This is a republication of an article I originally wrote for the HealthCareBlog.
Life is busy, yet we somehow find time to stay engaged on social media, remain engrossed in the 24/7 news cycle, and continue our futile efforts to resist clickbait. While social media can allow us to mindlessly scroll through feeds, it also provides an avenue to provoke vigorous dialogue, however diverse, controversial, or even rooted in unfettered biases. These exchanges have served as the primordial soup for a virtual friend or foe-ships. Tense and argumentative Twitter exchanges are especially entertaining given the challenges in justifying a position in fewer than 280 characters. Thus, tweetorials have emerged to explain a point of view via a thread of comments since it is not always easy to do so in 1 or 2 tweets. The longer the tweetorial, the more heated the debate. What I am trying to get at here, somewhat obtusely, is the concept of surrogates.
I have already suggested a surrogate. Length of a tweetorial is a surrogate for degree of controversy of the topic. Meaning, length is a surrogate, a proxy. We are surrounded by surrogates. Longer wait lines at restaurants and bars imply a hipper joint or tastier menu. My child being extra nice to me is a surrogate for him wanting more time on electronics. Not a day goes by without folks arguing about surrogate endpoints. I wanted to dig deeper into surrogates and since I am a physician, I’m focusing on surrogates in medicine. Apologies to those who thought I would be discussing restaurants or exotic trips.
I want to make sure my definition of surrogates is accurate: Merriam-Webster dictionary for enlightenment. The first use of the word “surrogate” was in 1533, B.T. (“Before Twitter”). A surrogate is defined as “one appointed to act in place of another” or “one that serves as a substitute”. We use surrogate endpoints in clinical trials as a substitute for other end points.
How ubiquitous are surrogates? Interrogating PubMed with the phrase “surrogate endpoints” returns 831,224 articles. Don’t bother verifying – I’m sure you’ll get a higher number as academia relentlessly churns out papers involving “surrogates.” I conducted a very informal survey of what people think of surrogate end points. It won’t surprise you that surrogate end points are heavily criticized. But, at the risk of being shouted at on Twitter, or losing followers, I believe the rage against surrogate end points is a bit unfair.
The FDA defines a surrogate endpoint as
“a marker, such as a laboratory measurement, radiographic image, physician sign, or other measure, that is not itself a direct measurement of clinical benefit, and (A) is known to predict clinical benefit and could be used to support traditional approval of a drug or biological product; or (B) is reasonably likely to predict clinical benefit and could be used to support the accelerated approval of a drug or biological product in accordance with section 506(c)”.
Note, implicit in the FDA-definition of surrogate end-points is a recognition of their limitations. I understand surrogates can be misused and abused. I understand that sometimes the wrong surrogate endpoint is studied. I know that they don’t always correlate with real outcomes and that some endpoints might not be relevant to patients. Using the wrong surrogate endpoint can lead to overtreatment and harm. Nevertheless, surrogate endpoints aren’t always bad. I mean, I know that fast heart rate (like I am experiencing now) is not always a surrogate for an arrhythmia, but sometimes it is. Pitfalls of surrogate endpoints apply to the real endpoints. How many real endpoints have been abused and how many are truly meaningful to patients?
Measuring real outcomes, such as all-cause mortality, takes longer time than measuring a surrogate, such as median survival. Trying to show incremental improvement of outcome in a disease that has a survival of 90% at 5 years is challenging. A study attempting to show the efficacy of a drug might be irrelevant by the time it is published given evolving therapies and changing understanding of mechanisms of disease. It may take us years to show that lowering lipids reduces subsequent heart attacks in a patient who has already had a cardiac event. However, if we can correlate lower serum lipids with fewer heart attacks, perhaps all we need to show is a drug’s ability to lower lipids for a few months after initiation and demonstrate that this effect is lasting.
Using surrogate endpoints can save biomedical research precious dollars which can be allocated to answer other pressing questions. If the surrogate endpoint is selected wisely and studying it is successful, a potential new entrant to market might reach patients in need sooner. Is there no welfare gain getting therapies to sick patients sooner?
Having cheap, available, and easily measured and monitored surrogate endpoints is an advantage that is valuable for clinicians and researchers. I know that many will cite where surrogate endpoints have failed, but I want to remind us all where we have been successful. Overall survival in colorectal cancer correlates with relapse-free survival at 3 years. Event-free survival at 24 months’ correlates with better survival in diffuse large B-cell lymphoma. If a PET scan is positive after 2 chemotherapy cycles for Hodgkin lymphoma, progression and relapse are more likely. A PET scan that is positive at the end of treatment of follicular lymphoma predicts early relapse. Detectable serum PSA in patients who have undergone radical prostatectomy is a marker for impending relapse. We may choose not to intervene on a slowly rising PSA, but we must acknowledge that measurable PSA after radical surgery indicates pending relapse. If a regimen delays detectable PSA after such surgery, then it likely delays relapse, an important endpoint for some patients. And yes, I know one thing or more outside of oncology. Lowering HgbA1C in patients with diabetes mellitus to below 7% markedly decreases complications. The list is long on both sides, but we must manage this political divide (which can be tough with current climate) and try to find a way to move forward.
I will concede that intervening based on a surrogate endpoint might not benefit all patients but will always argue that some patients do benefit, and our goal is to identify these patients. My colleague and friend Vinay Prasad is no fan of surrogate endpoints (as his >24k followers on twitter will attest), but even he argues, along with Robert Kemp, that surrogate endpoints have a role if properly validated. Interestingly, if you type “surrogate endpoints” in Google’s search bar, this paper appears as # 5, so mentioning it is a strategic move on my part to generate a smile from Vinay and Robert.
While acknowledging their limitations, let us not throw the baby out with the bath water. I will argue that advances in medicine depend on finding the best surrogate endpoints for every disease, not in abandoning them altogether. I would love to see us approving regimens only if overall survival is improved but this is not possible as patients are thankfully living longer because of advances. We simply do not have the resources or the time to always achieve the real endpoint, and we owe it to our patients not to shortchange them by attempting methodological purity.
I might vanish from Twitter for few days to avoid backlash.