Friday, June 10, 2011

Beware Surrogate Endpoints: Yet More Evidence

Thanks to my pals Rick Bukata and Jerry Hoffman over at Primary Care Medical Abstracts for telling me about this study (subscription required). A team from Naples, Italy led by Dr. Pierluigi Costanzo did a meta-analysis of 41 randomized trials, all of which allowed a comparison between carotid intima-media thickness (IMT) and cardiovascular outcomes.

Let's stop for a plain-English break. You can do a fairly simple, noninvasive ultrasound test and measure thickening in the wall of the carotid artery in the neck. This measure has for a long time been viewed as an aspect of atherosclerosis, or hardening of the arteries--and indeed, if you divide a large population into those with a lot of thickening and those with a little, the former group ends up showing more heart disease in the end. This has led investigators who want to test whether their drug or other treatment works to lower cholesterol and to prevent cardiovascular disease like heart attacks and strokes to measure IMT as their favored outcome measure. The theory is that it might take many years to see a difference among your treatment groups in heart attack, stroke, or death rates; but within a few months to a year you might be able to measure changes in IMT, so it's much quicker and cheaper to measure IMT than to wait for the "hard" clinical endpoints.

The Costanza group sugegsts that there's only one problem with this quick-and-cheap approach. Their meta-analysis showed that there is no consistent relationship between improvements in IMT and any of the outcomes that really matter. In short, IMT is just like too many other "surrogate endpoints" in medical research--just because it gets better, you cannot assume that what really is of interest gets better too. Costanza et al continued to agree with what we thought we knew in the past--that a high IMT at baseline is a risk factor for worse heart or vessel disease on a population basis. It's the later change in IMT that seems not to be correlated with anything of importance. (Just why this is so, they offer a number of possible explanations for, which need not concern us here.)

I have to note in fairness that these meta-analysis results can cut both ways. Guess who wants to do quick and cheap studies to show that a drug works for reducing your cardiovascular risk? Our old friends the drug industry, of course. So showing the lack of any linkage between IMT improvements and real risk reduction means that a number of studies that seemed to show great promise for any given drug are of no real scientific value--no matter how many drugs may have been sold to unwary docs based on those findings. But it also means that a drug that fails to improve IMT could, presumably, still end up being valuable in reducing heart risk. Consider Zetia or ezetimibe ( The manufacturer got stung because they put a lot of weight on a study, ENHANCE, trying to show that their drug reduced IMT, and it ended up maybe making it worse. We now can see that those findings may not have really told us much about whether or not ezetimibe is a good drug. (I understand that the manufacturer is now sponsoring a longer-term study to measure actual outcomes--which of course is what they should have done from the get go.)

In a way this is all a crying shame. It makes really good sense that IMT ought to be a reliable measure of the progression or improvement of atherosclerosis. Physicians and scientists who thought this has to be true are not smoking something; the hypothesis seemed to make excellent physiological sense. And this in turn illustrates a point that's becoming an old refrain in this blog. The Pharma marketers very seldom tell docs something that we all know to be untrue and get us to swallow it. They are, on the other hand, extremely adept at taking something we already believe to be true, even if it isn't, and then using that belief to manipulate us into a course of action that ends up with more revenue in their pockets. So the problem is us fooling ourselves as often as it is them fooling us--or more accurately, us helping them to fool us.

Costanzo P, Perrone-Filardi P, Vassallo E, et al. Does carotid intima-media rthickness regression predict reduction of cardiovascular events? A meta-analysis of 41 randomized trials. Journal of the American College of Cardiology 56:2006-20, 2010.

No comments: