Open Mike Perspective: Healthy Skepticism when Focusing Solely on Surrogate Endpoints in Clinical Research


I recently wrote an essay for the NIH’s Science, Health, and Public Trust series to encourage a healthy bit of skepticism about clinical studies that solely involve surrogate end-points (e.g. changes in “biomarkers” like blood cholesterol levels or findings on an electrocardiogram).

To make my point, I described experiences with a well-known cardiovascular trial — one that focused on the risk of sudden death among heart attack survivors. I encourage readers in this essay to not just assume that treating a surrogate endpoint will automatically treat the underlying condition. Sometimes it does.  But sometimes it may not and may even cause more harm than good.

As I mention in the essay, one way to deal with this problem is to conduct more trials that focus on clinical endpoints. We should consider conducting them in innovative, efficient, cost-effective ways such as using pre-existing large-scale databases or “registries. I am open to hearing other thoughts and approaches which have the potential to address this challenge.

The NIH’s Science, Health, and Public Trust series is intended to provide perspectives, tools, and resources to improve the quality and usefulness of information about science and health for the public. It seeks to share strategies and best practices that might contribute to public understanding of the nature of biomedical research and its role in health.


  1. do large , pragmatic , SIMPLER RCT. NIH trials are too complex , collect too much data and cost too much. For RCT, the simpler, the larger , the better focus on a primary substantial endpoint. there are clinical networks worldwide that can contribute.

  2. Thank you Dr. Lauer for you thoughtful essay. One of the challenges all of us engaged in clinical research face is how to explain why a given intervention may help a subset of subjects in a trial who suffer from some complex syndrome. Rightly or wrongly this has led the field to focus on potential biological findings that we assume are associated with the pathophysiology of at least one of the causes of the syndrome we are studying be it myocardial disease, depression or Alzheimer’s Disease. I wish we would spend more time answering some of the clinically relevant questions that may not be as scientifically sexy but are important, be it how to prevent a recurrence of major depression, or can we develop remote monitoring devices that allow patients with Alzheimer’s Disease to remain in less restrictive environments, like their homes, for longer lengths of time.

  3. That being said, turning our back on informative biomarkers that provide mechanistic information leaves us vulnerable to not understanding how to improve further on therapies, or how to explain unanticipated adverse “off-target” effects.

  4. This is an interesting, and probably, controversial point. While it makes intuitive sense that a biological variable alone cannot act a surrogate for more clinically meaningful outcome, it is also logical that once a biological endpoint has been well-established as having a tight link with a clinically meaningful endpoint there would have to be good rationale for not accepting it – at least for early phase studies and in some circumstances registration trials. The bigger issue is the validity, precision, and stability of the biological marker and whether in fact it might be equal to, or superior than, verbally reported or even objective outcomes. So. I don’t think one can just dismiss this issue in a categorical way. One example might be: suppose someone with lung cancer has a chest lesion by X-ray or CAT scan that regresses during treatment, but say there is a genetic blood test or other marker proven to suggest activity somewhere in the body. The patient might feel good, but still has disease. Another example, closer to my work, is the idea that reported alcohol use cannot substitute for life/health changes measured in other ways, despite a wealth of research spanning decades that reductions in alcohol use from high levels to lower levels lead to improved morbidity and mortality. Furthermore reported alcohol use is likely to be at least 30% inaccurate in clinical trials – therefore the bigger issue is whether biomarkers of alcohol use can assist in improving validity of reporting. So, in this later case what is the surrogate, the self-reported alcohol use or the biomarker validating that report? This is a complex issue and one overlapping with FDA concerns. It is worth a high level discussion and not a dogmatic approach. Glad you brought this important topic forth for discussion!

  5. Thank you Dr. Lauer! I think we need to focus more on Surrogate Endpoints, but one thing is that while things are changing day by day, it’s easy to make mistakes or just after a short time then everything is not. applicable. That makes us always work continuously or fail

Leave a Reply

Your email address will not be published. Required fields are marked *