What Does It Mean to Call Something a Cause?

The Cause That Isn’t Destiny - Part 2

This is Part 2 of The Cause That Isn't Destiny, an eight-part series on cholesterol, causality, and the limits of any single measurement to predict an individual's future.

You can find Part 1 here, in case you missed it.

If we strike a match and it catches fire, we would not doubt that the striking caused the fire. The connection is obvious, unambiguous — one event causes another, predictably and repeatedly. If we do the same thing tomorrow, we will almost certainly get the same result.

Biology rarely offers this degree of certainty. Most people know the connection between smoking and lung cancer. Although it is considered well established, other biological factors mean that the variability within this relationship is considerable. We know of individuals who have smoked for sixty years without developing lung cancer, while others may develop it after a much shorter smoking history. High blood pressure increases the likelihood of stroke, but many people with elevated blood pressure never have one. The same applies to diabetes, obesity, and most of the diseases we deal with every day.

In medicine, the word cause is often used differently than we use it in everyday speech. When we say that smoking causes lung cancer, or that LDL cholesterol causes atherosclerosis, it doesn’t mean the disease is unavoidable. What is usually meant is that the factor drives a biological process that makes the disease more likely. Even when a factor is accepted as causal, we often encounter it clinically as a risk factor because it changes probability rather than determining destiny. This is a subtle distinction in emphasis, but it changes almost everything about how we understand possible biological causal factors.

The woman I discussed in the last chapter shows clearly why this distinction matters. Her LDL cholesterol was very high, but her coronary arteries were remarkably healthy for her age. Her brother, however, died of a heart attack at fifty-two with much lower LDL cholesterol. These two stories don’t disprove the role of cholesterol. They remind us, instead, that a biological cause usually depends on many variables and is most often not a single, simple cause—the way the match was above.

But how do we go about deciding whether some factor is a risk factor or a genuine cause? There is little point in looking at a single patient and drawing conclusions from what we see. A hundred patients isn’t enough either. We need a much larger group.

A few individuals can surprise us; it is harder to deceive a large population. Much of our knowledge in cardiology has come about when researchers notice a pattern among thousands of individuals. Such a pattern, though, can never tell us that one factor causes another, only that the factors occur together. This is what is called an association.

In the latter half of the twentieth century, researchers began following large groups of healthy people for many years, measuring blood pressure, cholesterol, smoking habits, diabetes, weight, and much else. People were followed in order to see who went on to develop cardiovascular disease. The best-known study of this kind is undoubtedly the Framingham study, which began in 1948 and has followed several generations of people for more than seventy years. The same pattern appeared there, again and again: people with higher LDL cholesterol had heart attacks more often, and cardiovascular disease was generally more common than among those with lower levels. Studies of this kind are called observational epidemiological studies.

Heart disease and LDL cholesterol are a typical example of an association. An association, however consistent, can easily deceive us. People who carry matches or a lighter in their pocket are more likely to develop lung cancer. The reason, of course, isn’t the matches themselves, but that they’re associated with smoking. Heart disease is more common among people with grey hair, which obviously has nothing to do with the hair itself.

Medicine has repeatedly been misled by associations of this kind. Hormone therapy once appeared to protect women against heart disease. Vitamin E supplements seemed to be good for health. In this way, many kinds of treatments seemed to work when scientists were simply observing individuals who had chosen, or not chosen, to take them. It was not until later that clinical intervention trials revealed that many of these treatments had none of the effects that had been expected, despite the association.

Some are born with a genetic variant that causes them to have low LDL cholesterol their whole life. Others inherit familial hypercholesterolemia and have very high LDL cholesterol from childhood. Unlike diet, exercise, and other environmental variables, these differences are determined before birth. Nature has, in fact, randomized for us.

When these different groups are examined, it turns out that people who live their whole life with low LDL cholesterol get less atherosclerosis and fewer cardiovascular diseases, while those born with very high LDL cholesterol are much more likely to get coronary disease early in life. Studies like these are called Mendelian randomization studies, a name drawn from the fact that genetic variants are distributed more or less at random, just as when treatment is allocated in a randomized intervention trial. We cannot choose our own genes; the genetic variant is fixed long before lifestyle habits take shape. This way, various variables that could affect the results can be avoided.

Clinical, randomized intervention trials have also been carried out to examine whether an association contains a causal relationship. If LDL truly plays a causal role, lowering it should reduce cardiovascular disease. The effects of various drug treatments that lower LDL cholesterol have been examined in trials of this kind, and I would first mention the many trials on statins, and later the trials on PCSK9 inhibitors. A comparison group receives a placebo, and chance decides who gets which.

Most of these trials have shown that drugs that lower LDL cholesterol reduce the risk of heart attacks, strokes, and deaths from cardiovascular disease. Although the mechanism of action of the drugs examined is not the same, the results have generally pointed in the same direction. Lowering LDL cholesterol positively affects outcomes in patients with atherosclerotic heart disease. Many experts have viewed this as important evidence. It is harder to dismiss findings that align across different mechanisms than the isolated effect of a single drug. There is, admittedly, still much that calls for further explanation, particularly how we apply these findings to individual patients. Reality has a tendency to surprise us.

At the beginning of this century, it became difficult to avoid accepting this conclusion: LDL cholesterol is not only associated with atherosclerosis. It is one of its causes. This finding, and the interventions that came from it, have saved a great many lives, and are the reason doctors like me spend so much time discussing cholesterol with our patients.

But it does not explain the woman I discussed earlier. Nor does it explain her brother’s fate. Today we understand fairly well the processes that cause atherosclerosis. What we do not understand well enough is why it developed in him, and not in her. That question hasn’t been answered. It may matter more than the one we just spent this chapter answering.

---

Part 3 - The Number We Trust Too Much