Thursday, April 28, 2011

Let Us Honor Ancel Keys, Our Patron, As We Cherry Pick Studies to Bash Fructose (Revised and Extended)

My apologies to anyone who received this in their RSS feed on Thursday as a teaser.  I have now revised and extended it to include several studies showing that diet-induced obesity can be achieved in rats and mice without using any sugar at all, and have included a clearer conclusion.

Ancel Keys is best known in the assorted real food communities, or at least the assortment of real food communities friendly to dietary fat, for his infamous cherry picking.

Keys had presented data from six countries, purporting to show a clear linear relationship between the amount of fat consumed in a country and its incidence of heart disease.  This graph is shown on the left below.  The one problem was that data was available for 22 countries at the time, and including that data demolished the relationship.  The "true" graph is shown below on the right.

December 22, 2011 Update:  It would be more appropriate to say that including the data "greatly diminished" the relationship than that it "demolished" the relationship because the positive relationship still exists.  See this excellent analysis by Denise Minger.

If we are going to criticize Keys for cherry picking, it behooves us to root out cherry picking from our own communities as well.  This requires constant self-criticism, because bias is human nature, and anyone who isn't engaged in a devoted battle to overcome their own bias will be its prisoner.  I myself cannot claim to always be victorious in this battle, but I do try.

I consider Dr. Robert Lustig an ally in the fight for real food.  He makes as his primary enemies junk foods rich in refined sugar and high fructose corn syrup, and I agree these should be targeted as likely the most important causes of chronic disease.  Dr. Lustig and many in the blogosphere, however, are circulating the claim that high-fat diets only induce obesity in laboratory animals if they are also high in sucrose.  This simply isn't true.

Let's take a look at these diets.

About seven minutes into a recent interview on sugar and health, Dr. Lustig stated the following:
Those animal studies are very clear in terms of what's going on.  If you wanna do diet-induced obesity, people say you put the rat on a high-fat diet.  Garbage.  Here's why.  You can't get a rat to eat a high-fat diet, unless you add 20 percent sucrose to the diet.  Basically what you're doing is you're giving them cookie dough.  But if you give them lard, they won't eat it.  They actually lose weight on that because that's the Atkins diet for them.  It's not palatable, they don't like it, and they actually lose weight and their metabolic parameters improve.  The only way to get an animal, rat, monkey, or any other animal for that matter, to eat a high-fat diet, is to lace the fat with sucrose.  So the question is, which is doing the damage?  The fat, or the sucrose in that case?  And the answer is, both.

Research Diets marketed the first high-fat diet specifically designed and marketed to induce obesity in laboratory animals, and currently estimates that it ships enough high-fat diet to keep 50,000 mice worldwide on the march toward obesity at this very moment.

Their original high-fat diet contained 45% of calories from fat, mostly lard.  It did contain 17% of calories from sucrose, but the control diet it was matched to contained twice as much sucrose

Nowadays, their 60% fat diet is most popular.  Again, most of the fat comes from lard.  The diet contains less than 7% of its calories from sucrose, which translates to about 3% of calories from fructose.

The diet also contains 13% maltodextrin, which is partially digested starch.  The maltodextrin used in this diet averages ten glucose units per molecule, so can hardly be called a "sugar."  A "sugar" would contain one or two units per molecule.  In any case, maltodextrin does not contain fructose.

As I pointed out in "They Did the Same Thing to the Lab Rats That They Did to Us," the American Institute of Nutrition (AIN) recommended the inclusion of 13-15% maltodextrin broken down even further than this as well as 10% sucrose in order to reduce the amount of heat damage that would occur during the pelleting process.  Thus this diet can hardly be said to be high in maltodextrin and it is pretty low in sugar.

Here's an example of a recent paper that used this 60% fat diet and compared it to the 10% fat, 35% sucrose control diet over the course of 8 weeks in mice:
Kirpich IA, Gobesjishvili LN, Bon Homme M, Waigel S, Cave M, Arteel G, Barve SS, McClain CJ, Deaciuc IV.  Integrated hepatic transcriptome and proteome analysis of mice with high-fat diet-induced nonalcoholic fatty liver disease.  J Nutr Biochem. 2011;22(1):38-45.
Compared to the mice eating the low-fat control diet relatively high in sucrose, the mice eating the high-fat, low-sucrose diet had 55% greater body weight and seven times as much body fat.  The mice eating the high-fat diet developed fatty livers, whereas the mice eating the low-fat diet did not.

This all happened even though the mice eating the low-fat control diet were consuming over five times as much sucrose as the mice eating the high-fat diet.

More to the point, however, there are several studies available showing that high-fat diets can induce obesity in mice without any sugar at all.

Here is what happens to two different strains of mice consuming high-fat or low-fat diets with or without sucrose for four months:

We can see the effect of fat if we compare the black and white bars to the patterned bars.  Clearly, fat increased adiposity more than sugar in each strain of mouse, but the effects were much more pronounced in the strain of mouse shown on the right.  The white bars represent a 58% fat diet that had no sucrose in it whatsoever.  This diet clearly increased adiposity, especially in the susceptible strain.

Lest we be afraid the maltodextrin is the culprit, let's take a look at this study by the same authors:

After 16 weeks, fat increased adiposity in both strains (this time adjusted for body weight).  The white bars represent the 58% fat diet that had no sugar.  On top of this, none of the diets contained maltodextrin.  Clearly high-fat diets can induce obesity in mice even without any sugar.

One interesting study showed that including sucrose instead of starch on a similarly high-fat diet can increase adiposity in rats by 23%.  Unfortunately, this study did not include a low-fat control group so we can't really tell whether the sugar-free, high-fat diet induced obesity.  Fortunately, however, the same authors performed a very similar study that compared starch to sucrose on a low-fat diet, which allows us to make a rough comparison:

The red and blue bars represent rats on diets that were 64% fat and 14% protein, with 21% carbohydrate as either starch or sugar.  The green and purple bars represent rats on diets that were only 16% fat, 20% protein, and 54% carbohydrate as either starch or fat.

By comparing the blue and red bars, we see that a very high-fat diet produces 23 percent more adiposity in rats when it is accompanied by sucrose instead of starch.  The sugar-free, high-fat diet shown in red, however, produced 74% more adiposity than the sugar-free, low-fat diet shown in purple.  This would suggest that very high-fat diets produce adiposity in rats even when they do not contain any sugar at all.  Or maltodextrin, for that matter.

Am I about to blame fat as the cause of obesity?  Not in a million years.

There are a few reasons.

First, Research Diets boasts that they sell over 10,000 kilograms of their lard-based high-fat diet every month and that they have sold it to thousands of individuals at over 420 institutions in 33 countries.  The company estimates that 50,000 mice are currently getting fat by eating it at this very moment.

What does this mean?  It is well known that the response of rats and mice to these diets is highly variable.  Two of the studies I cited above showed very large differences between strains of mice.  Similar differences have been seen in rats, and in some strains of rats there is a very clear division between "susceptible" and "resistant" rats.

As I pointed out in the "High-Fat Diets and Choline" section of my article, "The Sweet Truth About Liver and Egg Yolks -- Choline Matters More to Fatty Liver Than Sugar, Alcohol, or Fat," the first researchers who used high-fat diets to induce fatty liver noted that they could achieve this in some but not all strains of animals, and that they could achieve it during the winter but not the summer.  This could, perhaps, also be true for obesity. This means there are likely to be genetic, environmental, and other factors that determine the success of the diet.  If these factors confound rodent experiments, they would confound any generalization from rodents to humans to a much greater degree.

It also means that with the sheer quantity of people using this diet to induce obesity, there is plenty of room for failure to go unpublished.

In any case, fat or fructose may well operate very differently in whole foods than in purified diets.  It is overall dietary patterns that produce health or disease, not isolated dietary components.  This will relate to the overall nutrient-density of the diet, the diet's effect on satiety and the metabolic rate, and as Stephan Guyenet's new series is revealing, the effect of the foods within that diet on the reward centers of the brain.  All of these effects are determined by interactions between different components.

As I pointed out in "Reductionism and Holism Go Hand in Hand," looking at isolated components is important for understanding how things work, but if we don't re-synthesize the big picture our reductionism is pointless.  Blaming fat because of its effect in these purified diets would be an example of "pulling a Campbell."

There are many different factors that affect the ability of a high-fat diet to induce obesity in rodents and reviewing all of them is a gargantuan task for another day.

In any case, we can see quite clearly that fructose is not required for high-fat diets to induce obesity in rodents.   While I do want to help steer our society away from junk foods full of refined sugar and high-fructose corn syrup (and fructose-free white flour!), I don't think that contriving fictitious myths about the effects of dietary fat on rodents is going to help us uncover the true causes of obesity.

That's not to say that you can't produce weight loss if you restrict carbohydrates to less than one percent and protein to less than ten percent.  You can, as was done here.  Although this is not because the rats won't eat the diet.  But that's a topic for the next post.

Read more about the author, Chris Masterjohn, PhD, here.


  1. I checked the specs on the 60% fat D12492 diet. 775 kcal of the total 4057 come from sugars (maltodextrin and sucrose) -- that's 19%. Still rather high carb for a "high-fat" diet, no?

  2. GK, yes it contains these sugars, probably to aid in pelleting (you need ~10% maltodextrin or so to get proper pelleting) and palatability. And yes it may affect the metabolic context. However, I was addressing a point specifically about fructose. Claiming the high-fat diet is harmful because of its fructose content is absurd, because this content is very low (~3%) and because the control diet contains five times more fructose. Chris

  3. I just revised and extended this post.

  4. Interesting that the high-fat high-sugar fed A/J mice actually had less fat than the high-fat no-sugar mice, and that the low-fat high-sugar diet produced less obesity than the low-fat low-sugar diet in both strains. This implies that sugar can actually be protective against obesity, no?

  5. Collden, I mean, yes, in rodents anyway.

  6. Perhaps if the rats' diet included the level of fructose akin to HFCS, then a correlation with obesity might be found? I understand that because sucrose is 50/50 glucose/fructose, it has different effects on the body.

    I'm wondering if the HFCS that is so prevalent in our culture might be a trigger for obesity, while sucrose, as shown above, is not.

  7. Hi wildweekend,

    Actually sucrose is half glucose and half fructose, so once digested is quite similar to high-fructose corn syrup. I actually do think that HFCS (and refined sugar) play a role in the obesity epidemic, but my point here is just that we need to start with the truth, and it is not true that fructose must be included in a high-fat diet in order to make a rodent get fat.

    Studies on the ability of fructose to make rodents fat are quite variable and in totality not very convincing, though it probably plays a role in smaller amounts by increasing palatability and in larger amounts by contributing to metabolic problems, when combined with high fats. In humans it probably plays a role in hyperpalability in some contexts and also in cheap unsatiating liquid calories.

    But I'll tackle those questions more comprehensively in later posts. I think Stephan's current series on palatability will turn out to be very important.


  8. Nice work Chris. Time to put some of this sugar bull to bed :)

  9. Very interesting. Seems to be following a little of what Don on Primal Wisdom has been posting.

  10. Chris, i think i love you! not only are your posts singularly informative, but i got a great belly-laugh from the headline today! :-)

  11. Good to put this up. The Lustig lecture was very useful in promoting concern about excessive sugar, so I like it for that. But you're right that not all of the science was thoroughly researched. The rats studies also are important because they remind us that various combinations can be more damaging than a specific nutrient alone. On the other hand, I do still believe that sugar has been the primary cause of obesity in humans. I won't detail all the reasons for that here, but I'll send you a link to some articles once I have them up.
    I often wonder why people even bother doing studies on obesity in rats. Of all the mammals you might use, you'd be hard pressed to find one less suitable to studying human obesity than rodents. Among other things, rodents aren't adapted to high fat diets (as far as I'm aware, there are no rats that naturally eat a diet similar to the Inuit diet, for example), they express uricase (which means they react to fructose very differently), and they evolved to be able to hibernate, unlike humans. I haven't done a search, but Stephan was unsure whether there was any evidence that mice and rats even absorb as much fructose as humans from the GI tract. What all this means is that we basically have no controlled experiments that show long term effects of fructose on humans. I'm actually appalled by that -- how is it that we can spend billions of dollar on scientific research, and yet there is not one study available to show us whether something people eat daily is ultimately toxic?

  12. Chris,

    I'm assuming these mice were overfed a fixed calorie amount daily? Or was this an ad libitum study? I say overfed because both strains gained fat mass. I also assume that the strains do not have the similar fate of zucker rat which are almost essentially doomed to a life of obesity. I do not have access to the full study to acquire the information myself.


  13. The formula also contains soybean oil. I wonder if the results would be the same without it...

  14. Responses to Grok, Katy, Anonymous, Jeremy, Evan, and Tracey

    Hi Grok, thanks.

    Katy, thanks, what post of Don's did you have in mind?

    Anonymous, I'm taken, but thanks. We should all love each other anyway. ;-)

    Jeremy, good points. I'll post more about fructose, rats, and humans, and I'll include some stuff on absorption. I think rodent models are useful to study some of the principle of obesity that should be shared across species, but not necessarily specific dietary factors. In any case, these "high-fat diets" are not the same thing as eating fatty foods.

    Evan, these were all ad libitum studies, as is typical. I would consider it dishonest to provide data from a force-feeding study and not explicitly note that.

    Tracey, true. The soybean oil is added to ensure there is no essential fatty acid deficiency. In the AIN-93 diets formulated in 1993, they replaced the corn oil with soybean oil so that it would contain a better balance of omega-3 to omega-6, and they increased the vitamin E to protect the fatty acids from oxidation. I explained this in They Did the Same Thing to the Lab Rats That They Did to Us: Not all of these studies used soybean oil. I guess it's a give and take. It's difficult to perfect things with a purified diet like this. I suppose the most natural thing would be to feed beef tallow with a tiny amount of fish oil. There is, in fact, evidence that the omega-3 to omega-6 ratio has an effect. Stephan has compiled a lot of information on this and I'm sure he'll blog about it in the future more than he has already. In any case, my point here wasn't to address why the high-fat diets are effective, just to dispense with a particular myth that sugar is required for the effect, which isn't true.


  15. Seems to me rats and mice are probably not going to explain obesity increases. Seems that rats and mice that are susceptible to obesity do not have weight set-points as tight as humans. They eat calories opportunistically. Obese humans in general do not overeat, or at least not much. Once you reach your setpoint, obese people eat not much more than non-obese:

    It doesn't really seem to matter what the diet components are, once the set-point has been established, the only thing that negates obesity is starvation. (Yes, even atkins diet seems to have only short-term success)

    I do suspect that high-fructose diets may be able to shift a person a bit above their normal weight set-point - which may be associated with poor health. However this weight shift is probably small compared to natural weight variations in humans. If fructose does drive obesity rates, then it must be some effect early in life, while the epigenetics is still malleable.

    Having read Gina Kolata's book (Rethinking Thin) and much of Sandy Swarks blog, I am quite convinced obesity is misunderstood.

    My theory is, obesity is mostly a natural variation in humans, but some environmental conditions tend to make it express earlier in life. Adipose tissue is mostly harmless if not protective against morbidity, however because it can be induced by certain stresses early in life, there can be some conditions that will be associated with it - very poorly I note.
    I note there is scant evidence to show that obese people losing weight improves total mortality results - which would be essential for promoting weight-loss for health reasons, you would think? If someone knows of such evidence then I would be interested - as I have put this challenge around a bit!

  16. @gordon

    I'm pretty sure that you don't spend money at the dentist because you imagine it would increase your lifespan.

    I'm pretty sure that the amazing boom in folks spending money on youthful-appearance rejuvenative cosmetic surgery/dermatology, is not because they imagine it would increase their lifespans.

  17. Greetings. I applaud the sentiment of being fair rather than being a biased advocate.

    Leaving aside for now the matter of whether Taubes is the worst cherry-picker of our time, I have a question about the accuracy of the posted Y & H graph:

    Why are the six countries plotted twice, yet with disparate values?

    Take Japan (the country of *processed* rice eaters that weren't fat or prone to MI for many generations, until they adopted Western food). It's plotted as 'Japan' and then again up a little higher as '13'. Do those two points represent different years? Or is the graph just very flawed?

    (Hopefully, you get notice of and reply to comments on older articles :)

  18. Lerner,

    Good question. I was reproducing the graph from another analysis, so I'm not sure why that value is duplicated. Eventually I'll reevaluate it myself.


  19. We study mice because of the detail available on the underlying biochemistry and the ability to bioengineer mutants. The mistake we make is to think that results with mice translate to humans as a positive rather than a negative model, that is, we should ask what are the features of (particular strain) mouse biochemistry that allow them to get fat without carbohydrate that do not let that happen in humans.

  20. Hi Chris,
    As thorough and analytical as your observations and research are, what if we came at the issue from the other side, as it were: anecdotally, aren't the people in the Western world who are either suffering with obesity, or are most likely to become obese, maintaining a diet that is either: 1)high in fat; 2) high in sugar, or 3) high in both?
    When it comes down to solutions, aren't we ultimately going to be advocating that those with obesity risks, or already there, should be re-hauling their diet with an intake of low sugar quantities and a surplus of healthy fats with minimal unhealthy fats?
    Many thanks-

  21. Hi Michael,

    I wasn't suggesting that diets high in sugar aren't a problem, I was just making the point that if we argue this, it should be done without cherry-picking the evidence. I think it's a bit more complicated than "too high in sugar" or "too high in fat," but certainly high-sugar high-fat junk foods are an abundant problem.


  22. If you are obese and you cut out carbs and increase fat you will get thin (as I and many others have). The problem comes when people, who have never been obese debate the topic, because for them it is a purely theoretical problem.

    1. I think Simon is right. I maybe not in the right place to tell what is right and wrong cause I never been obese. I'm just one of those people who based on theory and articles I've read. I think those obese people who lose weight can prove what would be the best or most effective diet for them.

  23. I added this paragraph:

    December 22, 2011 Update: It would be more appropriate to say that including the data "greatly diminished" the relationship than that it "demolished" the relationship because the positive relationship still exists. See this excellent analysis by Denise Minger.

    Within the last sentence is a link to Denise's latest post:

  24. Chris, it was always very apparent that the upper left quadrant was practically empty. That might be because lower-fat generally = lower calories ---- and maybe leanness is really the key.

    Also, if Keys were really intent on deliberate deception, then why wouldn't he have include those other points which also would fall close to his trend line?

    Lastly, Minger seems to give the impression that accounting for a country's prosperity wrt mortality is something new. But that was all thrashed out years ago, e.g.:

    Dietary Factors and Coronary Heart Disease

    Masironi 1970

    wrt per capita income and per capita
    energy consumption

  25. Is this result really so suprising. Correct me if I have this wrong, but doesn't the amount of protein have an impact? According to Steven Phinney, a protein intake of above 30% of calories stops ketosis (and presumably weight gain below in a diet containing below that 30% level would be very unlikely). So why couldn't these rats have a lower threshold? Protein does have an insulin response and and some amount of conversion of protein to glucose does occur.


To create a better user experience for everyone, comments are now moderated. Please allow up to one business day for your comment to post. In order to avoid the appearance of spam, please avoid posting links, especially to commercial destinations, and using all-caps.