The death of a Hypothesis: Cholesterol and Heart Disease (Ezetimibe edition)

Recent events have cast doubt on the well established "lipid hypothesis" of heart disease. In particular, the failure of Vytorin, aka ezetimibe, to lower intima media thickening. Instead the cholesterol absorption inhibitor appeared to worsen thickening, if it had any effect at all, in this rather small study. Of course, the lipid hypothesis is backed by other lines of evidence (1) but the most reliable evidence always comes from interventional studies.

So the key point of this controversy has been that no lipid lowering drug has ever reduced CVD (usually measured as a composite endpoint*) in a large study, with the exception of statins. This means it would be conceivable that some other effect of statins is responsible for their protective effects, the so called pleiotropic effects of statins.

The controversy can be put to rest given the results of the secondary-prevention study IMPROVE-IT with n ~ 18 000. I will spare you the details and just link to two other reviews instead. (Note, that the arseholes from medscape may require you to register before you can read the summary)

Over a period of 7 years, the addition of ezetimibe to simvastatin 40 mg reduced the primary end point—a composite of cardiovascular death, MI, unstable angina requiring rehospitalization, coronary revascularization, or stroke—by 6.4% when compared with patients who received simvastatin alone (P =0.016). The absolute reduction in risk over 7 years was 2.0%, with 32.7% in the ezetimibe/simvastatin arm experiencing a primary end point compared with 34.7% in the simvastatin arm.

Instead I want to make a few points:

1. Cholesterol was lowered from a very low starting level. LDL-C under a statin was approx. 70 and went down to 53 mg/dl. Obviously, one would expected diminishing returns. It is all the more impressive and a clear win for the "lower is better" hypothesis.

2. There was no benefit on all-cause mortality. This could easily be due to low statistical power. The result on the composite endpoint is driven by changes in hard outcomes, MI, ischemic stroke; and stroke (the only one at p > 0.05 with a p = 0.052). I find it hard to believe, very, very hard to believe that hard endpoints won't translate to reduced death rates.
As always, we'll see what we can get from secondary analyses, e.g. regarding cancer.

3. This study can be regarded as the death of the contrarian hypothesis (the claim that cholesterol is neutral, perhaps even healthy).


1. I notice that I cannot find a coherent write-up of this issue. For now, I'll link to a discussion I had a while ago about cholesterol and cancer risk

*Come to think of it, this seems like sloppy wording on my part. The gist of the argument is that most of the drugs barely affected the "hardest" endpoints, if they had any effect at all. From soft to hard: myocardial infarct - composite CVD - CVD mortality - all-cause mortality.