Surrogate endpoints in medical trials

What if I tell you that, in those who survive heart attacks, the presence of abnormal electrical conductions from the lower chambers of the heart (ventricular arrhythmias¹) is a risk factor for sudden death? That the risk of dying is in fact two to threefold higher in those patients after heart attacks with ventricular arrhythmias and no or mild symptoms? You’d insist we find a way to reduce those ventricular arrhythmias. And if I tell you that we already have two drugs, anti-arrhythmics called encainide and flecainide, that are quite good at doing just that, suppressing those arrhythmias? Perfect, you’d say.

You’d prescribe the medications to your patients and be reassured shortly afterwards on observing a reduction in their ventricular arrhythmias. Admittedly, small studies have not yet shown an increase in survival, but you’d accept that as a limitation of small studies. The reduction in ventricular arrhythmias is real, and larger studies collaborating your view that the drugs thus must also improve survival seem inevitable.

Such was the scenario in late 1989, and for the reasons above doctors regularly gave the anti-arrhythmic drugs encainide and flecainide to patients with ventricular arrhythmias after heart attacks. Then, the preliminary results of the Cardiac Arrhythmia Suppression Trial (CAST) trial were published in the New England Journal of Medicine (NEJM), and the medical community was stunned.

(The vertical line (y-axis) represents survival, and the horizontal line (x-axis) the passage of time.)

N Engl J Med. 1989;321:406-412.

CAST trial

As you would expect, in viewing the graphs, the CAST trial as it was initially designed had to be stopped early.

In the trial, patients after heart attacks with ventricular arrhythmias and no or mild symptoms whose arrhythmias would suppress with encainide or flecainide were given either the anti-arrhythmic agents or a placebo (a sugar pill). They were then followed for an extended period of time. The purpose of the study was to finally answer whether the suppression of these ventricular arrhythmias would truly increase survival.

Although the CAST trial was groundbreaking in this regard - the first well-designed trial to study the question of whether antiarrhythmic drugs safely and effectively reduce the risk of sudden death - the outcome for many physicians seemed to be a foregone conclusion. At the time a few probably even flinched at the idea of giving some of the patients a placebo.

Instead, those patients given encainide and flecainide had about a three-fold increase in overall mortality. They clearly did worse no matter how you analyzed the data: deaths from arrhythmias, nonfatal cardiac arrests, deaths from non-arrhythmic cardiac causes, death from any cause, etc. As Dr. Jeremy Ruskin wrote in an editorial that accompanied the results of the CAST trial, the “results … astounded most observers and challenge[d] much of the conventional wisdom about antiarrhythmic drugs and some of the arrhythmias they have been used to treat.”

Where had doctors’ thinking gone wrong?

Doctors began to equate the indisputable clinical endpoint of survival with the surrogate endpoint, the presence or absence of ventricular arrhythmias. Rather than studying survival of patients as the endpoint in their trials, doctors used the surrogate endpoint, ventricular arrhythmias, and then began in their minds to blur the distinction between reductions in ventricular arrhythmias and survival.

Simply put, the lessons of the CAST trial are that studies are still essential to answering clinical questions, despite what the medical community may believe on faith or common sense may tell you, and you shouldn’t rely heavily on surrogate endpoints.

Next up, Zetia and Vytorin?

A similar scenario potentially exists today with the cholesterol medication Zetia or ezetimibe. Zetia was approved in 2002 by the FDA and is sold both alone and in combination with a statin (the current standard in cholesterol medications), as Vytorin. Zetia works via a novel mechanism and as a result allows for further reductions in cholesterol in patients already on statins. It also is available for people who don’t tolerate statins.

Historically, drugs that lower cholesterol have been proven to prevent heart attacks and save lives, but will that necessarily always be true? Zetia received its FDA approval in large part based on its ability to lower cholesterol, the surrogate endpoint, but does Zetia prevent heart attacks and saves lives? There is no data to prove that. Despite FDA approval of Zetia six years ago, that data is not expected for at least another few years when a large trial called the IMPROVE-IT trial (critics argue that Merck and Schering-Plough were slow to initiate such a trial) studying about 10,000 patients is finally expected to be completed. Until then what should doctors recommend to their patients?

Could we discover that lowering cholesterol is not a sufficient surrogate? That Zetia does not in fact prevent heart attacks or save lives? Could the medical community find itself again with its pants down as it did when the results of the CAST trial were revealed? Possibly. Estimated sales of Zetia and Vytorin in 2007 close to $5 billion suggest that the lessons of the CAST trial perhaps no longer ring loudly. Growing uncertainty, however, has led guidelines to recommend more and more that Zetia not be used as a first line agent in lowering cholesterol.

How much should we rely on surrogate endpoints?

Despite historical failures of surrogate endpoints, such as with the CAST trial, surrogate endpoints have an important use in medical research, particularly with clinical endpoints that are rare or do not occur for protracted periods of time. For example, approximately a quarter of patients with chronic hepatitis C go on to develop cirrhosis (the replacement of the liver by scar tissue), but the process occurs over several decades. Imagine evaluating medications for preventing cirrhosis in hepatitis C and waiting over thirty years to determine whether the drugs are effective. Instead, most trials today look for the eradication of the hepatitis C virus from patients, a surrogate endpoint which has been shown to be associated with the lack of progression to cirrhosis.

Satisfying public demand for new medications and public demand for shortened approval processes for drugs require the use of surrogate endpoints; however, with a shortened approval process, post-marketing surveillance for key clinical endpoints and adverse events become even more crucial.

Our better understanding of science and medicine at the molecular level continually generates surrogate endpoints, but these surrogate endpoints must be meticulously and vigilantly selected. In addition, the medical community must always underscore the distinction between clinical and surrogate endpoints. The lessons of the CAST trial should not be forgotten.

.....

¹The heart is made of four chambers, two small chambers on the top (atria) and two large chambers on the bottom (ventricles). The atria merely feed the ventricles extra blood at the end of the filling cycle, but the ventricles are the workhorses of the heart sending blood to the lungs and the rest of the body. The normal electrical conduction of the heart begins in a node of the atria and spreads to the rest of the heart in a systematic fashion. Ventricular arrhythmias are abnormal spontaneous electrical conductions originating from the ventricles (six or more per hour).