Richard David Feinman

I was walking on a very dark street and I assumed that the voice I heard was a guy talking on a cell phone. Apparently about a dinner party, he was saying “Remember, I don’t eat red meat.” Only a few years ago, that would have sounded strange. Of course, a few years ago a man talking to himself on the street would have been strange. He would have been assumed to be deranged, more so if he told you that he was talking on the telephone. But yesterday’s oddity pops up everywhere today. Neo-vegetarianism affects us all. It’s all described very well by Jane Kramer’s excellent review of veggie cookbooks in the April 14 New Yorker,

“…from one chili party to the next, everything changed. Seven formerly enthusiastic carnivores called to say they had stopped eating meat entirely…. Worse, on the night of that final party, four of the remaining carnivores carried their plates to the kitchen table, ignoring the cubes of beef and pancetta, smoky and fragrant in their big red bean pot, and headed for my dwindling supply of pasta. “Stop!” I cried. “That’s for the vegetarians!”

Illustration by Robin Feinman. Reference: http://en.wikipedia.org/wiki/Carrie_Nation.

The New Yorker review provides great insight into the different forms of vegetarianism and the various arguments in favor of vegetarianism, arguments of variable force; The treatment of animals while alive, more than their slaughter, is probably most upsetting. Just the review, in the London Review of Books, of “Farmageddon. The True Cost of Cheap Meat” and “Planet Carnivore” is sufficiently scary as to be unreadable. Most of us just live with it. Personally, I think of it by analogy with the announcements on airplanes that, under conditions of low pressure, you should put on your own oxygen mask before helping others. When we start treating people better we can help the animals. Maybe it’s a rationalization. In any case, one cause that doesn’t sit well is “…the health argument (doctors and nutritionists, alarmed by the rise in illness and obesity in a high-fat Big Mac world)…”

It’s not a high fat world any more than it’s a vegetarian country. A better description might be: “Doctors and nutritionists, alarmed by the reduction in funding for anything but the party line, and imbued with a missionary zeal, try hard to find something wrong with meat, especially red meat.” A lifeline for bloggers: there is a “meat kills” article every couple of months, usually an epidemiologic study with an odds ratio of 1.4. Odds ratio is what it sounds like: Your odds of getting a disease with the intervention vs your odds under control conditions. These relative risk indicators are commonly around 1.4 for the usual “is associated with” article. For comparison, the odds ratio for getting lung cancer if you smoke is about 20 compared to not smoking. If you are a heavy smoker, odds are about 30 to 1.  A little elementary algebra will tell you that odds ratio of 1.4 comes out around 60:40 distribution for getting better with meat or getting worse with meat. That’s what it means. Statistically, you are looking at a two-tailed distribution. That means that if you don’t get better, you will get worse. Without other knowledge, more meat may be better. Not compelling odds to me.

Since we don’t really know what causes cancer or even heart disease or especially all-cause mortality, most of us let the “meat will kill you stories” fade into oblivion like other “breaking” news stories. However, when the dictum is “meat causes diabetes,” it is hard to ignore. Far-fetched and dangerous for it’s obscuring the elephant in the room, carbohydrate, such a preposterous notion should be addressed.

One of the worst of the meat scares was Pan, et al (JAMA Intern Med. 2013;173(14):1328-1335.) from the Harvard School of Public Health, a major supplier of these studies. This kind of stuff has been deconstructed by several bloggers but this particular paper had a new twist. The study, rather than gathering data at a single time-point considered the changes in consumption over time.  Strange, because most of these studies are based on food questionnaires which, by themselves, have a good deal of error. Measuring the differences between two time points with significant error only increases the error.  Error in a parameter that already has some uncertainty. It is the equivalent of weighing the captain by weighing the ship before and after they are on board. In any case, Pan’s paper concludes:

“Increasing red meat consumption over time is associated with an elevated subsequent risk of T2DM [type 2 diabetes mellitus]….”

You have to read the original in order to properly evaluate papers like this. First off, as in much of the medical literature, you see only one figure but several mind-numbing tables. Masses of numbers instead of graphics is a sufficiently serious problem that a whole book, “Medical Illuminations”  (recommended), has been written about it. The tables, however, do give you the raw data (at least as averaged into big groups) and the outcome from “corrected models.” Without going through the calculation here, the tables tell you how many people from each group (broken up by meat consumption) have diabetes. Divide the number of cases by number in the group and you can see the risk. When you plot the raw data you see that the reduction in red meat intake, which is really the ultimate recommendation of the paper, leads to an increase in diabetes. What? That’s the opposite of the authors’ conclusion.

The raw data is in the tables but the authors do not calculate the risk from that data. You have to do that yourself. It is just arithmetic but it seems like the authors would tell you. The table shows results from “models” that have been “corrected” for confounders. What does that mean? Well, when you get a positive result for an association, you have to make sure that there weren’t underlying factors (other than the one you are interested in) that account for the outcome. So, for example, if you say that increase in a particular food is associated with a disease, you are expected to subtract out the effect of any increased calories. On the other hand, if your primary data don’t show an effect then you are, more or less, out of luck. You can try, however, to “correct” with something known to cause the disease, something expected to make things worse. If this makes things better, you may have shown a benefit in your outcome but it becomes very far-fetched unless there is a very small number of variables. Generally, if your “confounders” improve the correlation, it is likely that it is the confounders that are the controlling variables.

I wrote a letter to the editor of JAMA Internal Medicine pointing out that “the authors measured the effect of reducing meat consumption, which increased the frequency of diabetes in all the cohorts studied, opposite to the expectation of a consistent dose-response curve.” The journal published the letter along with the authors’ answers (they get the last word). The journal has a strict policy on brevity and you are not allowed to use any figures, like the one above, to show what things are really like.  The authors answer to the dose-response question:

 “Figure 1 in our article1 showed that increasing red meat intake within a 4-year period was positively associated with T2DM in the subsequent 4 years in a dose responsive manner, not “the effect of reducing meat consumption, which increased the frequency of diabetes in all the cohorts studied,” as claimed by Dr Feinman.

Astounding. That’s the problem. They presented only part of the data. That increasing red meat had a positive association (first highlighted phrase) does not contradict the statement that reducing red meat also showed a positive association. My figure shows that increasing red meat or decreasing red meat increased diabetes. How is this possible? Red meat decreases risk of diabetes and then increases it? How could that work?  It is not possible. It is simply that the the data have too much randomness to be meaningful at all. In fact, if we platted

So how do they justify their conclusion? Simple, they correct the data for confounders. They correct for initial red meat intake which makes the effect of an increase in meat stronger as you would expect. They then correct for age but they don’t show you what that effect is. In fact, they correct for “race, marital status, family history of T2DM, history of hypertension, history of hypercholesterolemia, smoking status, initial and changes in alcohol intake,…” — I’m not making this up — “physical activity, total energy intake, diet quality, postmenopausal status and menopausal hormone use plus initial body mass index and weight change.” Mirabile dictu, they are able to get the answer to come out the way they want.

It would probably be hard to explain to the authors why this doesn’t make any sense. If you have to do so much work to get the answer, it can’t mean anything. It’s all like the old joke about the woman who calls the police because the guy next door is exposing himself. When the cops come, she shows them the window.  The cop says “Lady, that window is too high to see anything.” She says “Sure. Where you are. But stand on this chair and you will see what I’m talking about.”

So, does all this mean anything at all? Well, it means that diabetes is not correlated with red meat unless you include many other factors and maybe those factors are what we should be warned about. But it is the big picture that is simple. This is not done in a vacuum. There are big epidemiologic studies. The real point is, as in my Letter to the Editor, “Red meat consumption decreased as T2DM increased during the past 30 years.” The data are not about a couple of percent. The data are compelling:

The authors’ answer was “this ecological relationship cannot be used to argue against the causal relationship between red meat intake and T2DM because many other factors have changed over time.”

This statement stands as the embodiment of the total lack of common sense and the irrational perspective of the epidemiologist. (Okay. Just these epidemiologists ). Epidemiology is not more sophisticated than ecology and the data are epidemiologic any way. And there are always more factors. You have to say what the factors are. You can’t just say something else might be involved.  The end of common sense. The end of science.

But why do they do this? I am not sure why you would think that red meat had much to do with diabetes but the study showed that, if you did, you would be wrong. Research gives you a lot of failures. You just go on to something else. Nobody knows about motivation, nobody knows what was on their mind. Seven possible reasons are NIH grants P01CA087969, R01CA050385, U19CA055075, R01DK058845, P30DK046200, U54CA155626 and K99HL098459. Nonetheless, one has the sense that the authors really believe their conclusion and that there is a general emotional and puritanical reaction to red meat and its agents.

“Components in red meat that may contribute to T2DM…”

“…The time has been

That, when the brains were out, the man would die,

And there an end…”

— William Shakespeare, Macbeth.

A big problem: the underlying mechanism. What might actually be the agent that confers such danger on red meat? Pan, et al say “Components in red meat that may contribute to T2DM risk include heme iron, high saturated fat and cholesterol, added sodium and nitrites and nitrates in processed meat, etc.”

This list is notable for the presence of saturated fat and cholesterol. Isn’t that dead? The latest report about evidence that saturated fat does not pose a risk has a certain degree of squabbling but it is only one in a long line of individual studies and meta-analyses that drive a stake through the heart of cholesterol and saturated fat as a risk. Walter Willett, an author on Pan, et al just couldn’t face the result and wanted the paper withdrawn, but the history of risk of saturated fat and cholesterol is demonstration of one failure after another, some from his own lab. The idea never dies. One interesting part of the squabbling was the statement, “A 2009 review concluded that replacing saturated fats with carbohydrates had no benefit, while replacing them with polyunsaturated fats reduced the risk of heart disease. Several scientists say that should have been mentioned in the new paper.” Presumably it is the second part, rather than the first that they want mentioned.

But underneath it all is the moralistic, puritanical mindset. In trying to face the evidence in the original report, Alice Lichtenstein said, “It would be unfortunate if these results were interpreted to suggest that people can go back to eating butter and cheese with abandon.”  Abandon? I guess we are supposed to think of the gutted pig scene in Fellini’s Satyricon.

All such moralistic proscriptions have the risk of what pyschologists call counter-control, where people specifically do the opposite to what they are told because they don’t like to be pushed around.  I personally rarely eat meat before 6 PM, but when I found out that Mark Bittman says that that is what we all must do, it made me get out left-over spareribs for lunch.

Now going back to the original question of Carrot Nation, it is heartening to see that in her review of Deborah Madison’s Vegetable Literacy, Kramer points out that, in the preparation of cardoon risotto “there is permission to simmer it in a ‘light chicken stock,’ and even an acknowledgement that vegetable stock might ‘overwhelm’ the flavor of that delicately bitter member of the sunflower family.…” And, in the end, Kramer says, “The book is sly. Think of it as a pro-choice cookbook decorously wrapped in carrots and beans and lettuce leaves. Apart from the chicken broth, you won’t find anything ‘animal’ listed but read what she has to say about some of those recipes, and you will detect the beginning of a stealth operation — a call to sit down at the dinner table together and put an end to the testy herbivore-carnivore divide.” This suggests that they might both be in tune with my own philosophy. I am not arguing for meat but rather I am opposed to inaccurately attacking meat, a philosophy that I call antidiscarnivorianism.