Surrogate End Points in Drug Trials: Caveat Emptor

It seems like such a good idea.  Before a patient takes a new drug, they would like to know that it is going to improve a health outcome they really care about.  Will it make me live longer? Will it lower my risk of becoming disabled?

But the problem is that it often takes a long time for a study of a new drug to show that it has meaningful impacts on patient outcomes.  Enter the brilliant idea of surrogate outcomes.  A surrogate outcome is an outcome that is associated with the health outcome a patient may really care about.  For example, a patient may want to take a drug to reduce their risk of getting dementia or Alzheimer’s Disease.  They may care so much about preventing dementia that they will even take a drug that gives them side effects.  But, it may take a pharmaceutical company years to conduct a trial to determine if a drug prevents dementia.

Surrogate endpoints seem like a brilliant solution to this problem.  We know that biomarkers such as amyloid, that is found on a brain scan, or tau protein, that is found in the spinal fluid, are associated with dementia.  A patient may not care about reducing their level of amyloid or tau protein in and of itself.  But, the theory is that a drug that reduces amyloid or tau should also reduce the risk of dementia.  It will take a lot less time to prove that a drug has an effect on biomarkers than to prove it has an effect on dementia.  Testing a drug and treating a patient on the basis of these surrogate markers makes it possible to bring a drug to market much more quickly.

Doesn’t that sound great?  Well, as eloquently described by Svensson and colleagues in JAMA Internal Medicine, it may not be so great after all.  While surrogate endpoints sound good in principle, in practice they often do not work.  Not only do they not work, but there are numerous examples where reliance on surrogate end points had disastrous consequences and harmed patients.  Svensson notes several notorious examples in the e-table of the article.  For example :

• Clofibrate reduced cholesterol in persons at risk for heart disease.  Lower cholesterol is associated with a lower risk for heart disease.  Unfortunately, patients who took clofibrate were more likely to die.
• Encainide reduces the number premature heart beats (PVCs) is persons who have had heart attacks.  PVCs are strongly associated with a higher risk of death after a heart attack.  Unfortunately, patients who took encainide after heart attacks were much more likely to die.  It is estimated that encainide caused thousands of excess deaths.
• Rosiglatazone lowers the glycohemoglobin level in persons with diabetes.  Diabetes is a risk factor for heart disease and lower glycohemoglobin levels indicate better diabetes control. Unfortunately, patients who took rosiglitazone had more heart attacks.

It is curious why there is so much enthusiasm for the use of surrogate end points in dementia drug trials when there are so many examples of how the use of surrogate endpoints in other diseases led to such awful public health outcomes.  Hopefully, history will not repeat itself.

by: Ken Covinsky (@geri_doc)