You are here
April 27, 2018
Be Skeptical About Biomarkers and Surrogate Endpoints
By Mike Lauer, M.D.
NIH Deputy Director for Extramural Research
Many of us remember where we were when a sentinel event occurred—when President Kennedy was assassinated or planes crashed into the World Trade Center towers. For cardiologists, we remember where we were when we heard the preliminary results of the “Cardiac Arrhythmia Suppression Trial” (otherwise known as CAST) on August 10, 1989.
Let me explain. In the 1980s, a large body of research showed that patients who survived a heart attack (the technical term is myocardial infarction) and who had lots of skipped beats—heart beats that occurred out of synch with the regular heart beat (the technical term is ventricular premature beat or VPB)—had an unusually high risk of sudden death from an electrical storm that effectively disables the heart within a matter of minutes. In other words, among heart attack survivors, the presence of VPBs was a “biomarker” or “surrogate endpoint” of increased risk of sudden death.
As “luck” would have it, scientists and doctors had developed and tested drugs (called “anti-arrhythmics”) that could suppress VPBs—make the skipped beats go away. Doctors then reasoned: VPBs mean higher sudden death risk; anti-arrhythmics suppressed the VPBs; hence, anti-arrhythmics should reduce risk of sudden death. This might seem reasonable to you. So many doctors believed it that they dispensed hundreds of thousands of prescriptions for anti-arrhythmics.
There were a few doubters—doctors who wondered whether there should be a clinical trial to test the hypothesis that anti-arrhythmics could prevent sudden death and save lives. But many doctors felt it would be unethical to do such a trial since it was so obvious that anti-arrhythmics work. Nonetheless, the skeptics prevailed and the trial—CAST—was launched. Over 2,000 patients who had survived a heart attack and had VPBs were randomized to receive anti-arrhythmic drugs or to receive a placebo (inert drug).
Now we come to August 10, 1989. On that day, the preliminary results were announced. To everyone’s amazement, patients who were randomized to anti-arrhythmics were more than twice as likely to die. This made no sense, but the data were overwhelming.
Over the next few years, more trials were conducted and completed, and most showed the same: giving anti-arrhythmics paradoxically increased the risk of death. Over time, doctors were convinced and therapies began to shift. Today, many patients receive implantable defibrillators as the primary way to prevent sudden death. And trials have shown that implantable defibrillators do indeed save lives.
The CAST story was a sobering one. It was an early example of what happens when we rely too much on biomarkers or surrogate endpoints. There are lots of surrogate endpoints: blood pressure, cholesterol levels, certain proteins in the blood, tumor size, or findings on an X-ray or MRI. We might think that if we can figure out a way to treat the surrogate endpoint we treat the problem. Sometimes it’s true—for example, many drugs that lower blood pressure or lower cholesterol levels also lower risk of stroke, heart attack, or even premature death. But sometimes it’s not true, and we risk making serious—even deadly—mistakes like we did with anti-arrhythmics.
The next time you hear about the results of a clinical study—especially a clinical trial—ask whether or not the endpoint was a clinical endpoint or a surrogate endpoint. A clinical endpoint might be premature death, heart attack, stroke, disability, pain, or an unplanned hospitalization.
Now you might ask: why do we conduct trials that focus on surrogate endpoints? After all, don’t we really care about clinical endpoints? Of course, we do! The reality is that surrogate endpoint trials are easier, faster, and cheaper because they typically require fewer patients—sometimes a lot fewer. But we need to exercise caution. We’ve had a number of experiences like CAST, where the surrogate endpoint trials were later found to give us the wrong answer.
Some of us believe that the best way to deal with this problem is to conduct more trials that focus on clinical endpoints, but to conduct those trials in innovative, efficient, cost-effective ways. One way to do this is to design trials using pre-existing large-scale databases or “registries.” In the meantime, it’s important to have a healthy skepticism when reading the results of trials that focus on surrogate endpoints.