The field of research on communication interventions is heavily littered with studies that are too small, too indirect, and too riddled with…
As results for the big phase 3 trials for the BNT-Pfizer and Moderna were looming, a debate ramped up: when would the thousands of people in their placebo groups get vaccinated? That debate got really intense when we saw how effective those vaccines looked, and how bad the Covid pandemic was getting.
How do you weigh up the value of the critical information you would get from continuing the blinded trials – we only had very short-term efficacy and safety data when they delivered first results – versus not penalizing the group of people who had volunteered? If you didn’t offer them vaccination, they’d be waiting up to two years, at the end of the vaccine queue. And while this debate was going on, a person in the placebo group of the Moderna vaccine trial died of Covid-19, making the stakes excruciatingly clear.
What’s more, you couldn’t completely control this, of course: some or many of the people on the trials would get vaccinated as the vaccines rolled out anyway, even if you didn’t unblind the trials. (Unblinding the trials to the participants means letting people know if they had the vaccine or placebo.)
Enter the compromise: turning big, regular randomized trials into crossover trials, and inviting the participants to join this new extended trial. “Regular” randomized trials are parallel trials: you randomize people into 2 or more groups, and they run alongside each other. The groups are the controls for each other.
In a crossover trial, instead of people getting intervention A or intervention B, and then staying in parallel, at a certain point they essentially swap places – they cross over to the other group. So what you’re randomizing is not which intervention people get – A or B – but which order they get them in: A then B, or B then A. Each person becomes their own control – and the groups are comparing immediate intervention with deferred intervention. (I’ve written more about this in an explainer at Statistically Funny.)
Now, this has real limits. And it’s not what you expect from vaccine trials. It’s done for treatments with effects that fade away – “wash out” – before you cross over. Turns out, vaccine crossover trials are not completely unheard of, though. I quickly found a couple from the 1990s – one tested a single-dose live cholera vaccine against placebo, and the other assessed the effects of influenza vaccination on asthma symptoms. Still, the scale of this – and the information we’re relying on them for – makes the first Covid vaccine crossover trial a real landmark.
I first saw this possibility discussed in early December last year, in a New York Times article by Carl Zimmer and Noah Weiland. Anthony Fauci proposed it as a way to get large-scale randomized data on how long immunity lasted, without leaving people on only placebos. The idea was if the group that was vaccinated earlier starting to get Covid-19 while the later-vaccinated group didn’t, you’d get the signal that immunity in the first group was waning. (Immunity is expected to wane, because immunity to coronaviruses generally does – though it’s not guaranteed, of course.)
Later that month, a preprint by Dean Follmann and colleagues was posted outlining the thinking in the “Operation Warp Speed” camp on going to blinded crossover trials. A few days later, statistician Steven Goodman presented on this – peer review, really – at an FDA meeting discussing it. Here’s what we could learn from a Covid vaccine trial crossing over to crossover:
- Large-scale data on duration of immunity (as already mentioned);
- More data will be gathered on subgroups of people, like elderly people and racial/ethnic groups;
- There are very few vaccinated people who got Covid-19 in the early results – with more, the theoretical risk that some could experience what’s called vaccine associated enhanced respiratory disease (VAERD or VAED) could be better assessed;
- Assessing the impact of boosters on people who already in a trial and ready to go would be quicker and easier; and
- More vaccinated people who got Covid-19 would help with the across-trials exercise of trying to establish correlates of immunity – we need to know if you can detect a difference in laboratory tests between the vaccinated people who were protected and those who still got sick (called “breakthrough cases”).
That last point is critical. If we can find out reliable markers in blood tests that tell you if a vaccine has worked, that means future studies don’t need to randomize so many people. That kind of large and long trial to get an efficacy answer won’t be viable once large outbreaks have been eliminated, and isn’t feasible for quick response to virus variants. As Follmann and colleagues put it:
Following a strong positive efficacy result, an immune correlates analysis may be underpowered due to relatively few breakthrough cases in the vaccine recipients. However, placebo crossover could effectively double the
sample size for assessing immune correlates of risk and protection, a notable benefit of crossover.
You can see the impact of going with very early results. The trials for the BNT-Pfizer and Moderna vaccines were designed to get an answer as quickly as possible, whereas the design of the trial for the Johnson & Johnson (J&J) vaccine set the bar for numbers higher. Here’s how that stacked up:
|Vaccine||Vaccine group||Placebo group|
What are the limitations of a crossover for these trials, though? Well, for starters, the groups after the crossover will be different at the outset, even if everybody signed up and everyone stayed blinded. That’s because, as they point out, the people in the first placebo group who were most vulnerable to Covid-19 would have gotten it, whereas those people in the vaccinated group going into placebo already have protection. Changes in the virus across time could make a difference, too.
These kinds of factors will have to be considered when results are eventually interpreted, and a lot of supporting data will be vital – like what’s happening with signs of immune response, not just who got infected.
So much for the theory. What actually happened? Which vaccines are going on this journey?
By the time this proposal was fleshed out and on the table, the die had already been pretty much cast for the first 2 vaccines: the pressure to offer the vaccine to people in the placebo groups was high.
BNT-Pfizer decided to offer everyone in the placebo group the vaccine at 6 months after their second dose, but participants could ask to be unblinded straight away and get vaccinated. Moderna offered everyone in the placebo group vaccine, telling the FDA they were considering a crossover option – this didn’t apparently go ahead. Both are now looking to the people in their earliest phase trials for assessing the need for boosters, but there were only a few hundred people each in those trials (more for Moderna than BNT-Pfizer). The picture for J&J is less clear. They reported to the FDA that they’d offered unblinding and vaccination, but at the time, relatively few had requested it (mostly in the US).
However, Novavax took the plunge, first with their UK trial, then the big US/Latin American trial. That’s about 45,000 people altogether. No word yet on how many signed up to keep going, and how many unblinded.
The Canadian company, Medicago, announced their 30,000-person phase 3 trial has been a crossover trial from the outset. The French company Valneva is reportedly considering it, when their inactivated vaccine gets to phase 3.
We don’t know how many people are continuing in the Novavax crossover trial. Here’s an excerpt from Kate Pullinger’s story on being a participant, that shows how complex this gets:
By early February, the trial had proven that Novavax was effective and began to move toward being authorized for use. When I went in for my blood test, I was informed that because of this success, everyone on the trial would be offered a second two-dose vaccine in April 2021, while remaining “blinded”: meaning, if you’d been given the placebo, you’d now be vaccinated, and if you’d been given the vaccine, you’d get the placebo. This was great news — in less than three months, I’d definitely be vaccinated.
Despite the good news about the trial, the atmosphere in the clinic was grim. The nurse told me he’d spent the previous month working in intensive care. Through the look on his face, I caught a glimpse of what he had seen. He kept his voice low and said, “If you’re offered the NHS vaccine before April, you must take it. Do not feel obliged to stay on the trial. The risk is too great.”
When the NHS offered her vaccination, she requested unblinding: turned out she had been in the vaccine group. And, as she writes, her anxiety had been unnecessary all along. It’s a reminder of just how much we owe all these trial participants. Meanwhile, the advent of major vaccine crossover trials may be yet another way Covid-19 changes aspects of our lives.
A giant, heartfelt “thank you!” to everyone who volunteered for Covid vaccine trials! And any other trial, too. We’d be lost in a dark, dark place without your vital contribution.
See also my post on whether being in clinical trials is a good deal for participants.
To learn a little more about crossover trials, check out my explainer at Statistically Funny. For in-depth reading, check out statistician Stephen Senn’s book, Cross-Over Trials in Clinical Research. This link will help you find it in a library near you.
Disclosures: My interest in Covid-19 vaccine trials is as a person worried about the virus, as one of my sons is immunocompromised: I have no financial or professional interest in the vaccines. I have worked for an institute of the NIH in the past, but not the one working on vaccines (NIAID). More about me.