Link: Taubes Contra Guyenet on Sugar

On Thursday I linked to Stephan Guyenet’s negative book review of Gary Taubes’ work on sugar.

Taubes asked that I pass on a counterargument he’s been working on for a while. In early January, he posted a short version of his case against sugar to Cato Unbound. It got three response essays – one by Terence Kealy, one by Yoni Freedhoff, and one by Stephan Guyenet, where he made some of the same points as in the linked article.

So now Taubes has now posted a response, The Case Against Sugar Isn’t So Easily Dismissed.

All of the essays are interesting and informative, and everyone involved deserves praise for some high-level debate that’s helped me clarify my thinking on this.

This entry was posted in Uncategorized and tagged . Bookmark the permalink.

21 Responses to Link: Taubes Contra Guyenet on Sugar

  1. Whatever Happened To Anonymous says:

    Is the fact that Scott’s being asked to publicized this a sign of his growing popularity, or of the lack of popularity of this debate?

  2. hls2003 says:

    Thanks for the links. I think Taubes is at his least convincing when dealing with the graph cited by Guyenet. He cites his own prior comment that there should be approximately a 20-year gestation period, then criticizes Guyenet for essentially cherry-picking on the top of the graph to make a temporary divergence seem like a trend. But the “temporary” divergence is about 17 years. We’re getting pretty close to the point where we should see reductions. So, good, I guess, in that we can check on Taubes’ theory in another 3-5 years? Presumably in response to that, Taubes’ goes on to comment that there is still “too much” consumption of sugar to see a downturn – over 70 lbs per capita compared to recommendations below 50 per capita. (The analogy of 17 cigarettes instead of 20 per day). But the per capita number would include all segments, and a declining per capita rate (even if still too high) would almost certainly indicate a greater portion cutting sugar out entirely or reducing radically. So I don’t think that “we’re still too high per capita” should offset Guyenet’s evidence (assuming it sticks over the next 3-5 years, or 10 years) that obesity rates look unrelated to declining sugar intake. At least there should be a connection with the rate of change, but that’s not what Taubes argues.

    • Briefling says:

      Huh, I thought the graph part was super compelling, especially Guyenet’s manipulated sugar-consumption axis — which IMO essentially proved Taubes’s claim, that Guyenet is “more interested in appearing to win an argument than […] dealing with what may be the single most important public health issue of our era.” Guyenet’s graph was not exactly cooked, but it sure as hell didn’t prove the point he wanted.

      And your statement, that we are going to see if Taubes is right over the next few years, is just silly. Surely you concede that it is possible for sugar consumption to drop moderately, but for diabetes to keep going up, even if sugar is the primary cause of diabetes? Like, there are a million ways that could happen. I would point, in particular, to intergenerational effects, where obesity in the parent increases the likelihood of obesity in the child, controlling for diet. Then you’d see “obesity momentum” in the data.

      Really, your argument has a lot of parallels to Guyenet — I guess I’d describe it as “over-logicking,” but basically you’re under-appreciating how difficult it is to figure out causation in complex systems, and as a result you think it’s much easier to refute some position (“sugar is the primary driver of obesity and diabetes”) than it actually is. In particular you think it’s sufficient to poke holes in Taubes’s handling of some small section of the data, in order to refute his entire argument — when in fact it’s essential to look at all the data, holistically, before you make any kind of pronouncement.

      Because one of the first things you learn in the social sciences is, there are a million ways that one particular view of one particular data set can end up looking weird. To prove or disprove a broad pattern, you must take a broad view of the data.

      • formid0 says:

        As you note, it’s hard to prove causation, but it sounds like you’ve made up your mind for Taubes?

        Have you read Guyenet’s book review? It’s pretty hard to argue that he doesn’t take a holistic view of the data on diet science.

      • Jliw says:

        As I read it, hls2003 isn’t saying “well I guess I’ve disproved Taubes’ ideas entirely!”, but rather that one piece of his proposed evidence is looking flimsier than expected. Any hypothesis can always be rescued by invoking more confounding factors, and I’m sure hls2003 is aware of this; there are always a million reasons a motivated reasoner (heh) can point to that might explain it all away; but that doesn’t mean that counter-inuitive results — such as, e.g., “diabetes rates aren’t tracking sugar consumption as closely as expected” — aren’t reason for skepticism.

        There’s a lot of extremely broad criticism in the rest of that post, which can be hard to address (“you’re thinking wrong” — what do you say, “am not!”?).

      • hls2003 says:

        Sounds like your problem is with Taubes, not with me. I haven’t even read Taubes’ book, so I have no strong opinion one way or the other on the ultimate topic. But Taubes himself, in the linked response to Guyenet, basically admits that, if his thesis is right, the sugar consumption / diabetes graph should look a lot like the smoking / lung cancer graph. In fact Taubes uses that exact graph as part of his refutation – pointing out that you can zoom in on portions of the lung cancer graph and make it look like the Guyenet graph. (The manipulation of scale I agree was well-pointed out in terms of visual impact).

        Taubes’ response, then, as I read it, was twofold: (1) Guyenet zooms in too far, too fast when in fact any effect would be delayed by a substantial amount (due to things like the inter-generational effects) – give it enough time and a broad enough scope, and the sugar / smoking graphs will begin to look similar; and (2) there may be less effect than you might otherwise think because sugar consumption is still high despite the reduction trend – his example is that this is like “17 cigarettes instead of 20” (the 15% drop) – which will mask the trend.

        The first point is fine as far as it goes. But it doesn’t dispute the underlying utility of Guyenet’s proposed evidence; it simply says that it’s too early to judge. That stance can’t last forever. At some point, Taubes’ refutation on that point has to morph from procedural to substantive. Taubes’ rebuttal suggested a 20-year period; I’d have to go back and review if he agreed or was paraphrasing Guyenet’s argument there. But at some point, the time period is long enough that it’s not “zooming in” or cherrypicking anymore. Whether that’s 3, 5, 10, 20 years I’m not sure, but Taubes doesn’t appear to dispute that it will be meaningful at some point.

        The second point I find less persuasive. Again, Taubes doesn’t appear to be disputing that the sugar graph eventually should look like the cigarette / lung cancer graph – that’s his thesis, after all, so he has to support a pretty big causal effect or else he’s making an unremarkable argument. If he wants to say there may be less effect or even continued growth due to ongoing usage, I don’t think he provided basis for that opinion, at least not in the quick refutation piece (maybe it’s in his books).

        His stated rationale for why ongoing consumption would be an effective mask is that it’s still at 70 lbs/yr when it should be below 50 lbs/yr or lower under current guidelines. But that’s not very convincing to me because the only way the average would be going down 15% is if significant numbers of people were doing the radical reductions he suggested. It’s implausible to think that you get a significant reduction without mixing in a lot more people at very low levels. If he’s saying that almost everyone does use, at close to the average, he didn’t support that (again, maybe it’s in his books). So you should be seeing some effect, even if it’s just a correlation with the rate of change at first. They can’t diverge forever and have his thesis be supportable, I would think. But that’s not what Taubes argued. He just said, without support, that there could still be an increase despite declining sugar usage. But at some point, Taubes seems to acknowledge that the graph should look like the lung cancer graph, and it doesn’t right now; if the two don’t correlate at some point, I thought he implicitly granted that as valid evidence against his theory.

        I’ve no idea in the end whether Taubes’ theory is right (I’m probably inclined to say no based on my heuristic that ‘simple explanations in biology are usually wrong’) and TBH I’m not going to do all the necessary research to find out. But that’s why I wasn’t terribly wowed by his argument on the sugar/diabetes graph.

        • pdbarnlsey says:

          I was also pretty unpersuaded by Taubes’ response to the graph.

          First point to note is that Taubes used “deaths from lung cancer” and compared it to “incidence of diabetes”. We’d expect decline in deaths to lag the decline in incidence, simply because diseases (even lung cancer) take a while to kill.

          Then he talks about the lag between peak cigarette consumption and the decline in cancer mortality, when the relevant lag is between when consumption actually started to fall and the drop in cancer. Eyeballing the graph, that’s a much shorter period.

          Finally, if there is some kind of threshold effect – over 60 pounds/year means the bad effects kick in – then that would show up in the data when sugar consumption was on its up – we’d expect to see a step-change in the diabetes rates when individual sugar consumption crossed that threshold. And we’d need to hear that that the decline in sugar consumption had been an across-the-board reduction by high consumers, insufficient to bring many of them below the critical threshold. Basically, if this were even on the right track you could provide some proof, rather than just speculating about it.

  3. ShemTealeaf says:

    It’s unfortunate that Taubes responded to this particular essay by Guyenet, because the book review that Scott linked contained significantly stronger points. The historical arguments are interesting, but they’re not terribly relevant to determining the truth of Taubes’s argument.

    Taubes has a strong rebuttal of Guyenet’s sugar intake vs obesity graph, but I’d also like to see him address the points from the book review, particularly the metabolic ward studies.

  4. jimrandomh says:

    A major point of Guyenet’s response is that the federal government has in fact been telling people to eat less sugar, as evidenced by the 1980 Dietary Guidelines for America booklet which he shows the first page of. He says:

    In the section titled “Maintain Ideal Weight,” the document lays out its four-part plan for weight loss: Increase physical activity, eat less fat and fatty foods, eat less sugar and sweets, and avoid too much alcohol. Again, the advice to limit fat and sugar intake receive equal attention, and the Guidelines clearly implicate excess sugar intake in obesity.

    So I tracked down the pamphlet (link) to see what it *actually* said. Here’s what it says in section 5 under the title “Avoid Too Much Sugar”:

    The major health hazard from eating too much sugar is tooth decay (dental caries). The risk of caries is not simply a matter of how much sugar you eat. The risk increases the more frequently you eat sugar and sweets, especially if you eat between meals, and if you eat foods that stick to the teeth. For example, frequent snacks of sticky candy, or dates, or daylong use of soft drinks may be more harmful than adding sugar to your morning cup of coffee — at least as far as your teeth are concerned.

    Overall, contrary to Guyenet’s summarization, the pamphlet seems more strongly anti-fat than it is anti-sugar.

    • davismx1 says:

      Overall, contrary to Guyenet’s summarization, the pamphlet seems more strongly anti-fat than it is anti-sugar.

      agreed, but i think what was going on in the zeitgeist at the time is more important than what a given pamphlet said (although government pamphlets and whatnot don’t exist in a vacuum and affect the overall awareness of a concept).

      anyone who lived through the 80’s and 90’s can tell you that “low fat” was emphasized above all else in popular media and the culture. everyone talked about needing to cut back on saturated fat. hell my mom even started buying skim milk! on the other hand, i don’t remember hearing similarly damning messages regarding sugar consumption. but everyone lost their shit over snackwells.

  5. JayMan says:

    For the record, Taubes didn’t offer much in defense of his case.

    I didn’t read his book, so I don’t know if he applied his best points there, but there isn’t a case to be made in this essay.

    Taubes and his critics are all arguing in circles.

  6. onyomi says:

    The disanalogy between sugar and smoking is that you can’t quit eating. You can eliminate refined sugar, but unless you literally eliminate all plant products (not a recommendation I think anyone makes for heart health), you’ll still be getting some carbs. Eliminating refined sugar, then, is not equivalent to quitting smoking, but to substituting wine or beer for whiskey or vodka–i. e. getting a less artificially concentrated and purified version of the same thing.

    Once distillation became a thing, alcoholism went way up because alcohol, which people had been consuming in a much more diluted form for millennia, becomes a lot more addictive when you concentrate it. Like the coca leaf becomes a lot more addictive when you concentrate it. Like carbs and fat both become more addictive when you concentrate them.

    • louisjburke says:

      I don’t think the disanalogy is accurate. I’d say the appropriate level would be analogous to living in the Alps to avoid air that isn’t fresh.

      • onyomi says:

        No, because refined sugar is just refined carbohydrate. It’s a more pure version of the thing you’d be getting in a more adulterated (and therefore better for you) version anyway. Avoiding polluted air is analogous to smoking because you don’t need any pollution. Sugar in your food, unlike pollution in your air, isn’t a poison or adulteration of something good for you. It is something good for you, but not when consumed in a highly purified and concentrated form.

        Sugar, when it comes packaged with a bunch of fiber, water, vitamins, phytonutrients, etc. is not only not bad for you, it’s good, even essential (your brain runs on sugar, after all, and must produce an approximation of it by ketogenesis in its absence). Fat is the same way: packaged in a nut, or an avocado, or a piece of salmon, you probably won’t get too much fat and a certain amount is even necessary for health. Packaged as a bottle of purified canola oil you pour into the brownie mix full of white flour and refined sugar, you easily can consume too much (note that I love brownies, but I’m pretty sure they’re highly addictive and bad for me, and not just because of the sugar).

  7. onyomi says:

    One other thing I don’t see mentioned (though I haven’t looked carefully): the possibility that the spike in obesity in the 80s is at least partially a result of the drop in smoking since the late 60s, given smoking’s well-known appetite-suppressing effects and what seems to be at least partial fungibility of addiction in general (people seem to be able to trade one addiction for another).

    • NostalgiaForInfinity says:

      This seems to me like part of a weakness in Taubes’ argument generally:

      We need a hypothesis that holds the promise of explaining the epidemics everywhere.

      Do we? My working assumption for the obesity crisis has been that it’s a whole bunch of things working in concert: increased alcohol consumption, changes in sleep patterns, central heating, more sedentary lifestyles, dietary changes, the decline in smoking…

      I don’t see why it can’t be a combination of factors.

      Even the fact that better healthcare is saving overweight heart attack patients could mean people who previously would have died are still here, and changing the % of the population that’s overweight.

  8. formid0 says:

    I would specifically note this part of the debate, from Guyenet[1]:

    “Here’s a third example. Taubes upbraids the research community for its belief that body fatness is determined by calorie intake, rather than the impact of foods on insulin. He supports the latter proposition with semi-anecdotal observations from Africa suggesting that a group of people eating a high-sugar diet supplying “as little as sixteen hundred calories per day” were sometimes obese and diabetic.

    “A person who actually wants to get to the bottom of this question should conduct their investigation in a very different manner. The first order of business is to look up the relevant metabolic ward studies, which are the most tightly controlled diet studies available. These studies consistently show that calorie content is the only known food property that has a meaningful impact on body fatness. This is true across a wide range of carbohydrate-to-fat ratios and sugar intakes, and a correspondingly wide range of insulin levels (17).

    “What makes Taubes’s oversight so extraordinary is that he was involved in funding one of these metabolic ward studies, which compared two diets that differed more than tenfold in sugar content. The results showed that a 25 percent sugar, high-carbohydrate diet caused slightly more body fat loss than a 2 percent sugar, very-low-carbohydrate (ketogenic) diet of equal calories (18). Despite these clear and consistent findings, Taubes continues to insist that calorie intake is not an important determinant of body fatness, and he offers the reader questionable evidence in support of this while omitting high-quality evidence to the contrary. All while exuding righteous indignation about the scientific community’s misguided beliefs.”

    While his latest book is against sugar specifically, I believe Taubes is against carbs in general. He subsists on a low carb diet, if I’m not mistaken.

    Kitavan’s live on potato starch and have no diabetes. If a Kitavan starts eating a Western diet, he develops metabolic syndrome. Taubes has specifically defended against this criticism before by claiming they are genetically adapted to potato starch. Then what about the Irish from a couple hundred years ago? Did they have time to adapt to starch? Medical records are poor, but historical accounts of the Irish don’t paint a picture of obesity and metabolic syndrome.

    Isn’t the far simpler explanation (while explaining all phenomena) for the problems of the Western diet that the food has grown in palatability and dropped tremendously in price over the last few decades? Everywhere there’s been a crisis of obesity, relative to the recent past for that particular culture, there’s been an improvement in palatability or drop in food prices.


  9. Michael B says:

    I have to admit, even though it has been years, I am still dumbfounded by the evident animosity between Guyenet and Taubes. Their dispute is more than just a professional disagreement, it is personal.

    It’s a real shame because parts of their respective arguments are more orthogonal than opposite.

    Guyenet has a real point about palatability, I believe, for simple reasons of wealth, technology, and economics. It is undeniable that food production technology (or food-like substance production technology) has changed greatly over the past century, as a result of societal wealth and of technological progress. And the economic component comes in on grocery store shelves – some of the most costly real estate in the world, and therefore a place ripe for evolution by selection. Products that sell are the ones that are able to maintain their place on store shelves. Other beneficial qualities include longer shelf-life and reduced wastage, but in the end sales is what keeps the food like substances in the grocery stores. With all of that in mind, all it really takes, in conjuction with decades of competition in supermarkets, is for it to be possible to create foods that people are more prone to overeat and those foods will start taking over our food supply through simple selection.

    But I think Taubes is on point in his response essay. Guyenet’s own response kins of straw-manned Taubes’ argument – as someone who was a teenager for most of the 80s I can well remember “low fat, high carb” being the conventional wisdom – to dismiss that because the 1980 guidelines suggested moderating intake of both sugar AND fat is to miss the point. Guyenet’s mortality-based argument is exceptionally weak because he did not address the decline in mortality related to improvements in medicine. Beyond that, if palatability is a key issue, sugar surely plays a role in that.

    To some extent they both seem like they are overly committed to one cause, sugar vs palatability, but Taubes seems at least to recognize the complexity.

  10. caryatis says:

    >And the economic component comes in on grocery store shelves – some of the most costly real estate in the world, and therefore a place ripe for evolution by selection…With all of that in mind, all it really takes, in conjuction with decades of competition in supermarkets, is for it to be possible to create foods that people are more prone to overeat and those foods will start taking over our food supply through simple selection.

    Is that Guyenet’s point or yours? Either way, I disagree. It’s a plausible explanation for something that simply doesn’t happen. Unhealthy, hyperpalatable foods taking over store space to the extent that healthy ones disappear is not something we observe today, and not a likely future outcome. We know most of us are obese, but even obese people still eat some amount of vegetables, lean meats, beans and so forth. And health nuts are a small but passionate and valuable market segment. Grocery stores seem to be following a strategy of offering a wide variety of foods, rather than JUST the most popular ones. Even a store like Aldi, with its limited selection and focus on prices, has plenty of basic, unprocessed foods.

    Now, if your argument was about restaurant food, you might have a better point.

    • Michael B says:

      “Is that Guyenet’s point or yours?”

      My point (though I’m sure not unique to me).

      “It’s a plausible explanation for something that simply doesn’t happen. Unhealthy, hyperpalatable foods taking over store space to the extent that healthy ones disappear is not something we observe today, and not a likely future outcome. ”

      I wasn’t talking about unhealthy replacing healthy, really. Instead of wondering whether the produce section shrinks to make room for more cookies, just think about the cookie aisle in isolation. It’s full of cookies of different types, different brands, etc. Over time, new types/brands appear and disappear, and sales are a key driver of that. If palatability is part of what drives sales, and food science can increase palatability, then over the long term palatability should rise as more palatable brands (or formulations) replace the less palatable.

      On the topic of palatable junk crowding out healthy food in stores, I agree there’s a imit to how far that could go and we are more or less there.

  11. Michael B says:

    Gary Taubes was just interviewed by Russ Roberts on a new episode of Roberts’ podcast, EconTalk:

    I haven’tlistened yet but this promises to be good.