- Why Eating Low Carb Won’t Kill You
- 1. The Data Were Observational, and Significant Caveats Apply
- 2. The Study Data Came From Questionnaires, Not Observation of What Participants Ate
- 3. Confounding Factors Were Not Adequately Controlled For
- 4. Macronutrient Quality Is More Important Than Quantity
- 5. It’s Possible to Follow a Diet That Is Both Low in Carbohydrates and High in Fresh, Nutrient-Dense Foods
- 6. Humans Can Thrive on a Variety of Macronutrient Ratios—as Long as They’re Eating Whole Foods
- 7. Meta-Analyzing Data From Multiple and Heterogeneous Sources Opens the Door to Confirmation Bias
Last week, a new study was published in The Lancet Public Health that claimed to find that both very-low- and very-high-carb diets shorten our lifespan. Predictably, the mainstream media jumped on this finding without doing a shred of due diligence—more on that below—and we were subjected to splashy headlines like this:
- Low-carb diets could shorten life, study suggests (BBC News)
- Low and high carb diets increase risk of early death, study finds (CNN)
- Low-carb diet may cut years off life, study suggests (Newsweek)
- Your low-carb diet could be shortening your life (Fast Company)
- Paleo fail: meat-heavy low-carbohydrate diets can shorten lifespan, researchers say (South China News)
I’ve been writing about health and nutrition for more than a decade now, and without fail, at least once a year a study like this is published. I could set my watch to it.
Understandably, my Twitter, Facebook, and email accounts blow up with messages from concerned readers, who want to know if the diet they are following is going to kill them.
Each year, my response is the same: no, your nutrient-dense, whole-foods diet that includes animal products is not going to give you a heart attack, increase your risk of cancer or other chronic diseases, or shorten your lifespan. In fact, it’s likely to have the opposite effect.
Why Eating Low Carb Won’t Kill You
This year, it’s no different. In this article, I’m going to give you seven reasons why you should take the recent Lancet study with a huge grain of salt. If you’ve been following my work for some time, I hope you’ll recognize many of the shortcomings of the study, because you’ve seen them before:
- Using observational data to draw conclusions about causality
- Relying on inaccurate food frequency questionnaires (FFQs)
- Failing to adjust for confounding factors
- Focusing exclusively on diet quantity and ignoring quality
- Meta-analyzing data from multiple sources
Unfortunately, this study has already been widely misinterpreted by the mainstream media, and that will continue because:
- Most media outlets don’t have science journalists on staff anymore
- Even so-called “science journalists” today seem to lack basic scientific literacy
Most people—medical professionals and the general public alike—will just read the sensationalized headlines and assume that they are true. The percentage of the population that will find their way to a critique of the study like this, read it in its entirety, and comprehend it, is disappointingly low. This is what we’re up against. So, if you are reading this, please share it with anyone that you think would benefit.
Are you concerned about recent news regarding low-carb diets? Don’t be. Here’s why eating low carb isn’t likely to shorten your lifespan.
With that in mind, let’s take a closer look at the issues with this study.
1. The Data Were Observational, and Significant Caveats Apply
As I explained in a podcast called “A Beginner’s Guide to Scientific Research,” an observational study is one that draws inferences about the effect of an exposure or intervention on subjects where the researcher or investigator has no control over the subject. It’s not an experiment where they are directing a specific intervention (like a low-carb diet) and making things happen. Instead, the researchers are just looking at populations of people and making guesses about the effects of a diet or lifestyle variable.
An example of an observational study would be in comparing rates of lung cancer in smokers and nonsmokers. They might look retrospectively at groups of people who smoke and groups of people who don’t smoke, see what the rates of lung cancer are in each of those groups, and then draw some conclusions.
Repeat After Me: Correlation Is Not Causation
One of the key things to understand about observational studies is that you can’t establish causation from observational studies. You can establish a correlation or an association between two variables, but you can’t establish causation conclusively.
If you take a class on research methodology, you’ll often hear some silly examples of how observational data can be misinterpreted. Consider the statement, “The more firefighters that are sent to a fire, the more damage gets done.” Obviously that’s not how it works. It’s not that more firefighters are causing the damage. It’s that when fires are worse, more firefighters are required to fight it, so the causation there is reversed.
Another one would be, “Children who get tutored get worse grades than children who don’t get tutored.” Again, the causality is reversed. Children who are not getting good grades are more likely to hire tutors, or their parents will.
Consider a more relevant example. For decades, observational research suggested a correlation between dietary cholesterol intake and heart disease. This led to public health recommendations to limit cholesterol in the diet and generations of people unnecessarily torturing themselves with egg white omelettes (or even worse, Egg Beaters!), boneless, skinless chicken breasts, and (gasp!) margarine. Today, we now know that dietary cholesterol does not contribute to an increased risk of heart disease, and virtually all industrialized countries in the world—including the United States as of 2016—do not suggest limits to intake of cholesterol in their dietary guidelines.
To do that, you need a randomized, controlled trial (RCT). In an RCT, study participants are randomly assigned to two groups: a treatment group that receives the intervention being studied, and a control group that does not. The participants are then observed for a specific period of time.
This Lancet study was observational, not experimental. They simply observed participants over a 25-year period and assessed outcomes. As we’ll discuss below, this creates significant potential for error when attempting to draw conclusions.
2. The Study Data Came From Questionnaires, Not Observation of What Participants Ate
Do you accurately remember what you ate on March 15th, 2014? How about during the month of November 2015? I didn’t think so. Yet this is exactly the methodology in the studies analyzed in this report to determine participants’ carbohydrate intake.
More specifically, the underlying studies used FFQs. In an FFQ, researchers ask participants how much they ate of certain foods over a given time period. Not surprisingly, FFQs have been criticized for their inaccuracy for several reasons: (1, 2)
- People tend to underreport foods socially considered “bad,” like red meat and alcohol
- People overreport foods socially considered “good,” such as vegetables and fruits
- People may not know all the ingredients in restaurant or prepared foods
- People don’t weigh or otherwise measure portion sizes
- People find tracking every bite and meal inconvenient
- People are human and just can’t remember every little thing they eat
- People’s diets tend to change over long periods of time
Also, as you might suspect, the further back in the past participants are asked to recall their diet, the less accurate an FFQ will be. In the Lancet study, the subjects’ diets were only assessed twice throughout a 25-year period, separated by an interval of six years.
This means that people were asked to report on what they ate over a previous six-year period. And even then, the FFQs only covered 12 years of that 25-year period.
A perfect example of how inaccurate FFQs are can be found in the Supplementary Appendix of the study. As you can see from the screenshot below, those in the lowest quintile of carbohydrate intake (i.e. those that ate the fewest carbohydrates) had a self-reported calorie intake of around 1,500 calories/day. This is extremely unlikely, to put it mildly. Larger studies of self-reported calorie intake have found that males eat an average of 2,475 calories/day, and females eat an average of 1,833 calories per day. (3)
What’s more, research has shown that calorie intake is often significantly underreported—by as much as 25 percent in one study of women. (4) This suggests that the average female intake is likely greater than 2,000 calories/day and the average male intake is likely greater than 2,500/day. If we combine the male and female averages, we get an overall average of 2,250 calories/day—much higher than the 1,500 calorie/day figure reported by the participants in the study that were in the lowest quintile of carbohydrate intake.
While it is possible that people on a low-carb diet would temporarily reduce their calorie intake—which is one of the reasons it’s effective for weight loss—studies have shown that this effect does not persist over time. It is beyond implausible to assume that these so-called “low-carb” dieters were eating only 1,500 calories, and this throws the entire dataset into doubt.
While we’re looking at this screenshot, it’s worth pointing out that those that were on the “lowest-carb” diets were still eating 38 percent of calories as carbohydrate. Most nutrition experts, even critics of low-carb diets, would not call getting almost 40 percent of calories from carbohydrates “low-carb.”
3. Confounding Factors Were Not Adequately Controlled For
One of the biggest problems with observational studies is that it can be difficult, if not impossible, to isolate the influence of a single variable. Human beings don’t live in highly controlled environments, and there are numerous factors that impact our health and lifespan, ranging from genetics to air and water quality, from socioeconomic status to lifestyle and behavior.
This is why most nutritional studies are met with heavy criticism. A recent article from the Mayo Clinic Proceedings even claimed that because nutrition studies “cannot be reliably, accurately, and independently observed, quantified, and confirmed or refuted,” they do not follow the scientific method and should be regarded as “pseudoscience” at best. (5)
Let’s use a simple example. Imagine you’re a scientist and you want to find out whether eating red meat increases the risk of heart attack. You recruit participants and ask them to track how much red meat they consume over 20 years. Then, you measure how many heart attacks occurred throughout the study period.
When examining the data, you notice a strong correlation between red meat consumption and heart attack. In other words, the people who ate the most red meat were the most likely to have a heart attack, and the people that ate the least red meat were the least likely to have a heart attack.
Case closed, right? Not so fast. What if the people who ate the most red meat were also more likely to smoke cigarettes, have high blood pressure and diabetes, eat more refined carbohydrates and sugar, not eat vegetables, and not exercise? In this scenario, it’s impossible to know whether the higher rate of heart attacks was caused by eating more red meat, any of these other single factors, or a combination of some or all of them.
The Healthy User Bias
The scenario I just mentioned is not hypothetical—it’s incredibly common. It’s so common, in fact, that it even has a name: the “healthy user bias.” I discussed this in detail in a 2014 podcast called “Heart Attacks and Red Meat—Correlation or Causation?,” but here’s the short version. People who engage in a behavior perceived as healthy are more likely to engage in other behaviors that are also perceived as healthy, and vice versa.
So, because red meat has been perceived as “unhealthy” for so many years, on average, people that eat more red meat are more likely to:
- Drink too much
- Eat too much sugar
- Not exercise, etc.
Of course, most researchers are well aware of the influence of confounding factors and the healthy user bias, and the good ones do their best to control for as many of these factors as they can. But even in the best studies, researchers can’t control for all possible confounding factors, because our lives are simply too complex.
In the Lancet paper, researchers included a study if it controlled for at least three of the following factors:
- Smoking status
- History of cardiovascular disease
- Family history of cardiovascular disease
That’s a step in the right direction. However, it still leaves huge room for confounding factors and healthy user bias. For example, say one of the studies controlled for age, sex, and whether participants were obese. That still leaves many factors—smoking status, diabetes, hypertension, high cholesterol, history of cardiovascular disease—that could affect the outcome.
It opens up the possibility that people who were following a very-low-carb diet were more likely to have an underlying health condition like diabetes, hypertension, or high cholesterol, or that they were more likely to engage in unhealthy behaviors like smoking. And in fact, that’s exactly what happened in the Lancet study. According to the authors:
Participants who consumed a relatively low percentage of total energy from carbohydrates (i.e., participants in the lowest quantiles) were more likely to be young, male, a self-reported race other than black, college graduates, have high body mass index, exercise less during leisure time, have high household income, smoke cigarettes, and have diabetes. [emphasis added]
That’s not surprising, is it? People who follow diets—whether very-low-carb or very-high-carb—are far more likely to have some kind of health problem that led them to start the diet in the first place. Unfortunately, this study didn’t adequately control for this almost certain fact.
This is bad enough. But it gets worse when you consider the confounding variables that weren’t even on the researchers’ list, such as:
- The amount of fresh fruits and vegetables consumed
- The amount of sugar they consumed
- The quality of protein, fat, and carbohydrate they consumed
- How much physical activity they engaged in
The question of diet quality—whether the person was eating primarily fresh, whole, nutrient-dense food or highly processed, refined food—is especially important. In the United States, we know from other research that the majority of Americans eat mostly processed and refined food. For example, a study published this year found that 60 percent of the calories Americans consume come from not just processed food—but ultra-processed food. These foods do not impact the body in the same way that fresh, whole foods do. I’ll discuss this more below.
4. Macronutrient Quality Is More Important Than Quantity
Researchers have long debated whether low-fat or low-carbohydrate diets are best for weight loss and overall health. Regardless of the macronutrient content, however, most long-term studies have reported little success in achieving and maintaining significant weight loss. In 2016, I wrote an article called “Carbohydrates: Why Quality Trumps Quantity,” in which I argued that the answer to obesity and metabolic disease lies not in how much carbohydrate we eat, but rather what types of carbohydrate we eat.
Earlier this year, a landmark study published in JAMA supported this argument and suggested that the same principles apply to fats. The researchers found that on average, people who cut back on added sugar, refined grains, and processed food lost weight over 12 months—regardless of whether the diet was low carb or low fat.
I wrote about this study in detail in an article called “Why Quality Trumps Quantity When It Comes to Diet.” Here’s the TL;DR: when the subjects focused on real, whole foods and cut processed foods out of their diet, they lost significant weight, without having to count calories or restrict energy intake.
Now, this study focused on weight loss, but it’s ludicrous to assume that the same distinction between real, whole foods and processed, refined foods wouldn’t apply to a study looking at longevity.
Consider two hypothetical people:
- A person on a low-carb diet that eats primarily refined fats like industrialized seed oils (found in most processed foods and in foods cooked in restaurants)
- A person on a low-carb diet that eats primarily natural fats from fresh, whole foods (meat, fish, avocados, nuts, seeds, etc.) prepared mostly at home
Is it logical to predict that these two people will enjoy the same health, protection from disease, and lifespan? Of course not. Yet that is exactly what the Lancet study did assume.
Decades of nutrition research have myopically focused on the quantity of protein, fat, and carbohydrate we eat, without considering the quality. In my mind, this is perhaps the single biggest shortcoming of the bulk of nutrition research.
5. It’s Possible to Follow a Diet That Is Both Low in Carbohydrates and High in Fresh, Nutrient-Dense Foods
It should be clear by now that the participants that were following a low-carb diet were not following a Paleo-type low-carb diet that is rich in natural, whole foods. The researchers themselves point this out:
By contrast, the animal-based low carbohydrate dietary score was associated with lower average intake of both fruit and vegetables (appendix pp 9, 10).
But of course it doesn’t have to be that way. A common misconception of the Paleo diet is that it’s “meat heavy,” rather than “plant based.” But consider someone who is abstaining from eating grains, dairy products, and processed and refined foods.
What might their plate consist of? A serving of protein (fish, poultry, meat), and typically two to three servings of non-starchy vegetables. Depending on their carbohydrate intake, they may also eat whole fruits (especially those lower in sugar, like fresh berries) and even starchy tubers like sweet potatoes and yams that are relatively low in carbohydrates. These foods will often be supplemented with healthy fats like nuts, seeds, avocados, or olives.
This is NOT the diet that was studied in the Lancet paper. Therefore, if this is the diet that you’re eating, the results in that paper do not apply to you.
6. Humans Can Thrive on a Variety of Macronutrient Ratios—as Long as They’re Eating Whole Foods
The Lancet study suggested that the optimal range of carbohydrate intake for a lengthy lifespan is between 50 and 55 percent of calories. Is it plausible to assume that humans can only live for a long time within such a narrow range of carbohydrate consumption? No.
That would have put us at a significant evolutionary disadvantage. Humans evolved in diverse environments around the world, and studies of contemporary hunter–gatherer populations demonstrate that we can thrive on a broad range of macronutrient ratios as long as we are following a traditional, whole-foods diet.
Carbohydrate Intake Varies in Ancestral Diets
For example, the Kitavan Islanders of Melanesia live as horticulturists, with little access to Western foods. Carbohydrates make up 60 to 70 percent of their energy intake (higher than the recommended 50–55 percent range in the Lancet study), much of that coming from fruit or tubers with a fairly high glycemic index. (6) Their saturated fat intake is also high.
Yet despite obvious similarity between Kitavan and Western diets in both macronutrient composition and glycemic index, Kitavans boast levels of fasting insulin and blood glucose that are even lower than the levels deemed healthy in Western populations. (7, 8) They also have lower levels of leptin and a virtual absence of diabetes, atherosclerosis, and excess weight. (9, 10, 11)
On the other end of the spectrum, analyses of hunter–gatherer populations, including the Masai, Kavirondo, and Turkhana, suggest that a low-carb diet (between 22 and 40 percent of calories, again lower than the 50 to 55 percent range in the Lancet study) with high intake of unprocessed meat and saturated fat does not result in poor cardiovascular or metabolic health. (12) (For more on this, see my special report on the truth about red meat.)
Critics of the Paleo diet and ancestral nutrition claim that there’s no point in studying what hunter–gatherers eat, because they all die when they’re 40 years old. This is incorrect, as I explain in this video.
While it is true that, on average, hunter–gatherers have shorter lifespans than people living in the modern, industrialized world, those averages don’t consider important challenges that are largely absent from modern life: high rates of infant and early childhood mortality (30 to 100 times higher) and deaths to trauma, warfare, and exposure to the elements, most of which are caused by a complete lack of emergency medical care.
Yet anthropological studies of modern hunter–gatherers have shown that when they have access to even the most rudimentary form of medical care (think a half-day’s walk to a rural clinic), they live life spans roughly equivalent to our own. (13, 14) But in contrast to us, they reach these ages without acquiring many of the chronic, inflammatory diseases that characterize our old age—like diabetes, cardiovascular disease, and Alzheimer’s.
Consider two articles recently published in The New York Times examining the absence of chronic disease in the Tsimané, a subsistence farming and hunter–gatherer population in Bolivia.
The first article, Learning from Our Parents’ Heart Health Mistakes, reported on a study showing that the Tsimané have a prevalence of atherosclerosis 80 percent lower than ours in the United States and that nine in 10 Tsimané adults aged 40 to 94 had completely clean arteries and no risk of heart disease. (Note that the study included adults between 40 and 94 years of age; clearly they are not all dying when they’re 40!)
In a follow-up article, researchers even put to rest the old canard that hunter–gatherers don’t have “diseases of civilization” like diabetes and cardiovascular disease because they don’t live long enough to develop them:
The Tsimané suffer from high infant-mortality rates, but those who reach adulthood live about as long as most other people, making it possible to measure their health outcomes up to age 90 and beyond.
This in spite of the fact that the Tsimané have high rates of infection with parasites, and consume 72 percent of calories from carbohydrates—far higher than the 50 to 55 percent range suggested in the Lancet paper.
7. Meta-Analyzing Data From Multiple and Heterogeneous Sources Opens the Door to Confirmation Bias
A meta-analysis is the statistical procedure for combining data from multiple studies. They play an important role in research, but they’re also plagued with several disadvantages, which Wikipedia does a good job of summarizing. They include publication bias, statistical challenges, and, most relevant to this discussion, an “agenda-driven bias”:
The most severe fault in meta-analysis often occurs when the person or persons doing the meta-analysis have an economic, social, or political agenda such as the passage or defeat of legislation. People with these types of agendas may be more likely to abuse meta-analysis due to personal bias. For example, researchers favorable to the author’s agenda are likely to have their studies cherry-picked while those not favorable will be ignored or labeled as “not credible.” In addition, the favored authors may themselves be biased or paid to produce results that support their overall political, social, or economic goals in ways such as selecting small favorable data sets and not incorporating larger unfavorable data sets. The influence of such biases on the results of a meta-analysis is possible because the methodology of meta-analysis is highly malleable.
Another term for agenda-driven bias is “confirmation bias.” This is defined by Wikipedia as “the tendency to interpret new evidence as confirmation of one’s existing beliefs or theories.”
Was this an issue in the Lancet paper? While we can’t be sure, it’s certainly a possibility. The paper was published by a research group that included Walter Willett, a physician and researcher at the Harvard School of Public Health who is notorious for his advocacy of a low-fat, plant-based diet. This alone is not necessarily cause to suspect confirmation bias.
However, in an unprecedented turn of events, Willett was censured in an editorial and feature article in the prestigious journal Nature for “promoting over-simplification of scientific results in the name of public health and engaging in unseemly behavior towards those who venture conclusions that differ to his.” (15)
Willett co-authored a study claiming to link aspartame with cancer, but the study was retracted by Harvard at the last minute because the data did not support that conclusion. Meanwhile, the damage had already been done by sensational media headlines like “Aspartame Causes Cancer.” Sound familiar?
In an interview with NBC News about this incident, Dr. Steven Nissen, chair of Cleveland Clinic’s Cardiovascular Medicine Department, said:
Promoting a study that its own authors agree is not definite, not conclusive and not useful for the public is not in the best interests of public health.
What’s more, it later became clear that this study had been rejected by six journals, before finally being published in the American Journal of Clinical Nutrition, where—surprise, surprise—Willett is a member of the editorial board.
Unfortunately, this is the reality of medical research today. I’ve written extensively about how financial conflicts of interest and fraud impact scientific findings (see “Behind the Veil: Conflicts of Interest and Fraud in Medical Research”).
But don’t take it from me. In a 2009 article called “Drug Companies & Doctors: A Story of Corruption,” a physician, the former editor of The New England Journal of Medicine, said:
It is simply no longer possible to believe much of the clinical research that is published, or to rely on the judgment of trusted physicians or authoritative medical guidelines. I take no pleasure in this conclusion, which I reached slowly and reluctantly over my two decades as an editor of The New England Journal of Medicine.
Consider, also, a paper called “Why Most Published Research Findings Are False,” by John Ioannidis, a Professor of Medicine and of Health Research and Policy at Stanford University School of Medicine and a Professor of Statistics at Stanford University School of Humanities and Sciences.
Ioannidis explains that in many research papers, “Claimed findings may be accurate measures of the prevailing bias.” Clearly, he struck a nerve; this paper is now the most widely cited paper ever published in the journal PLOS Medicine.
In other words, most published research findings support the status quo; they’re not necessarily based on solid evidence. Often, the research that builds on an initial study ends up perpetuating questionable findings. It’s like building a house of cards: a paper gets published that references another paper; then, a third paper gets published that references that second paper, which referenced that first paper, and so on. The assumption is that the evidence in that first paper was correct—but what if it’s not? The edifice of peer-reviewed research is not as perfect as we tend to believe.
If you’re still with us, congratulations! You now have a clearer grasp of the problems with most nutrition studies than the vast majority of journalists working today. My hope is that, armed with this knowledge, you can protect yourself from sensationalized headlines that are based on agenda-driven, poorly designed studies—and continue to follow whatever version of a nutrient-dense, whole-foods diet works best for you.