May 7th, 2012

Another Surrogate Endpoint Falls Short

A recent paper in the Lancet explores whether carotid intima–media thickness (CIMT) is an adequate surrogate marker of risk for cardiovascular events. Matthias Lorenz and colleagues combined data from 16 longitudinal observational studies of CIMT progression, involving about 37,000 patients. In their patient-level meta-analysis, less CIMT progression was not associated with a lower likelihood of subsequent cardiovascular events. An accompanying editorial by Vijay Nambi and colleagues strikes the right chord, particularly with respect to heterogeneity in techniques for assessing CIMT and measurement errors (or noise) associated with its use.

As a backdrop to this newly published information, current guidelines make a class IIa recommendation for a single CIMT measurement to enhance risk-stratification in asymptomatic adults. Single CIMT “snapshots” may indeed be useful in certain clinical settings, but it’s unclear what CIMT progression over time means for an individual patient’s clinical outcomes, as the Lorenz study makes clear. Furthermore, that study is consistent with two recent meta-analyses that used similar data from clinical trials on novel therapeutic agents; one was led by my colleague at the University of Michigan, Zach Goldberger. Collectively, the three studies should discourage reliance on CIMT progression as a surrogate endpoint.

Some have argued, though, that CIMT avoids the substantial costs and lengthy follow-up required for trials that focus on “hard” but uncommon outcomes such as death. Presumably, many of these folks will remain unconvinced by the accumulating data and point to the inherent limitations of meta-analyses while highlighting some strengths of CIMT.

Overvaluing surrogate endpoints is nothing new, of course. Cardiologists have learned this lesson many times, beginning with the infamous CAST trial, which demonstrated that successful suppression of premature ventricular contractions by encainide and flecainide was not just ineffective at improving outcomes but also associated with higher rates of death.

Nevertheless, unlike serum biomarkers, CIMT (along with IVUS) has achieved something of a unique status as a tool in clinical trials. That’s because of its advantage in actually visualizing atherosclerosis directly in the vessel wall, making it appear to be less of a “surrogate” endpoint. However, oncologists would not be surprised that imaging does not correlate with hard outcomes. They have long understood that tumors that initially melt away with chemotherapy can come back. Cardiologists, too, must be willing to admit that although our ability to view changes in atherosclerosis in the vessel wall over time is better than ever, we still don’t know what those changes mean in terms of the outcomes that matter most to patients.


1. What do you think about the use of CIMT, IVUS, and other imaging modalities as surrogate endpoints in clinical trials?

2. Can studies be better designed to identify surrogate endpoints that might be valuable early in a drug or device evaluation?

3. How do we reconcile our need for tools to evaluate potentially useful therapies early in their development, given that it is impossible to imagine conducting large clinical trials with hard outcomes for every promising drug?

3 Responses to “Another Surrogate Endpoint Falls Short”

  1. Robin Motz, M.D., Ph.D. says:

    As I often tell my medical students and residents, the map is not equivalent to the actual terrain, and the test may not indicate what is wrong with the patient. After all, don’t all cervical xrays of patients over the age of 40 show “osteoarthritis” even though the patient may have no such cervical symptoms? And don’t all color Doppler echocardiograms show trace MR because the leaflets take a finite time to close? The problem with surrogate endpoints is that we are assuming a cause-and-effect relationship where none may exist, similar to the chronic brouhaha about the utility of lowering TGL. Just because the metal arrow in the lobby indicates the floor where the elevator is, it does not therefore follow that forcing the arrow down to “L” will bring the elevator down to the lobby level.

  2. Henry Green, A B, MD says:

    If our guidelines were irrefutable, they would never need to be updated. In spite of this, they are the best we can do in informing our clinical judgement. Unfortunately, they have also become the basis for policy decisions, reimbursement, and litigation.

    We have been deluded by surrogate end points for millenia. Often, the surrogate is not in the same causal path as the disease process. The reason they are used is that hard end points may take years to occur. Studies that rely on them would take much longer and cost more.

  3. It still may be unclear what CIMT progression over time means for an individual patient’s clinical outcomes. However, it is still a good non-invasive screening device that helps individualize treatment as well stratify cardiovascular risk.

    The new machines are produced by Panasonic (CardioHealth® Station) and they have taken operator error out of the equation though automation.

    For example; I had two 60 y/o patients last week with moderately elevated lipids. Both were give a prescription for Rosuvastatin. Neither felt like filling it. CIMT demonstrated in one patient that he was in the 75th percentile. He will take his Rosuvastatin(along with CoQ10 to reduce side effects)and also get a heart scan now. The second patient was below the 25th percentile. He will get diet and lifestyle intervention and just start Niacin. Both will follow up 4 times a year for lipids.