Red Meat & Type 2 Diabetes

Flawed research stirs up old fears

Recently, The American Journal of Clinical Nutrition (AJCN) made available online, an article titled “Red meat intake and risk of type 2 diabetes in a prospective cohort study of United States females and males.”

Media outlets went crazy over the study’s conclusion that “[the] study supports current dietary recommendations for limiting consumption of red meat intake and emphasizes the importance of different alternative sources of protein for T2D prevention.”

I’m sure those outlets believed getting the word out was going to save lives.

I went crazy over the study for a different reason — the study is complete bull$#!+. Following its advice was going to likely do more harm than good.

The Study & My Critique

Note: The full study is behind a paywall, so I can only present and critique what the AJCN published online. Text in italics are quotes from the online-access document. Also, I will mention my background (Master’s degree in Research Methods and Applied Statistics from the University of Illinois at Chicago) to show I have a deep knowledge and a lot of experience critiquing research studies.

The study background statement was: Studies with methodological advancements are warranted to confirm the relation of red meat consumption to the incidence of type 2 diabetes (T2D).

As stated, it seems that some sort of significant relationship has been found in the past and that this study looked to apply more advanced analytical techniques to confirm this previously found relationship.

But this is not the case.

Research in the area of nutrition is typically based on observational data. The methods of data collection used and the lack of experimental controls in this type of research prevent any valid causal claims. So, no — there has not been any sort of significant relationship found between red meat consumption and type 2 diabetes.

Keep in mind that type 2 diabetes is a dysfunction in the management of blood glucose — and that red meat doesn’t contain glucose. From a physiological perspective, red meat should only have a positive impact on glucose management in that eating it instead of other things would put less glucose into the blood that needed to be managed.

It’s hard to imagine how eating something that contains no glucose could act to cause a dysfunction in blood glucose regulation.

This also illustrates a problem with the way data are collected, which I’ll discuss in detail in the section on Methods.

The listed study objective was to assess the relationships of intakes of total, processed, and unprocessed red meat to risk of T2D and to estimate the effects of substituting different protein sources for red meats on T2D risk.

While the observational nature of the study prohibits causal claims, it does allow one to calculate various measures of association, or relationship. We just need to remember that finding a relationship doesn’t mean we have found a cause.

My objection here has to do with the researchers’ attempt to estimate the effects of meat protein substitution. Observational research is descriptive, not prescriptive. I wrote an article that goes into the details on this here:

In a nutshell, the lack of rigor in the research design does not allow for valid conclusions related to change in one aspect of the data. Making such a conclusion is to say that a causal relationship has been found — but we already know that valid causal conclusions cannot be drawn from observational data. So, one can’t make a valid estimate of what would happen if meat protein was substituted with (i.e., changed by using) other sources of protein.

The methods used for the study are the weakest form of research. And what the authors of the study being critiqued did was really just an analysis of data collected for studies they had no connection with (the Nurses’ Health Study (NHS), NHS II, and Health Professionals Follow-up Study).

The weakest link in observational studies like those mentioned above is in the way data are collected. These authors say red meat intakes were assessed with semiquantitative food frequency questionnaires (FFQs) every 2 to 4 years since the study baselines.

I love the use of the word “semiquantitative” as a way of describing the food frequency questionnaires. These questionnaires are wildly unreliable. For example, they will ask questions like “How many apples have you eaten in the past year?” Or, “How many blueberries have you eaten in the past year?” Some of these questionnaires are over 100 pages long.

The researchers do point out that unlike many observational studies that administer the questionnaire once every 10 years or so, the data they used was collected every two to four years. But more crappy data doesn’t make for better data.

When it comes to questionnaire items about meat, things get inconsistent. Typically, when referring to meat in a sandwich, beef, pork and lamb are lumped together, along with stews, lasagna and frozen dinners. As a main dish, pork gets its own category.

The big thing such questionnaires miss is a sense of the whole meals people eat. It’s relevant here because what is typically eaten along with red meat — bread, pasta, potatoes — does contain glucose, which, in turn, is relevant to glucose management and type 2 diabetes. And including lasagna as a meat is crazy given its base of pasta.

So, you have to at least consider that it’s not the meat that’s related to type 2 diabetes, but the carbs eaten along with the meat. But, because there are no experimental controls in observational research, we can’t untangle any carbs eaten along with the meat. This is something the researchers should mention — but they don’t. That they don’t mention alternative hypotheses suggests the researchers would rather the reader not consider anything other than the anti-meat hypothesis they offer.

The results section is complex for those without a statistics background, so I’ll just touch on the biggest red flags.

The first is that the researchers present what’s called a Hazard Ratio. A Hazard Ratio looks at the relationship between the risk of some negative event (a hazard) for one group of people relative to another group.

In this study, the researchers compared the prevalence of type 2 diabetes in the people who (based on the food frequency questionnaire data) ate the least amount of red meat to those who ate the most. And they did this calculation for Total Red Meat, Processed Red Meat and Unprocessed Red Meat.

The problem with the Hazard Ratio is that because it is a relative measure, you are one step removed from the actual data. And this is where things can get misleading.

Looking at Total Red Meat, the researchers report a Hazard Ratio of 1.62. That means those eating the most red meat had a prevalence of type 2 diabetes that was 62% higher than those who ate the least amount of meat. Not 62 percentage points higher — just 62% higher than what was found in the Low Meat Group.

While that sounds impressive, because the number seems big, it doesn’t tell us anything of real value.

It would be something to take note of if the prevalence of type 2 diabetes in the low-meat group was, say, 30.0% versus 48.6% in the high-meat group. That 18.6 percentage point difference works out to a Hazard Ratio of 1.62–48.6% ÷ 30.0% = 1.62.

But if the prevalence of type 2 diabetes in the low-meat group was 0.000003% versus 0.00000486% for the high-meat group (a difference of 0.00000186 of a percentage point), that’s also a Hazard Ratio of 1.62–0.00000486 ÷ 0.000003 = 1.62.

But the researchers don’t list the actual prevalence rates — just the Hazard Ratios. That, to me, suggests that the researchers are afraid their results will look insignificant in the eyes of anyone reading the paper and have presented only the Hazard Ratio to hide this.

The second red flag in the results is that the Hazard Ratio for Total Red Meat is larger than either that of Unprocessed or Processed Red Meat. Unless there are types of red meat that neither fall into the Processed nor the Unprocessed groups, the Hazard Ratio of Total Meat should be somewhere between the two subcategories. This suggests the researchers are unaware of basic math principles.

In the discussion section of the paper, the researchers state that intakes of red meat, including both processed and unprocessed red meats, were strongly associated with higher risks of T2D.

Well, not exactly.

Observational research, as we have noted already, can, at best, only suggest causation. Back in 1965, Sir Austin Bradford Hill provided a framework that is used to determine if it is worth conducting a randomized controlled trial (a research method that CAN identify causal relationships) on variables of interest from observational research results.

The Bradford Hill criteria include nine criteria by which to evaluate human observational evidence to determine if causation is likely to be found in a more rigorous study — the most important being the strength of the association found in the observational research.

According to the Bradford Hill criteria, any Hazard Ratio under 2.0 is not considered to be a strong association. The researchers here reported the highest Hazard Ratio as being 1.62 — which is under 2.0 and therefore should not be considered strong.

The researchers’ conclusion that Our study supports current dietary recommendations for limiting consumption of red meat intake and emphasizes the importance of different alternative sources of protein for T2D prevention is, in reality, not at all supported by their analysis.

As an aside, one of the researchers involved in the research paper being critiqued here (Walter C. Willett) is reported to have strong ties to the processed food industry, which has a lot to gain from people eating less meat. But that’s another story for another day…

Thank you for reading this article — hopefully you found it interesting!

Follow me to see more of my work in your feed or sign up for email notifications when I publish new articles.