[PART I – MYSTERIES]
[PART II – CURRENT THEORIES OF OBESITY ARE INADEQUATE]
[PART III – ENVIRONMENTAL CONTAMINANTS]
[INTERLUDE A – CICO KILLER, QU’EST-CE QUE C’EST?]
[PART IV – CRITERIA]
[PART V – LIVESTOCK ANTIBIOTICS]
[INTERLUDE B – THE NUTRIENT SLUDGE DIET]
[PART VI – PFAS]
[PART VII – LITHIUM]
[INTERLUDE C – HIGHLIGHTS FROM THE REDDIT COMMENTS]
[INTERLUDE D – GLYPHOSATE (AKA THE ACTIVE INGREDIENT IN ROUNDUP)]

People have been asking us if we’re going to review the literature on seed oils. 

We weren’t aware this was such a popular theory when we wrote the series, so we didn’t think to address it. Briefly, this theory says that some of our food oils — usually soybean, corn, canola, cottonseed, and safflower oils, but sometimes others — are behind the modern meteoric rise in obesity rates. Clearly this is relevant to our interests. 

People are split over exactly what to call the culprit — seed oils, industrial seed oils, vegetable oil, etc. These terms may mean slightly different things to different people, but ultimately these are all a closely related set of theories. In this post, we will go with the term “seed oils” because it is the shortest. 

Here is just a small subset of the sources people sent us:

• How Industrial Seed Oils Are Making Us Sick by Chris Kresser
• Death by Vegetable Oil: What the Studies Say by Jeff Nobbs
• The Case Against Processed Vegetable Oils by The Organic Consumers Association
• The SCD1 Theory Of Obesity Archives / The Croissant Diet Archives from Fire in a Bottle

We haven’t read everything ever written on seed oils because there are a truly enormous number of pieces on this idea. But we do think it’s worthwhile to do our own review of the theory, so here goes. 

Plausibility

There are a number of things that make this hypothesis attractive. For one, these oils truly are in everything. 

One of the authors of this blog went on a couple dates with a girl who had a serious corn allergy. Unfortunately for her, corn oils and other corn byproducts are all over the damn place. She couldn’t eat out at restaurants or buy most prepared foods. She couldn’t have chocolate. She couldn’t even eat citrus, because corn proteins are a component in the pesticides/preservatives sprayed on citrus trees and fruit. She could eat apples if she peeled them, but even that was a risk. She mostly lived on rice, rice noodles, a small number of safe vegetables, and potatoes (which, to be fair, are the prince of food). 

And this is just corn! We’re looking for a factor that is omnipresent in the environment and difficult to escape regardless of one’s diet, and seed oils do seem to fit the bill.

We should also note that the contaminants theory more or less predicts something like seed oils. Any contaminant that causes obesity probably bioaccumulates in plants and animals. If it bioaccumulates in plants, then highly concentrated byproducts of those plants might contain higher concentrations of the contaminant. This would vary based on the species of plant (some will bioaccumulate more than others) but we would expect to see especially high levels of contaminants in some extremely concentrated food products. 

Seed oils are also highly processed (check out the cool/disturbing video at that link) so they have many opportunities to come into contact with industrial contaminants. “I thought it was going well,” says YouTube commenter Sam De Francesco, “up to the bit about using a chemical solvent to remove the remaining oil from the canola cake. It went downhill from there quickly.”

The two theories are complementary, and it’s possible that both could be true. We can even look at specific contaminants to see if there might be a connection.

There is some work on measuring the levels of PFAS in cooking oils, and they do record some PFAS being detected in “Edible Oil Samples from Markets in Beijing, China, in 2013”, including corn oil, soybean oil, etc. But these levels of contamination are in about the same range as PFAS found in other foods, like milk, not thousands of times higher.

We can’t find any actual measurements of lithium levels in different food oils. This paper finds no relationship between vegetable oil consumption and lithium in the blood plasma of 922 Germans, but the food consumption measure was all self-reported, and people are probably not fully aware of how much vegetable oil is in many foods. Some secondary sources (here, here) say that lithium is most concentrated in vegetables, cereals, and grains, which would sort of line up with the seed oil hypothesis, but we can’t find the primary sources for these claims, so take this one with a grain of salt for sure. 

Glyphosate’s Hail Mary

In a previous post, we took a look at glyphosate and concluded that it doesn’t seem like glyphosate could be a primary driver of obesity, and probably doesn’t contribute to obesity at all. There is one longshot hypothesis, however, of how glyphosate might play a role in the obesity epidemic, or at least the increase we’ve seen since the 1990s. 

When reviewing the list of Roundup Ready crops, varieties designed to be sprayed with huge doses of glyphosate, we couldn’t help but notice that soybeans, corn, canola, and cotton are all on the list (though not safflower). As a result, these crops are sprayed with much more glyphosate than normal. This seems like kind of a weird coincidence. You’ll also recall that in 2016, the EPA found glyphosate in 63.1% of corn samples and 67.0% of soybean samples they tested. 

The level of contamination was quite low in all of these samples, but seed oils are highly concentrated, so the levels of contamination in corn oil could potentially be hundreds of times higher. That said, we haven’t been able to find any data on glyphosate levels in any of these oils, so it’s hard to tell. There’s some evidence that glyphosate is in more foods than officially recognized (thanks to commenter Louis for finding this one!), but that’s not a direct connection to seed oils per se. We also found one claim that glyphosate “does not mix with oil, so there are negligible glyphosate residues in vegetable oil from treated crops such as soybeans, canola and corn”, but this is from gmoanswers.com.

Challenges

The seed oil account of obesity also has several weaknesses, which make the hypothesis seem less likely.

First of all, it seems unlikely that there is a problem with the sources of the oils themselves. Soybeans and corn have been eaten for centuries without a wave of obesity. Rapeseed (you know this as the source of canola oil) is yet another Brassica, which people have been eating forever. 

It could all be in the dose, though. These seed oils are highly concentrated — hundreds or thousands of servings of the relevant vegetable are needed to make one serving of the oil — so if there is some corn enzyme that is harmless at a small dose but dangerous when the dose is 100x higher, that could lead to corn being harmless but corn oil being an issue. A common saying among seed oil proponents is “the dose makes the poison,” and that certainly might be the case.

We notice that the seed oils under suspicion were developed around 1900-1930, which seems a little too early to match the timeline for obesity. People have been eating more and more of them over time, but this trend took off around 1940 and doesn’t show a clear inflection point around 1980. This isn’t damning evidence, but it doesn’t quite fit the pattern. 

In addition, seed oils face the same problems faced by every food-based explanation for the obesity epidemic.

Seed oils have a hard time accounting for patterns like obesity being related to altitude, and wild animals also becoming obese. Maybe wild animals are eating scraps of human food out of dumpsters behind the 7-11, but it’s not clear why people at low altitudes would be eating more seed oils than people at high altitudes.

Seed oils also seem unable to account for the big differences between obesity rates in different professions. It doesn’t seem like people in different professions are likely to eat different amounts of seed oils, but they do seem likely to be exposed to different contaminants at work.

Similarly, there is a large amount of variance in obesity rates between countries, and there has been for a long time. Why is the Middle East so obese? Why has Kuwait always been one of the most obese countries in the world? Why was Kuwait 20% obese in 1979, back when the obesity rate for the US was only 13%? It doesn’t seem like seed oils could be behind the high rates of obesity in the Middle East, which limits the possible role they could play in driving the obesity epidemic.

Health Claims

Of course, what we’re really interested in is whether or not seed oils are related to weight gain and obesity.

It’s unfair to look through a literature and start with some random papers. You want to let the theory’s supporters point you to the evidence they feel is strongest, the evidence they think is most important. We wanted to look at the strongest possible argument, the sort of case that only a supporter could put together. So we decided to start by looking at one person’s case against seed oils.

For this purpose, we decided to focus on work by Jeff Nobbs, because his series was the best of the many arguments we saw, and we want to engage with the strongest version of the theory. Nobbs cites lots of primary sources, links directly to them, and tells you which parts of the findings he thinks are relevant. He doesn’t hit you in the face with a bunch of biochem claims right out of the gate. He doesn’t engage in name-calling and doesn’t try to scare the reader by going on about how dirty or unnatural seed oils are, something we can’t say about some of the other sources we saw. He just shares the evidence that convinced him and explains why he thinks it’s important.

In Part 1 of his series, Nobbs does a really great job describing how levels of chronic disease and obesity are increasing, and reviewing the evidence for just how weird this is. He hits a number of the same points we made in our review of the mysteries surrounding obesity, and often he makes a more compelling case for their weirdness than we did. In particular, he does a better job describing how exercise rates are increasing. We especially like this figure: 

Part 2, called Death by Vegetable Oil: What the Studies Say, reviews a number of studies that suggest that seed oils (Nobbs prefers the term vegetable oil, but it’s essentially the same thing) contribute to health conditions like cancer, dementia, and weight gain. In this series we are primarily interested in obesity, so we will stick to the weight gain results.

In this post, Nobbs references four studies showing an obesity-seed oil connection:

In one study, rats were divided into different groups receiving diets identical in fat, protein, and carbohydrate calories but differing in the source of the fats. The rats in the group receiving fat from safflower oil had a 12.3% increase in total body weight compared to the rats eating traditional fats [6].

In a randomized trial on rabbits, three groups of rabbits were given access to identical foods, with only one difference: the first group of rabbits was fed unheated vegetable oil, the second group was fed vegetable oil that had been heated once, and the third group was fed vegetable oil that had been repeatedly heated multiple times. Everything else about their diets was kept the same [7].

The outcome? Compared to the group of rabbits eating unheated oil, the group eating single heated oil gained 6% more weight, and the group eating repeatedly heated oil gained 45% more weight!

A 2020 study in mice showed that consumption of soybean oil leads not only to weight gain, but also to gene dysregulation that could cause higher rates of neurological conditions like autism, Alzheimer’s disease, anxiety, and depression [8].

In another mouse study, feeding mice the equivalent of 2 tablespoons of canola oil per day is associated with worsened memory, learning ability and weight gain, along with “considerable neuronal damage” and increased formation of beta-amyloid plaques, the signature of Alzheimer’s disease [10].

Links 8 and 10 are to news sites, not to the original research papers, but as far as we can tell, 8 is probably this paper and 10 appears to be this paper.

We do want to mention at this point that all of these studies were conducted on animals, not humans. (Nobbs also knows this, of course.) We recognize the value of animal studies and we often cite them ourselves, but it’s curious that there seem to be no studies showing that seed oils cause weight gain in humans — more on this in a bit.

These sources describe what Nobbs says they do, showing connections between seed oil consumption and weight gain. And in fact there are many other studies that also show this connection — this study shows soybean oil led to more weight gain than other oils (in mice), this study shows soybean oil led to more weight gain than fish oil and palm oil (in rats), and this study shows safflower oil led to higher levels of leptin (but not body weight) compared to animal-based fats (also in rats).

However, there are also studies that show essentially no difference between seed oils and other oils in terms of weight gain. For example, in this study soybean oil led to an equal amount of weight gain as lard and palm oil, and canola oil led to less weight gain than the other three (still in rats), and in this study lard, sunflower oil, and palm olein oil led to nearly identical weights (mercifully, in vervets this time). 

Many other studies find the exact opposite, that seed oils cause less weight gain than other fats, including animal fats. For example, this paper found that a butter-rich diet led to more weight gain than a canola-rich diet (in rats), this paper found that a lard diet led to greater weight gain than safflower oil (in mice), and mice that were started on a lard diet and then switched to safflower oil actually started losing weight, this paper reviewed the literature and found that pretty much any high fat diet leads to some weight gain (in rats and mice), and the paper Body Fat Accumulation Is Greater in Rats Fed a Beef Tallow Diet than in Rats Fed a Safflower or Soybean Oil Diet found what it says on the tin. 

Even this is only a small selection of the large number of papers on this topic. Some of these papers are probably better than others — some we should take seriously, and others we should discount. Probably a bunch of these studies are crap. If someone wants to do a deep dive into these papers, and all the other ones out there, to try to figure out the state of the literature, we would be interested in reading that piece. But even if someone did a deep dive, we doubt there would be a clear result. It’s hard to look at this literature and say anything more than “wow that looks complicated”.

This is just one of those very large literatures where you have to be careful about drawing a conclusion from only a few studies. “Some studies where seed oils lead to more weight gain than other fats, some studies where they lead to less, and a couple studies where there’s no difference” is what we should expect to see if there is a small effect in either direction, or no effect at all.

There certainly doesn’t seem to be a strong effect here. If the effect were strong, it would be detected more reliably. So if seed oils have an effect on weight gain (in mice, rats, and vervets), it’s a weak effect at most, and it’s not clear from the evidence whether seed oil causes more or less weight than other fats.

Another problem with the weight gain results in rodents and monkeys is, why do none of the human studies find the same thing? Nobbs cites a number of studies on people, but none of the ones we looked into show any evidence that seed oils cause weight gain in humans. 

Let’s take a look. The Los Angeles Veterans Administration Study specifically mentions, “the unrestricted consumption of the two diets had no significant effect on average body-weight.” The Minnesota Coronary Experiment reports BMI in both conditions and finds them to be nearly identical, 24.6 versus 24.5. 

The MARGARIN Study reports BMI for all conditions and finds no difference in people who ate diets with margarine made with more fat from vegetable oil versus margarine made with less fat from vegetable oil. 

The Sydney Diet-Heart Study had two groups start out with similar BMIs (25.4 vs. 25.1) and BMI in both groups went down by a similar, small amount (ending up at 24.5 vs 24.3) after 12 months, regardless of whether they were eating mostly safflower oil and margarine, or mostly olive oil and butter.

These studies generally run for several years and have much higher sample sizes than the studies run on animals, and of course they’re looking at humans. If there’s no evidence in these studies that people gain weight when eating seed oils, then any effect of seed oils on weight would have to be very small or very subtle. Or, there might be no effect at all.

In fact, this is pretty strong evidence against seed oils causing obesity. If seed oils cause weight gain when people eat them, why didn’t seed oils cause weight gain when people ate them? Sure we didn’t review every study out there, but these are four rather large studies (a couple hundred to a couple thousand people each) and all of them show an effect of seed oil on weight that is absolutely zilch, bupkis, nothing. This is more than “no smoking gun”, this is “suspect was in another country giving a speech in front of 10,000 people on live TV.”

Rigor

And also… ok, we do want to pick on the rigor here, just a little.

Nobbs says, “In the Los Angeles Veterans Administration Study, the group of participants who increased fat from vegetable oil–while keeping total fat the same–were 82% more likely to die from cancer compared to the control group that didn’t increase fat from vegetable oil.” 

This sounds very impressive — but as far as we can tell, it is referring to this result: “31 of 174 deaths in the experimental group were due to cancer, as opposed to 17 of 178 deaths in the control group.” This is about 82% more in the experimental group, but as you can see, the absolute numbers are quite small. 

This difference is not even statistically significant, which the authors note, giving p = .06. In addition, they begin the discussion by noting, “The experience of other investigators using similar diets has not been the same.” They mention a couple other studies, including one from London which tested a diet high in “soya-bean oil”, where there were 6 cancer deaths in the control group and only one cancer death in the group eating a diet high in “soya-bean oil”.

We see something similar in the other studies. Nobbs mentions that “the group consuming more vegetable oil had a 62% higher rate of death during the seven-year study compared to the group eating less vegetable oil” in the Sydney Diet-Heart Study. Again, this sounds like a lot. We’re not entirely certain which result he’s referring to, but as far as we can tell this is also not significant, p = .051. In fact, every p-value reported in this paper seems to be between .02 and .13, which does not inspire confidence and is probably an indicator of p-hacking. Either way, the majority of these results are not statistically significant, and if we were to apply a simple correction for multiple comparisons, none of them would be.

In the MARGARIN Study, the “seven times higher” figure refers to one death in the less-vegetable-oil group and seven deaths in the more-vegetable-oil group. The authors note that this is “not significant because of the small numbers.” They don’t report a p-value here, so we double-checked their analysis just to be sure and found the same thing, p = .096.

The evidence in the Minnesota Coronary Experiment is a little more mixed, but the authors themselves do not seem to have access to the raw data, and in reference to the apparent increased mortality in adults over 65, say, “in the absence of the raw data, however, we cannot determine the statistical significance of this finding.”

Nobbs did a good job reviewing the literature and providing direct references to the studies he draws his conclusions from. But the literature on weight gain in animals is far larger and more mixed than many people realize, and doesn’t clearly point to seed oils causing obesity. The literature on seed oil consumption in humans consistently shows that seed oils cause no more weight gain than other fats. When we took a closer look at some of these studies, we found serious problems with several of the analyses. The evidence here is weak at best. 

This doesn’t mean that seed oils, or vegetable oils, or whatever you want to call them, are good for you. They may still be very bad for you, and the case for other health effects (including a connection with cancer) seems stronger. But it doesn’t look like they could be a major cause of the obesity epidemic, and probably, they play no role at all.

[Next Time: PARADOXICAL REACTIONS]

[PART I – MYSTERIES]
[PART II – CURRENT THEORIES OF OBESITY ARE INADEQUATE]
[PART III – ENVIRONMENTAL CONTAMINANTS]
[INTERLUDE A – CICO KILLER, QU’EST-CE QUE C’EST?]
[PART IV – CRITERIA]
[PART V – LIVESTOCK ANTIBIOTICS]
[INTERLUDE B – THE NUTRIENT SLUDGE DIET]
[PART VI – PFAS]
[PART VII – LITHIUM]
[INTERLUDE C – HIGHLIGHTS FROM THE REDDIT COMMENTS]

Glyphosate is a herbicide and all-around general weed killer that you probably know as the active ingredient in Roundup *whip cracks*

We were put onto the trail of this here varmint by Justin Mares, who pointed out that there are several reasons to be suspicious of glyphosate. We’ll start with the fact that this stuff is pretty much everywhere. “Because of its widespread use, glyphosate is in water, food and dust, so it’s likely almost everyone has been exposed,” says PBS.

Most sources say that glyphosate has a relatively short half-life in both soil and water, though apparently this “can vary widely based on environmental factors” and “values between 2 and 197 days have been reported in the literature.” Many sources downplay this contamination, emphasizing how quickly it degrades, but it does seem to end up in groundwater, as there is a whole water treatment literature about how to remove it (see also this other review paper). 

Relevant to our interpretation of the altitude mystery, one study that examined glyphosate contamination in two rivers in Mexico found that the rivers were generally (though not always) more and more contaminated as they flowed downstream.

Glyphosate was patented in 1971 and first sold in 1974, but the FDA didn’t test for glyphosate in food until 2016, which seems pretty weird. The results of these tests are publicly available — in the report from 2016, we see that they tested 274 samples of corn, 267 samples of soybeans, 113 samples of milk, and 106 samples of egg for glyphosate. No glyphosate at all was detected in the milk or eggs, but 63.1% of corn samples and 67.0% of soybean samples showed some glyphosate contamination, though none contained glyphosate levels above the limit set by the EPA. It’s clearly in some foods, though apparently less in animal products and not at especially high levels overall.

Glyphosate can affect the growth of microorganisms that rely on a specific pathway for growth, but this paper finds that most gut bacteria don’t have that pathway, and so glyphosate probably doesn’t affect their growth. On the other hand, this paper says that “54% of species in the core human gut microbiome are sensitive to glyphosate.”

There’s also some evidence that glyphosate interferes with various enzymes, at least in rats. Some of these enzymes are related to the metabolism and clearance of many drugs, and interference with these enzymes does seem to sometimes lead to negative drug interactions. This suggests that even if glyphosate doesn’t cause obesity by itself, it could potentially make people more susceptible to other compounds that do.

In addition there is at least one speculative paper arguing explicitly that glyphosate’s suppression of cytochrome P450 enzymes leads to many modern diseases, including obesity. However, we should mention that the journal that published this paper found it so weird that they added a note about “potential bias in opinions and bias in the choice of citation sources used in this article” and issued “an Expression of Concern [emphasis in the original] to make readers aware that the approach to collating literature citations for this article was likely not systematic.” Also see this pushback, and this pushback from a group funded by the American Chemistry Council, American Petroleum Institute, etc. etc.

When we look at a map, the distribution of glyphosate use in the United States matches county-level obesity data pretty darn well: 

It’s not a perfect match, but it’s pretty striking. Granted, this also looks a lot like the distribution of other pesticides, for example our old friend 2,4-D:

And Atrazine, another common pesticide:

Maps of pesticide use mostly highlight agricultural regions, regardless of the pesticide you look at. The big difference is that unlike these other pesticides, glyphosate is used more and more every year, matching the rise in obesity. Again from the USGS:

Glyphosate (interestingly, the increase seems to have stalled around 2012?):

2,4-D:

Atrazine: 

Glyphosate was developed to fight weeds in the early 1970s and was first brought to market in 1974 as Roundup. Since then, glyphosate use has increased pretty much every year, both in the US and worldwide.

This is actually a little on the late side, especially since glyphosate saw relatively limited use before the 1990s. Things really kicked in with the introduction of genetically engineered Roundup Ready crops. Roundup kills most plants, so it kills most crops too, and in the beginning the only way to use it was to spray selectively. But in 1996 Monsanto introduced Roundup Ready soybeans, which are resistant to glyphosate. Now farmers could dump Roundup on the whole field and kill everything but the soybeans. This was followed by Roundup Ready corn, canola, sugar beets, cotton, and alfalfa. These new resistant varieties led to a huge increase in glyphosate use during the 1990s and onward.

(In addition, we should note that Kuwait was already 18% obese by 1975, and it’s hard to see how glyphosate could be responsible for that.)

Round Down

There’s some circumstantial evidence that if contaminants are responsible for obesity, at least one of those contaminants is related to agriculture. As we see from the maps above, the most obese parts of America are largely farm country. If we were to suspect an agricultural chemical, glyphosate would be at the top of our list.

But there are many more signs that the main contaminants are not agricultural. If the most widely used herbicide in the United States were the cause of obesity, then we would expect agricultural workers to be especially obese, as a result of their high level of exposure. Instead, we see that the rate of obesity among agricultural workers is pretty average — sometimes slightly higher than average, sometimes slightly lower.

Agricultural workers are clearly exposed to glyphosate. One study of 48 farmers, their spouses, and their 79 children found that 60% of farmers had detectable levels of glyphosate in their urine. Farmers who wore rubber gloves had less glyphosate in their urine, and farmers who made skin contact with the glyphosate had more. A small percent of spouses and children also had low levels of glyphosate in their urine, but this mostly seemed to happen when the spouses or children helped prepare the glyphosate formulation or apply it to the fields. Despite this direct exposure, farmers are not especially obese.

Maybe glyphosate (or some other pesticide) degrades over time, or combines with something else in the environment, and ends up forming some byproduct that causes obesity? Again this seems unlikely. If some byproduct of a chemical used at farms caused obesity, you would think that the people who spend all their time at the farm would get the most exposure. There are some exceptions — farm workers probably don’t get much exposure to the antibiotics given to livestock, for example — but it seems unavoidable in the case of contaminants that are sprayed directly on fields.

Maybe eating glyphosate-laced food is different from inhaling glyphosate or absorbing it through your skin. This seems possible, given that there are some signs glyphosate might disrupt the microbiome, and this could potentially explain why farmers are not all that obese, since they’re not eating the stuff. But once again this seems unlikely, because there are major differences between obesity rates for other professions.

Motor vehicle operators, healthcare workers, and law enforcement workers are some of the professions with the highest levels of obesity. Teachers, design workers, and legal workers are some of the professions with the lowest levels of obesity. Some of these differences are very robust. 

This seems hard to reconcile with an account where glyphosate, or a glyphosate byproduct, causes obesity. Granted, glyphosate is used other places than just on farms. But why would truck drivers and police officers be exposed to so much more glyphosate than everyone else? How could they be exposed to more than farm workers? It’s possible that there is something about driving a truck that unlocks the harmful potential of glyphosate (or something about glyphosate that unlocks the harmful potential of being a truck driver) but at the moment this seems unlikely.

Like farmers, forestry workers also work with glyphosate. But despite this, forestry workers involved in spraying don’t show any glyphosate in their urine, despite clearly getting it on their clothes and even on their skin. It’s hard to see how law enforcement workers or truck drivers could be getting an appreciable dose of glyphosate when forestry workers who handle the stuff directly don’t seem to get any in their body. And, we can note, forestry worker is a pretty lean profession as well.

While the geographic match is pretty good in the US, the international match is mixed. We haven’t been able to find the clearest sources, but here’s a map from one paper: 

Compared to:

If glyphosate were a major driver of obesity, we would expect Central and South America and the Middle East to be much less obese. We would expect Spain, Germany, Poland, and much of Africa to be much more obese. Food imports and differences in water treatment techniques may be able explain some of this, but it’s not an immediate hit for glyphosate. 

Direct Evidence

If you look into a potential glyphosate-obesity connection you will eventually find a 2014 paper titled Genetically Engineered Crops, Glyphosate and the Deterioration of Health in the United States of America, or sources based on it, so we need to address it here. 

This paper reports an absolutely comically huge correlation of 0.962 between glyphosate application and obesity, and similarly enormous correlations with other diseases. This correlation is implausibly large, and it is not accurate — it is the result of a common error people make when working with time series data.

A good explanation of this error can be found in Avoiding Common Mistakes with Time Series Analysis by Tom Fawcett, where he shows that adding a slight trend to two totally random, unrelated time series ends up making them appear extremely correlated despite the total lack of a relationship. In his example, the apparent correlation ends up being .96, despite the true correlation being zero. What this paper reports is not evidence that glyphosate causes obesity, it is only evidence that glyphosate use has increased since 1996 (we already knew that) and that obesity has also increased since 1996 (we already knew that too).

There is some evidence of weight loss, or at least “decreased body weight gain”, in animal studies. All these studies are in rats and mice, however, and this seems to happen only at the highest doses. When we say “highest doses”, we mean really high — in one study the highest dose was 1183 mg/kg/day, in another it was ​​3500 mg/kg/day, in a third it was 4945-6069 mg/kg/day. 

In comparison, the U.S. Environmental Protection Agency reference dose for glyphosate in humans, “or estimate of daily exposure that would not cause adverse effects throughout a lifetime”, is 2 mg/kg/day. In that study from before, the highest estimated systemic dose in farmers who were working directly with glyphosate was only 0.004 mg/kg.

This is a tiny bit of evidence for glyphosate causing weight change, but it’s 1) in animals, 2) weight loss rather than weight gain, and 3) only at doses about 1000 times higher than recommended and about 250,000 times higher than the doses found in people who work with glyphosate directly.

The best evidence for glyphosate causing weight gain that we could find was from a 2019 study in rats. In this study, they exposed female rats (the original generation, F0) to 25 mg/kg body weight glyphosate daily, during days 8 to 14 of gestation. There was essentially no effect of glyphosate exposure on these rats, or in their children (F1), but there was a significant increase in the rates of obesity in their grandchildren (F2) and great-grandchildren (F3). There are some multiple comparison issues, but the differences are relatively robust, and are present in both male and female descendants, so we’re inclined to think that there’s something here.

There are a few problems with extending these results to humans, however, and we don’t just mean that the study subjects are all rats. The dose they give is pretty high, 25 mg/kg/day, in comparison to (again) farmers working directly with the stuff getting a dose closer to 0.004 mg/kg.

The timeline also doesn’t seem to line up. If we take this finding and apply it to humans at face value, glyphosate would only make you obese if your grandmother or great-grandmother was exposed during gestation. But glyphosate wasn’t brought to market until 1974 and didn’t see much use until the 1990s. There are some grandparents today who could have been exposed when they were pregnant, but obesity began rising in the 1980s. If glyphosate had been invented in the 1920s, this would be much more concerning, but it wasn’t.

Conclusions

It doesn’t look like glyphosate can be a major contributor to the obesity epidemic. It was introduced slightly before obesity began to skyrocket, but it didn’t see much use until decades later. It seems to have arrived too late to be responsible. There may be a slim chance that glyphosate contributes in some small way to obesity, and might be responsible for some of the increase in obesity since the mid-90s, but it doesn’t look like it could have started the epidemic.

Glyphosate exposure doesn’t seem to match international patterns of obesity, and glyphosate doesn’t seem to be able to explain the variation in obesity rates by profession. If glyphosate caused obesity, we would expect farm and forestry workers to have high levels of obesity, but in fact their obesity rates are pretty average, or even below average. Finally, there’s very limited evidence of glyphosate exposure causing weight change in animals, and no evidence in humans, at least none that we can find.

To us, it doesn’t seem likely that glyphosate plays any serious role in the obesity epidemic, and it probably doesn’t play any role at all. If you disagree or have any evidence to the contrary, however, please let us know!

[Next Time: SEED OILS]

[PART I – MYSTERIES]
[PART II – CURRENT THEORIES OF OBESITY ARE INADEQUATE]
[PART III – ENVIRONMENTAL CONTAMINANTS]
[INTERLUDE A – CICO KILLER, QU’EST-CE QUE C’EST?]
[PART IV – CRITERIA]
[PART V – LIVESTOCK ANTIBIOTICS]
[INTERLUDE B – THE NUTRIENT SLUDGE DIET]
[PART VI – PFAS]

Lithium is the third element on the periodic table, the lightest metal, a hit Nirvana single, and a mood-stabilizing drug often used to treat bipolar disorder.

Lithium isn’t synthetic, of course, but it can still be an environmental contaminant. While it occurs naturally in small concentrations in groundwater, human activity might have led to serious increases over the past few decades.

Unlike the other contaminants we’ve reviewed, we don’t need to spend any time convincing you that lithium makes people gain weight: it does. Almost everyone who takes lithium at therapeutic levels gains some weight. About half of them report serious weight gain, on average 22 lbs (10kg), and about 20% of patients gain more than that. Weight gained is correlated (r = .44, p < .001) with dosage. Unsurprisingly, weight gain on lithium is related to an increase in leptin levels.

We’d love to tell you whether lithium concentrations in groundwater have increased over time. But while lithium is easy to detect, assessing lithium levels is not a part of the standard analysis of drinking water, so we don’t have reliable historical data to work with. There aren’t even EPA standards for lithium levels in drinking water.

We’d also like to tell you about how much lithium people are exposed to, and whether that has increased over the past several decades. Measuring serum lithium is relatively easy, and people who are starting lithium treatment get checked frequently to make sure that their blood levels aren’t too high. Despite this, there doesn’t seem to be any data on serum lithium levels in the general population. The NHANES data has records of how much uranium there was in your urine every year from 2000-2016, but not a single measure related to lithium. Great job, guys.

We can’t talk about trends in the groundwater or in people’s bodies directly. But what we can do is look at other trends that we would expect to be related. For example, this figure shows a graph of USGS-reported records for global lithium production since 1900:

This graph is pretty telling. Almost no lithium was produced before 1950, so human activity couldn’t have been adding a meaningful amount to the groundwater back then. Serious lithium production started around 1950, which could help explain why obesity went from about 3% in 1890 to about 10% in 1980, but we see that lithium production truly spikes around 1980. While there have been a few ups and downs, production has as a rule continued to rise ever since. This graph only goes up to 2007, but USGS and other sources confirm that production has continued to increase, at about 11% per year from 2007 to 2017.

Lithium definitely ends up in the groundwater. It’s there in small concentrations naturally, but human activity adds more. In Seoul, South Korea, lithium concentrations sextuple as the Han river passes through the city’s densest districts. This is reflected not only in the river, but also in the tap water in the city — tap water from sites further along the river’s course have similarly elevated lithium levels.

This suggests that rivers pick up lithium along their course and generally have higher lithium levels as they flow downhill. This is supported by data from Austria, which shows that lithium levels in drinking water vary systematically with altitude, with higher concentrations of lithium found in districts at lower altitudes:

We should note that a paper looking at groundwater in the United States from 1992-2003 found the opposite effect: higher levels of lithium at higher altitudes. “However,” they say, “these findings should be interpreted with caution.” We agree. There are 3,141 counties in the United States, and they only looked at data from 518. They only examined data from 15 states, most of them states at relatively low elevation. These weren’t randomly selected, either; they were the sites with the highest number of lithium samples in the years 1992-2003.

We’ve already discussed the issues that come up when you conduct analysis on a restricted range of data. Further, the Seoul data shows that lithium levels spike around urban areas. If some of the high-altitude measurements were near or immediately downstream from cities or manufacturing areas, that might make it look like higher-altitude locations have higher levels of lithium on average.

Dosage

The therapeutic dose of lithium in blood serum is usually considered to be in the range of 0.8 – 1.2 mmol/L, though some sources suggest that lower doses are more effective, with the “minimum efficacious serum lithium level” being possibly as low as 0.4 mmol/L. To translate these to more familiar terms, 0.8 mmol/L is about 5600 ng/mL, and 0.4 mmol/L is about 2800 ng/mL.

That’s quite a lot. In comparison, lithium levels in groundwater rarely exceed 200 ng/mL. But perhaps surprisingly, even very low levels can have an influence on our health and mental states. One study examining data from 27 Texas counties between 1978-1987 found that rates of suicide and homicide (as well as other forms of violent and impulsive behavior) were negatively correlated with lithium in drinking water, over water lithium levels ranging from 70 to 170 ng/mL. Another study looking at various cities in Lithiuania (no relation to lithium) found a negative relationship between lithium exposure and suicide. The lithium levels in the public drinking water systems they examined ranged from 0.5 to 35.5 ng/mL, with a median level of 3.6 ng/mL. In general, reviews of this literature find that trace levels of lithium have a meaningful impact on behavior.

There’s only one randomized controlled trial examining the effects of trace amounts of lithium, but it finds the same thing. A group of former drug users (heroin, crystal meth, PCP, and cocaine), most of them with a history of violent crime or domestic violence, were given either 400 µg per day of lithium orally, or a placebo. For comparison, a normal clinical dose is 300,000 – 600,000 µg, taken two to three times per day. Even on this comparatively tiny dose, everyone in the lithium group reported feeling happier, more friendly, more kind, less grouchy, etc. over a four week period, “without exception”. 

(These former drug users didn’t have normal moods to begin with — one said “I am always extremely moody and fight with my girlfriend frequently” before he was treated — so it’s not clear if trace amounts of lithium would improve mood in everyone else. Similarly, therapeutic doses are only given to patients with bipolar disorder, and it’s not clear what the effects would be on someone without this diagnosis.)

In the placebo group, people were just about as grouchy as before. When they switched the placebo group over to lithium, these people responded in exactly the same way.

This is pretty strong evidence that even very small doses of lithium can have meaningful effects. So should we be surprised that they don’t mention any weight gain? There isn’t much data on the time course of weight gain in lithium treatment, but it seems to come on pretty fast. In one study with normal therapeutic doses, 15 bipolar inpatients gained an average of 13 lbs (5.9 kg) over six weeks. While the sample size is quite small, this tells us that sometimes a lot of weight gain can happen fast.

We don’t think this is a huge problem. The randomized controlled trial on trace exposure only lasted a couple of weeks. Even if patients were gaining weight at the same rate as patients on a therapeutic dose, they might not have noticed. The researchers didn’t intend to examine weight gain, and probably didn’t measure it. They don’t report any other side effects. Heroin, crystal meth, PCP, and cocaine all make people lose weight, so it’s possible that weight gain in the former drug users would be seen as a sign of health. It’s also worth noting that while these trace amounts do appear to have real, consequential effects, the dosage was about 1,000 times smaller than a therapeutic dose. In this situation, it’s not crazy to think that weight gain might take a few weeks or even months to manifest. We see a version of the main effect — improved mood — in this tiny dose. It seems reasonable that we might see a version of this side effect as well.

This study can give us a lower limit not only of the dosage, but by extension, of the minimum effective serum level. As a ballpark estimate, therapeutic dosages are in the range of 1,000,000 µg per day and they lead to serum levels of around 5000 ng/mL. This means the 400 µg per day dose from the randomized controlled trial would lead to a serum level of around 2 ng/mL. Other sources suggest that blood levels may end up slightly higher on low doses. For example, doses in the range of 385 to 1540 µg per day lead to serum levels around 7 to 28 ng/mL. One study testing serum lithium as a compliance marker for food and supplement intake found that giving people lithium-tagged yogurt with a dose of about 1000 µg per day over a period of six weeks led to an increase in serum levels from 6 ng/mL to 46 ng/mL.

These papers disagree on the specifics but they agree on the general picture. A serum level in the range of 10 ng/mL is enough to influence mood, and a dose of about 400 µg per day is enough to get you there. Whether lower doses have an effect is unclear, but we should certainly be interested in numbers in these ranges and up.

Trace Exposure

There aren’t any other randomized controlled trials, but if trace amounts are enough to cause obesity, then we should see relationships between trace lithium levels and obesity rates.

In Texas, a survey of mean lithium levels in public wells across 226 counties (Texas has 254 in total) found lithium levels ranging from 2.8 to 219.0 ng/mL. Now Texas is not one of the most obese states — but it tends to be more obese along its border with Lousiana, which is also where the highest levels of lithium were reported.

West Virginia and Alabama, on the other hand, are two of the most obese states in the nation. We haven’t been able to find any reliable groundwater concentration data for these states, but we can certainly note that both of them have problems with lithium contamination coming from local mining operations. And it’s not just Alabama and West Virginia — other states are facing similar contamination.

In Greece, lithium levels in drinking water range from 0.1 ng/mL in Chios island to 121 ng/mL on the island of Samos, with an average of 11.1 ng/mL. Unfortunately there’s not much data on the prevalence of obesity in Greece, but we can conduct some due diligence by checking a few of these endpoints. Samos, with the highest levels, is the obvious place to start. On Samos, 10.7% of children aged 3-12 are overweight, compared to 6.5% on the island of Corfu. A full 27% of high schoolers on Samos island were overweight in 2010, and 12.4% were obese. In comparison, about 12.5% of American high schoolers were obese in the same period.

In the Caspian Sea, lithium concentrations are 280 ng/mL. As we’ve already reviewed, some of the most obese provinces in Iran border the Caspian. The Dead Sea has concentrations even higher, at 14,000 ng/mL, but for the obvious reasons, we don’t think people are getting a lot of their drinking water from the Dead Sea. On the other hand, obesity in the West Bank is pretty high — as high as 50% in men in 2003!

Very high concentrations of lithium have also been reported in Austria. For the most part, Austria has normal amounts of lithium in its drinking water, around 13 ng/mL. But in the east, the concentrations are much higher. In the Mistelbach district, the average level of lithium in the drinking water was 82 ng/mL, and the highest single measurement was near Graz, at 1300 ng/mL. Both of these are in eastern Austria, where obesity levels are highest. Mistelbach in particular is one of the most obese districts in the country.

Chile and Argentina are the most obese countries in South America (28% each) and are two of the biggest exporters of lithium in the world. Unsurprisingly, this is reflected in their groundwater.

In northern Chile, lithium levels in the groundwater can reach levels 10,000 times higher than normal. The rivers running through many small valleys see lithium concentrations of 600 – 1600 ng/mL. The drinking water in many towns has levels up to 700 ng/mL. And in the headwaters of the Rio Camerones, lithium hits concentrations of 24,880 ng/mL. In 1980, Zaldivar reported similar levels of lithium in the groundwater, stating that they were “the highest reported in the world”. He also measured serum levels in people at these different sites, and found values ranging from 22.3 ng/mL to 85.8 ng/mL.

(“We tried to find Zaldivar to learn more about his work,” says a later paper, “but to no avail. He left Chile when Pinochet came to power and effectively disappeared.”)

In Argentina, lithium can reach up to 1000 ng/mL in drinking water, and the locals end up with a lot of exposure. At the site with the highest levels, lithium concentrations reached an average of 4,550 ng/mL in people’s urine. The highest level in someone’s urine was actually a whopping 14,300 ng/mL. The locals seem to be getting much of their lithium exposure from their tap water, as the amount of lithium in urine was correlated with the number of glasses of water consumed per day (r = 0.173, p = 0.029).

These are all freshwater levels. In seawater, lithium concentrations are reliably quite high, ranging from 100 ng/mL to over 1000 ng/mL. Now, most people are not actually drinking meaningful amounts of seawater, but if you live near the ocean, you might still be exposed indirectly.

One mystery we haven’t mentioned yet is that the Middle East is extremely obese (see map below), one of the most obese regions on earth. Jordan, Qatar, Libya, Egypt, Lebanon, and Saudi Arabia all barely trail the United States in terms of obesity, and Kuwait is actually slightly more obese than the United States, about 38% obese compared to 36% in America. These countries are very dry, and so all of them get a lot of drinking water from desalinated seawater. Saudi Arabia gets about half of its drinking water from desalination and is one of the most obese nations on earth. Kuwait built its first desalination plant in 1951, and has actually been one of the most obese countries in the world for a long time. Back in 1975, when the rate of obesity in the United States was around 10%, the rate of obesity in Kuwait was about 18%.

Desalination removes all trace elements from seawater, but because distilled water corrodes metal pipes and trace elements are important to health, the desalinated water is remineralized by blending it with 5-10% brackish water. This means that desalinated water could easily have lithium concentrations of up to 100 ng/mL. Unlike contamination in some forms of drinking water, which might vary with factors like rainfall and industrial activity, we would expect lithium levels to be reliably high in desalinated water, because they are inherent to the source.

We even have some data about lithium levels in the waters of the Persian Gulf. Near Qeshm island, at least, seawater concentrations vary somewhat by season but are usually around 300 ng/mL. This is reasonably high for seawater and much higher than the levels usually observed in groundwater. Not all of that makes it back into the water after desalination, of course, but even if only 10% got back in, 30 ng/mL is still a pretty high dose to be receiving regularly.

(If you are a Kuwaiti or Saudi desalination engineer, please contact us! This blog has gotten 15 views from Kuwait and 26 from Saudi Arabia, we know you’re out there!)

In any case, if lithium from desalinated seawater doesn’t explain why the Middle East has such incredibly high rates of obesity, then some other explanation will have to be found for this extremely striking observation.

On average, people drink about 3 liters of water per day. If they’re drinking from a normal freshwater source with 1-10 ng/mL, they’ll get a dose of about 3-30 µg / day. If they’re drinking from desalinated seawater, with concentrations of about 30-100 ng/mL, they’ll get a dose of about 90-300 µg / day. If they’re drinking from sources like those found in Texas, Greece, and Mistelbach, with concentrations of about 100-200 ng/mL, they’ll get a dose of about 300-600 µg / day. If they’re drinking from sources like those found in Graz, Chile, and Argentina, with 1000 ng/mL, they’ll easily get a dose of 3000 µg / day or more:

In comparison, therapeutic doses are in the range of 1,000,000 µg per day, but remember the randomized controlled trial showed effects at only 400 µg per day. Many people are getting doses of similar amounts from their drinking water alone. And this is assuming that they’re not also exposed to lithium in other ways.

Common Uses of Lithium

Which they probably are, because lithium has a wide variety of applications. In 2017, the USGS estimated that 48% of the global market for lithium was batteries, 26% was ceramics and glass, 7% was lubricating greases, the remainder being industrial uses like polymer production and air treatment. They also mention a couple uses like “agrochemicals”, airbag ignition, aluminum alloys, cement and concrete additives, and dyes and pigments.

You’ll remember from our review in the PFAS section that some of the most obese professions include firefighters, cooks, food workers, cleaning workers, motor vehicle operators, vehicle mechanics, transportation and material moving, healthcare support, health technicians, and some construction occupations (“Helpers, construction trades” and “Other construction and related workers”).

If you go and see your local auto mechanic, the black smears covering his hands and forearms might be engine oil. But they might also be lithium grease. This grease is ubiquitous in auto engineering, routinely applied to hinges, joints, and pivot points. It’s used in aviation and on many kinds of heavy machinery, including logging and construction equipment, trains, and tractors. It also has a number of household applications. You might put it on your garage door, or the hinges on the gate of your fence. About 7% of the global supply of lithium goes into lubricating greases of one kind or another. That’s a lot of grease.

In addition to any lithium they’re exposed to in their food and water, vehicle mechanics, truck drivers, and transportation workers are also constantly exposed to lithium grease at work. They may literally be rubbing the grease all over their hands. Like any grease, this is hard to get off your skin, in most cases requiring a special soap. Hopefully they keep it out of their eyes and mouth, but even so, it doesn’t seem like it would be great for you.

Construction workers also use lithium grease to lubricate their tools and equipment. They may be exposed through lithium added to concrete and cement. Lithium is used in agrochemicals like pesticides, though information on exactly what agrochemicals this includes is spotty. It’s possible that this explains the higher levels of obesity in cooks and food workers.

Lithium grease isn’t considered food-safe, and in theory it shouldn’t be used on food-handling equipment. In practice, however, manufacturers and restaurants don’t always follow regulations. A quick Google search reveals incidents like lithium grease being stored with food equipment, lithium grease being stored next to garlic bulbs, and lithium grease being stored above mustard.

It’s not clear if there’s a connection with healthcare professionals, firefighters, or cleaning workers. Maybe there’s a hidden lithium connection out there. But there doesn’t have to be. Lithium may just be part of the story — the rest could be explained by other contaminants, like PFAS.

And speaking of PFAS, there’s actually a connection. Lithium greases often include other substances to improve their performance, including Teflon, aka PTFE.

Bioaccumulation

One strike against lithium as an explanation for the obesity epidemic is that it only stays in the body for a few days, with a half-life of 18-36 hours. The medical consensus seems to be that it probably doesn’t bioaccumulate.

But this may not be a problem, for a few reasons. First, lithium may be able to affect weight without accumulating in the body. One possible mechanism by which environmental contaminants could cause obesity is by interfering with the microbiome. If exposure to lithium changes the composition of your gut microbiota, then exposure to lithium could have serious impacts on your weight without any bioaccumulation. Even brief exposure to lithium could have long-lasting effects. And in fact there is evidence that dietary lithium affects the microbiome, at least in rats.

Second, the medical consensus might simply be wrong. While lithium is traditionally measured in the serum, this may not be the best way to evaluate bioaccumulation. “In contrast to other psychotherapeutic drugs,” says one paper on the pharmacokinetics of lithium, “Li+ is fairly evenly distributed in the body, but in tissues such as the white matter of the brain, the bones, and in the thyroid gland the concentrations per kg wet weight are about twice those in the serum.” 

A particularly interesting example is a case study of a patient who died after lithium poisoning. Researchers found that most tissue samples (“liver, spleen, kidney, lung, muscle, cardiac muscle, pancreas”) contained about the same concentration of lithium as found in the serum, in the range of 0.4-0.6 mmol/kg. But in the thyroid gland and in brain tissue (especially white matter), lithium concentrations were nearly twice as high as in serum, in the range of 0.7-0.8 mmol/kg. They suggest that this is to be expected, saying, “lithium has an increased affinity to thyroid tissue,” and, “high concentrations of lithium in brain tissue – especially in white substance – agrees with investigations that reveal the lithium elimination from brain tissue to be slow.”

This is particularly interesting not only because it seems to show evidence of bioaccumulation, but because of the particular tissues in which concentrations were found to be highest. The thyroid gland is very important to weight regulation, and so to find that lithium concentrations in that organ were second-highest in the body is very neatly in line with our expectations. Maybe it shouldn’t be a surprise, because lithium therapy is associated with thyroid disease.

For a while this was the best evidence of lithium accumulation in the brain, but a few weeks ago our friend Dr. Grace Rosen sent us this Nature paper published in March 2021. In this paper, the authors took 139 brain samples from three deceased individuals and took them to a nuclear reactor in Bavaria, where they were placed in a chamber and shot with a “well focused beam of cold neutrons with a neutron flux of ɸ = 1.2 × 10¹⁰ cm⁻² s⁻¹”. The lithium-neutron interaction gives a unique “associated coincident energy pattern”, which allowed them to measure the amount of lithium at different sites in the brains. Unsurprisingly, the patient receiving lithium therapy at the time of death had the highest concentration, but the brains of the other two individuals did as well, presumably from trace exposure.

Just like the case study above, they found that lithium was more concentrated in white than in gray matter. Additionally, the authors note that lithium concentrations were especially high in the thalamus and Brodmann Area 25. This is interesting for our purposes because Brodmann Area 25 “influences changes in appetite and sleep” and the thalamus governs “sensory relay in visual, auditory, somatosensory, and gustatory systems.” Both of these brain regions are related to eating behavior.

Sadly the authors did not examine any samples from the hypothalamus, which governs eating more directly, but this is still evidence that trace lithium bioaccumulates in brain regions that are important to the regulation of food intake and body weight.

[Next Time: HIGHLIGHTS FROM THE COMMENTS]

[PART I – MYSTERIES]
[PART II – CURRENT THEORIES OF OBESITY ARE INADEQUATE]
[PART III – ENVIRONMENTAL CONTAMINANTS]
[INTERLUDE A – CICO KILLER, QU’EST-CE QUE C’EST?]
[PART IV – CRITERIA]

Obesity in the United States – Dysbiosis from Exposure to Low-Dose Antibiotics? suggests that the obesity epidemic is driven by population-wide exposure to residual antibiotics from livestock cultivation, and the resulting impact on gut microbiota.

This is one of the most similar proposals to the theory presented in A Chemical Hunger, though they don’t go quite as far as we do, still attributing some of the influence to diet and exercise: “Most reports attribute the obesity epidemic to factors such as excess food energy intake, changes in diet and eating behavior, and increasing sedentary life style. Undoubtedly, these factors contribute, but can they all account for the rapid increase in this problem that occurred over the last two decades?”

They make a pretty compelling case. A large percentage of antibiotics are excreted in animal waste and end up in the water supply, where they affect natural microorganisms. Relevant to our interests, antibiotics are more and more prevalent in rivers as they make their way towards the ocean: “The only site at which no antibiotics were detected,” they write, “was the pristine site in the mountains before the river had encountered urban or agricultural landscapes. By the time the river had exited the urban area, 6 of the 11 antibiotic compounds that were monitored were found in the samples. At Site 5, which had both urban and agricultural influences all five of the TCs monitored were detected.”

Exposure has increased in the US over time, closely matched with the increasing prevalence of obesity — “practically overlapping with the counties with the highest obesity prevalence in the US.” Similar trends can be observed in other countries. For example, in Great Britain, by 1958, around 50% of British pigs were fed antibiotics, and nearly all piglets were given food containing tetracyclines. In West Germany in the 1960s, an estimated 80% of feeds for pigs, calves, and poultry contained antibiotics.

The Effects on Human Health of Subtherapeutic Use of Antimicrobials in Animal Feeds opens by saying, “The food-producing animal and poultry industries have undergone a dramatic change that began around 1950.” That’s a little earlier than we would expect, but depending on how you measure things, it took until the 70’s or 80’s before things really got rolling.

In meat animals, antibiotics often lead to weight gain, sometimes as high as 40% weight gain compared to control, and there’s reason to suspect that this might be linked to the microbiome. Gut microbiota influence energy intake and body weight in mammals, and even short courses of antibiotics can reduce gut microbiota and increase BMI in humans (though the BMI effect was only seen in some antibiotics).

There’s even a study where they put fecal matter from human twins into germfree mice. (This is one of the more creative study designs we’ve seen.) They started by finding pairs of twins where one twin was fat and the other twin was lean. This is pretty uncommon — normally, twins weigh the same amount. They transplanted fecal matter from the twins into mice and found that mice that got fecal matter from the obese twin gained weight — unless it was housed with one of the mice who got fecal matter from the lean twin.

However, there is also evidence against this picture. For one thing, Germany, Spain, Italy, and Japan all use a lot of antibiotics in their meat, and none of these countries is particularly obese. Australia and South Africa are both pretty obese, but both of these countries use less antibiotics than usual. This could maybe be reconciled if these countries use different kinds of antibiotics, but we would need to see that case made to evaluate it.

There’s also some evidence in favor of this theory that this paper didn’t review.

For one thing, people who eat fewer animal products have lower BMIs, and the effect seems to be dose-dependent. In a sample from 2002-2006, average BMI was lowest in vegans (23.6) and incrementally higher in ovo-lacto vegetarians (25.7), pescitarians (26.3), semi-vegetarians (27.3), and nonvegetarians (28.8). We can note that the BMI for vegans is about the same as that found in hunter-gatherers and in Civil War veterans in the 1890s. That said, everyone in this sample was a Seventh-Day Adventist, so they may not be all that representative.

India and Japan are the least obese of the developed countries. Both have obesity rates below 5%. India is the most vegetarian country on the planet and Japan, while not especially vegetarian, mostly consumes seafood in place of meat products.

This would mean that vegan diets would work really well for weight loss, right? Well, maybe. As we previously reviewed, all diets seem to work a little, and no diet seems to work all that well. We see something similar in vegetarian and vegan diets. A 2015 meta-analysis found that people assigned to vegetarian diets lost more weight than those assigned to nonvegetarian diets. People on vegan diets lost a little more weight than people on vegetarian diets, about 5.5 pounds (2.5 kg) to 3.3 pounds (1.5 kg). The studies differed quite a bit in the size of the effect, but all of them had similar conclusions. The other meta-analysis from 2015 found the same general pattern, and individual studies comparing different types of vegetarian and vegan diets seem to confirm this dose-dependent trend.

This looks a lot like other studies, where the differences between diets are technically reliable but so small as to be basically meaningless, but the possible dose-dependent effect is interesting.

The most interesting study might be this one, that compared a vegan diet to a conventional low-fat diet. So far so standard, but unlike most diet studies, which end after 12 or 18 months, this one followed up two years later. The vegan group not only lost more weight (4.9 kg versus 1.8 kg), they kept it off better at the two-year followup (3.1 kg versus 0.8 kg). On most diets people lose a little weight but then gain it right back, so the fact that people kept most of the weight off for two years is interesting. Even so, the amount of weight lost in an absolute sense is still quite small. It could take more than two years on a vegan diet for you to see all the effects — but if this were the case, you’d think people would have lost even more weight by year two, but that’s not what we see.

None of these are smoking guns. At best, they are consistent with the idea that some of these contaminants are more prevalent in animal-based foods. And we know that this can’t be about the animal products themselves, because hunter-gatherers and our ancestors in 1890 ate lots of meat and didn’t experience modern levels of obesity.

Consider:

Environmental contaminants tend to build up in animals through the plants they eat, so any contaminants in the environment will bioaccumulate, and concentrations will be higher in animals than in groundwater or in plants. Compounds in a farmer’s fields will end up in the corn or alfalfa fed to their cows, and the cows will end up getting an even larger dose, which will be passed on to the person who eats the resulting cheeseburger. So the fact that meat consumption is linked to obesity doesn’t necessarily implicate antibiotics. It could be something else in the meat.

[Next Time: THE NUTRIENT SLUDGE DIET]

[PART I – MYSTERIES]
[PART II – CURRENT THEORIES OF OBESITY ARE INADEQUATE]
[PART III – ENVIRONMENTAL CONTAMINANTS]
[INTERLUDE A – CICO KILLER, QU’EST-CE QUE C’EST?]

Evidence strongly suggests that the obesity epidemic is the result of environmental contaminants.

However, it’s not entirely clear exactly which contaminants are responsible.

If we’re lucky, a few compounds are entirely responsible for the increase in obesity over the past forty years, and we can ban those chemicals.

If we’re not lucky, the obesity epidemic is the result of dozens or even hundreds of different contaminants, each with a small effect, which when combined lead to extreme obesity. In this case we can try to ban or regulate them all, but it will be much more difficult to find ways to get all of them out of our food, water, and homes.

While more work will be needed to pin down exactly what contaminants are responsible, we can make some very educated guesses, because we already know what kind of contaminants we’re looking for.

4.1    Invention and Introduction

The big inflection point for the obesity epidemic was around 1980, so we should be looking for compounds that entered the environment slightly before then. Either they were discovered around 1960-1970 and were immediately introduced, or they were discovered some time before and went into widespread use just before 1980.

These contaminants may be synthetic, but they don’t have to be. These could also be naturally occurring compounds that are introduced into the environment through human activity.

We’re aware that correlation doesn’t imply causation. By itself, showing that the time course of some contaminant is related to the upward trend in obesity rates isn’t enough to show that the contaminant is responsible for the obesity epidemic. But it is suspicious. We should keep in mind that the link between smoking and lung cancer was largely established through correlational data.

If we don’t see a relationship between a contaminant and the obesity epidemic, it’s harder to make the case for that particular contaminant. For example, some people have proposed that certain milk proteins might be behind the obesity epidemic. But if you look at dairy consumption by country, you immediately see that some of the leanest countries are near the top and many of the most obese countries are near the bottom. As a result, we don’t think this is a serious candidate and we don’t discuss it in this paper. A relationship is one of the first signs we should look for in a proposed explanation, even though, by itself, it’s not enough to be convincing.

4.2    Between-Group Differences

There are large differences in rates of obesity between counties, states, countries, and even professions.

Some of this is due to differences in factors like altitude or genetics. For example, in a group of about 43,500 patients from the San Francisco bay area, the rate of obesity in European-Americans was about 26%, but the rate of obesity among Asian-Americans was much lower, at 12% obese. This also differed quite a bit by country of origin. Among Asian-Americans, Filipino-Americans (24%) and Indian-Americans (17%) were the most obese, and Chinese-Americans (7%) and Vietnamese-Americans (6%) were the least obese.

This suggests that even if these environmental contaminants were just as widespread in China as they are in the United States, the rate of obesity would only be about 7%. In this light, it’s not surprising that the rate of obesity in China is currently around 6%. That isn’t a mystery that we need to explain, it’s about what we would expect based on what we know about the genetics of obesity.

Despite this, some of the difference in obesity rates between countries is probably due to differences in exposure to contaminants. This is something we should keep an eye out for. In contrast to China, Indian-Americans are about 17% obese, less obese than European-Americans but much more obese than people living in India, who are only about 4% obese.

4.3    Dose Dependence

An obvious smoking gun would be if the amount of exposure to a contaminant were related to obesity. If people who are exposed to a high dose of the contaminant are fatter than those who are exposed to a low dose, that would be a strong indication that the contaminant is responsible.

Right? 

Well, maybe.

We are all living in a fattening environment. A 2012 meta-analysis of 115 studies concluded that around 75% of individual difference in BMI is genetic. This is entirely compatible with the idea that contaminants are responsible for the difference between obesity rates in 1970 and 2020. It just means that, now that everyone has been exposed to more or less the same contaminants, 75% of the variation in the population is genetic.

That only leaves 25% of the variance to potentially be explained. If we assume that all of that variance is due to differences in dose, then some simple math tells us that the correlation with dose would be r = 0.50. This is a reasonably large correlation, but we also shouldn’t expect things to be so simple. If we expect that there are ten contaminants, the correlation between dose and BMI for each would be about r = 0.16. If dose explains only 15% of the remaining variance rather than 25% (leaving a reasonable 10% as the result of noise), the correlation for each of the contaminants would be r = .12.

With a large enough sample size, this would certainly be detectable. But dose alone almost certainly can’t explain all the remaining variance. We should be aware that even if there is a strong dose-dependent effect, the effect size might appear statistically to be quite small.

At some point in the past, these contaminants weren’t in the environment at all. Now, they’re so widespread that almost everyone in the industrialized world is getting a dose. This means we’re working with a restricted range. No one, or almost no one, has a dose near zero. Most of us are probably getting similar doses — that’s part of what it means when we see that 75% of the variation in the current population is genetic.

When the range of a variable is restricted, the correlation always ends up looking smaller than it really is. That link leads to a paper where they show that, for a dataset with a true correlation of r = .82, with different range restrictions the apparent correlation can be .51, .47, or even as small as .18! In some cases, a restricted range can even make a positive correlation appear to be negative.

We know that in most samples, everyone will have similar levels of exposure to the contaminants. We’ll be working with a restricted range, and any correlation we see will be smaller than the true correlation. It’s not clear how restricted our range is, but the correlation we see may be much smaller than the true relationship. It might disappear altogether, or even appear slightly negative. And remember, if genetics is 75% of the variance, then the largest dose-dependent correlation we can expect to see is only .50, which is not all that large to begin with.

(This is a known issue in studying public health. E.g. Lindeberg: “Another difficulty is that the variation of dietary habits in the population being studied may be too small to allow for demonstration of a possible relationship with health. Salt consumption among a particular ethnic group may not show much variation, and the majority has an intake that is much higher than what was practically feasible during evolution. This problem can be compared to studying the importance of smoking for myocardial infarction without having access to non-smokers. (In the case of smoking, often no relation is apparent in epidemiological studies.)” )

One pattern you might expect to find is that dose-dependent effects only become apparent in samples with a wide range of dosages — that is, in cases where the range isn’t restricted. They might be especially pronounced in groups that have extremely high levels of exposure, such as people who work with these contaminants directly, because that increases the range of dosages.

There might also be diminishing returns. Let’s imagine that today, people on average get a dose of 100 units. Back in 1970, everyone got a dose of 0 units. Now, the first 20 units might be very fattening indeed. But in general, the human body can only get so fat. The first 20 units might make you gain 20lbs on average. But the next 20 units only make you gain 10lbs. And the next 20 units make you gain only 5lbs. Now that everyone is at 100 units, every additional unit of exposure leads to a nearly undetectable change in body mass.

We don’t know that there is diminishing weight gain from greater doses of these contaminants, but diminishing returns are pretty common in pharmacology (see for example here). If there are diminishing returns, we may be near or past the ceiling effect, in which case we might not be able to detect a dose-dependent effect. If this were the case, however, we might expect to see dose-dependent effects in samples with lower average doses; perhaps samples from the 1980s or 1990s, or samples from developing countries which don’t yet have doses in the same range as industrialized countries.

A further complication is that being exposed to contaminants doesn’t make you gain weight that very same day. Even on Olanzapine, which makes people gain an average of 13.7 kg after 48 weeks, you generally don’t see your first kilogram of weight gain until after 8 weeks. The dose in your system today will be less correlated with your weight than the dose you were on 6 months ago. The dose will be a lagging indicator, and this will also reduce any correlation.

This is further complicated by the fact that these compounds might have paradoxical reactions, which is what we call it when a drug sometimes has the opposite of its normal effect. This would cause some portion of people to actually lose weight, and would further reduce the apparent correlation between the contaminant and obesity.

Finally, there are a priori reasons for us not to expect there to be strong correlations in the existing literature. If there were a compound or contaminant that was correlated with BMI, even a relatively small correlation like r = 0.20, someone probably would have noticed. This means that either 1) the contaminants are compounds that we don’t usually measure, so no dataset exists where we can compare them to measures of obesity, or 2) the relevant contaminants are commonly measured, but for statistical reasons like the ones above, there isn’t an obvious correlation with obesity.

Evidence of a dose-dependent relationship would be a smoking gun in favor of a contaminant being one of the causes of the obesity epidemic. But the lack of a dose-dependent relationship isn’t evidence against the contaminant being involved.

4.4    Environmental Interactions

In the following sections, we do our best to identify which of the contaminants we’re putting into the environment could be the cause of the obesity epidemic, and we believe that we have found some likely candidates. Unfortunately, this search is complicated by the fact that chemistry and biology allow for bewildering interactions, and sometimes seem to be working against us.

There are two general ways this can happen. The first is that when chemical contaminants end up in the environment, they can be transformed into different compounds. This can occur as a result of interactions with minerals in the groundwater, from exposure to sunlight, from exposure to radioactivity, or from chemical interactions with other contaminants. Since contaminants can sit in soil and groundwater for decades, there’s a lot of time for these transformations to happen.

In her book Silent Spring, Rachel Carson describes how contaminants “pass mysteriously by underground streams until they emerge and, through the alchemy of air and sunlight, combine into new forms that kill vegetation, sicken cattle, and work unknown harm on those who drink from once pure wells.”

She provides a few illustrative examples. One case involved a manufacturing plant in Colorado. In 1943, the Rocky Mountain Arsenal of the Army Chemical Corps began to use the plant, located near Denver, to manufacture war materials. After eight years, the same manufacturing plant used to make these war materials was leased to a private oil company for the production of insecticides. Even before this point, however, there were already reports from miles away of mysterious sickness in livestock, crops dying and turning yellow, and even human illness, possibly related. A thorough investigation eventually revealed that the groundwater between the arsenal and the farms had become contaminated, but it had propagated so slowly that it took several years for the contamination to reach the farmland.

Analysis of the farms’ shallow wells revealed contamination with arsenic, chlorides, and other dangerous substances. This was enough to explain the majority of the reports of illness and crop damage. But most mysterious was the discovery of the weed killer 2,4-D in some of the wells:

Certainly its presence was enough to account for the damage to crops irrigated with this water. But the mystery lay in the fact that no 2,4-D had been manufactured at the arsenal at any stage of its operations. After long and careful study, the chemists at the plant concluded that the 2,4-D had been formed spontaneously in the open basins. It had been formed there from other substances discharged from the arsenal; in the presence of air, water, and sunlight, and quite without the intervention of human chemists, the holding ponds had become chemical laboratories for the production of a new chemical—a chemical fatally damaging to much of the plant life it touched. And so the story of the Colorado farms and their damaged crops assumes a significance that transcends its local importance. What other parallels may there be, not only in Colorado but wherever chemical pollution finds its way into public waters? In lakes and streams everywhere, in the presence of catalyzing air and sunlight, what dangerous substances may be born of parent chemicals labeled ‘harmless’?

While we can try to identify the contaminants that cause obesity, the disturbing fact is that the contaminants responsible may be compounds which we are unfamiliar with, because they weren’t created in a lab and have never been examined for safety. Again, Rachel Carson puts it better than we could:

Indeed one of the most alarming aspects of the chemical pollution of water is the fact that here—in river or lake or reservoir, or for that matter in the glass of water served at your dinner table—are mingled chemicals that no responsible chemist would think of combining in his laboratory. The possible interactions between these freely mixed chemicals are deeply disturbing to officials of the United States Public Health Service, who have expressed the fear that the production of harmful substances from comparatively innocuous chemicals may be taking place on quite a wide scale. The reactions may be between two or more chemicals, or between chemicals and the radioactive wastes that are being discharged into our rivers in ever-increasing volume. Under the impact of ionizing radiation some rearrangement of atoms could easily occur, changing the nature of the chemicals in a way that is not only unpredictable but beyond control.

To make matters worse, something quite similar can happen inside our bodies. As surprising and chaotic as the interactions between contaminants can be, their interactions with human biochemistry can be even more complicated.

“A human being,” writes Carson, “unlike a laboratory animal living under rigidly controlled conditions, is never exposed to one chemical alone. Between the major groups of insecticides, and between them and other chemicals, there are interactions that have serious potentials. Whether released into soil or water or a man’s blood, these unrelated chemicals do not remain segregated; there are mysterious and unseen changes by which one alters the power of another for harm.” She goes on to describe several such interactions in gory detail.

The organic phosphates, “those poisoners of the nerve-protective enzyme cholinesterase,” become much more dangerous if a person has previously been exposed to chlorinated hydrocarbons that injure the liver. Pairs of different organic phosphates themselves can also interact with each other, “in such a way as to increase their toxicity a hundredfold.” Organic phosphates also have the potential to interact with all sorts of other things in the environment, including prescription drugs, synthetic materials, and food additives.

Similarly, a person exposed to DDT is much worse off if they have already been exposed to another hydrocarbon that causes liver damage — “so widely used as solvents, paint removers, de-greasing agents, dry-cleaning fluids, and anesthetics.” As a result, a dose of DDT that is survivable for one person may be devastating to someone else. “The effect of a chemical of supposedly innocuous nature can be drastically changed by the action of another,” Carson tells us. “One of the best examples is a close relative of DDT called methoxychlor”:

Because it doesn’t accumulate in the body to any great extent when given alone, we are told that methoxychlor is a safe chemical. But this is not necessarily true. If the liver has been damaged by another agent, methoxychlor is stored in the body at 100 times its normal rate, and will then imitate the effects of DDT with long-lasting effects on the nervous system. Yet the liver damage that brings this about might be so slight as to pass unnoticed. It might have been the result of any of a number of commonplace situations—using another insecticide, using a cleaning fluid containing carbon tetrachloride, or taking one of the so-called tranquilizing drugs, a number (but not all) of which are chlorinated hydrocarbons and possess power to damage the liver.

Another good example is malathion, an insecticide that at the time was commonly used by gardeners. The name of this product sounds so evil that we’re surprised it passed the review of the corporate public relations people, but apparently the name comes from the smell and means “bad sulphur”, so there you go. Malathion is extremely deadly to insects but is “safe” for mammals, including humans. But malathion is only “safe” because the mammalian liver detoxifies it with an enzyme, rendering it harmless. “If, however, something destroys this enzyme or interferes with its action,” we are warned, “the person exposed to malathion receives the full force of the poison. Unfortunately for all of us, opportunities for this sort of thing to happen are legion.”

In particular, Carson relates the story of how a team from the FDA found that when malathion was administered at the same time as some of the other organic phosphates, “a massive poisoning results—up to 50 times as severe as would be predicted on the basis of adding together the toxicities of the two.” This led them to test the combination of many different organic phosphates, and found that many pairs of these compounds are exceedingly dangerous in combination.

The reason for this appears to be “potentiation” of their combined action — when one of the compounds destroys the liver enzyme responsible for detoxifying the other. “The two need not be given simultaneously,” we are warned. “The hazard exists not only for the man who may spray this week with one insecticide and next week with another; it exists also for the consumer of sprayed products. The common salad bowl may easily present a combination of organic phosphate insecticides. Residues well within the legally permissible limits may interact. The full scope of the dangerous interaction of chemicals is as yet little known, but disturbing findings now come regularly from scientific laboratories.”

Carson relates several more examples in a similar vein. Malathion also appears to become much more dangerous when a person is exposed to certain plasticizing agents. Just like its combination with other organic phosphates, “this is because it inhibits the liver enzyme that normally would ‘draw the teeth’ of the poisonous insecticide.” Similarly, exposure to malathion seems to increase the effect of certain prescription drugs, including muscle relaxants and barbiturates.

In addition, we found that malathion can, under some conditions, transform into malaoxon, which is 61x more deadly. One of these conditions is when malathion is exposed to chlorine, as it might be in some drinking water.

Our concern in this paper isn’t the toxicity of different insecticides, of course. The point is that the contaminants that cause obesity may not have a straightforward profile. It’s possible that two (or more) well-known and relatively safe contaminants combine in groundwater to form an unknown new contaminant that causes weight gain, and a host of other problems. It’s possible that a single contaminant becomes something else entirely when it is exposed to sunlight in fields, ponds, and rivers. It’s possible that there are two contaminants, neither of which cause obesity in isolation, but which in combination overwhelm the body.

4.5        Three Possible Contaminants

In the next parts, we propose some contaminants that we think might be responsible for the obesity epidemic. But we should make it clear upfront that the theory itself doesn’t hinge on these compounds. Even if it turns out that none of these compounds could possibly be responsible for the modern rise in obesity, we still think that the evidence is very strong that environmental contaminants are responsible.

We take a close look at three contaminants, and examine the evidence for each.

[Next Time: SUSPECT NUMBER ONE]