“What gets measured, gets managed” is an adage that doctors know all too well. Physicians routinely monitor diseases by measuring biological factors such as cholesterol or blood pressure. And clinical trials are starting to use such indirect measures as 'surrogate markers' of a treatment's effect on the disease in clinical trials (see page 510). The practice has helped to get drugs to market quickly by allowing researchers to perform rapid and easy measurements rather than waiting for long-term clinical outcomes such as heart attacks or strokes.

But every so often, a surrogate deceives, as in the unexpected results from two recent clinical trials. One found that the drug ezetimibe, which lowers cholesterol levels, did not seem to slow the steady march of atherosclerosis in patients with high cholesterol (J. J. P. Kasterlein et al. N. Engl. J. Med. 358, 1431–1443; 2008). In the other, one part of the trial was halted when the researchers found that using a combination of strict diet and insulin to lower diabetics' blood-sugar levels to those of healthy people actually increased mortality rates in some patients.

These results conjure up unpleasant memories of past surprises with surrogate markers. Several drugs that quelled irregular heartbeats in patients who had had a heart attack shocked cardiologists when they were found also to increase the risk of a subsequent heart attack. Another candidate drug for controlling cholesterol, torcetrapib, raised levels of the 'good' cholesterol — high-density lipoproteins — yet increased the number of deaths. Such events are troubling, and regulators face a high-stakes tug-of-war between the risk of relying on a surrogate marker, and the harm of delaying a potentially useful therapeutic from being used in the clinic.

Cardiologists at this week's annual meeting of the American College of Cardiology urged their colleagues to try other medications before resorting to ezetimibe. The warning should not have been necessary. Other cholesterol-lowering drugs have been on the market for 20 years, and have been vetted by large-scale clinical trials that evaluated direct outcomes such as heart attacks. Ezetimibe was approved five years ago in the United States solely on the basis of its ability to lower lipid concentrations. Large clinical trials to look at its effects on the cardiovascular system were not initiated until several years after it was approved. Some of the results aren't due until 2011. The choice between ezetimibe and the older drugs should have been clear, yet the statistics suggest that many doctors favoured ezetimibe. In 2006, ezetimibe accounted for 15% of the prescriptions for cholesterol medication in the United States. Sales of the drug climbed above $5 billion in 2007.

Why did the medical community embrace ezetimibe with such enthusiasm? There are two likely explanations. The first is an aggressive and effective marketing campaign. The second is a firm belief in the predictive power of a trusted surrogate. Previous cholesterol-lowering drugs had proven successful, and in heart disease as in diabetes, the belief came to be: 'the lower, the better'.

We won't know for some time whether ezetimibe works for the end points that matter: heart attack and death. But it is clearly time to check our assumptions. Surrogate end points are valuable, but are not always accurate indicators of the important outcomes. Drugs approved solely on the basis of surrogate end points come with a risk that should not be ignored, no matter how trusted the surrogate — or how persuasive the marketing campaign.