The Harvard Red Meat Studies: Some Comments

Sensational newspaper headlines highlighted the results of a recently released Harvard study, ominously The first thing I saw when I took a cursory look at the study's extract was the name of Walter Willett, chairman of the department of nutrition at Harvard's School of Public Health. Willett, of course, is a widely-known nutritionist, known for his influential criticism of the traditional USDA pyramid. Willett is a well-known proponent of the Mediterranean diet, which limits red meat to a few servings a month. In response to the new USDA My Plate metaphor, Willett criticized the approach for focusing on portion size at the expense of nutritional specifics: "It really makes a difference whether the grains you eat are whole grains or refined grains. It makes a huge difference what kind of proteins are being consumed — to be healthy, we need to be replacing the meat with a mix of chicken, nuts and legumes."


Let me think about this: Dr. Willett has spent the last several years publicly advocating for eating less beef; are we really surprised to find his name on this paper? Did we expect him to say, "Whoops! Never mind... Excuse me; I have a reservation at Ruth's Chris Steak House"?


In the interests of fairness, I've embedded the second video below summarizing the studies and the authors' conclusions below. But what are the laymen like me to think? In fact, just two years earlier, a different meta-analysis (basically a study that looks for patterns across various empirical studies) found issues with processed meats (hamburger, chicken nuggets, kebabs, bacon, cured or dried meats, corned beef, various cold cuts, hot dogs and sausages, etc.), but not with fresh meat cuts (steaks, roasts, etc.)


I have not done empirical research in the nutrition or more general health areas, but I have conducted applied behavioral research and have served as a peer reviewer for various journals; there are a number of things that we researchers look at in studies, including not restricted to, the following:

• the type of research (experimental design or exploratory: in experimental studies, we are generally trying to establish causation of a treatment/variable under rigorously controlled conditions (e.g., regular consumption of standard unprocessed meat portions results in some predicted outcome (weight, lifespan, or catastrophic disease)). Generally we prefer the more direct evidence of  experiments over exploratory study results, which yield correlation or suggestive results. Correlations are often incidental or spurious in nature, linked to some other, more salient, but unspecified variable.
• the nature of the data collection (qualitative versus quantitative; objective vs. subjective, independently observed or self-reported: there are known biases with dependence on self-reported measures.  J. Stanton, in his critique of the current meat study, cites an amusing observation from another study where nearly two-thirds of self-reported vegetarians ate meat at least one time during 2 days of observation.)
• use of reliable, valid, standard measures
• questionable statistics for low volume outcome measures (e.g., study mortality) 
• use of standardized materials/dosages (for example, if we were testing the effectiveness of a nutritional supplement, we wouldn't want results to be confounded by the amount or purity of the investigated substance)
• the nature and extent of study participants (ideally we want randomized participants, not self-selected study participants, or with some unspecified disproportionately represented characteristic with a moderating affect on outcomes; moreover, we prefer greater sample size for purposes of statistical power)
• interpretation of results: do the study's authors generalize beyond what the study design and results legitimately allow?

What do we know about the latest study? The most important point is that what we have is correlation, not causation. Study limitations include, but are not restricted to: self-reported data collection (this goes beyond various social desirability, halo effects, etc., but in terms of the use of faulty memory, the sheer length of categories and fuzzily defined categories); poorly operationalized variables (e.g., hamburger, nearly half of the beef consumed in the US, or composite dishes (say, meat sandwiches, curries or kebabs) as unprocessed vs. processed meat: this is a problem in the sense that you have other ingredients which could obfuscate meat's relationship with outcomes); and the nature of study participants (primarily white, healthy; there were also a number of characteristics that had to be controlled for, e.g., active vs. sedentary lifestyle, calorie consumption, smoking, drinking, weight, etc.).

I won't go into a detailed review of study methodology and analyses here, but I can recommend the following relevant posts for interested readers: Predator Nutrition; Denise Minger/Mark's Daily Apple; Zoe Harcombe; and Gnolls.org.

There are certain things that these authors point out that the study authors failed to take into account and I've noted in past posts: the quality of red meat can vary by source, and the method of preparation can affect nutritional value.

First, the way cattle are raised and the nature of their food will affect the balance of Omega-3 to Omega-6 fatty acids. (The saying "you are what you eat" also applies to animals: the more natural/unprocessed, the better.) Generally speaking, the ratio is more balanced and natural when cattle are able to eat from pastures versus foods they can't metabolize very well (e.g., corn), commonly used in factory farms to fatten cattle for slaughter. We generally think the standard American diet has too much Omega-6 from a number of processed foods in our regular diet. What many natural food proponents point out is that Americans over the past century  have shifted from eating more unprocessed foods (like whole grains and grass-fed animal meat, e.g., venison) to processed foods, many of which are high-carb or contain harmful trans fats.

Second, we need to be very careful of how we prepare meat: cook it low and slow, avoid charring the flesh.

What was I thinking about as I read the Pan et al. study extract? The time I went to visit my first married sister in east Texas. My brother-in-law's father raises a few head of pasture-raised cattle on his nearby ranch. My sister prepared a nice roast that came from "Blue Eyes". I think she spread a can of cream of mushroom soup over it. Very tasty; I think it's been too long between visits...

2010 Meta-Analysis Study Renata Misha, Dariush Mozaffarian, and Sarah Wallace, Circulation, 5/17/10


2012 Study Pan, Sun,  Bernstein, Schulze, Manson, Willett, & Hu,