Latest Low-Carb Study: All Politics, No Science
The only evidence to be found in this research is evidence of bias.
Posted Sep 05, 2018
Recently, the journal Lancet Public Health published a study conducted by researchers at Harvard University and the University of Minnesota warning people that low-carbohydrate diets can cause early death.
The paper, entitled “Dietary carbohydrate intake and mortality: a prospective cohort study and meta-analysis” has enjoyed broad media coverage and ignited passionate debate in nutrition circles around the world.
Why the uproar?
Researchers claim that:
- They tracked the diet and health of more than 15,000 people for up to 30 years.
- People eating diets lower in carbohydrate died sooner than people eating a “moderate” amount of carbohydrate.
- Individuals in the lower carbohydrate group seemed to live a little longer if they ate less animal protein.
Their observations led them to conclude:
“…animal-based low carbohydrate diets, which are more prevalent in North American and European populations, should be discouraged.”
Given the rising popularity of low-carbohydrate diets, this is a bold recommendation with potentially far-reaching implications for public health. Many people will take this study at face value because it is a very large, decades-long, Harvard-affiliated study that passed scientific scrutiny by peer reviewers and was found worthy of publication.
So, why bother to go beyond the headlines to question its validity? Because it seems curious that low-carbohydrate diets, which make so many people healthier, should somehow simultaneously hasten their demise.
Low-Carbohydrate Diets in the Real World
A growing number of clinicians are successfully prescribing low-carbohydrate diets in their practices to address obesity, type two diabetes, and other serious metabolic disorders. An expanding body of scientific literature supports the safety and effectiveness of low-carbohydrate diets, finding that they are at least as good, if not better, than other diets for weight loss. Remarkably, low-carbohydrate diets have the power to put type two diabetes into remission and reverse signs of “metabolic syndrome” such as high insulin, high blood sugar, high blood pressure, high triglycerides, low HDL, and inflammation. Stricter versions of low-carb diets called “ketogenic” diets have been used for nearly a century to treat epilepsy. Emerging science is exploring the potential for ketogenic diets to help manage other neurological disorders, psychiatric disorders, dementia, and even cancer. People around the world are discovering the health benefits of low-carbohydrate diets for themselves and sharing their progress with their friends, family, and on social media. [Full disclosure: I have eaten a low-carbohydrate diet for the better part of the past decade and recommend it as an option to many of my patients.]
If you have improved your own health, lost weight, or been able to cut back on medications by eating a low-carbohydrate diet, should you worry that you are sacrificing years of life for the benefits you have seen?
Of course not.
Let’s take a closer look at the study so you can see for yourself why there is absolutely, positively nothing to be afraid of . . . except bad science.
Where's the Evidence?
Ludicrous Methods. The most important thing to understand is that this study was an "epidemiological" study, which should not be confused with a scientific experiment. This type of study does not test diets on people; instead, it generates guesses (hypotheses) about nutrition based on surveys called Food Frequency Questionnaires (FFQs). Below is an excerpt from the FFQ that was modified for use in this study. How well do you think you could answer questions like these?
How is anyone supposed to recall what was eaten as many as 12 months prior? Most people can’t remember what they ate three days ago. Note that “I don’t know” or “I can’t remember” or “I gave up dairy in August” are not options; you are forced to enter a specific value. Some questions even require that you do math to convert the number of servings of fruit you consumed seasonally into an annual average—absurd. These inaccurate guesses become the “data” that form the foundation of the entire study. Foods are not weighed, measured, or recorded in any way.
The entire FFQ used contained only 66 questions, yet the typical modern diet contains thousands of individual ingredients. It would be nearly impossible to design a questionnaire capable of capturing that kind of complexity, and even more difficult to mathematically analyze the risks and benefits of each ingredient in any meaningful way. This methodology has been deemed fatally flawed by a number of respected scientists, including Stanford Professor John Ioannidis in this 2018 critique published by JAMA.
Missing Data. Between 1987 and 2017, researchers met with subjects enrolled in the study a total of six times, yet the FFQ was administered only twice: at the first visit in the late 1980s and at the third visit in the mid-1990s. Yes, you read that correctly. Did the researchers assume that everyone in the study continued eating exactly the same way from the mid-1990s to 2017? Popular new products and trends surely affected how some of them ate (Splenda, kale chips, or cupcakes, anyone?) and drank (think Frappucinos, juice boxes, and smoothies). Why was no effort made to evaluate intake during the final 20-plus years of the study? Even if the FFQ method were a reliable means of gathering data, the suggestion that what individuals reported eating in the mid-1990s would be directly responsible for their deaths more than two decades later is hard to swallow.
There are other serious flaws to cover below, but the two already listed above are reasons enough to discredit this study. People can debate how to interpret the data until the low-carb cows come home, but I would argue that there is no real data in this study to begin with. The two sets of "data" are literally guesses about certain aspects of people's diets gathered on only two occasions. Do these researchers expect us to believe they accurately represent participants’ eating patterns over the course of 30 years? This is such a preposterous proposition that one could argue not only that the data are inaccurate, but that they are likely wildly so.
Low-carb diets were not studied. Yes, you read that correctly, too. The lowest-carbohydrate group in the study reported consuming 37% of their approximately 1,558 calories per day as carbohydrate. This 37% translates to a whopping 144 grams of carbohydrate per day. Nowhere else would this be considered a low-carbohydrate diet. Most low-carbohydrate practitioners recommend between 20 and 50 grams of carbohydrate per day. Truly low-carbohydrate diets were not studied. Instead, researchers simply assumed that diets containing even less than 37% carbohydrate would lead to even shorter lives. On the surface of it, this may sound like it makes sense: if lowish is bad, then shouldn't even lower be worse?
The problem with this reasoning is that low-carbohydrate diets tend to have a "threshold effect" on metabolism. This means that most people must drop their carbohydrate intake below a particular “sweet spot” in order to reap benefits. For many, lowering from 150 grams per day to 75 may not make much difference, but dropping below 25 grams per day can bring significant improvements in appetite, weight, blood sugar and insulin levels. Therefore, even if eating 144 grams of carbohydrate per day were dangerous (which this study does not demonstrate), eating 20 grams isn’t necessarily worse, and may for some be better.
No Substitutions. The authors imply that people who eat low-carbohydrate diets can delay meeting their Maker by replacing animal protein with plant protein:
“…mortality increased when carbohydrates were exchanged for animal-derived fat or protein and mortality decreased when the substitutions were plant-based.”
This is rather misleading, as nobody substituted anything for anything else in this study—this was not an experiment. These substitutions took place only in the researchers’ minds.
No Explanations. By saying that low-carbohydrate diets are "associated with greater mortality risk" and that they should therefore be "discouraged," the authors are insinuating that low-carbohydrate diets kill people. If they truly believe this, they have a responsibility to themselves as scientists, to the peer reviewers who accepted their paper for publication, and to the public to explain how carbohydrate restriction endangers lives—particularly given that there is now so much clinical trial evidence demonstrating that carbohydrate restriction can improve the signs and symptoms of some of the deadliest chronic diseases we face.
The only clear attempt made by the authors to offer a plausible mechanism by which low-carbohydrate diets snuff people out is represented by this unreferenced hypothetical statement:
“Long-term effects of a low carbohydrate diet with typically low plant and increased animal protein and fat consumption have been hypothesised to stimulate inflammatory pathways, biological ageing, and oxidative stress.”
By contrast, it is well-documented and widely agreed that sugar (a carbohydrate) is a powerful promoter of inflammation and oxidation [Brownlee M. Nature 2001, 414: 813-820; Ghasemi R et al. Mol Neurobiol 2013, 47: 1045–1065; Kiecolt-Glaser JK. Psychosomatic Medicine 2010, 72: 365-369]. Curiously, the very real dangers of sugar aren’t acknowledged anywhere in this paper. That the authors ignore the scientifically undisputed virtues of sugar elimination while focusing on the purely hypothetical hazards of meat and fat strongly suggest potential bias . . . or at least a significant blind spot.
The Great Debate: Low-Carb vs. Low-Fat
Wherever there are people, there are politics, and the world of nutrition science is no exception. There is a paradigm shift occurring in nutrition, and a changing of the guard may be on the horizon. For decades, nutritional epidemiologists at prestigious institutions like the Harvard School of Public Health have occupied virtually all of the seats at the nutrition power brokers' table, and most of them have used their influence to promote low-fat, low-cholesterol, high-plant diets. Their recommendations have been enshrined into the dietary guidelines of the U.S. and many other countries as gospel, even though their arguments have rested almost entirely on "data" generated from Food Frequency Questionnaires such as the one highlighted above, originally developed by renowned Harvard nutrition researcher Dr. Walter Willett, one of the authors of this paper.
The low-fat philosophy upon which these giants have built their reputations has been called into question in recent years by low-carbohydrate clinicians, researchers and community members who witness and document examples every day of low-carbohydrate, high-fat diets improving health. Similar challenges to low-fat, plant-based principles have come from the paleo and carnivore communities which renounce "heart-healthy" whole grains and legumes and embrace animal foods with their naturally-occurring fats as the foundations of a healthy diet.
This summer, the BMJ medical journal together with global re-insurance company Swiss Re hosted a groundbreaking summit in Switzerland entitled Food for Thought: The Science and Politics of Nutrition, intended to foster open dialogue between prominent figures within the low-fat, plant-based community and prominent figures within the low-carb, healthy fat, meat-positive community. One of the central questions posed by the organizers was this:
"What evidence can we trust?" [video of panel discussion]
It was clear that our hosts, Dr. Fiona Godlee of the BMJ and Dr. John Schoonbee of Swiss Re—highly-respected authorities from science and industry—were taking epidemiology to task. What I was privileged to have witnessed over the course of those two days was nothing short of a watershed moment in nutrition history. We critics of epidemiology-based dietary guidelines—the illogical, unfounded, hopelessly complex guidelines that have been destroying our health for decades—were finally being given seats at the big kids' table.
Perhaps the authors of this new paper (including Dr. Walter Willett, who was a panelist at the Swiss conference), were hoping headlines about their research would be passively accepted by the public without scrutiny and delay the washing away of their whole-grain sand castle.
Escaping Epidemi-illogical Escapades
The field of nutritional epidemiology has a dismal track record when it comes to the validity of its guesses—more than 80% of its hypotheses are later proved wrong in clinical trials (human experiments). This is why nutrition headlines are so confusing—one day eggs are bad for us (epidemiology), the next day they are perfectly fine (clinical trials). In my opinion, this study’s thinly-veiled attempt to deter people from embarking on or continuing a healthy low-carbohydrate diet by using smoke-and-mirrors methodologies and invoking images of the Grim Reaper takes this study out of the realm of science and places it squarely in the arena of politics.
These researchers did not conduct a study of low-carbohydrate diets. They dug up some old "data" from epidemiological studies of heart disease, desperately picked through the rubble looking for anything they could find that might support their dying hypothesis, and then repackaged their observations for public consumption.
Don’t be afraid of this paper tiger. There is no evidence of any kind in this study that low-carbohydrate diets—whether they include meat or not—will have you pushing up daisies any sooner than anyone else.