The Vaccine Tightrope

We’re getting closer to having to deal with a number of tricky issues around the first Emergency Use Authorizations (EUAs) for coronavirus vaccines. These have never quite come up in this way before, because (for one thing) EUAs for vaccines are relatively rare events, and (for another) we’ve never had so many simultaneous vaccine trials against the same disease before.

So let’s just stipulate that Somebody (be it Pfizer, Moderna, AstraZeneca, whoever) asks for an EUA before all the other Somebodies, and that this request is granted. I don’t know when this is going to be – December? January? Whenever. Someone is probably going to be first, and it could well happen somewhere around then. The actual date doesn’t matter so much as what happens relative to the event.

At that point, it means that there is a coronavirus vaccine that has been authorized for human use. Immediately we have all the rollout concerns about who gets it, in what order, where, how much vaccine is available and how it’s going to be distributed. We’ve talked a bit about that here, and there seems to be a lot of planning going on around these questions (as there had better be!) But what if you were in the placebo group of the very vaccine trial that led to the EUA? Will this trial be unblinded, or not, and will such participants be given the chance to be given the actual vaccine? What if you’re one of the (many thousands) of people who are in the placebo group for the other vaccine trials, the ones that have not yet been authorized for use?

Steve Usdin goes into these questions here at BioCentury (article is free to read). He notes this recent FDA guidance document that goes into some of these issues:

It is FDA’s expectation that, following submission of an EUA request and issuance of an EUA, a sponsor would continue to collect placebo-controlled data in any ongoing trials for as long as feasible and would also work towards submission of a Biologics License Application (BLA) as soon as possible. FDA’s recommendations regarding the safety and effectiveness data and information outlined below are essential to ensure that clinical development of a COVID-19 vaccine has progressed far enough that issuance of an EUA for the vaccine would not interfere with the ability of an ongoing Phase 3 trial to demonstrate effectiveness of the vaccine. . .

“As long as feasible” is a well-crafted way of putting it. But we need to ask just how long that might be. You can see from the latter part of the quote that one of the considerations for issuing an EUA at all will be whether it might fatally disrupt the ongoing Phase III, and indeed, the FDA goes on to say that they don’t consider a vaccine EUA itself as grounds for unblinding. This leads us to the various clinical trial designs and interim data analyses, and how they fit in with an EUA request. It’s important to understand what the interim data readouts are designed to be able to say, which (translated into English from statistics) is something like “the distribution of coronavirus cases observed so far is not inconsistent with the vaccine having an eventual efficacy at the end of the trial of at least X per cent, at a certain pre-defined confidence level”. This sounds rather less definitive than what I think the general public might imagine the unblinding of clinical trial data looks like, and I should also add that it can take longer than you’d think (once you have the unblinded data) even to be able to say that much. This is not like rubbing off a lottery scratch ticket, that’s for sure. Rarely does a trial read out in a way that you look at the raw data and say “Holy guacamole, I think that stuff kicked butt”, although it should be said that obvious butt-kickings in the other direction are somewhat more common,

But I’m not expecting any of the latter, to be honest. Clinical trial success rates for vaccines against infectious diseases are (according to these estimates) the absolute best in any therapeutic area in the whole industry. Now, that means that a full one-third of those trials are successful, as opposed to about 3% of the oncology trials (to pick the therapeutic area at the other end of the scale!) But we’re going to do even better than that. Thanks to the SARS and MERS epidemics, we had a real advantage on this coronavirus, in terms of what the most likely antigen might be for a successful vaccine, what to look out for in animal models, and so on. If we get some weirdo pandemic from a less-studied group of viruses, it’s going to be a lot harder, and let’s hope we never get a chance to find out just how much. No, I expect the current vaccines to all work to some degree, and the whole point of running the trials is to figure out what that degree is and how it compares to what we need.

Thus all the statistics, and thus all the hard decisions. We may get into a situation where an interim readout of the data show that a vaccine may well be working, but that granting an immediate EUA has a real danger of blowing the statistics for the complete trial. That is truly the worst outcome: ending up with something that might be useful, but being unable (despite all the time and money and effort) to able to say if it really is. We’ve got to avoid that.

Update: turns out that Pfizer is apparently already saying that if their vaccine candidate gets an EUA that they will vaccinate eligible placebo group patients. This is something that they’re going to have to thrash out with the FDA before any such authorization is given. . .

But the patients involved in all these trials may have other ideas. Each individual that decides to leave the trial protocol may feel that loss of their own data is not enough to affect the overall result, but if enough people think that way, that result will most certainly suffer – a tragedy of the clinical commons. It’s important to remember that the “tragedy of the commons” doesn’t have to happen every time it could: there are plenty of examples of it being avoided, and we have to make sure that this is going to be another one. The considerations run the other way as well – it may end up being incumbent on the trial organizers, from an ethical perspective, to break the blind and offer the vaccine to all participants. “We should have such problems” is my first reaction to that possibility, but we’ll have to make that call carefully, not ruling out such a decision but not leaping to it, either.

But let’s get back to the question of what happens to the other vaccine trials. In the same way that you can’t force the participants of the emergency-authorized vaccine trial to stay in it, you also can’t force the participants in the other trials not to get the newly authorized one. An additional problem is that I strongly suspect that many participants across all the US trials have a good idea of whether they’re in the placebo group or not. AstraZeneca has been giving a meningitis vaccine in the control group in the UK (and other countries?), but their US trial (and Moderna’s, Pfizer’s, and J&J’s) are using saline control. I’d have to figure that someone who got notable site-of-injection reactions knows they got the real vaccine, while someone who hasn’t (especially after a second injection) figures that they’re in the control group. What’s to keep the latter from going out and getting vaccinated, if another vaccine is available to them?

And how will all this affect the coming clinical trials for other vaccines? Novavax is an obvious example – they have a recombinant protein vaccine coming along, which is a method that hasn’t been into human trials yet for the current coronavirus, and has potential advantages for storage and distribution. Then you have various companies investigating things like nasally-administered vaccines, which also could have advantages that will only be made clear in large controlled trials. How will anyone run them, if there’s another vaccine (or two, or three) available? One answer, as the BioCentury article goes into (based on the FDA documents) is that you might have to switch over to using the authorized vaccines as the control group and run “non-inferiority” trials instead of placebo-controlled ones. That’s going to be a tricky gear shift, though. Here’s Usdin:

It is far from clear, however, that non-inferiority trials of COVID-19 vaccines would be feasible. Moreover, the difficulty of doing this could preface pressure on FDA to accept external control arms or real-world data as controls. Demonstrating non-inferiority to an effective intervention can require very large trials or the acceptance of large confidence intervals around the results. The placebo-controlled Phase III COVID-19 trials already include at least 30,000 participants, and there will be little acceptance of uncertainty about the efficacy of a vaccine to prevent a life-threatening disease.

These considerations are getting less hypothetical all the time. As noted in the BioCentury article, J&J is already openly asking the FDA and the other vaccine developers to work together and ensure that an EUA and unblinding event doesn’t create a de facto monopoly in the vaccine space. There will be a number of ways to get this wrong, and I’m glad that the issues are at least out there being talked about. Next come actions.

Update: see also Jon Cohen’s article here, which I had missed last week! He has a lot of good quotes from experts in this area that illustrate that even these people don’t always come down on the same side of these issues. Everyone can agree that we need to thrash these issues out, but the thrashing itself could get messy (and see the Pfizer-related update above for an illustration!)

Update II: signs that the FDA is worried about these issues and that it may well affect the entire idea of issuing an EUA at all.