Earlier this year a study by Harvard University sparked worldwide media headlines declaring that red meat consumption increased the risk of dying at a younger age. However what was largely overlooked was that the study was observational, not controlled, and therefore could not be used to imply causation. In the following article, reprinted with permission from his blog site “The Meat of the Issues”, Colorado State University meat science PhD candidate Travis Arp explains why the results of the study should be taken with a grain of salt.
No one has ever claimed science is easy to understand. As a consumer, we are bombarded with “science” on a daily basis; on the nightly news, in the paper, on websites, in blogs, and in everyday conversation.
“A study on child car seats reported ______.”
“Research shows drinking a glass of red wine each night helps prevent _______.”
“In a study conducted by _______, it showed laughing makes you healthier.”
As people who watch the news, we can become desensitized to phrases like that because we hear them nearly on a daily basis. That is, until we get bombs dropped on us, like the study from the Harvard School of Health, telling us that if we eat red meat and processed meats, increases risk of mortality.
Whew! That is scary stuff. You mean if I have higher than normal meat consumption, I’m going to die earlier? The mortality rate is increased by 20pc for red meat eaters? Carnivores everywhere shriek in horror.
I’ve read the entirety of the new Harvard study, and while the media has been very comfortable with saying that the study shows that undoubtedly you will increase your chance at mortality by eating red meat by X-precent…lets divulge into how this study is put together so you can get a better idea of how their conclusions are drawn.
Behind the science
First, to understand this study (and many of its ilk), let's loosely define “science.” There are many different types of sciences and ways that we can study a topic. In grad school, we are taught to strictly abide by the “scientific method,” which is asking a question, designing a well-controlled study to reduce statistical error, and drawing conclusions based ONLY on observations from the study.
On the other hand, many human health studies are epidemiological studies, in which observations are collected over a period of time from people who may be at risk or have contracted a health issue; the researchers look at trends from this population and draw conclusions. This is an effective method for looking at health problems, because we really can’t design a study in which we tell human subjects that they may contract a disease from our treatment.
So we look at the population, study common factors that may be related to the issue of concern and draw conclusions. Effective…but plenty of room for error. I’ll get to that in a second…
How the data was collected
In regard to the Harvard beef study, they collected data from a large group of people from two different studies. They have data on foods consumed by these people dating back to 1980 and 1986 for each study. The data is collected using “Food Frequency Questionnaires” (FFQ’s) in which these people would essentially estimate their daily food intake (proportion or amount of red meat, vegi’s, fruits, whole grains, etc) at the end of the year. This is point of contention #1 with the study.
Can we honestly expect people to accurately tell us what they eat every week for a whole year? Can we even expect accuracy over the course of a week? It leads to a lot of “well, I think I ate about x-amount of fruits this year.”
Plus, do I really want to admit that I go home and eat two spoonful’s of peanut butter on a biscuit every day and chase it with a Coke? Probably not. But if I eat one slice of wheat bread every week at my girlfriend’s house does that mean I’ll put a check next to “consume normal daily value for whole grains?” Absolutely. The whole concept of FFQ’s leads to mis-reporting, under-reporting, over-reporting, and just blatant lies to save yourself some embarrassment.
What data was collected
While reports of the study indicated a direct causation of red meat intake to mortality (I’ll come back to the causation term again), they don’t indicate some other factors that are mentioned in the study. Along with segregating “red and processed meat consumption” into 5 different quintiles (the main segregation factor), subjects also listed many other health and lifestyle factors in the study. These include: body mass index, level of physical activity, alcohol consumption, smoking status, current use of multivitamins, total caloric intake, and total servings of fruits, vegetables, and whole grains per day.
Now, as expected these are all variables that could likely have an impact on healthfulness, and risk of getting cardiovascular disease, coronary heart disease, hypertension, cancer, or just flat will kill you. When looking at this data, people in the fifth quintile (highest red meat consumption) versus the first had:
- Higher BMI
- Lower physical activity
- More likely to smoke and higher alcohol consumption
- Higher caloric intake
- Consumed less fruits, vegetables, and whole grains
- Used less multivitamins
What does this tell you? These people that have extremely high red meat intake probably aren’t very healthy people on the large. But yet we are drawing a direct conclusion that eating meat increases your risk of dying… hmmm… so how do we analyse these statistics?
The biggest point this paper tries to make, and what they claim separates it from other similar studies, is that they control the statistical analysis to account for those aforementioned variables. They use them as “covariates” which take a variable that could be a confounder in the data set and control it, thus reducing statistical “noise.” This is done quite often in research of all types.
What they don’t discuss in this paper is what a covariate actually does to the data. When controlling for covariates it does not eliminate it from the data, as the Harvard study would suggest. Rather, it pulls the variable (e.g. smoking status) to a centralized mean. For example, if the average smoking for all people in the study is five cigarettes per day, it drags everyone to that centralised mean. Thus, people that don’t smoke at all would be averaged to five cigs a day, and those that smoke two packs also get pulled to the average as well.
So while it is accounting for it, it is still artificially skewing the data when in fact a covariate is used to normalise data sets. Also, when you have a variable included in a study looking at death rates… can we really exclude smoking status out of the analysis?
What can we conclude?
If you watch the news, it seems the only logical conclusion is that eating more red meat will increase your risk of dying. This has been in many other reports (several from Harvard School of Health) that those same eating habits increase risk for CVD, heart disease, and various forms of cancer. If you read those studies, much of the analysis is done similar to the previous discussion. Really though, the main take-away should be: "an unhealthy lifestyle and diet increases your risk of mortality."
The role of beef in the diet is well recognized. A 3 oz. serving of lean beef provides less than 10 percent of the calories in a 2,000-calorie diet, but more than 10% of the value for: protein, iron, zinc, Vit B6 and B12, niacin, riboflavin, choline, selenium, and phosphorus. Also, while beef gets a bad rap for higher saturated fat levels in comparison to plant derived protein sources, a third of that is steric acid, which has a neutral effect on LDL or “bad cholesterol.” It also is a natural source for trans-fats that are different from those in partially hydrogenated vegetable oils, and those in beef have a neutral or positive effect on HDL’s or “good cholesterol.”
Like any food, the U.S. beef industry advocates intake in moderation. It’s a common misnomer that the beef industry combats studies like this saying that ad libitum beef consumption is alright. If you eat a 12 oz. Prime Ribeye steak every night for dinner, that isn’t healthy and you are increasing your risk for health issues associated with high fat and cholesterol diets. However, beef plays an integral role in a healthy diet, especially for growing children.
While science is hard to understand, these studies need to be taken with a grain of salt. Given the nature of how the study was conducted, it’s unfair and bad reporting on the researcher’s part to imply causation of red meat consumption to death.
You cannot draw parallels with skewed food reporting and confounding variables and say that a single factor of the data set undoubtedly, indefinitely, no questions increases your risk to die. Doing so is irresponsible and unnecessarily causes fear to the consumers.
Institutions like Harvard School of Health publish piles of papers like this every year. And they get published and reported on because they are Harvard…so they MUST be right.
At the same time though, they have published papers that directly contradict everything they said in the recent mortality paper. This doesn’t represent good science, and hopefully people will become more critical of the studies as they come out flip-flopping on their stances.
Until then, we cannot settle for the face-value of these studies and encourage reaching out to people you may know associated with science to get a better understanding of the research itself.
Statistics are easy to report on, understanding how they got there, though, can tell an entirely different story.