Meat, Potatoes, and Mortality: How Understanding Dies in a Cyberspatial Car Crash

Q: When are car crash fatalities more likely: when a population has more cars and drives more, or when a population has fewer cars and drives less?

A: Hold that thought…*

A study just published in Food and Nutrition Research reports that meat intake in general, and saturated fat intake in particular, though associated with elevated blood cholesterol levels, are not associated with cardiac disease. The authors go on to say that it is actually potatoes and cereal grains that are associated with heart disease, and that “dietary recommendations regarding CVDs should be seriously reconsidered.”

This conclusion, and almost everything else in the paper, is egregiously misrepresented, memorably misguided, and extravagantly wrong. 

Before getting into the particulars that justify that harsh indictment, a few words of preamble. First, I mean no disrespect to the study authors; I’m sure they meant well. They simply didn’t do well. 

However, they did do an incredible amount of work. It was a rather massive effort to read through this paper, running to dozens of pages and well over 30 figures. I can only imagine how much effort went into conducting this sadly pointless analysis and writing it up. I am also inclined to imagine that the editors who decided to publish it took pity on the authors in deference to that volume of work, knowing full well the paper was mostly nonsense. Either that, or they simply never managed to read through it all in the first place.

Second, when I am debunking in this manner, it is generally because a paper has been covered copiously, and misleadingly, in mainstream media. That is not the case here. To the best of my knowledge, there has been almost no media coverage of this paper, and that is very much to the credit of the editors and producers who knew to give this study, whatever its titillation value, the thumbs down. 

Unfortunately, that lack of conventional media attention no longer suffices to contain the echoes of such a provocation; they reverberate through social media, and cyberspace. I first found out about this paper, in a relatively obscure journal I don’t read, when celebratory noise about it in the so-called “low-carb, high-fat” community (LCHF) was retweeted enough hundreds of times to make it into my social media feed. When I discovered this study was getting lots of love from the “just put butter in your coffee and all will be well” crowd, I decided to take a closer look.

One final thing before addressing the particulars of this new study: I fully support the value of well-conducted observational epidemiology, interpreted in the context of all relevant evidence. This is in no way an indictment of observational research. It is, as you’ll see, an indictment of poorly conducted and poorly interpreted research (observational or otherwise); and, in particular, of double standards and hypocrisy.

The study in question examined the mean intake of 62 food items in 42 European countries, and compared the variation in those foods with variation in cardiac risk factors, and cardiovascular disease- or so the authors claim in their abstract. It’s not quite true. They did not measure dietary intake at all; they had information about food availability per capita, by country, and simply assumed that available food reliably indicated food consumed.

That’s already a big problem, but it’s not THE big problem. The big problem is that this study was ecological, meaning it looked only at information in the population, not individuals. Ecological studies are a notoriously weak kind of observational research, famously prone to the “ecological fallacy,” in which both A and B occur in the same population, but are true, true, and unrelated. So, for instance, countries with good Internet access don’t have polio. The ecological fallacy invites the conclusion that Internet use, rather than the immunization provided by developed countries, prevents polio.

In this study, we have no information about individuals at all. Rather, there are estimates of the mean intake of certain foods- at the population level. We have population level estimates of prevalence for cardiac risk factors, and for behaviors like smoking. This study did not, and cannot, examine any direct links between behaviors and outcomes in the same people. There is a good reason, in other words, why this mammoth study was published in a very low impact, obscure journal- and not The Lancet. 

The findings are nothing less than bizarre. Meat intake and saturated fat intake were associated with cholesterol, but not heart disease. Tobacco actually seemed to protect women against heart disease.

Rather than pick at the findings, let’s deal with the methods. How can a study like this produce massively misleading results?

Consider the possibility that more affluent countries eat more meat in general (this is an established fact), and have higher cholesterol levels (also an established fact), and also have better, modern health care in general (yet another established fact). Well, then, you might see the following: high intake of meat, high levels of cholesterol, and a relatively low rate of heart attacks. Is this because high blood cholesterol is suddenly irrelevant? No, it’s because modern countries with modern medical care identify and treat cardiac risk factors, and prevent heart attacks pretty adeptly.

The 42 countries included in this analysis encompassed both Western and Eastern Europe. The issues of livelihood, stress, mental health, housing, family size, and poverty- among others- were not addressed. Which countries in Europe have the lowest intake of meat, relative to potatoes and cereal grains? Poor countries in Eastern Europe, for the most part, where poverty, duress, stress, and disadvantage are hyperendemic. Past the many pages of mind-numbing number crunching, this paper may simply show that poor, unhappy populations have more heart disease.

Not convinced? OK, then here are some actual data about European countries to help get you there.

The highest rates of death from cardiovascular disease in Europe are seen in Bulgaria, Romania, and Serbia. Meat intake in Romania, Bulgaria, and Serbia is, indeed, lower than in, say Germany, France, and Italy. It is, in general, higher than in Japan. It’s lower than in Switzerland and Sweden. 

But, life expectancy at birth is higher in ALL OF THOSE COUNTRIES than in Bulgaria, Romania, and Serbia. In fact, Japan- with particularly low meat intake- has one of the highest life expectancies in the world (in the top 3). In contrast, Bulgaria, Romania, and Serbia all have more than 100 countries ahead of them on a total list of 224.

While exonerating meat, the authors of the new study indict potatoes and cereal grains. Depending on how potatoes are prepared and grains are processed, they can have a relatively high or low glycemic load, and a low or high overall nutritional value. It is certainly plausible that potatoes processed into chips and fries, and grains stripped of their bran do contribute to cardiometabolic diseases. But this study is simply too flawed to count on any of the findings, whether they make sense or not.

Let’s have a closer look at potato intake. It is, indeed, high in Eastern Europe, notably Latvia – where mortality is high and life expectancy low. But potato intake is also impressively high in the UK, which enjoys about 6 years more average life expectancy than Bulgaria, and is almost 90 spots higher up on the global life expectancy list.

So, what does all of this really mean? It means meat and potatoes cannot explain anything about mortality in an analysis that overlooks the elephant in the room: the massive influence of social factors that differ by country. 

Just in case you are not yet thoroughly convinced of the flagrant liabilities of ecological research run amok, let’s change the subject and talk about cars. 

Bulgaria, Romania, and Latvia are near the bottom of Europe’s list in terms of car ownership per capita. Maybe the take away message, then, is: if you don’t want to have a heart attack, buy a car! But wait, it gets a whole lot weirder.

Romania is near the bottom of the list for number of cars per 1000 citizens, and near the top of the list for total number of annual car crash fatalities. This leads ineluctably to the conclusion that the best way to avoid dying in a car crash is to own and drive more cars. Or, the more cars your country has, the fewer car crash deaths you’ll have. Or, the fewer cars your country has, the more likely someone living there is to die in a car crash.

Perhaps you are thinking that Romania is just a fluke, but it’s not. Latvia also has one of Europe’s lowest rates of car ownership per capita, and has had the highest number of car crash deaths over recent years. 

I can do this all day, but let’s not. The case is clear. Unless you are willing to accept that car ownership prevents both heart attacks and car crashes, you have to accept that ecological research is glaringly prone to the “garbage in, garbage out” phenomenon. I am genuinely sorry for the authors in question that so much work went into the manufacture of epidemiologic garbage- but alas, it is what it is.

Whatever the value of observational studies in general- and again, it can be quite high in my opinion- ecological studies such as this are generally useful only for generating hypotheses, not testing them. But that’s what you do before you have answers from far more robust methodologies. Here, it is a flagrantly backward step, since intervention studies, even at the level of a whole population, have tested and affirmed the hypothesis that shifting from meat-centric to plant-centric diets, reducing saturated fat intake, and lowering mean cholesterol levels causes heart disease rates to plummet, and life expectancy to soar.

As for the endless round of tweets that brought this rubbish heap to my attention, it reeks of hypocrisy. Certain members of the “eat more meat, butter, cheese” tribe have pretty much built their careers and reputations by maligning observational epidemiology. Their favorite target, the work of Ancel Keys, is of embarrassingly (for them) higher quality and vastly greater scope than this study they are now busy celebrating.

One would think, and hope, that if seats on the LCHF train were filled by something other than hypocrites, then this new study would have been disqualified as a source of their raptures under the “what’s good for the goose is good for the gander” clause. Observational studies are either of potential value, or not. I certainly think they are of value, or at least can be when done well and interpreted appropriately- but celebrated members of the LCHF club have built their rarefied reputations claiming they are not.

Alas, people with strong biases, who favor their ideology over epidemiology, don’t play fair. Observational studies are bunk if they reach a conclusion they don’t like; but are cheered, unchallenged, and retweeted a gazillion times if they reach a conclusion they do like.

Note that the remedy to this tendency of amplifying one version of junk science is certainly NOT to amplify another. Those of us who know that diets of wholesome foods, mostly plants, are best may also be tempted to amplify bad science that happens to support that position. We should not, or we encourage that bad practice. Fortunately, we have no need to do so: the remedy is to follow the overall weight of evidence, based on quantity, quality, and diversity. 

The case for any finding in biomedical science is best made when mechanistic studies in vitro and in animal models; observational epidemiology in all of its guises; diverse intervention studies; randomized trials; and outcomes at the level of whole populations over generations align- as they do for a diet of mostly plants. This very conclusion has been reached by diverse authorities around the globe again, and again, and again, and again, for very good reason.

Just how inappropriate is it to use an ecological study of this kind to claim that “dietary recommendations regarding CVDs should be seriously reconsidered”? It’s nearly as preposterous as using passages from the bible and an abacus to argue that “heliocentrism and the shape of the earth should be seriously reconsidered,” despite the modern evidence involving advanced math and physics, orbiting satellites, and actual images from space. It is backwards to the point of bizarre. Doing the best you can with an abacus when an abacus is the best you’ve got makes good sense. Going back to it in the age of Hubble speaks ulterior motives, and a thinly veiled effort to cook the books.

For the many LCHF enthusiasts with no actual knowledge of epidemiology, ignorance is the likely excuse for retweeting nonsense (i.e., some people, presumably, actually don’t know that the earth isn’t flat, and revolves around the sun). But this ludicrously extreme example of it should at least serve as a precautionary tale for next time: you can’t assume a study is of value just because you like the conclusion.

For the experts, and perhaps pseudo-experts, also busily retweeting this rubbish: well, shame on you. Either you didn’t recognize it as rubbish, in which case your ability to interpret the literature is very much in doubt. Or, you knew it was rubbish, but pretended it wasn’t because you liked the conclusion- in which case, your shame is greater, because that hints at charlatanism.

When, in other words, science was mangled and understanding killed in a hit-and-run on the “information” superhighway, the traffic cameras show clearly – that you were driving the bus.

*A: As the statistics cited above show, when a population has fewer cars and drives less! This, of course, is not because there is a paradoxical protection against car crashes that comes from driving a car. Rather, it’s because societies that can afford more cars per capita can also afford, and have, better cars, better roads, better law enforcement, better emergency response systems, better hospitals, generally lower rates of alcoholism, and so on- all leading to lower car crash fatality rates. The relationship between meat intake and mortality is subject to much the same contextual influences.