Landmarks, Surrogates and Strategies for Accelerating Clinical Trial Results

The standard of care for castration-resistant prostate cancer (CRPC) has changed so much in the past decade that it is sometimes hard to keep up with what might be coming down the pipeline. The TROPIC Study, when it became public back in about 2011, was a complete surprise to me. The first I heard of the study was when the trial reported out the survival data and almost overnight we had a new chemotherapy option for people and a series of potential new clinical trials. The ALSYMPCA radium trial was similar. I really didn’t see it coming.  Of course, that was because I wasn’t really involved in the design, conduct or enrollment in any of these studies.

There are now so many studies that are “out there” it is likely that there will be some surprises. In the hormone-sensitive prostate cancer space, new trial results – like the LATITUDE, STAMPEDE and CHAARTED trials, have proven to be particularly challenging to predict.  One of the challenges is the long follow-up period of studies in HSPC. Most patients obtain disease control with ADT alone or ADT plus abiraterone, often for several years. Thus, waiting for the data to mature to the point of demonstrating a difference between two treatment arms can take a long time.

In the classic model of drug development in advanced cancer, it is customary to ask whether the treatment has a beneficial effect on overall survival and the readout for that is to look at MEDIAN survival – the time until 50% of the subjects on either arm have died from any cause.  The (good) challenge that we have is that the median OS for mCRPC is about three years and increasing. It is much longer in the hormone-sensitive setting – between 4-8 years depending on the risk level of the patients.

Can we overcome this time barrier and design trials that demonstrate clinical benefit in a shorter period of time? We could use earlier timepoints and surrogate endpoints. Here’s some thought on how we could do that.

  1. Surrogate Endpoints: The best example of a surrogate endpoint that would change our standard of care is progression-free survival. An ongoing conversation in the field is whether PFS is a true “surrogate” for overall survival or not. It is generally accepted that PFS is a reasonable surrogate endpoint for OS in the mCRPC setting, as this was the basis for the expanded approvals of enzalutamide and abiraterone. In the case of abiraterone, we actually did a formal analysis of how closely the PFS and the OS data were in a concordance and the correlation coefficient was 0.7- suggesting that a finding of an OS improvement is highly and positively correlated with overall survival. 
So that means if we do a trial and we demonstrate an improvement in PFS for one therapy, there is likely an improvement in OS. That might be enough to get a drug approved. Especially if it is an expanded approval as occurred with Enzalutamide and Abiraterone in the chemotherapy naïve studies.

Formal statistical criteria – the Prentice criteria- exist for establishing a link between two endpoints and establishing surrogacy. They are like a Koch's postulates for drug development. The short version is you have to prove that your endpoint ( e.g. decline in viral load in HIV patients) is caused by the drug and that the survival benefit is a function of that change in endpoint (drop in viral load IS the response) and you have to show it prospectively. Viral load in HIV is a bit like PSA in prostate cancer, you can follow it in the short run, but there is a key difference ( outlined below)

  1. Landmark analyses:
A landmark analysis would be another way for us to shorten the clinical trial timeline. Note that in the LATITUDE study that OS was the primary endpoint but an analysis of the outcomes at 2 or three years after randomization demonstrate clear benefits to the use of abiraterone. So, one could conceivably design a study that looks at PFS but not the median PFS but rather looks at the three-year progression-free interval. That way you wouldn’t have to wait for the median time to anything – in other words, you could read a trial out as ‘positive’ well before 50% of the patients had died or experienced progression. If, in the scenario, only 10% of patients on the new treatment had experienced progression of disease after 2 years whereas 35% had on the standard treatment in that time frame, then one could conclude that the new treatment was effective, even though the ‘median time to progression’ had not been reached, and your study only took two years to read out.

As stated above, one surrogate endpoint that we do not have at our disposal is anything to do with PSA, as you may have noticed. This was the discussion of the prior decade ( in fact almost two decades ago now) in which the question that was posed was whether a 50% decline in PSA was a surrogate for improved survival. It met many, but not all, of the Prentice criteria for surrogacy. 

In mCRPC studies we have shown that patients who experience a 50% or greater decline in PSA on a given therapy are more likely to survive longer – and in practice, I follow this logic, as do most clinicians I’m sure. If the PSA is down it generally indicates that the disease is under control. 

The challenge of surrogacy has to do with the fact that not all therapies that improve survival do so with a concomitant reduction in PSA (examples are Radium and Sipuleucel-T) and the biology of prostate cancer is such that some patients can experience disease progression and death even when the PSA is down. This is particularly due to the transformation of the disease to neuro-endocrine or small cell phenotypes or other forms that may not drive up PSA. 

Keep in mind that surrogate endpoints and landmark analyses are different. You can use a landmark analysis ( e.g PFS at three years) as a surrogate endpoint. But to do so you would have to show that PFS at three years is associated with an improvement in median survival – and preferably do that prospectively. The way to do that is to formally embed these analyses into the design of trials we are putting together today while at the same time analyzing in great detail the trials of yesterday. Both of these things are happening.

So, there is continued movement in the development of new therapies for prostate cancer and we can expect to hear more about various studies, positive or negative, in the coming year. As a clinical trialist, the possibility of a bombshell trial result dropping in front of us is a continuous challenge, but a good challenge to have. We are faced with designing studies that are based on improving on the existing standard of care with the full knowledge that the ground may shift under our feet in the form of a new standard of care “control”. I’ll take that challenge on gladly.

Written by: Charles Ryan, MD, B.J. Kennedy Chair in Clinical Medical Oncology, Director and Professor of Medicine, Division of Hematology, Oncology and Transplantation, University of Minnesota, Minneapolis, Minnesota

Published Date: July 30th, 2018

Further Related Content: 
Watch: Practice Changing Events in Prostate Cancer- Charles Ryan