I feel Robert Shewfelt’s pain, I really do.
This past month has been a difficult one for defenders of processed food.
Four recently published studies make it abundantly clear that there is a link between how much ultra-processed food one eats and a variety of health outcomes. Note: I’m being very clear here about using “ultra-processed” and not the simpler “processed”, precisely because I think it is only if you willingly conflate the two that there’s anything worth defending. More on that later. In the meantime, the studies.
Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain: An Inpatient Randomized Controlled Trial of Ad Libitum Food Intake started the ball rolling. Kevin Hall of the National Institute of Diabetes and Digestive and Kidney Diseases and his colleagues studied 10 men and 10 women for 4 weeks. For two weeks, half of them got a diet of mostly ultra-processed food while the others got a diet of mostly minimally processed food. The two groups then switched for the final two weeks. The diets were matched for calories, sugar, fat, fibre and macronutrients, and the people were free to eat as much or as little as they chose. People eating the ultra-processed diet ate about 500 calories a day more and, unsurprisingly to me, gained a bit of weight. Perhaps the main take-away is that calories do matter to weight gain, and that ultra-processed foods somehow make it easier to eat more calories.
Then there were two papers in the British Medical Journal looking at ultra-processed diets and health: Association between consumption of ultra-processed foods and all cause mortality: SUN prospective cohort study from Spain, and Ultra-processed food intake and risk of cardiovascular disease: prospective cohort study (NutriNet-Santé) from France.
Both were prospective cohort studies, meaning that people were enrolled into the study before any kind of investigation of their health. In both, participants reported their diet at regular intervals, and the researchers classified the foods they ate according to the four NOVA categories (which is where this discussion began). In both, participants were divided into four equal groups, based on the amount of ultra-processed foods in their diet. And in both, those who ate the most ultra-processed foods had poorer health outcomes than those who ate the least.
Specifically, in the Spanish study of almost 20,000 people, the top quartile was 1.62 times more likely to die from any cause than the bottom quartile. And the more ultra-processed food, the greater the chance of dying. The French study, of more than 105,000 people, looked at causes of death in more detail. Overall, those who ate lots of ultra-processed food were 1.14 times more likely to die of cardiovascular disease. Broken down, coronary heart disease deaths were 1.13 times more likely and cerebrovascular disease 1.11 times more likely in the top quartile of ultra-processed food eaters. In all cases, the researchers carried out various tests to ensure that the differences were not caused by some variable other than the amount of ultra-processed food in the diet.
And then, although it doesn’t seem to have featured (yet?) in Shewfelt’s mense horribilis, a large group of researchers centred on Darriush Mozaffarian’s group at the Friedman School of Nutrition Science and Policy at Tufts University in Boston published details of the Preventable Cancer Burden Associated With Poor Diet in the United States. They estimate that in 2015 around 80,000 new cases of cancer could be attributed to “suboptimal diet”. That represents about 5% of all new cases of cancer. Of the cancers, colorectal cancer accounts for almost 40% of all diet-related cancers. And the most important dietary factors were low intake of whole grains and dairy and high intake of processed meats. The authors don’t define processed meats; I imagine all would be classified as ultra-processed. They do point out that “[u]nlike the red meat consumption that showed a decreasing trend, the consumption of processed meats remained unchanged in the past 15 years. US adults consumed, on average, about 1 oz of processed meats daily, more than twice the recommended intake by the American Heart Association.”
As far as I am concerned, then, the evidence is that eating a lot of ultra-processed food is not good for you. Why is that so hard to accept?
One reason might be that ultra-processed food is more profitable than minimally processed food, so one might expect those who profit by making the stuff to take an interest in muddying the waters. Mostly, that takes the form of criticising the NOVA classification, and only by implication the conclusions based upon it. At least some critics who take this tack have strong industry ties.
Robert Shewfelt adopts a similar approach, focussing for the most part on poking holes in his versions of the NOVA classification scheme. For example:
Remember, ultra-processing has nothing to do with how it has been processed or how much it has been processed. It is only about the ingredients.
The beauty of this classification scheme is that it is easy to distinguish between an ultra-processed food from one that is either unprocessed or just processed. All one has to do is to be able to count to six.
On the contrary. “Ultra-processed” is all about how it has been processed or how much it has been processed. And nowhere in the most recent explanation of the NOVA food classification and the trouble with ultra-processing is the number of ingredients considered important. Likewise, the NOVA classification says nothing about whether ingredients are unpronounceable, which Shewfelt considers vitally important.
Of course, as any right-thinking statistics undergraduate will tell you, correlation is not causality, and the right thinking reasearchers agree. “Causality remains to be established,” say the French. Equally, there is clearly something about ultra-processed foods worth investigating. Or, as the Spanish put it: “Similar results in different populations, with different age ranges … and diverse methods for assessing dietary exposures support a causal association”.
There are hints at possible mechanisms. One review of the changing microbial landscape of Western society points out that “Western” diets tend to be associated with a gut microflora that is both less numerous and less diverse. It seems that fibre is an important food for “good” gut bacteria. Added sugar, by contrast, may be good for “bad” bacteria. A diet low in fibre and high in sugar could thus be a double whammy. Getting down to details, a group at Georgia State University, Atlanta, has shown that small amounts of two commonly used (and easy to pronounce) ingredients of ultra-processed foods, carboxymethylcellulose and polysorbate‐80, speed the onset of colitis in mice that have a predisposition to the disease. In normal mice, the same ingredients cause low-grade inflamation and mild obesity. Gut microflora underpin the effects, which are not seen in germ-free animals, until they are given a faecal transplant. There are also hints that inflamation of the gut can interfere with leptin, the hormone that makes us feel full, so leading to overeating. I’m willing to bet that we will soon see lots more studies showing ill effects of both ultra-processed diets in general and specific ingredients in particular and, frankly, one doesn’t need a mechanism to act on an association. What’s the downside?
Arguments about the cost of eating a minimally processed diet, in specie and in time, have a lot going for them. The NIH study that demonstrated weight gain on ultra-processed diets estimates that 2000 calories a day of the ultra-processed diet cost $106 a week compared to $151 for the minimally processed. That difference alone, to say nothing of lack of cooking skills and equipment, will make it hard for many people to abandon ultra-processed foods entirely. I’d like to see a more nuanced comparison, including the externalities associated with each type of diet and the role of subsidy in the cost of ultra-processed food. In other words, why is ultra-processed food both cheaper and more profitable?
In that connection, I was intrigued by an item in Jess Fanzo’s Food Bytes, in which she drew attention to a study that showed that “55% of the global rise in mean body mass index since the mid–1980s – and more than 80% in low- and middle-income regions – was due to increases in body mass index in rural areas”. Eyeballing the map she included, in Latin America, Africa and much of Asia, the excess weight is in the cities. In North America, Europe, Australia and New Zealand, rural areas predominate. As the study authors say: “In high-income and industrialized countries, we noted a persistently higher rural BMI, especially for women.” I wonder whether that reflects rural poverty and a reliance on ultra-processed foods.
But to return to Robert Shewfelt’s difficult month, I’m really pleased to see that he acknowledges that the NIH study “is an extremely well-planned and well-executed study”. On the weight aspect, he concludes that “[a] diet eliminating ultra-processed foods appears to be useful for losing weight”. So far so good. But he can’t leave it at that: “[S]uch an unprocessed diet may not be adequate to maintain weight in healthy individuals. A balanced diet that includes fruits, vegetables, whole grains and lean proteins incorporated into a diet also containing ultra-processed foods could provide an appropriate blend of health and convenience for a modern family.”
Precisely. These days any diet in a halfway-advanced economy is bound to be mixed to some extent. What matters is the relative proportions.
I can’t leave it at that either. In bashing the use of NOVA classification, Shewfelt and a commenter somehow got hold of the idea that “whole milk [was] considered ultra-processed while 2% milk [was] considered unprocessed”. The NOVA categorisation is absolutely clear that “fresh or pasteurized milk” is “unprocessed or minimally processed”, so in an attempt to gain clarity I looked at the specific menus in more detail. On the ultra-processed diet, whole milk appears for breakfast on day 1 and partially skim milk (2% fat) for dinner on day 7. The unprocessed diet includes skim milk for lunch on days 2, 4, 5, 6 and 7, with partially skimmed for breakfast on day 3. At no point does the Hall et al. paper say that one diet was exclusively unprocessed or exclusively ultra-processed, only that they were designed to be “widely differing in the percentage of calories derived from ultra-processed versus unprocessed foods”. I’m thus not at all surprised to find milk, an unprocessed food, in two of the 21 “ultra-processed” meals.
What to do? A big problem is that ultra-processed foods are very precisely calibrated to make them absolutely yummy. Warning labels probably aren’t much help. Higher levels of general nutrition literacy (along with the prerequisites to make use of that learning) might be useful, though I am somewhat dubious. Most likely, it will take wholesale changes to the entire food system, for which, alas, there seems to be very little appetite among governments or their electorates.
p.s. I wish I could be in London on 2 July to hear Mike Gibney, of University College Dublin, give the UK’s Institute of Food Science and Technology lecture on Ultra-processed foods: The science and the policy. Gibney is not a fan of the NOVA classification, so it will be interesting to hear what he makes of the new results. I’m sure this time he will also be sure to declare any conflicts of interest. The website says the event will be “streaming live to various satellite receptions around the UK”. I guess that means the stream won’t be public. Perhaps they will publish a recording eventually. #IFSTlecture is the tag to watch.