPF): Ultra processed foods are basically confections of Group 2 ingredients, typically combined with sophisticated use of additives to make them edible, palatable and habit forming. They have no real resemblance to Group 1 foods, although they may be shaped, labeled and marketed so as to seem wholesome and “fresh.” Unlike the ingredients Included in Group 2, ultra processed foods are typically not consumed with or as part of minimally processed foods, dishes and meals. They are designed to be ready to eat sometimes with addition of liquids such as milk or ready to heat, and are often consumed alone or in combination, such as savory snacks with soft drinks, bread with burgers. It really feels like a very vibey definition.

The definition Aubrey read seems straight-forward to me: UPFs are foods that are primarily composed of industrial ingredients, combined with additives to enhance palatability and shelf-life, designed for immediate consumption, and engineered to replace rather than complement whole foods.

As an exercise, let’s compare this to the DSM-5 criteria for major depression, which includes symptoms like "depressed mood," "loss of interest or pleasure," "feelings of worthlessness," and "diminished ability to think or concentrate." These are based on subjective interpretation and clinical judgment, but they would never dismiss depression as a "vibey" diagnosis. The UPF definition is actually more concrete, because it’s based on observable ingredients and processes, not subjective experiences.

Maybe more importantly, they’re critiquing this definition as if it were meant to be the final word on food classification, when the paper they’re citing explicitly says: "A short commentary cannot be comprehensive, and a general proposal such as that made here is bound to have some problems and exceptions... Readers' comments and queries are invited." This is how scientific inquiry works - researchers propose frameworks, acknowledge their limitations, and invite refinement through peer review and further research. As you know if you listened to this episode, the NOVA classification system has since been refined. What Michael presents as vagueness is, in my mind, actually scientific humility and invitation for collaborative improvement.

16:55 Michael: The core definition is that ultra processed foods are made up of Group 2 substances to which either no or relatively small amounts of minimally processed foods are added. So, the idea is that these products are majority, like oil and fat. Something like Nutella, which is 13% hazelnuts. And effectively everything else is just oil and sugar.

This is a great time to point out that the opinion piece that Michael is relying heavily on here was written by Michael Gibney, who was a nutrition researcher in Ireland and served on the Nestlé Nutrition Council for over 30 years (talk about a conflict of interest!). Given how often Michael and Aubrey mention conflicts of interest in pharma trials, it is an oversight for them to ignore this.

It’s also entirely incorrect to say that UPFs are “majority” oil and fat (not to mention that oil is a fat). The definition is very clear that they contain starches, sugars, hydrolyzed proteins, etc. Nowhere does this paper suggest that UPFs are mostly fat.

Lastly, small and unimportant correction, but an easy Google search shows that Nutella is only 50% sugar and palm oil.

Aubrey: But Michael, that's not typically consumed with or as part of minimally processed foods, dishes and meals [Michael laughs] which means “I'm sorry, you never have dipped a banana into Nutella.”

This is a misrepresentation of the definition of UPFs. The definition is very clear that typically UPFs are not eaten with minimally processed foods. That doesn’t mean something that is eaten with minimally processed foods occasionally (or even regularly) is ipso facto not a UPF.

17:30 Michael: You know, earlier he said, “The problem with ultra processed foods is that they're very high in sugar, they're high in fat, they're very calorie dense.” But then he includes things in ultra processed here that are not particularly high in sugar or fat or energy dense, like all breads.

False. Michael is conflating nutrient density and energy density. The paper says, “One important factor is food nutrient density (nutrient per energy unit) and food energy density (energy per volume). Commonly consumed foods with low nutrient density (of protein or vitamins, for instance) or high nutrient density (of saturated fat or sodium, for instance), as well as with extreme energy densities, unbalance diets and cause either nutritional deficiencies or chronic diseases (for example, obesity, dyslipidaemias and hypertension), or both.” Both are issues - UPFs are not necessarily energy dense.

Michael: You're including tortillas. This also excludes a lot of foods. This definition does not include potato chips because, I looked at Lay's potato chips, like Lay's original potato chips have three ingredients… Potatoes, oil and salt. And lots of French fries also are just three ingredients. Right. It's potatoes, oil, and salt. That's the canonical food you shouldn't be eating so many potato chips, you shouldn't be eating French fries. But those actually count as minimally processed foods under this definition… You're saying ultra processed foods are bad for you. But then you have this definition of ultra processed foods that includes a lot of foods that are not particularly bad for you. And then you have this definition of unprocessed foods that includes a ton of foods that are bad for you or like, at least they're calorie dense, energy dense.

Michael is essentially arguing that because the system doesn't perfectly align with his personal intuitions about which foods are "good" or "bad," the entire framework must be flawed. He is misunderstanding ultra-processing as a proxy for nutritional quality, when it's actually measuring the degree of industrial transformation. Lay's chips may have a simple ingredient list, but this doesn't make them nutritionally equivalent to a baked potato, for example. And, importantly, eating behavior also differs when comparing a baked potato to Lay's chips. The NOVA classification system isn't claiming that all minimally processed foods are “healthy” or that all UPFs are “unhealthy” - its purpose is to identify foods that have undergone extensive industrial manipulation, often involving ingredients and processes unavailable in the average home kitchen.

I’m also not entirely sure what Michael means by calling potato chips and french fries "bad" and dismissing foods as problematic because they're "calorie dense." Historically, Aubrey and Michael correctly argue against food moralization and the reductive focus on calorie density as a measure of food's worth. I’m not sure why ‘good’ and ‘bad’ foods are suddenly in play when examining food through an industrial lens rather than a nutritional one.

From an uncharitable perspective, it seems like a podcast that purports to challenge simplistic nutritional thinking is rejecting a nuanced classification system because it doesn't conform to conventional diet culture wisdom about which foods deserve moral condemnation.

I was thinking, because he includes milk in the unprocessed category, like a Crème Brûlée would count as like, unprocessed even though it's extremely calorie dense, it's still mostly cream, right?

18:43 Ice cream always makes the list of ultra processed foods. But I looked this up Häagen-Dazs ice cream has five ingredients. It's like cream, sugar, vanilla. There's nothing you can't pronounce in there.

This seems to be tied into the confusion about good vs bad foods: to reiterate, this classification system is measuring the degree of industrial processing a food has undergone. Sure, traditional crème brûlée made with cream, eggs, sugar, and vanilla would be classified as processed (not unprocessed, as he suggests), but a store-bought crème brûlée loaded with stabilizers, artificial flavors, and preservatives would be ultra-processed. Same with ice cream. Sure, ice cream is generally considered a UPF because it contains stabilizers, etc. That does not mean all ice cream is a UPF. The classification depends on the actual ingredients and preparation methods, not the food category. Just because Michael can find an example of an ice cream without additives doesn’t mean that most ice creams don’t contain additives. You can look at almost every other Häagen-Dazs flavor and see what I mean. For example, the ingredients for their Rocky Road (my personal favorite flavor of ice cream) are: Cream, Skim Milk, Sugar, Corn Syrup, Almonds, Cocoa Processed With Alkali, Egg Yolks, Corn Starch, Safflower Oil, Egg Whites, Cream Of Tartar, Salt, Pectin, and Natural Flavor. (Also, Häagen-Dazs is definitely a higher-end ice cream - at least where I live.)

Michael does seem to understand this on some level, despite the previous quote. Right before minute 42:00 he says, “Cake can either be ultra processed or not ultra processed. It depends on the f**king cake.”

18:56 Michael: But then the second problem with this definition is that it includes these concepts that are just not related to human nutrition. So, he goes into this whole thing that the real problem with ultra processed foods is that they're profit maximizing. They're produced by large corporations. They're these international commodities.

Nowhere does the article say that the “real problem” is about profit maximizing.

19:18 Michael: So, he says “Ultra processed products are typically branded, distributed internationally and globally, heavily advertised and marketed, and very profitable. But the problem with this is that fucking fruits and vegetables and food that is good for you is also very profitable and [laughs] also produced by fucking international global corporations.” If you go to the grocery store and get strawberries, they're going to be from fucking Driscoll’s. Driscoll's is a massive corporation.

Driscoll’s has a yearly revenue of approximately $5 billion USD while Mars Snacking (the company behind Snickers, Skittles, Twix, Kind Bars, Wrigley gum, etc.) has a yearly revenue of approximately $18 billion USD (over one third of the broader company’s $50 billion USD per year). Nestlé’s Confectionery division brings in approximately $10 billion USD per year, and Unilever’s ice cream division brings in about that much, too. Calling Driscoll’s a “massive corporation” in the context of the snack behemoths is a bit of an overstatement.

No one is argues that “healthy” food can’t come from global corporations - it’s about certain corporations making billions of dollars by manufacturing highly palatable, convenient, shelf stable, cheap foods that replace less processed options while targeting marketing to lower income and marginalized populations (ref 1, ref 2), compounding pre-existing health disparities.

19:44 Aubrey: I mean, I think here's the interesting thing here. There is a critique to be had about the behavior of any number of multinational corporations. That criticism of corporate behavior isn't the same thing as proving that there are negative health effects as a result of that bad corporate behavior. Again, they're trying to ride the coattails of this makes sense to you, right? Yeah, you know that corporations are bad, so they're probably also bad for your health.

Sure, you can critique any number of multinational corporations - that doesn’t mean that the critiques are necessarily invalid. The prevalence of these foods and the associated negative health effects are intrinsically linked to the behaviors of the corporations: they are engineering foods that make people want to eat more of them (ref 1, ref 2, ref 3).

Aubrey’s statement is akin to saying that while it was shitty that Juul marketed their vapes to teenagers, the fact that teenagers used them so extensively and had negative health consequences is unrelated. That would be an absurd argument, no?

20:13 We found this in the Michael Pollan book too, that people keep presenting these dietary choices as somehow a break from capitalism or somehow virtuous in all of these other larger economic ways. And they just aren't. Food can be healthy and produced by miserable corporations. This is the way that we've chosen to structure our economy. You cannot escape from this by buying virtuous food. I think this is a huge mistake in the way that people frame this stuff.

Dividing individual foods into “healthy” and “unhealthy” categories isn’t a particularly helpful way to approach nutrition. The point of the NOVA framework is to assess overall dietary patterns and the resulting physiological implications. The way Michael and Aubrey chose to engage with this topic is not what I would have expected, given that they have previously pointed out systemic obstacles to accessing fresh foods, storage and food preparation spaces, and opportunities to walk or bike or participate in activities. Dismissing the role of corporations in driving changes to food consumption patterns as “this is how we’ve chosen to structure our economy” is akin to dismissing attempts to create more walkable cities by saying “we’ve chosen to structure our transportation system around cars”. It’s also relevant to note that a company that manufactures food has two main ways to increase profit (aside from increasing profit margins): get more people buying the food, or get the same people to buy more food - and UPF engineering serves both goals.

20:39 Aubrey: And if you can, then the escape is only an escape that is available to people who can afford it. Maybe the push then should be, a, we first have to establish that there is hard and fast evidence that this is like actively uniquely bad for you. And B, then I think the task becomes, then you regulate healthier products.

Aubrey is entirely correct here about point A; she just doesn’t realize that we already DO have extensive research suggesting that consumption of UPFs is associated with negative health outcomes. Interestingly, Aubrey’s point here is eerily similar to the historical discourse around smoking cigarettes from the tobacco industry, namely their denial that there was any “hard and fast evidence” that smoking was “actively uniquely bad” for people. Despite overwhelming observational evidence and even a Surgeon General’s warning, the tobacco industry used their money and influence to double down and manipulate the media and public perception. This is from a 1954 Tobacco Industry Research Committee statement: “...statistics purporting to link smoking with the disease could apply with equal force to any one of many other aspects of modern life. Indeed the validity of the statistics themselves are questioned by numerous scientists.” We know that cigarette companies bought food companies and applied their existing research and marketing knowledge to this new area of industrial ready to eat food production.

When Aubrey says “And if you can, then the escape is only an escape that is available to people who can afford it,” she seems to be suggesting that discussing the potential harms of UPFs is disparaging to lower income individuals. A more productive path would perhaps be to engage with the existing knowledge around the effects of dietary patterns high in UPF, help people understand potential risks, advocate for systemic changes to ensure that UPF is not the only option they have available, and then let individuals make their own (informed) choices.

21:00 Michael: So, as we just covered this-- this definition is not all that useful. I think this paper is like actually quite bad, like shockingly bad considering it like began this entire field, it just obvious contradictions. You know that I love a petty, academic paper, Aubrey. I love like peer reviewed, hmm, interesting.

What Michael calls “contradictions” are actually just Michael incorrectly representing the paper. And despite his dismissive assessment of NOVA's utility, the classification system has proven quite “useful” for what it was designed to do: identify patterns in the relationship between food processing and health outcomes. It’s also worth noting that Michael's own description of the research literature later in this episode directly contradicts his claim here. He says, “We then of course get a huge wave of observational studies. You can look up on like Google Scholar, there are dozens of studies that measure ultra processed foods versus non-processed foods. And, and like they all basically find the same thing. It's like higher rates of cancer, higher rates of cardiovascular disease. Like it's all the stuff that you would expect."

This is precisely what you'd expect from a useful classification system. If NOVA were truly as arbitrary and meaningless as Michael suggests, we wouldn't see consistent associations across multiple studies, populations, and health outcomes. NOVA has been successfully employed in hundreds of peer-reviewed studies across different countries, cultures, and population groups. It's been utilized by national dietary guidelines in Canada and Brazil, integrated into public health recommendations, and the World Health Organization is currently working on developing guidelines for UPF consumption.

If a tool consistently produces meaningful, replicable results, I’d argue it’s pretty useful. That being said, there isn’t scientific consensus about whether NOVA is capturing a latent variable (some underlying characteristic we can't directly measure) or an actual property of ultra processed foods themselves (Hall’s study, covered later in this episode, was one way of starting to investigate this). Some research has looked specifically at the additives in foods, rather than the entire UPF category, as well (spoiler alert: greater intake of emulsifiers appears to be associated with higher risk of Type 2 diabetes). As Drs. Mendoza and Tobias so perfectly wrote: “Although public health will continue to learn from the evolving research landscape as it disentangles the relevant features and pathophysiologic mechanisms of exposure to UPFs, the perceived inconsistency of the current evidence is overstated and largely explained by research methodology rather than uncertainty of true biological harm.”

What’s ironic is that Michael admits that studies using NOVA consistently find health associations, but then dismisses the classification system as useless. The least charitable explanation for this contradiction is that his objection is more ideological than methodological. Perhaps he's uncomfortable with what the research is revealing, so he's attacking the tool rather than engaging with the findings. The most charitable view is that he is just not well versed in this field and ‘doesn’t know what he doesn’t know’.

I think this quote Ultra-Processed People captures part of what Michael and Aubrey seem to be missing:

“Let’s imagine Monteiro’s team had started with something explicitly arbitrary, like star signs. Instead of UPF, they might have suggested that being a Leo is the cause of obesity – in scientific terms, it just doesn’t matter, so long as you can back it up with evidence.

Let’s imagine someone had observed that Leos did have more obesity. Well, then the researchers would have to build an intellectual model to explain why that might be: seasons, weather at conception, maternal diet, circulating viruses at birth and so on. They could have done animal experiments, breeding mice to be born between 23 July and 22 August, and then comparing them to mice born on other dates. Then imagine that, having tested their model to destruction, they found that being born when the sun is transiting the 120th to 150th degree of celestial longitude, the constellation of Leo, made all the difference and nothing else mattered. Well, we’d have to live with it. It would be weird, but it would still be true, even though the starting point had been completely and utterly arbitrary.

Philosophers of science aren’t in total agreement about where knowledge comes from, but most accept that science starts with an observation, followed by building a model and then testing that observation. Sometimes the observational data feel very science-y: measurements of the movement of celestial objects, or numbers read out from a fancy machine like a linear collider. But other times a dog walker finds a dead goose in a park and it’s the first data point in a bird flu pandemic.

Of course, in reality, the arbitrary astrological proposition would be supported by none of the data and the model would collapse. The researchers would need to go and look for another cause, like a food system that compels people to eat lots of industrially produced food, or being a Sagittarius. The power of good science is that it can handle a bad, wrong or arbitrary hypothesis. That, really, is the defining characteristic of science.

Real-life science often starts with something arbitrary. Sticking things in boxes. Grouping things together. Naming them. We have to draw a line somewhere and describe the object of interest. In the physical sciences, the boundaries are often clearer. In physics, particles are grouped and described according to how they behave in gravitational or electromagnetic fields. In chemistry, elements are ordered in the periodic table according to their subatomic composition and chemical behaviour. The systems are objective and discrete.

Sometimes, in the biosciences, we also have well defined categories. HIV is, now, a binary diagnosis: you have it or you don’t. But many of the most pressing problems are much fuzzier. Obesity in adults is defined, arbitrarily, as having a BMI of 30 or above. It wouldn’t matter if the threshold was 29 or 31 instead. There’s no sudden change in health that happens at 30 exactly – the risks are just gradually increasing. Almost all biological measurements – blood pressure, haemoglobin, lung capacity – fall along a continuum. At some point we draw a line, somewhat arbitrarily, and say the people on one side of the line have high blood pressure or anaemia or obesity and those on the other side don’t.”

21:27 Michael: So, I found an article on all of the ways that they had to change the definition of this term over time. Between 2009 and 2017, they changed the definition of ultra processed foods seven times. And there's this article called Ultraprocessed Foods: Definitions and Policy Issues by Michael J. Gibney that follows all of these changes in like a super petty but also very useful way.

Not sure what is “petty” about tracking changes in a definition, but again, this is an opinion piece by a guy who served on Nestlé Nutrition Council for over 30 years, including at the time of writing this piece. And the article is full of factual errors!

21:59 Michael: So in 2010, the definition of ultra processed foods is updated to durable, accessible, convenient and palatable, ready to eat or ready to heat food products liable to be consumed as snacks or desserts or to replace home-prepared dishes.

Michael relies on a biased reporter rather than checking references. Gibney's citation for this supposed 2010 change is a 2018 paper - making no sense. Monteiro's 2010 paper aligns with the 2009 definition. Gibney selected different text portions to make definitions seem different.

2009 paper: “Ultra-processed foods are basically confections of group 2 ingredients, typically combined with sophisticated use of additives, to make them edible, palatable, and habit-forming.”

2010 paper: UPFs “...result from the processing of several foodstuffs, including ingredients from group 2 and unprocessed or minimally processed basic foods from group 1.”

Aubrey: I like that they keep throwing in highly palatable, which is just like, if it tastes good.

Unfortunately, this take dismisses an entire field of research. It’s like if someone said “depression” is just if you feel sad. “Palatability” is not just “if it tastes good”, it’s a concept in neuroscience related to hedonic reward. It has been a topic of research for decades, primarily in rat models, but also in humans.

Michael: The other like the first thing that jumped out to me about this the first time I read it “is this ready to eat thing” where like these are often ready to eat. But do you know what else is ready to eat? A fucking apple.

Aubrey: Totally. And some of that is a grilled chicken breast that's cut up and thrown in a package at the grocery store and then you pick up. So, to refer to a food like that while conjuring an image of just like a heap of like Hostess Cupcakes and Cheetos… Feels misleading in a way that really verges on deliberate here.

I don’t have much new to add here except that this is yet another example of injecting unnecessary confusion into this topic. Just because something is “ready to eat” doesn’t make it a UPF. That is not suggested anywhere, in any of the literature or the definitions of UPFs. The fact that many UPFs are ready to eat does not mean that all ready to eat foods are UPFs. The way Aubrey and Michael continue to discuss this could be misconstrued as “misleading in a way that really verges on deliberate”, to use Aubrey’s own wording.

As an aside, a cursory search for “grilled chicken breast that is cut up and thrown in a package at the grocery store” finds multiple options that all have a considerable number of additives thrown in (again, not a judgment on my end, it’s just very obvious that the additives are part of the definition of UPFs and I’m not sure why the hosts are not engaging with that piece of the puzzle). Here are the ingredients in Tyson Grilled and Ready Chicken Breast Strips: boneless, skinless chicken breast with rib meat, water, contains 2% or less of: dextrose, dried garlic, dried onion, grill flavor (from sunflower oil), maltodextrin, modified food starch, natural flavors, natural hardwood smoked sugar (water, natural hardwood smoked sugar), natural smoke flavor, salt, sodium phosphates, spice, sugar, vinegar, yeast extract. This isn’t just a one-off - many frozen chicken breast strips have similar ingredients. Of course you can find grilled sliced chicken breast that doesn’t have these additives, like at Whole Foods, for example, but that’s going to cost you a considerable amount more - which is where socioeconomic issues come into play.

Aubrey and Michael seem to be trying to disprove that UPFs exist by saying that you can find examples of processed foods that they deem “healthy.” As I have said repeatedly, this entirely misses the point. Dishes prepared at home are more likely to be free of certain additives like stabilizers that are used for preserving freshness, which is what this classification system is measuring.

Michael: There's also the thing, it says they're intended to replace home-prepared dishes, but also this is not a biological concept. The purpose of making food does not affect your body differently. If I'm eating a brownie to replace a meal that doesn't make the brownie like affect my body differently. It's again, we're just throwing in these concepts that are not actually related to nutrition.

No one is saying that the “purpose” of making food affects your body differently. The “purpose” of the food does, however, impact what ingredients are in the food, which matters. And it’s not just about what UPFs contain, but what people are now lacking in their diet because they are replacing things with UPFs. This is spelled out pretty clearly here (page 9): “Processes and ingredients used for the manufacture of ultra-processed foods are designed to create highly profitable products (low-cost ingredients, long shelf-life, powerfully branded). Their convenience (imperishable, ready-to-consume), hyper-palatability, and ownership by transnational corporations using pervasive advertising and promotion, give ultra-processed foods enormous market advantages. They are therefore liable to displace all other NOVA food groups, and to replace freshly made regular meals and dishes, with snacking any time, anywhere.”

23:15 Aubrey: At least let’s be honest about what we're grappling with… You already know which foods it is. It's the foods you already don't trust… And you already know who eats those foods and it's people who probably make less money than you do.

This is a strawman argument. UPFs are ubiquitous in western culture among all social classes. Who “doesn’t trust” UPFs? They can be a whole range of things: bread, spreads, plant based meat alternatives, pasta sauces, pancake or dessert mixes, prepackaged chicken kabobs, yogurts, dairy products, frozen dinners, frozen pizzas, gyozas, sausage, hot dogs, burgers, etc.

I think this take from Aubrey and Michael may inadvertently reinforce problematic assumptions. When we rely on the narrative that low-income families primarily eat UPFs due to affordability constraints, we risk using this stereotype to justify discouraging important health research. This perspective can actually maintain the status quo rather than advocating for the systemic changes needed to expand food access and options for all families.

There is also inconsistency in their class-based arguments throughout the episode. They acknowledge that UPFs are consumed more frequently by lower-income populations due to economic factors, while simultaneously downplaying the role of corporate systems in creating these disparities (as in their comment that "food can be healthy and produced by miserable corporations") and avoiding engaging with the idea that lower-income populations should be supported in accessing other food options. This raises a question: if UPFs are nutritionally equivalent to less processed alternatives, what explains the income-based consumption patterns?

The truth is that ultra-processing allows companies to use inexpensive commodity ingredients and transform them through industrial methods designed for palatability and shelf stability. The affordability of UPFs isn't coincidental - it's a direct result of this production model.

23:28 Michael: We then in 2012 get another update where ultra processed foods are defined as. These are formulated mostly or entirely from ingredients and typically contain no whole foods. So, this is yet another message that you see in this world even in like academic articles. It's like they're not even food. They're edible food-like substances.

Again, not true (why doesn’t Michael read the primary sources?). Gibney’s source for this supposed 2012 change is Monteiro’s 2010 paper which makes no sense. And nowhere in the 2010 paper does it say what Gibney suggests it says (if he even means this paper? Don’t trust a journal if it publishes articles without checking if the references make sense.).

24:10 Michael: These efforts go on for like a decade to try to come up with a classification. They finally in 2017 come up with what's called the NOVA classification, which is now what is used in all of the studies. And from three groups they've now made it four groups. There was a period where they're like 3A and 3B or whatever, but now they're just like, fuck it, there's four groups. So, I'm going to send you a JPEG of the current definitions and some examples.

Aubrey: Group 1 is unprocessed or minimally processed foods, naturally occurring foods with no added salt, sugar, oils or fats. Group 2 is processed culinary ingredients. Group 3 is processed foods defined as food products made by adding sugar, oil and/or salt to create simple products from unprocessed or minimally processed foods with increased shelf life or enhanced taste.

This is nuanced, but important to correct - Group 3 is actually adding salt, oil, sugar or other substances to Group 2 or Group 1 foods. MP’s definition misses that other additives are relevant, and that it’s not just adding things to unprocessed or minimally processed foods, but also to processed culinary ingredients. I recommend reading the actual paper in which these definitions are laid out.

Aubrey: And then the last one is like a brick. And that is the definition of ultraprocessed foods. “Industrially created food products created with the addition of multiple ingredients that may include some group 2 ingredients as well as additives to enhance the taste and/or convenience of the product such as hydrolyzed proteins, soy protein isolate, maltodextrin, high fructose corn syrup stabilizers, flavor enhancers, non-sugar sweeteners, and processing aids such as stabilizers and bulking and anti-bulking agents.”

Michael: They're industrially created food products created with the addition of multiple ingredients to enhance taste and/or convenience.

This is not the official definition. The actual definition is quite long, but the TLDR is: “Ultra-processed foods are formulations of ingredients, mostly of exclusive industrial use, typically created by a series of industrial techniques and processes (hence ‘ultra-processed’).”

25:45 Michael: The examples here are commercially produced breads, rolls, cakes, cookies, donuts, breakfast cereals, soy burgers, flavored yogurts, ready to heat meals such as frozen pizzas, soft drinks and candy.

Aubrey: Soft drinks are not ready to heat.

Michael: The people who are doing this defining are not writers. I'll say that.

This is not from the actual definition of UPFs, but I still think most readers would understand that soft drinks aren't part of "ready to heat meals" examples.

26:08 Michael: I think the greatest challenge that they come up with is this thing of there's processed foods which are fine and then there's ultra processed foods which are bad… This is like I think what they spent 10 years trying to figure out because obviously everything is processed and there's stuff cheese which is produced in a very processing kind of process. God damn it. But they don't want to call that bad for you because that's traditional or virtuous, I guess. So that's in group 3.

I don’t know if making cheese is virtuous, but that’s beside the point. There are different ways to make cheese, and some cheeses are considered UPFs. It’s about the ingredients and methods. Artisanal cheese and Velveeta are two very different products made with different ingredients, processes, and price points. One is accessible to almost everyone, the other less so.

26:37 Aubrey: Yeah. It just feels like so clearly such a line drawing exercise around like how do I keep in the things I like and cut out the things I don't.

This is a fundamental misrepresentation of the NOVA system.

26:46 Michael: Here's also the one that really stuck out to me the first time I saw this was in group 3, which again is good. You have freshly made bread. And then in group 4, which is bad, you have commercially produced bread. Again, these are not nutritional content concepts. Something can be made in very large batches and also be very good for you. And the other way around-- And like, at what point does bread become commercially produced on some level? All fucking bread is commercially produced. I bought it at the bakery.

Michael is still looking at the Gibney paper rather than the actual source. In actuality, Group 3 contains, “unpackaged freshly made breads” while Group 4 contains, “mass-produced packaged breads.” The difference is obvious - it’s the additives that are used for preserving shelf-stable breads that will sit in the grocery store for weeks (which Michael even called out in his 2013 HuffPost piece!).

Aubrey: Well, also, group 4 talks about high fructose corn syrup, but group 2 includes honey and maple syrup… So, what is the metabolic difference? It feels like they're trying to have a hundred scientific conversations at once. And I'm like, “No, dudes, you got to go through beat by beat and be like, here's the problem with emulsifiers and why they might be bad for your health. Here's the evidence for that. Here's why honey is different than high fructose corn syrup.” It isn't.

High fructose corn syrup is discussed as an ingredient in UPFs, not as a UPF itself. The point is not whether or not there is a metabolic difference between honey and HFCS. HFCS is an ingredient in many UPFs, not itself a UPF, per the NOVA classification.

Contrary to what Aubrey and Michael say here, there is a very robust body of evidence about food additives and health (ref 1, ref 2, ref 3, etc.). That’s how and why we have the FDA’s GRAS List (“Generally Recognized As Safe). There is a robust history of the FDA saying certain additives are not allowed because of this body of research!!

Michael: There's also, I mean, maybe this is me being annoying, but I also object to cakes and cookies being an ultraprocessed because some cakes and cookies are ultraprocessed, but some cakes and cookies, you bake at home with five ingredients. And surely the whole point of a fucking processing scale is to organize foods according to how processed they are. The processes by which they are made. You just have all cakes and all cookies are in here. Presumably because they're very high in fat and high in sugar. But then if we're just putting in all foods that are high in fat and sugar, then why isn't this just the fucking, how high are foods in fat and sugar scale?

Again, the definition very clearly states, “mass-produced” cakes and cookies, which makes the rest of this discussion irrelevant.

30:12 Michael: So, okay, so that was the decade long and I think unsuccessful effort to define what this term means. We then, in 2019, get the first evidence that this category of food is uniquely bad. So, this comes from a researcher named Kevin Hall, who was previously a physicist, but drifted into diet research. He was the guy that wrote The Biggest Loser study, the study that found that their metabolisms were still hella slow, like, years after they were on The Biggest Loser. He, in 2015, meets Carlos Monteiro at a conference. And Monteiro is like you're looking at this the wrong way. You shouldn't be looking at nutrients, you should be looking at processing. According to the lore, he's like, I don't really buy this. I don't know about this whole ultraprocess thing. I'm going to design a study to disprove this concept.” And then he accidentally ends up proving the concept.

This characterization of Kevin Hall's research is both professionally disrespectful and scientifically inaccurate. Hall didn't "drift" into diet research - he has been a leading researcher in nutrition and metabolism at the NIH for over two decades. His dissertation was about mathematical modeling of cardiac arrhythmias, and he transferred that skill set to modeling human metabolism and energy balance. Dismissing his credentials because he was originally trained as a physicist reflects a disturbing anti-interdisciplinary bias. Many advances in science come from researchers with diverse backgrounds. (I’d also like to note that in The Trouble With Calories episode, Kevin Hall’s work is misrepresented in an attempt to discredit the energy balance model.)

Anyway, this framing misses the point. The trial was designed to test a specific hypothesis about food processing and eating behavior. The research didn't "prove the concept" of UPFs being bad for health - that's not how rigorous science works, and it's not what Hall claimed. What the study did show is that when people were offered the opportunity to eat as much as they wanted (“ad libitum”), they consumed significantly more calories when presented with UPFs and gained weight accordingly. This does not invalidate the importance of nutrients, but adds a crucial additional layer to our understanding. In fact, Hall has been quoted saying, “[Just because a food is ultra-processed] doesn’t mean that nutrient profile isn’t important. … So, for example, even just within ultra-processed foods like the bread category, there are ultra-processed breads that have poor nutritional profiles. There are others that have much better nutritional profiles. So you don't have to ignore nutritional profiles. I think it makes a lot of sense, until proven otherwise, that one of them is going to be much more healthy for you than another. Now is a minimally processed version like an artisanal baked bread that's not precut or something like that, is that going to be even better for you? Maybe. But until we know that, we have to consider all these other factors like nutrient profile, as well as convenience and taste and expense, so all of these things are weighing into people's choices.”

This study revealed that nutrient composition is not the only factor behind overconsumption. The conclusion here is not that nutrients don't matter, but that food processing affects eating behavior even when controlling for macronutrient content.

31:23 Michael: So, the way that he does this is he gets a grant from the NIH to basically take 20 people and lock them in a room, not really, but metaphorically and monitor their diets. And so, what he does is he gives them for two weeks an unprocessed diet. So, completely whole foods. And then for the next two weeks he gives them an ultraprocessed diet and he switches this. So, 10 people start with unprocessed and then go to processed. 10 people start with processed and then go to unprocessed. So that way he's like flipping them around. They're given this food and they have 60 minutes to eat as much of it as they want.

And then when they're done, the researchers take it and they weigh it to see exactly like gram by gram, exactly how much of it did they eat. So that way they can measure their intake. And then of course, there's like a ton of tests at the beginning and at the end.

Aubrey: You and I have talked about sort of like there are a couple of ways to do nutrition research. And one is like in a lab, in a vacuum, it gets you much more limited in scope kind of data. Or you can go larger scale, more longitudinal, but that's usually dependent on self-reports and people sort of like adhering based on the honor code.

Michael: Yes.

Aubrey: So, much of what I've heard about processed foods is about long-term health effects.

Michael: Aubrey, are you saying you can't measure the long-term effect of lifestyle on health in two weeks with 20 people? Aubrey, I don't know.

This study was not intended to measure long-term effects, nor does it conclude anything about long-term effects. If you want a useful summary of the study that isn’t the primary source, I recommend the summary on the NIH site. From that summary: “The basic question for the researchers was - do people eat more on a processed diet than on an unprocessed diet?” Note that this is very different from how MP characterizes the intent of this study.

32:51 Michael: Let me just make you read the description of the results. This is from a New Yorker article by Dhruv Khullar.

Aubrey: When participants were on the ultraprocessed diet, they ate 500 calories more per day and put on an average of 2 pounds. They ate meals faster. Their bodies secreted more insulin. Their blood contained more glucose. When participants were on the minimally processed diet, they lost about 2 pounds. Researchers observed a rise in levels of an appetite suppressing hormone and a decline in one that makes us feel hungry.

Michael: So, this is very decisive. It's people who ate ultraprocessed foods gained a bunch of weight, they ate more. All of these markers got worse. The unprocessed people, they did great. They lost weight, they felt awesome. This study when it comes out is like, it's wild how popular the study was. The study has been cited 1200 times.

The actual conclusion from the paper is:

In conclusion, our data suggest that eliminating ultra-processed foods from the diet decreases energy intake and results in weight loss, whereas a diet with a large proportion of ultra-processed food increases energy intake and leads to weight gain. Whether reformulation of ultra-processed foods could eliminate their deleterious effects while retaining their palatability and convenience is unclear. Until such reformulated products are widespread, limiting consumption of ultra-processed foods may be an effective strategy for obesity prevention and treatment. Such a recommendation could potentially be embraced across a wide variety of healthy dietary approaches including low-carb, low-fat, plant-based, or animal-based diets. However, policies that discourage consumption of ultra-processed foods should be sensitive to the time, skill, expense, and effort required to prepare meals from minimally processed foods—resources that are often in short supply for those who are not members of the upper socioeconomic classes.

Again, this is very different from what MP characterizes.

33:45 Aubrey: Yeah. And I don't know, this feels like the glycemic index all over again, which is a teeny tiny group of people have a specific response to a food or group of foods and that then somehow becomes like conventional wisdom in really short order.

The comparison to the glycemic index is not appropriate. The glycemic index was developed from observing blood sugar responses in small groups, then extrapolated broadly without rigorous testing. Hall’s study, by contrast, used a controlled experimental design specifically to test a hypothesis about the effect of processing on dietary consumption, while controlling for other variables.

Michael: This study is so much worse than just the fact that it was two weeks. So, the whole point of a study like this is to hold everything else constant and only look at the effect of “processing.” But then if you read the fucking text of the study, that was not remotely true. The ultra processed diet had twice the energy density of the unprocessed diet. It had twice the saturated fat and it had 1.5x more sugar. So, these are not equivalent diets at the most basic level. And this is like in the fucking study.

The point of this study was not “to hold everything else constant and only look at the effect of ‘processing.’” It was to assess “whether people ate more calories when exposed to a diet composed of ultra-processed foods compared with a diet composed of unprocessed foods.” What he describes here as design flaws are actually necessary features of testing real-world ultra-processed versus unprocessed/minimally processed diets. When you create an unprocessed diet and an ultra-processed diet that people will actually eat, there will inevitably be some differences due to how these food categories are typically formulated. The reason why this study made such an impact is that it controlled for the variables that nutrition science traditionally focuses on - calories, macronutrients, fiber, and sodium - while allowing the diets to differ in ways that are typical when looking at processed vs unprocessed foods.

Aubrey: You're like one group had a green salad and the other one had like a value meal from Wendy's. And the ones who had a value meal from Wendy's gained weight. Would you believe it?

A green salad and a value meal from Wendy’s are not likely to have the same number of calories, sugar, fat, or fiber, which were controlled for in Hall’s study.

36:19 Michael: This is actually the reason why the energy density is so different between the two diets is because I think in an effort to hold a fiber content, there's no fiber in ultra processed foods. It's like one of the things that makes them ultra processed foods. So, the only way to get participants fiber was to basically give them sodas every day with like fiber supplements in them. So, these are diet lemonade. But oftentimes, it's just like juice or like a little smoothie or milkshake, something like that. But there are drinks with every meal for the ultra processed people. There are no drinks for the unprocessed meal. Every once in a while they get milk or something, but like they're not getting diet sodas or anything. So that's a huge difference between the two diets.

Remember that this is an ad libitum study, meaning that the participants were not required to eat anything - they were allowed to eat as much as they wanted. Given that, I certainly don’t think it is a “huge difference” that participants were sometimes offered beverages along with their meals.

37:00 Aubrey: If you were sitting down to make yourself a meal, the chances that like most people would make two whole sandwiches and eat an entire can of peaches. It just not representative of like how people eat.

Participants were not required to eat anything. The point of the study was to see how much people eat when presented with these foods. And if you look around you, most people have tons of food at their fingertips. That’s part of the issue with UPFs - they are readily available and people like to eat a lot of them!

37:17 Michael: Well, also the thing that really stuck out to me is that there are three desserts with this meal. There are cookies, there are Fig Newtons, and there's a can of peaches in syrup. Almost every single meal of the ultra processed comes with cookies or shortbread or like some sort of like pudding. There are no desserts with any of the unprocessed meals. Yeah, the other like really striking thing about this is that they're not the same meal. They're completely different.

Of course they are not the same meal - that’s the point. It’s interesting that Michael calls all of these foods, including canned peaches, “desserts”, because it reveals to me that he isn’t aware of how these foods are consumed by large swaths of the population. The Newtons website quite literally markets fig newtons as “snacks” (see below). And I don’t think most people consider canned peaches “dessert”, either.

37:46 Aubrey: If you want to do a one to one on your ultraprocessed, you get the like, I don't know, Stouffer's version of like beef and broccoli.

Michael: Exactly.

Aubrey: Okay, let's go head-to-head with similar dishes.

Michael: This is the thing is like what they're actually fucking doing. They're calling this a test of are ultra processed foods worse for you? But it's literally, it's like you give one group of people salads, it's a lot of salads. You give another group of people fucking cookies and they're like, “Oh my God, the people ate more cookies.” Therefore, ultra processed foods are bad for you.

Michael: I actually think there's a huge missed opportunity here because something people always talk about in this field is that like, well, pizza isn't necessarily bad for you. If it's a frozen pizza from the grocery outlet, but then it's bad for you. But if it's like artisanal and there's only three ingredients in the dough and it's like lovingly made, then it is good for you. It doesn't have to be bad. So, why didn't you fucking test that hypothesis?

Again, they aren’t calling this a test of, “are ultra processed foods worse for you?” This study is testing “whether people ate more calories when exposed to a diet composed of ultra-processed foods compared with a diet composed of unprocessed foods.”

Michael and Aubrey are essentially arguing that the study should have created artificial UPFs that don't exist in the real world to perfectly match unprocessed foods on every possible variable. But that would defeat the entire purpose - the question isn't whether you can engineer UPFs to behave like unprocessed foods, but whether the UPFs that actually exist in our food environment drive different eating behaviors.

39:02 Aubrey: The ultra processed menu for day seven dinner is made for a child.

39:09 Michael: It looks so bad. Also, again, there's fucking two desserts.

The meal in question is “two PB&Js that are packed to the gills” per Aubrey. Looking at the picture below, it looks like a pretty normal PB&J to me. If you Google image search for PB&Js, you’ll see a bunch of pictures of sandwiches with the same amount of filling (or more!). And it’s interesting, again, that Michael is classifying certain foods as “desserts” when the menus do not classify any of the foods as “desserts”.

Michael (describing the unprocessed Day 7 dinner): I wonder why people ate 500 fewer calories. Wow. We gave them the most boring food imaginable.

No judgment, but the processed food meals seem equally “boring” to me. Plain bagels with cream cheese and turkey bacon. A cheeseburger with American cheese and french fries. For a lot of people, these might be interesting, and for a lot of people, they may be boring! You could argue none of these meals are particularly “exciting” - probably so that they appealed to most study participants.

Aubrey: They're not listing the giant bowl of grapes, you gluttons.

It quite clearly says, “Grapes”:

40:51 Michael: But also, if the title of this stuff was like, people eat more cookies than grapes, I'd be like, yeah, I don't know that this says anything about processed versus unprocessed foods. These are different foods.

This is a classic straw man fallacy. This study is not investigating if people eat more cookies than grapes - it’s looking at how much people eat when presented with all of these foods over a period of time (not just one meal, not just cookies vs grapes, etc.). That wouldn’t be the title of the study, because it’s not what the study found.

41:03 Aubrey: But listen, if you scrape off this top layer of window dressing kind of stuff around, we're actually concerned about long term health conditions. We're actually concerned about blah, blah, blah. Most of those things are things like diabetes, like heart disease, and things that we associate with fat people. So, I think part of what you're seeing here is an assumption about how fat people eat and how poor people eat.

This study was very clear about what it was intending to investigate. It was cautious in its conclusions. If we want to discuss the socioeconomic and class issues at play, we also have to be willing to discuss that dietary patterns may have an impact on diabetes, heart disease, and other chronic conditions.

41:30 Michael: Okay. So that was the experimental study that was, that was the attempt to prove that ultra processed food is bad for you.

I’m repeating myself but, again, this wasn’t the point of the study.

Michael: We then of course get a huge wave of observational studies. You can look up on like Google Scholar, there are dozens of studies that measure ultra processed foods versus non-processed foods. And, and like they all basically find the same thing. It's like higher rates of cancer, higher rates of cardiovascular disease. Like it's all the stuff that you would expect. Mm-hmm . This is sort of the second way that you can measure the effect of food on health is like, you take these, these big studies of like hundreds of thousands of people. You give people food frequency questionnaires. How often are you eating something? Oftentimes they'll do this in two ways. At the same time they'll be like, how often do you eat these foods in general? And they'll also do a 24 hour recall, like, what did you do yesterday? This is like kind of as good as it gets. Although people are so bad at estimating what they're eating. And like, especially the amounts, right? Like if you, if you go to a restaurant, you have like pod Thai, can you say how many ounces you ate? So there's that sort of layer of just gathering the basic information.

This is a repeat topic for Michael - that food frequency questionnaires (FFQs) are unreliable. If you want to listen to some actual experts speak about the utility and limitations of FFQs, I highly recommend this episode of Sigma Nutrition with Dr. Deirdre Tobias. Dr. Tobias also wrote an article on the topic. Here are two quotes explaining the actual purpose of these tools:

“Despite its stunningly descriptive name, many incorrectly assume that FFQs are designed to measure a food quantity, rather than a food frequency.”

“A conclusion that an FFQ validly estimates coffee intake, for example, does not claim that the instrument captures the precise quantity of coffee an individual consumed on a specific day. Rather, validity means the investigator will correctly and consistently rank participants who “never or rarely” drink coffee below those who “occasionally” drink coffee, “occasional” coffee drinkers below “frequent” drinkers, and “1 cup per day” drinkers below “multiple cups per day” drinkers. In an extreme example of an FFQ being completely incapable of measuring coffee any more accurately than throwing darts to assign someone’s intake, a true “multiple cups per day” drinker would be just as likely to correctly report being a “multiple cups per day” drinker as being a “never drinker,” and the correlation of the FFQ-derived coffee intake against a gold standard reference measure would be zero. Further, epidemiologic investigators using this useless estimate of coffee intake exposure would have a difficult time ever reporting a significant relationship between coffee drinking and health outcomes, beyond that of random chance.”

But then on top of that, researchers will then go in and they will code people's answers for ultra processed food. So they get these answers, this is what I ate. I ate cake yesterday, cookies yesterday, whatever. And then researchers will go in and go, aha cake is ultra processed. So this person is Yes. Eating ultra processed food. Ah-huh. So there's two layers of errors with this. And the biggest thing with this is that they're not fucking measuring whether people are eating ultra processed food. They're just having these food categories. So again, fucking cake cake can either be ultra processed or not ultra processed. It depends on the fucking cake.

This is an oversimplification. Let’s look at a huge international study from 2023 that examined the relationship between UPF consumption and mortality:

As part of the PURE study, the Environmental Profile of a Community’s Health (EPOCH) substudy [15], we collected a sample of packaged foods (e.g., chips, cookies, sweets, chocolate, etc.) from each participating community in PURE [16]. In addition, for each country, local nutritionists and the first author (MD) reviewed the list of food items in the FFQ and identified UPFs. The classification of UPFs was country specific. For example, some types of bread in India and Iran were not classified as UPFs because bread is daily prepared at home (chapati, roti, and dosa) or freshly bought and consumed (Iranian bread), whereas mass-produced breads in Canada and Sweden were classified as UPFs…Overall, for the present analysis, we only included food items that were most likely to be UPFs; those that we collected a sample of packaged foods or there was no doubt in the level of processing (e.g., soft drink, processed meat, etc.).

It’s a lot more technical and nuanced than Michael made it seem! And that study is not a one-off. Here’s another example:

We classified foods by considering the NDNS variables ‘Food Number’ and ‘Subsidiary food groups’. When foods were judged to be homemade dishes, we applied the classification to the underlying ingredients in order to ensure more accurate classification. The NDNS database was provided with most food items systematically disaggregated into their individual components and the method adopted to disaggregate the food codes has been described in a previous paper [29].

In fact, these studies do often distinguish between kinds of cake. Here’s just one study’s list of types of pastries/cake classified as UPFs:

Cakes and pastries: tea cakes, pastries, whole grain pastries, cake, lady fingers/sponge fingers/sponge cake, tea buns/soft bread buns, chocolate filled buns, custard filled buns, cream buns, Swiss buns, choux buns, chocolate Danish pastry, churros, dried fruit sponge cake, croissant, chocolate filled croissants, cheese filled croissants, doughnuts, Mallorcan Danish pastry, fairy cakes/cupcakes, short crust pastry, puff pastry, lard, meringue, puff pastry filled with crème and custard, Danish pastry, puff pastry, Danish pastry filled with jam, chocolate cake, sponge cake, cream cake, pastry filled with custard, plum cake, homemade doughnuts, cake filled with cream, fruit cake, whisky cake, apple pie, cheese cake, egg pudding, dried fruit sponge cake, sponge cake, chocolate sponge cake, tea buns, butter sponge cake, Madeira cake, choux pastry with custard, fruit Danish pastry, raisin pudding, apple pie with custard.

The point is - there is a lot of nuance and detail here. Doesn’t mean it’s perfect, of course!

43:40 Aubrey: We've seen now multiple ways that processed and ultra processed foods are categorized. And we've already explored that reasonable minds can differ, right? That like two people could in good faith put the same thing in any of the four different categories or in two of the four categories or whatever, right?

Michael: And some of the studies do actually attempt to control for that. They'll have blinded-- one researcher will do it and then another researcher will also do it like independently. And they'll say like, “Okay, we have 95% agreement.” They're attempting to control for this. But then what the real problem is that it's not actually the coders. It's the actual designers of the research. They all say, like, “Oh, we use the NOVA classification system. But if you read studies, different studies have different classifications. So, we mentioned, honey before shows up in all different categories. I also notice alcohol. Some studies just remove it altogether. They're like, “We're not looking at alcohol consumption.” Some studies will put it as not ultra processed. Some studies will put it as ultra processed. I also found one that put wine and beer as not ultra processed, but vodka as ultra processed.

Michael: The same study also put it was bread was not ultra processed, but pretzels were ultra processed.

For a specific study, there will be a protocol for how to define each type of food (see above examples). As far as across studies - we absolutely should expect classifications to vary due to cultural and geographic factors. Since he doesn’t provide specific examples, I can’t speak to the studies Michael mentions have honey and alcohol inconsistencies. But, for example, the study I cited above did not just “remove” alcohol - they didn’t collect alcohol consumption because alcohol consumption in the countries they conducted the study in was either prohibited or very low. You can’t ask people about their alcohol consumption in a country where alcohol is illegal and the subjects could get in significant trouble if they were revealed to have consumed alcohol.

Michael: The other one that really bugs me is like, hamburgers are always in the ultra processed category. But I looked this up. A McDonald's hamburger is 100% beef. It's beef. It doesn't have a bunch of weird ingredients in it.

Aubrey: Do you think they're maybe counting on buns and American cheese and ketchup and all of that kind of stuff?

Michael: Well, that's the thing is, technically, yeah, you could put it in there, but again, it could be ultraprocessed or it could not be like, everything fucking else. Like the cookies and the cakes and the pizza. Like, pizza is always in ultra processed as well. But it matters for your whole thesis, whether it's frozen pizza or homemade, artisanal, whatever, “nice pizza.”

There is often a justification or rationale for the things he keeps mentioning. For example, from one study: “Hamburgers are classified as an ultraprocessed food according to NOVA (2). Hamburgers are not ground beef but are a dish made of industrial breads, sauces, reconstituted meat, processed cheese, bacon, etc. Therefore, we included hamburgers in the ultraprocessed food group. Some hamburgers can be homemade, and doubts may arise about their inclusion in this group. Unfortunately, our study could not differentiate between fast-food and homemade hamburgers. Even when we removed hamburgers from the ultraprocessed foods, the results were only slightly attenuated (HR: 1.23; 95% CI: 1.07, 1.41; P-trend = 0.003).”

Michael is making sweeping generalizations without understanding the body of literature he is speaking about.

45:18 Aubrey: Has anyone hazarded a guess at what the mechanism is here?

Michael: Yeah, this is something that you find in the critical literature, as a research field, we're actually missing a crucial component and we're leapfrogging over this because we thought it was saturated fat, then we thought it was sugar. Now we think it's processing, but no one can agree on what the fuck processing is.

There are a lot of things in medicine that we don’t understand the mechanism for! That does not mean the association does not exist or that we shouldn’t keep studying it. How would we know to investigate a mechanism if we didn’t first see that the exposure was associated with the health outcome? This is how epidemiologic research works! As I mentioned above - it’s possible that UPFs represent some latent variable. It doesn’t make this research useless.

45:55 Michael: Another problem with these studies is they don't distinguish between different types of ultra processed food, oftentimes it's just this weird binary distinction between ultra processed food and everything else. So, all three of the first three categories are just like good and ultra processed is bad. But there's a couple studies that actually look at different categories of ultra processed food. They're like, okay, breakfast cereal, candy bars, various other things. In this study that looked at 10 categories of ultra processed foods, the only ones that showed a clear and consistent association with worse disease was soda, processed meats and alcohol. All the other ones, like cookies, refined bread, all this other stuff, it was like too mixed to really say anything.

There are plenty of good critiques of many of these studies, but calling it a “problem” that they don’t distinguish between UPF types isn’t one. If my research hypothesis is that eating vegetables improves eyesight, I only need vegetable intake as an aggregate exposure. It’s an entirely different question to say, “how does consumption of different vegetables affect eyesight.” That’s when I would pay attention to different types of vegetables.

Aubrey: But even within those. If we're talking about breakfast cereals, my guess is that Grape-Nuts is going to have a different health effects… than like Lucky Charms.

See below for the nutrition panels for Grape Nuts (left) and Lucky Charms (right). If we convert to the same serving size (27g), the comparison would be:

Calories: 93 vs 110

Total Fat: 0.5g vs 1g

Sodium: 130mg vs 170mg

Total Carbohydrates: 22g vs 22g

Dietary Fiber: 3g vs 2g

Sugar: 2g vs 10g

Protein: 3g vs 2g

Not that different!

47:06 Michael: And the final thing I want to mention is the lack of a dose response. So, the way that they do these studies is they compare the people who eat the least ultra processed foods to people who eat the most ultra processed foods. And it's like a really wide gap. Some people are eating like 60% ultra processed foods and some people are eating 7%. When you compare the least versus the most, you do get these pretty large effects. However, there's some studies actually list the effect for each of the quintiles in between. And there's something weird that the death risk actually goes down sometimes if you eat more ultra processed foods.

47:43 A real effect should have a dose response like a little bit of ultraprocessed foods is a little bad for you. A lot is a lot bad for you. But we don't find that in the results. One of the papers found a 50% higher cancer risk if you're eating a ton of ultraprocessed foods. But then once they adjusted it for the dose response, they only found a 5% difference.

I haven’t even read every paper on this topic, and I’ve already found several with dose response relationships (ref 1, ref 2, ref 3), including the ones that Michael and Aubrey linked to. I can’t verify what Michael is saying about the risk of mortality decreasing with increased intake of UPFs, since he doesn’t cite his sources. And “adjusting for the dose response” isn’t a statistically intelligible concept, so I’m not sure what he means there. My best guess is he is misinterpreting the results when researchers report both an effect comparing the lowest quintile and highest quintiles of UPF intake as well as an effect for every 10% increase in UPF intake, such as in this study. But that’s not “adjusting for dose response”, it’s just an entirely separate way of measuring and reporting an effect.

Michael: Some people, if you eat a little bit of ultraprocessed foods, you're actually less likely to get cancer. That's something that never comes up. I don't think this is causal. And also, it's based on these fucking food frequency questionnaires, so who knows. But it's like to the extent that we have can give diet advice which we all know everybody's going to give diet advice on the basis of these fucking correlational studies to the extent we can give diet advice, it's like, “Well, yeah, if you're not eating any ultraprocessed foods, you should start eating some” because those people actually have a lower risk of dying.

They did not link to any papers looking at cancer risk, so I cannot fact check this, unfortunately. Would love to know which paper he’s talking about!

To be continued…