We have some more data to mull over with the Oxford/AstraZeneca vaccine. The situation so far has been pretty confused, with various efficacy numbers appearing from different people in different venues. It’s fair to say that the rollout of the clinical data has not gone smoothly, and that it’s done the effort no favors. As many will recall, the current big questions are whether a lower first dose of the two-dose protocol is more effective (as appeared from some of the earlier data) and what the interval between the two doses (lower dose or standard dose) should be, since the UK government has been looking at getting a higher percentage of the population vaccinated with the first shot by delaying the second.
The earlier report showed 54.9% efficacy in the group that got two standard doses four weeks apart (95% confidence interval of 32.7% to 69.7%). That number improved to 66.7% when the low dose/standard dose cohort was added in, because that smaller group itself showed 90% efficacy (95% CI of 67% to 97%). This new preprint reports on 1293 participants who had a 12-week interval between two standard doses. Efficacy in this group was 82.4% (95% CI of 62.7% to 91.7%), which would seem to be a notable improvement. That’s not a very large sample, and the confidence intervals between the four-week and the twelve-week group still overlap in the 60% efficacy range, but you can make a case (and AstraZeneca certainly is) that this shows better overall effects.
Another key piece of data is the efficacy seen during that 12-week period: 76% (95% CI of 59% to 86%), which is basically the same as when analyzed after the second dose. The preprint makes the point (and I agree with them) that the second dose is likely to be needed for longer-lasting protection, because it really does raise antibody titers significantly, but it certainly looks like the protection from a single dose with a longer delay is worthwhile, and that the delay will not hurt things (and may well make the overall efficacy higher).
Why should this be? The answer is “immunology”, and that’s not just the last refuge of scoundrels. Historically, it appears that longer delays in a two-dose regime can make things better, make them worse, or not make much difference, and the only way to be sure is to go out and get the clinical data. So even though this is not a large 12-week data set, I’m glad to see it. I think that the UK’s move to get as many first doses into the population as they can was the right one, and it’s good to see some data that at least don’t undermine it.
What about the low-dose/standard dose business, though? This preprint offers a possible explanation: it turns out that the cohort that got the lower dose at first also had a longer delay before getting the second dose. So it’s possible that the apparent increase in efficacy was driven less by the lower first dose than by the longer gap between the doses. We can’t rule out an effect from both, though – the data are just not in a shape to do that. Overall, the complaints that I (and many others!) have had about the data collection and rollout for this vaccine are still valid: we’re learning what could be important things about this candidate from analysis of small subgroups, some of which were themselves the results of mistakes and miscommunication during the trials. And the release of that data has been just as patchy and noisy – you really would have expected better from AstraZeneca.
But there is something good to say about their data collection: since the UK study that’s included in these numbers tested its subjects by nasal swab every week, regardless of any symptoms, we can actually get a read on something that everyone’s been wondering about: transmission. It’s become clear from all the successful trials that vaccination (whether by mRNA, the several different viral vectors, or recombinant protein) is extremely effective at keeping people out of the hospital and at preventing people from dying from the coronavirus. This is very good news, and it deserves to be highlighted. But are those severe cases just being converted to lesser ones, with other lesser cases then being converted to asymptomatic ones, and in that case has the number of people walking around shedding infectious virus really changed?
The swab data say that it has. It appears that the vaccine reduced the number of people showing PCR positivity by 50 to 70%. The actual numbers were -67% after the first dose and -54% overall, but I wouldn’t read anything into that difference, because the confidence intervals for those two measurements completely overlap. So it looks like everything is shifted: hospitalized cases end up being able to stay at home with more moderate symptoms, people who would have had moderate symptoms end up asymptomatic, and people who would have been asymptomatic end up not testing positive at all. Oh, and people who would have died stayed alive. There’s that, too.
If you just look at efficacy in preventing asymptomatic infection, you get a really low number (16% efficacy, confidence interval banging into the zero baseline). But my interpretation of that is that the overall number of asymptomatic patients didn’t change too much, because as just mentioned, the “would have been asymptomatic” group is not showing infection at all, and their numbers have been replaced by people from the “would have been showing symptoms” cohort, who are now just asymptomatic. And since transmission would seem to depend on viral load (among other factors), reducing viral load across the population (as shown by the significant decrease in PCR positivity) would certainly be expected to slow transmission. As Eric Topol noted at the time, this same effect had been noticed in the Moderna data in December. So with the numbers we have now, I feel pretty confident that yes, as one would have hoped, these vaccines also reduce transmission of the virus in the population. I believe that we should soon see this in a large real-world way in the Israeli data, where a significant part of the population has now been vaccinated.