Surrogate endpoints in EBM: What are the benefits and dangers?
Posted on September 4, 2014 by Mohammed Shabsog
This article aims to emphasize the importance of assessing clinical trials which use surrogate outcomes very carefully before applying the recommended intervention into clinical practice. To do so, we will explore the following questions: what are surrogate endpoints? What makes them so appealing to use in research? What are the dangers of their misuse? What determines whether using surrogate endpoints is the right approach to assess an intervention?
What are surrogate endpoints?
A surrogate endpoint is a physical measurement of a specific outcome which is considered to be a valid predictor (or representative) of the real outcome or final result. In simple words, a surrogate endpoint is like a measurable indicator that can help us know what the real result is. One example of this is cholesterol levels and the risk of having a heart attack. Generally, a high level of cholesterol in the bloodstream indicates a greater risk of having a heart attack. Thus, measuring higher cholesterol levels (which is the surrogate endpoint) allows us to predict that the patient has a greater risk of suffering from heart attacks (which is the real outcome). Another example is blood pressure and the incidence rate of heart attacks and strokes. Evidence shows that having high blood pressure increases the risk of suffering from strokes or from heart attacks. In this case, blood pressure is the surrogate endpoint as a higher blood pressure is a predictor of a greater risk of strokes and heart attacks. In another example, having an irregular heart beat (the surrogate endpoint) suggests that there is a greater risk of suffering from a sudden cardiac death (the real outcome). In other words, surrogate endpoints are useful since they can, in some situations, be representatives of the final clinical outcomes.
Benefits of surrogate endpoints:
Surrogate endpoints are very appealing to use in research for several reasons. Firstly, measuring real outcomes (unlike measuring surrogate outcomes) is often very time-consuming. For instance, it could take researchers several years to collect enough information to determine whether a new cholesterol lowering drug significantly reduces the incidence rate of heart attacks in the group being tested. It would be easier, instead, to measure cholesterol levels in patients a few weeks after starting the clinical trial and to judge the efficacy of the drug based on the degree of reduction in the cholesterol levels.
Secondly, using surrogate outcomes reduces costs for manufacturers since it reduces the duration of the clinical trial which means the new intervention could be introduced to the market faster.
In addition to this, surrogate endpoints could, on some occasions, be useful since they could give more balanced findings than using real outcomes. For instance, when testing a new painkiller, measuring the related surrogate endpoints (before and after administering the new painkiller) would probably give a less subjective finding than performing pain assessment on the patient. This means that, on some occasions, it would be better to consider and assess both the surrogate and real outcomes.
Despite these benefits, it is important to remember that using surrogate endpoints might be problematic on some occasions since they are not always good representatives of the real clinical outcome.
In a general sense, a surrogate outcome would only be a good representative of the real outcome if the surrogate endpoint itself is the sole or major contributor to the progression of the disease (or disorder) towards the real endpoint.
The dangers of using surrogate endpoints
To understand the dangers of the use, or perhaps the misuse, of surrogate endpoints, it is important to understand how researchers choose these surrogate endpoints to begin with. Researchers typically select certain bio-markers or chemicals as surrogate endpoints based on the existing hypothesis that they have about the mechanism of the disease. In other words, researchers choose those surrogate endpoints only if they believe that the selected bio-markers are important contributors to the mechanism of the disease. Although researchers formulate such hypotheses based on existing evidence, there is a good chance that there are some aspects of the disease mechanism which have not yet been discovered and which, thus, have not yet been accounted for in the hypothesis. This means that researchers have no choice but to choose a surrogate endpoint based on a hypothesis which may, or may not, accurately describe the mechanism of the disease or disorder. Consequently, researchers might select the wrong surrogate endpoints if the hypothesis itself was inaccurate. Historically, there were several occasions where patients unnecessarily died due to the misuse of surrogate endpoints.
The story about how measurements of surrogate endpoints mislead researchers in their assessment of a new group of anti-arrhythmic is one notable example of this. Patients suffering from arrhythmia (irregular heartbeat) are usually given anti-arrhythmic drugs to normalize the heartbeat or to ‘correct’ the heartbeat back to normal. Arrhythmia is a dangerous condition since it raises the risk of the patients suffering from sudden cardiac death. In the 1970s, a group of researchers were testing a new batch of anti-arrhythmic drugs and early results suggested that the new drugs were successful at normalizing the heart beat. Researchers believed that measuring the change in heartbeat would be a good representative of the risk of having a sudden cardiac arrest. Early results showed that the drugs did, in fact, normalize the heartbeats which means that the surrogate endpoint (that is the heartbeat measurement) predicted the drugs were successful. Nevertheless, results later showed there was a greater mortality rate in the group receiving the anti-arrhythmic drugs compared to the group receiving the placebo. Therefore, heartbeat measurements were misleading and the drugs had actually been doing more harm than good.
There are several situations where surrogate outcomes would not be regarded as valid predictors of the real outcomes. For example, the surrogate endpoint might not be a major contributor towards the progression of the disease. Rather, there could be other (unknown) contributors which have a more significant influence on the disease’s mechanism. Furthermore, the intervention might have a detrimental on the patient if it disrupts other biochemical pathways that are (unknowingly) important in the disease’s mechanism.
Surrogate endpoints are often used in clinical trials to check the efficiency of a new intervention or treatment. However, it is important to note that misusing surrogate endpoints can mislead researchers to come up with wrong conclusions which could, in turn, place patients’ lives at risk. It is, therefore, important to take the benefits and dangers of surrogate endpoints into consideration when withdrawing conclusions from those clinical trials that use them.
1. Tassinari D. Surrogate end points of quality of life assessment: have we really found what we are looking for? HQLO.[Internet] 2003 Nov 24[cited 2014 Sep 3];1(1):71 Available from: http://www.hqlo.com/content/1/1/71
2. Fleming TR, DeMets D L. Surrogate Endpoints in Clinical Trials: Are We Being Misled ? Ann Intern Med.[Internet]1996 Oct 1[cited 2014 Sep 3];125(7):605-613. Available from: http://annals.org/article.aspx?articleid=710042
3. Epstein A E, Hallstrom AP, Rogers W J, Liebson P R, Seals A A, Anderson J L, Cohen J D, Capone R J, Wyse D G. Mortality Following Ventricular Arrhythmia Suppression by Encainide, Flecainide, and Moricizine After Myocardial Infarction: The Original Design Concept of the Cardiac Arrhythmia Suppression Trial (CAST). JAMA. 1993 Nov 24[cited 2014 Sep 3];270(20):2451-2455. Available from: http://jama.jamanetwork.com/article.aspx?articleid=409358