Are Low-Carb Diets Dangerous?

A link from dearieme led me to a recent study that found low-carb high-protein diets — presumably used to lose weight — associated with heart disease. The heart disease increase was substantial — as much as 60% in those with the most extreme diets. (A critic of the study, Dr. Yoni Freedhoff, called the increase in risk “incredibly small“.) Four other studies of the same question have produced results consistent with this association. No study — at least, no study mentioned in the report — has produced results in the opposite direction (low-carb high-protein diets associated with a decrease in heart disease).

I find this interesting for several reasons.

1. I learned about the study from The psychological costs of obesity are huge. The popularity of low-carb diets probably has a lot — or everything — to do with the failure of researchers to find something better. I have never seen people who criticize low-carb diets appear aware of this. I disagree with a lot of Good Calories Bad Calories but I completely agree with its criticism of researchers.

2. There has never been a good explanation of the success of low-carb high-protein diets (why they cause weight loss), although this has been well-known for more than a century. (A good explanation would be a theory that made predictions that turned out to be true.) Such diets require a big change in what you eat. A big change is likely to have big health consequences in addition to the weight loss, and those side effects could be either good or bad. It now appears bad is more likely.  With a good theory of weight control, you should be able to find a much smaller change that produces the same amount of weight loss as a low-carb high-protein diet. Because the change is much smaller, it should have much smaller side effects. Much smaller side effects (unknown whether they are good or bad) are much less likely, if bad, to outweigh the benefits of the weight loss. I have never come across a low-carb advocate who seemed to understand this (that we don’t know why they work and it would be a very good idea to find out).

3. The Japanese are remarkably healthy (live very long), slim, and have very little heart disease, yet eat lots of rice. Which makes absurd the notion that all high-carb diets are unhealthy or fattening.

4. The comments on the low-carb study are mostly critical and the criticisms are terrible. For example, Dr. Yoni Friedhoff, who blogs about weight control, says, “The paper’s basing all of its 15 years worth of conclusions off of a single, solitary, and clearly inaccurate, baseline food frequency questionnaire”. The authors of the study correctly reply that inaccuracy would reduce the associations.

5. Until nutrition scientists do better research, our best source of nutritional guidance may be what we like to eat. Evolution shaped us to like foods that are good for us, at least under ancient conditions. We like carbs and we don’t like foods high in protein (lean meat is barely edible) so a low-carb high-protein diet is on its face a bad idea.  This is why I find it plausible that the low-carb high-protein association with heart disease reflects cause and effect (low-carb high-protein causes heart disease) and that in particular a high-protein diet causes heart disease. (Too little of the right fats?) We very much like fat. Under ancient conditions, the fat people ate was mostly animal fat and, before that, if you believe in aquatic apes, fish oil. It is quite plausible that lactose tolerance spread so quickly throughout the world because at the time everybody was starved of animal fat — high-fat mammals had been hunted to extinction — and dairy products were a good source of it.



15 Replies to “Are Low-Carb Diets Dangerous?”

  1. I am confused by this post. The vast majority of low-carb diet suggestions I see are not low-carb high-protein, but instead low-carb, moderate-protein, high-fat. In particular, the Paleo community would never [afaict] suggest just replacing all your carbs with lean protein.

    Seth: The article discussed low-carb high-protein diets. Their selection, not mine. Yeah, maybe the title of my post is confusing.

  2. From the study: “The questionnaire used in the study was self administered and recorded information on several lifestyle variables (including detailed smoking and alcoholic drinking habits), anthropometry, and history of diagnoses of major diseases and conditions, including medical diagnosis of hypertension. For the assessment of physical activity, women rated their overall level of activity (that is, activities in the house and occupational and recreational physical activity) on a five point scale with examples attached to levels 1 (low), 3, and 5 (high). Dietary intakes were assessed with a validated food frequency questionnaire…”

    Self-reported food studies are unreliable enough that they cannot be used to draw conclusions, even from an epidemiological perspective. The margin of error is often very wide, and there’s widespread recognition of this fact in the epidemiological literature.

    “The authors of the study correctly reply that inaccuracy would reduce the associations.”

    How can they know this? If all they have is inaccurate data, they can’t state that the better data would swing the results one way or the other. If they have better data, why didn’t they use it?

    “Limitations in the assessment of dietary energy intake by self-report”

    “…Comparisons of measured energy expenditure with energy intake from either weighed or estimated dietary records against energy expenditure have indicated that obese subjects, female endurance athletes, and adolescents underestimate habitual and actual energy intake. Individual underestimates of 50% are not uncommon….”

    “The most popular among these diets emphasise reduction of carbohydrate intake, thereby encouraging high protein intake,7 as high fat diets are generally avoided in most Western societies.”

    Atkins is a low-carb, high-fat, moderate-protein diet. Not high-protein. A diet that is too high in protein induces a state called “rabbit starvation”, which will kill. Rabbits are high-protein, low-fat. Adding fat to the diet cures the condition. It’s possible that lower, but still high levels of protein aren’t great for you, but you won’t find out via a study of this nature, as the error rate would be too high.

    “The Japanese are remarkably healthy (live very long), slim, and have very little heart disease, yet eat lots of rice. Which makes absurd the notion that all high-carb diets are unhealthy or fattening.”

    Too much rice and too little of everything else is certainly harmful: it leads to the condition known as “beri-beri” which killed many Japanese sailors before their navy figured out that it was caused by a high-rice diet. Moderate amounts of rice in the diet seems perfectly consistent with good health, as the Japanese demonstrate.

    “Until nutrition scientists do better research, our best source of nutritional guidance may be what we like to eat.”

    Agreed, within reason. “Hyper-palatable” foods that, for whatever reason, cause us to over-eat them are the primary cause of the obesity epidemic. Even the Japanese, as they adopt more Western-style foods, are seeing their obesity rates increase.

    Interesting point about dairy and the gene for lactose tolerance.

    Lactose tolerance isn’t seem necessary for consumption of dairy, however. Most Masaai, for instance, are not lactose “tolerant”, as they cannot break down lactose into glucose; yet they do just fine on a high-dairy diet.

    Great post! 🙂

  3. The authors of the study correctly reply that inaccuracy would reduce the associations.

    Random inaccuracy would dilute statistical significance, but there is no reason to think that inaccuracy in food questionnaires is random. Selection bias very likely plays a role — more responsible people who think more about their health are probably more likely to report having eaten certain foods (whatever is thought to be “healthy”), thus inflating any supposed relationship between those foods and better health outcomes.

    Seth: Your argument is reasonable and makes additional assumptions that may or may not be supported by evidence (I don’t know). It is not what the critic said.

  4. I see the anti-low-carb folks fighting an old battle – low carb – high protein (the old Adkins diet) vs the USDA high card food pyramid, when the low carb advocates are rapidly moving to a low carb – high fat paradigm (LCHF or HFLC).

    And, what is the per capita consumption of carbohydrates of the Japanese vs the USA? What about sweeteners? What about total calories? I would bet they’re all much lower. And what is the metabolic difference between short grain Japanese rice and American dwarf wheat, HFCS, and sucrose. Saying the Japanese eat lots of rice is misleading because it doesn’t say anything about what they don’t eat, and what they eat with it.


  5. While we can question whether “low carb” diets are dangerous, I think the better question, here, would have been “are high protein diets dangerous” – which the evidence suggests that they are.

    And even then, some high protein diets -.e.g. lots of processed soy – are likely to be more dangerous still.

    I think, as do many others, it is time to move past the discussion of macronutrient ratios as if they alone were the determinant of healthy diets.

    A malnourishing diet (i.e. deficient in essential micronutients) is bad for health, as is a diet full of toxins/antinutrients or foods that are not tolerated (e.g. wheat, soy, MSG, many processed foods)

    The evidence of successful hunter gather societies around the world shows that different ones had good health with widely varying macronutrient ratios – but they all had adequate nutrition.

    If we take the Paul Jaminet view of white rice – a toxin free, but micronutrient poor, starch, then eating lots of it is fine, as long as you are still getting adequate nutrition from other foods. Japan today seems to be following many of the other traditional societies, where rice and other traditional foods being replaced by, in the words of Weston Price “the displacing foods of modern commerce” – with predictable results.

    The continued obsession of the media, and many food/medical researchers with simple calories and macronutrients ignores the greater problem of malnutrition and toxic foods.

  6. I’m not going to repeat the arguments above, I mainly agree with them on the High Protein part. I personally am on the High Fat, low carb, moderate protein diet and am, by all measurements, healthier than I have ever been. More importantly, its not something I have to work at or worry about anymore.

    I don’t think that it should surprise anyone that different people from different parts of the world could have different dietary requirements. 23andMe puts my ancestory into Eastern and Northern Europe. For most of evolution, my ancestors probably didn’t have access to plentiful carbohydrate sources for large parts of the year.

    I agree with the statement that the body wants things that are good for it. Evolution never prepared us for buying sugar by the pound. Those signals could very easily lead you astray in our modern world. On that same note, you might desire carbs a lot because you need some carbs or some carbs re beneficial and they were harder to obtain. Your desire made you take the extra steps needed to get those nutritious carbohydrates that you needed.

  7. Be careful generalizing about who likes what. I love high protein foods like chicken & beef jerky. I did a high fat, low carb diet for a while & had to force myself to eat enough fat — after a while it lost the appeal.

    It is also probably important to look at the characteristics of the people who are eating a lot of protein. If they are average & immobile, maybe it has one effect. If they are a professional bodybuilder, maybe it has another.

    Personally, I eat a lot of protein, but only in the 48 hour window after a heavy weight lifting session. Same for carbohydrates. The rest of the time I am either fasting or eating some fat.

  8. the study involved Swedish women. I understand from a Swedish woman i knew, that the Swedish diet lacks vegetables, especially in the winter (or perhaps only in the winter, i can’t recall); so all they eat (at least for signficant part of year) is protein. if one doesn’t eat vegetables then, one will be subject to all sorts of diseases, including heart disease. Paul N alludes to this when he says “A malnourishing diet (i.e. deficient in essential micronutients) is bad for health, as is a diet”
    My understanding is that many micronutrients are provided by vegetable.

    also, Seth, i wonder if you eat a low carb diet, and i wonder what part of Taub’s book relating to carbs you find incorrect.

    Seth: I eat a fairly low-carb diet, yes, to keep my blood sugar low. The part of Taubes’s book I disagree with is the part where he explains why low-carb diets cause weight loss. The theory, in other words.

  9. Seth, I’m surprised you say “There has never been a good explanation of the success of low-carb high-protein diets.” Is there a flaw in the theory that less carbohydrate means less insulin, and lower serum insulin permits lipolysis to occur?

    (Other than the fact that protein also raises insulin, so low-carb moderate-protein high-fat diets are even more successful.)

    Seth: “Flaw in the theory”? Yes. It fails to explain dozens of facts. Here’s one: Increasing the water content of rats’ food makes them fatter. Another flaw is that it has failed to make correct predictions.

  10. I’m going to have to contest you on (4). I see what you’re saying, but the problem with basing something off a single observational study is this:

    If the study is randomly chosen, then you’re right that there’s a higher chance of error. On the other hand, because of the randomness of that kind of evidence, it’s also easier to find spurious correlations. Of course, I imagine that the doctor you mention could have made the point I made by finding observational data that contradicts the study in question.

    Seth: Adding noise does not make it easier to find spurious correlations. You can only find a spurious correlation if there is no actual correlation — that is, if what you have is pure noise. Adding noise to noise just gives you more variable noise. Tests for association adjust for amount of variation. I agree when you say that a good critic would have simply pointed to data that contradicts the study in question. None of the critics did, if I remember correctly.

  11. Seth, I agree with others that the Swedish study is so weak as to be almost uninteresting.

    Their statistical model adjusts for fat consumption, so it looks like they are only testing the effects of substituting carbs for protein. As demonstrated by the first line of the abstract, these researchers don’t understand that a low carbohydrate diet is high in fat (if they don’t know what it is, why are they studying it?) If there is a benefit to high fat consumption, we can’t tell from this research.

    There are a number of interesting biases reported in the way people respond to food frequency questionnaires. For example, those with medical conditions will report eating 1.5x as much meat as those without ( I didn’t see any evidence they adjusted for this sort of thing, despite claiming their questionnaires were “validated.”

    I eat a low carb diet, though I freely admit that I don’t have any idea why it works. The commonly cited theories are clearly wrong as you point out, and I wish more people were honestly trying to figure it out.

    Seth: They believed, like many people, that fat consumption affects weight. So they adjusted for it. I fail to see the problem. As for the bias problem, inaccuracy will reduce associations, not increase them. Sure, someone else with different assumptions would do a different analysis. Until we know for sure their assumptions are wrong — and we are far from that point — their work is worth paying attention to. If they got the wrong answer, the lack of contradictory evidence (studies that got the opposite answer) is striking and not explained by any of their critics.

  12. I think that, by controlling for fat, the paper did not test the hypothesis that replacing carbs with fat has an effect on cardiovascular events. They might have tested the hypothesis that replacing carbs with protein increases cardiovascular events, but I think that result is more easily explained by the fact that unhealthy people systematically overstate their meat consumption. Since the bias is directional, couldn’t it create, and not just mask, an association?

    I haven’t done an exhaustive search, but I think the Nurse’s Health Study does contradict the finding, and with a larger sample size ((, see Table 3).

    Seth: Yes, inaccuracy in the sense of bias can create a false association. I should have been clearer about that. I agree, the meat/poor health bias really does make plausible an alternative explanation that the authors should consider. I am less convinced than you. Is the bias large enough to explain the results? Maybe it is too small. Maybe there are offsetting biases. The study you point to did not specifically study the effect of protein replacing carbs, it studied the effect of carbs replacing everything. I used “low-carb” to mean “low-carb high-protein”. Again, I should have been clearer.

  13. The way I see it and according to what I’ve read from it is important to take track of a persons nutritional needs especially if you are in a low-carb diet. Since your are reducing those foods that you eat, the nutrition absorption is reduced and in this case we are prone to many deficiencies that may weaken our immune defense.

Comments are closed.