Skills Lab: Rethinking Surrogate Endpoints Approval

Medscape

3 days ago

This transcript has been edited for clarity.
Hello, everyone. This is Dr Bishal Gyawali, from Queens University, Kingston, Canada. Thank you for joining us in our Skills Lab video series. In the last lecture we discussed how progression-free survival (PFS) response rate and the surrogate endpoints are defined. We talked about the particularities with response rate and PFS, and in the adjuvant setting, with disease-free survival.
Today we'll talk about surrogate endpoints and their role in regulatory approval. One question that people commonly ask is: What strength of surrogacy is required for drug approval? Are different strengths of surrogacy required for accelerated approval (or 'conditional' approval) vs regular approval (or 'full' approval)?
We would expect that regular approval requires a higher bar of evidence as compared to accelerated approval. Unfortunately, there are no such published criteria, but based on the data we have seen, I think there are probably no unpublished criteria either.
If we compare the response rates of drugs that have received accelerated approval vs regular approval, then we don't see much of a difference. Also, in one of these studies from 2002 to 2021, we saw that 30% of approvals were based on response rates from single-arm trials that received regular approval rather than next approval.
You would believe that a response rate–based approval would usually be acceleratedor conditional approval so that the drug confirms clinical benefit in the future, but we found that in 30% of these cases, they received upfront, regular, full approval rather than accelerated approval. I call this upfront, premature, regular approval.
If you look at the trends of the FDA approval, you'll notice in recent years that the number of approvals based onsurrogate endpoints is increasing, while simultaneously, the number of accelerated approvals is decreasing. At first, it does not make sense because accelerated approvals are supposed to be surrogate-based approvals.
What this means is that nowadays we are approving more and more drugs on the basis of surrogate endpoints by giving them full or regular approval rather than accelerated approval, which I think is very concerning from a societal point of view.
If we are giving upfront, regular approval, then we can never be sure that they confirm clinical benefit because they don't have a mandate to do such a confirmatory trial. Also, even if these drugs were to fail in subsequent trials, we can't withdraw them. Only accelerated approvals get withdrawn.
I'll give you one example:the BOLERO-2 trial of everolimus in breast cancer. As you can see, the first graph is PFS, and there was a 6-month improvement in PFS, with a hazard ratio of 0.36, which led to the drug's approval. It was not accelerated approval; it was full approval.
In just 2 years, the overall survival results were published. As you can see in the second graph, the overall survival was negative. Overall survival did not improve, but because it was a full approval, the drug was not withdrawn. I call this premature approval, and we should be asking why we are granting full approval based on surrogate data.
One more example: Margetuximab in breast cancer was approved on the basis of PFS improvement. You can see how much PFS improvement there was. It was 5.8 vs 4.9 months, so it was just 0.9 months, or less than 1 month, of PFS improvement.
This drug was given full approval, not accelerated approval like it was supposed to. You can look at the date. This drug was approved in December 2020, and just 9 months later, in September 2021, we received overall survival results from this trial. The drug failed to improve overall survival.
If we had not given regular approval, then we would have withdrawn this drug, but because it was given premature full approval, the drug was not withdrawn. It took only 9 months to get overall survival data, so why did we not give accelerated approval? That is something we should be asking.
We recently published a paper in JCO about why PFS should not be used as a primary endpoint for registration of anticancer drugs. We list several reasons why PFS fails to capture clinical benefit in these trials. I think we should at least agree that, if we're approving drugs based on surrogate endpoints, it should be accelerated or conditional approval and not full approval.
The other common question is: If PFS did not lead to overall survival improvement, did it not improve quality of life? We tried to answer that question. First, of course, there is no reason for PFS to correlate with overall survival. As we discussed in the last lecture, nothing magical happens at 20% and not all progression events are the same.
Some progression events are more symptomatic than others, therefore the quality-of-life effectis not always predictable. Quality-of-life impact is not only a function of how big the tumor is but also a function of drug toxicities.
In recent years, we have seen that PFS has become the default primary endpoint. In the past decade or so, the frequency of trials using PFS at the primary endpoint has increased substantially and now exceeds the percentage of trials with overall survival as the primary endpoint.
The strange thing about PFS is that we talk about correlation of PFS with overall survival and how we need to do correlation analysis to confirm that this is a valid surrogate endpoint.However, as we saw with the case of bevacizumab in breast cancer, even after eight clinical trials were conducted, was PFS a good surrogate for overall survival in this setting? The answer was that we do not know. Maybe, maybe not.
What I'm trying to say here is that, even after eight clinical trials of the same drug in the same setting, we are still unable to conclude definitively whether PFS is a goodsurrogate for overall survival. This means that it is far more expedient to just measure overall survival than to continue to do multiple trials and continue to guess whether PFS is a good surrogate for overall survival.
Thank you very much. I'll see you in the next lecture, in which we will talk about physicians' and patients' understanding related to surrogate endpoints.