Tag Archives: Essays

A useful reminder that correlation is not causation from Good Cholesterol May Not Be What Keeps The Heart Healthy – Science News:

Doctors often use measurements of two types of cholesterol in the blood as a tool for identifying patients facing a higher risk of heart attack. Data collected from large studies show a consistent pattern: People with high levels of low-density lipoprotein, called LDL or bad cholesterol, stand a greater than average chance of having a heart attack. Those epidemiological data have been backed up by drug studies showing that lowering LDL levels also lowers heart attack risk.

But the case for HDL has been less clear. “As a predictor, HDL is perfect. It’s unequivocal,” says Benjamin Voight, a geneticist at the University of Pennsylvania in Philadelphia and a lead author of the new study. Yet drugs that raise HDL levels haven’t performed well in clinical trials. On May 7, the drug company Roche announced that a trial of its HDL-boosting drug dalcetrapib had been stopped because the drug was not effective.

“As a predictor, HDL is perfect.” This means that over the years, we have observed that higher HDL levels correlate with lower heart attack risk: the two things go together. They are associated. They tend to occur together. It does not mean, and does not in any way imply, that higher HDL levels cause lower heart attack risk. Two completely different things, and it is vitally important to understand how different these concepts are.

The mere fact that we can find predictive ability for a given outcome from a given factor — aka, “correlation” — does not establish a single iota of proof related to whether that factor directly causes that outcome, in the sense that it happens 100% of the time when the factor is present, and 0% of the time when it is not. To say that A causes B is to say both of these things: (1) when A is present, B always occurs, and (2) when A is not present, B never occurs. It is a very high bar to clear, in the real world.

Also, just because you can measure something doesn’t mean it is important. Decades ago, we learned how to measure two components of blood cholesterol, HDL and LDL, and because HDL showed predictive behavior for lower heart attack risk, medical science has now spent how many billions of dollars trying to manipulate HDL levels. Why, exactly? Knowing what we know now, has there ever been a good reason to do that?

This seems to happen over, and over, and over again. Man learns how to measure something, and sees a correlation of some kind with some desired outcome, and then proceeds to waste time and money trying to “fix” something with a pill, when instead we could have just said “hey, stop eating so much and get up off the couch” with much better results. Not to mention the side effects of messing with intricate processes inside our bodies that perhaps we ought not touch.

Considered separately, the human body and the world we live in are two incomprehensibly complicated systems. The number of interactions between those two systems is much too huge to contemplate, much less measure or explain. Maybe HDL works with other factors that we know something about, but have not yet researched thoroughly. Or, maybe there are other factors that we don’t know about, or know how to measure. And maybe one such unknown factor exists with a known factor, or some combination of other known and unknown factors, but not in everybody everywhere. Which means in those smaller tests showing 100% correlation for HDL, and then again in the larger tests that do not, it might actually be something else leading to the good outcome. Something we don’t know about or understand well enough yet.

The number of variables in how we live our lives is essentially limitless, and unknowable. There is always, always, something we don’t know about or understand well enough yet. This is the starting point for every discussion about science, health, or any other topic based on human endeavor.

We seem much too anxious to resolve mysteries rather than accept and embrace them. Rule #1: not everything is knowable. Accept this and move on. You’d be surprised how freeing that idea is.

That is really the larger point behind stories like this: a system as unbelievably complex as the human body, and how diet, exercise, and medication affect nutrition and health, is not a simple one to objectively and scientifically measure. Can’t be done. Compare, for instance, to actual scientific experiments in a chemistry lab, where you can control every single factor, and you can run the same trial over and over and over again, and compare the results, and then offer a theory based on those trials. But you can’t do that with real live human beings. You can’t control every factor, or even most of them, and you can’t run the same exact trial over and over again. You must rely on surveys, which means people have to be 100% honest and have perfect memories and recall. You must use the current state of understanding of the human body, and of the world around us, both of which we continue to learn more about every year, so obviously, we don’t know everything yet — and we can’t even pretend to know just how much we don’t yet know. It’s the unknown unknowns that get you every time.

Studying health is not like a chemistry lab at all, and we ought to learn to stop thinking of it as a “science” in any real, meaningful way. Again, Rule #1: not everything is knowable. Accept it and move on. Also, stop eating so much and get up off the couch. It’ll do you some good.

If you need more convincing . . .

Losing weight, stopping smoking and increasing exercise — all actions that boost HDL levels — also improve other heart disease risk factors. That has made it difficult to tease out just how protective HDL is on its own. To find out, Voight and his colleagues compiled blood test results and genetic data gathered in earlier studies from more than 116,000 people, including more than 20,000 who had suffered from heart attacks. About 2.6 percent of those studied carry a gene mutation that raises HDL cholesterol levels, on average, about 7 milligrams per deciliter over levels of people who don’t carry the mutation. That increase in HDL was expected to decrease the mutation-carriers’ heart attack risk by 13 percent.

“But surprise, we didn’t find that,” says study coauthor Sekar Kathiresan, a cardiologist and geneticist at the Broad Institute of MIT and Harvard in Cambridge, Mass. People genetically endowed to make higher levels of HDL didn’t have a lower risk of heart attack.

Got that? Even those whose genetics predispose them to creating more HDL do not show lower risk of heart attacks. And even if they did, why should we assume that boosting HDL artificially, via drugs, would yield the same results in other people? People are genetically predisposed to create more HDL are probably different in other important ways, too, ways that we do not even understand yet. Do we really believe that we can isolate just one solitary factor out of an amazing complex system like the human body, and transfer it into another body with completely different genetic makeup, and watch the heart attack risk drop as if by magic? Call me crazy, but I’m pretty sure it doesn’t work that way.

To make sure the result wasn’t a fluke, the researchers compiled a genetic score for 53,813 people based on the number of 14 other HDL-boosting genetic variants the person carries. Even people with the highest genetic score, and thus the highest HDL levels, weren’t protected from heart attack. To make sure the technique works properly, the researchers also compiled an LDL genetic score and got the expected answer: higher LDL levels were associated with higher heart attack risk.

Commit it to memory, and to heart: raising HDL levels does not, by itself, do anything to reduce heart attack risk. Ignore those who claim that it does, because there is no evidence to support it.