March 24, 2021 - 3:10pm

I wanted to say a few things about the Oxford-AstraZeneca vaccine. The US National Institutes of Health have released a statement saying that AZ’s press release claiming 79% effectiveness in its US trial was based on “outdated information”; AZ have responded. It’s caused a big furore.

Here’s what I have to say. About 28 million people have been given the first dose of a vaccine in the UK so far. How many of them have had the Ox/AZ jab isn’t completely clear to me, but let’s say half. (It started later than the Pfizer jab, but we have much more of it, so that’s probably a reasonable guess.) So about 14 million people.

That is an extraordinary dataset. The Phase III trials of the vaccines had a few tens of thousands of people. The rollout is three orders of magnitude larger than that.

It’s worth noting that true clinical trials have other advantages — subjects are randomly allocated into having either the vaccine or the control, which means that the two groups should be similar. In the real rollout, it’s slightly harder to compare the two groups: they’re non-random and might differ in some clinically important way.

That being said, there is plenty of evidence showing that the Oxford/AstraZeneca vaccine is effective. A study looking at 7.5 million (!) over-70-year-olds in England found that 35 days after their first dose, patients were 73% less likely to develop symptomatic disease, and those that did were a further 37% less likely to be hospitalised. (Which makes it about 80% effective against hospitalisation.)

There was quite a lot of uncertainty around those exact numbers, but they match other estimates. A study of five million Scottish people found that recipients of the AstraZeneca vaccine were 94% less likely to be hospitalised with Covid than unvaccinated people. A Public Health England analysis from February suggests a single dose of the Oxford vaccine led to a 60 to 70% decrease in symptomatic disease, and a greater than 80% decrease in severe disease. The Pfizer/BioNTech vaccine had broadly similar outcomes.

Also, there don’t seem to have been any serious side-effects. There are reports of some sore arms and short-lived flu-like symptoms in a reasonably large minority of people, but the “yellow card” system for reporting adverse reactions has found that, apart from a very small number of allergic reactions immediately after vaccination, there is no evidence of danger, including blood clots.

So let’s go back to the Oxford/AstraZeneca issue. It seems to arise because the trial released data taken up to a prearranged cutoff date. The trial’s Data Safety and Monitoring Board (an independent body) thinks the investigators should have included another month’s worth of data which includes more cases and brings the efficacy down a bit, to between 69% and 75%.

I’m completely happy to believe that AZ have buggered up here. They have done repeatedly, with confusing press releases, weird accidental dosing regimes and so on. If they’ve released data early to make themselves look good, that is amazingly stupid and harmful. If they’ve done it through incompetence, then that’s still not great. Even if it was a prearranged cutoff, it would have been good to be ostentatiously careful about not overstating the effectiveness, and to release the most inconvenient data they have.

But I do not understand why the DSMB and the NIH would want to then publicise a minor disagreement — is it 79% effective or 69% to 75% effective? — when the outcome of this trial is already kind of unimportant? We have evidence from tens of millions of real-world patients; a 30,000-subject clinical trial can’t tell us much that’s new. Besides, it finds roughly the same as everyone else anyway, whoever’s version of events you believe.

From a utilitarian point of view, it seems amazingly likely to me that this announcement, by decreasing public confidence in an obviously pretty good and very safe vaccine, will do a lot more harm than it does good. Maybe the NIH and DSMB have some ironclad regulatory requirements that mean they have to behave like this, but my guess is that some non-trivial number of people will die as a result.


Tom Chivers is a science writer. His second book, How to Read Numbers, is out now.

TomChivers