Results from an observational study came out a couple weeks ago that followed almost 200,000 participants over more than 20 years in an attempt to correlate different diet and lifestyle factors with disease incidence. The most popular headline that came out of this study was that a gluten-free diet was correlated with a higher incidence of Type 2 diabetes. But before you freak out, despite the headline, the study did not show that gluten-free diets cause diabetes. The quality of this data, and even more so the quality of reporting that came out of it, are fraught with issues.
What was missing in the reporting of the results of this study? Let’s first look at what type of study this was. Researchers used three groups of long-term study data to evaluate correlations between health and lifestyle factors: the Nurses’ Health Study, the Nurses’ Health Study II and the Health Professionals Follow-up Study. This type of observational study is referred to as an epidemiological study. Data from epidemiology studies can be helpful at identifying correlations between different factors, but they have been largely ineffective at reliably identifying a cause-and-effect relationship.
For example, you can observe that Louisiana has the highest obesity rate in America. But would you ever say that living in Louisiana causes obesity? No, that would be ridiculous. An epidemiological study is the exact same. One can establish a correlation (a random person is Louisiana is more likely to be obese than a random person in another state), but not a cause (living in Louisiana causes obesity). In the case of this study, they established a correlation (a low-gluten diet is correlated with a higher rate of diabetes), but it would be equally ridiculous to use that information to claim a cause (a low-gluten diet causes diabetes).
Additionally, when datasets become this large, it becomes increasingly easy to draw inaccurate correlations on almost anything. In fact, there’s a whole website devoted solely to spurious correlations. Here’s an example:
As this paper published in the journal Cancer Epidemiology, Biomarkers, & Prevention aptly states, epidemiological studies like this one lead to false findings and results over 95% of the time. “False-positive findings dominate the epidemiological literature under the current standards of practice, with empirical evidence indicating that there are at least 20 false-positive results published for every true-positive finding. This is not surprising in view of the increasing number of epidemiology journals, the enormous and ballooning number of published epidemiology articles, and the pronounced incentives on the part of researchers and journals to publish positive results.” So, there’s less than a 5% chance that this correlation between low-gluten diets and diabetes is accurate. Yup, that’s right. 5%. So, that’s our first problem.
Another issue we have with this study is confounding variables. If someone was eating less gluten than the standard population, what were they eating instead? It’s noted that people who eat less gluten are also less likely to eat cereal fibers, which are known to help prevent Type 2 diabetes. What other variables could be confounding these results? We don’t know because the study didn’t mention any. Considering how many factors play into Type 2 diabetes (exercise, weight, tobacco use, alcohol use, and a whole host of dietary factors), it’s highly likely that there are many more variables that could play into a higher rate of Type 2 diabetes.
Completely separate from the methodology of the study is the issue of the quality of reporting on this study. As noted above, it is impossible to establish a cause-and-effect relationship from an epidemiological study. The press release makes this clear with the title they used (emphasis mine), “Low gluten diets may be associated with higher risk of Type 2 diabetes.” (Side note: that sounds like an extremely confident assertion. Yes.. that was sarcasm.) And yet, reputable news sources like Newsweek picked up the news and published a headline, “Gluten free diets actually increase risks of Type 2 diabetes.” This headline is downright false and highly misleading reporting. I’m calling out Newsweek here, but they’re not alone. Almost every article I’ve read on this study has some level of misinformation, some worse than others.
To top it off, this study did not actually look at people on a gluten-free diet at all. The study started in the 1970s when there was no such thing as a gluten-free diet. This study only looked at low or high gluten levels in a person’s diet. There is not one single data point about a person being on a gluten-free diet. Even if there were people on a gluten-free diet, it simply wasn’t tracked. The Newsweek headline about “gluten-free diets” is inaccurate on another level now too.
So, do gluten-free diets cause Type 2 diabetes? We don’t know, based on this study.
However, this 2013 study came out, clearly showing a cause-and-effect relationship between gluten consumption and incidence of Type 1 diabetes in mice. Yes, that’s right. Researchers have shown that eating gluten can actually contribute to Type 1 diabetes in mice. We can’t say that applies to humans (that would be just as irresponsible), but it’s food for thought for anyone rethinking their gluten-free diet.
-Scott Sundvor, co-founder and CPO, Nima