November 10, 2020

The last few months have been grim. Not just the fear of the virus itself, but the various miserable intrusions we’ve all had to endure as a result — the lockdowns, the restrictions, the masks and the handwashing. The purpose of all these measure has been not just to stop the spread, but to buy time — to keep the virus down while we prepare our response.

The big response, of course, is vaccines. All the rest of it, all the rapid testing, track and trace stuff, can help control the virus with fewer restrictions on our daily lives. Only vaccines hold out the realistic possibility of getting rid of it altogether, and (eventually) returning to normal.

So it is potentially extremely good news that Pfizer-BioNTech have announced that early results from their trial show that their vaccine candidate, an mRNA vaccine, is around 90% effective. How close, then, is that return to normal?

Boris Johnson says that we have awaited “the distant bugle of the scientific cavalry coming over the brow of the hill” and that now “that toot of the bugle is louder”, but that it is still some way off. The turn of phrase may be tooth-grindingly irritating, but he is basically right.

In short: this is very exciting; the vaccine almost certainly works, although it would be nice to have a bit more clarity about what “works” means; and it’s worth remembering that, beyond just proving that it works, producing it in large quantities and getting it to where it’s needed are real challenges as well. But, mainly, it is completely okay to be excited, as long as you realise that it’s still a few months to wait. 

I think, though, we can be pretty confident that the vaccine actually works. It would be understandable to be sceptical because there has, as yet, been no scientific paper released, or even very much data, beyond one simple number: 94. 

That figure is the total number of cases in both arms of the trial. There were 43,538 people on the trial; according to the study protocol, half of them were given the real vaccine, and half a placebo. Three weeks later, all but 4,000 or so received a booster shot, of the same jab (either the vaccine or the placebo) they had before.

According to the Pfizer press release, there were 94 cases in total, and “the case split between vaccinated individuals and those who received the placebo indicates a vaccine efficacy rate above 90%”.

What that means is that the vaccine prevented at least 90% of cases, so there were 10 times as many in the placebo group as the real vaccine group. Say there were 100 cases in the placebo group and 50 in the other: that would mean that it was 50% effective; if there were 100 in the placebo and ten in the real vaccine group, that would mean it was 90% effective.

In this case, if there are 94 cases in total, there can’t have been more than eight cases in the real-vaccine group, and 86 in the placebo group, for it to be “above 90% effective”. And we also know that the vaccine is astonishingly safe, given no major adverse events in the 20,000 people who were given it.

Again, we have no further data, and it’s always sensible to be wary when ‘pharma company puts out press release saying pharma company’s product is great’. But the results weren’t released by Pfizer — they were put out by an independent monitor, the Data and Safety Monitoring Board (DSMB), so it would be pretty astonishing if they have just made this up.

What might seem more worth worrying about is whether this could be a fluke result. This isn’t, after all, the final analysis, and if you release data from a trial before it’s finished, it’s reasonable to be worried that it’s wrong. 

Let me explain, with this excellent analogy from Tim Harford’s new book How to Make the World Add Up. Imagine that two basketball teams are playing a game, and you want to determine which team is the best. You could just say “whoever wins is best” but it might just be a fluke. So you say beforehand: let’s draw an arbitrary line. If one team wins by a given number of points — let’s say 20 — we’ll say they’re the best. Otherwise we’ll say the results are “not statistically significant”. The higher that number, the more certain you can be when someone does reach it, but also the more times you’ll have to say “we don’t know” because they don’t.

More from this author
Will a vaccine stop Covid?

By Tom Chivers

That’s pretty much how scientific studies work. They say: “If the vaccine does better than the placebo, that could be fluke. So we’ll draw an arbitrary line and say if it does this much better, then it works well enough and we’ll use it.” The arbitrary line was 50% — so if there were 50% fewer cases in the real arm than in the placebo, they’d say it worked well enough.

But if, in the basketball game, one team was 20 points ahead in the second quarter, and then that team’s manager said: “Okay, let’s end the game there”, you wouldn’t think that was a fair assessment. Showing results from the trial early could be doing the same as that.

However, that’s not quite what has gone on here. The plan was always to release data after 32 cases were detected, then again after 62 and 92. (See Table 5.) For reasons that they don’t make clear, but which were “after discussion with the FDA”, Pfizer decided not to release its findings after 32 cases; by the time those discussions were complete, they’d already reached 94 cases. The study will continue until they reach 164 cases, but releasing early data was always part of the plan — so I don’t think it being released now should undermine our confidence at all. 

The other possibility would be that this result is a freak of some kind, that more people happened to get the virus in the placebo group than in the vaccine group. But that’s probably not a serious concern, and we can pretty much rule out the possibility that the vaccine does nothing. If it did, then seeing an 86-8 split would be as unlikely as tossing at least 86 heads in 94 coin-tosses, which is so unlikely that it’s essentially impossible. A one-in-some-number-of-quadrillions chance. 

The real number could be slightly more or slightly less than 90%, but realistically, not very much more. We’re talking no worse than 80%, at the outside. So I think it’s fair to say that the vaccine probably works, and is very effective. The next question is: what exactly does it do?

That might sound stupid — it stops Covid, of course. But what’s not completely clear to me is whether it stops Covid from infecting you or it stops Covid from making you ill

More from this author
Has Covid become less dangerous?

By Tom Chivers

In the trial I’m on, of the Oxford-AstraZeneca vaccine, I have to do a nasal and throat swab every week to see if I have the virus. So it’s very unlikely that I’d get infected and not know about it.

That’s not what happens in the Pfizer-BioNTech one. They seem to only swab people when they come in for follow-up appointments every few weeks or months, or when people get symptoms.

Robin Shattock, who runs Imperial College London’s own vaccine programme, describes it as “a symptom-directed endpoint that requires virological confirmation”. Nilay Shah, a professor of chemical engineering at Imperial, agreed: “My understanding is that the 94 cases must be both symptomatic and RNA-positive to be counted.”

What that could mean is that lots of people in the vaccine group are becoming infected — as in, some detectable amount of virus has entered their bodies and started replicating — but that they are not getting ill, so they’re not getting tested. Shattock told me that even people with mild symptoms, “runny nose, headache, fatigue etc”, would be tested, but still, it could easily miss some infected people.

Obviously it would be better if people were definitely being protected against infection as well as disease, since asymptomatic people can spread the virus. But even if the vaccine has literally no impact on infection rates at all, stopping 90% of symptoms is a big deal. What’s more, asymptomatic people probably spread the disease much less, and besides, it’s very unlikely that the vaccine will have no effect on infection. 

I spoke briefly to Vishal Gulati, a biotech investor and former virologist, and he said that even if the vaccine only protects against disease, it would be “nice to have”, but that he thinks it’s amazingly unlikely that it doesn’t have an effect on the viral load itself. “It’s just not possible that someone who is not sick is producing the same amount of virus over a significant period as someone who’s getting sick,” he said.

It may be that the vaccine is a really good one to give to older and more vulnerable people, to keep them safe, and that later vaccines that are more effective against infection should be given to younger people to stop the spread. According to the protocol, the study is doing antibody tests, so hopefully when the full data is released we’ll know more about the impact on infection as well as disease.

So it probably works well, and most likely reduces the effects of the disease as well as infection. The next problem is going to be making enough of it.

RNA vaccines like the Pfizer one, although completely untried, do have some big advantages. 

One is that making RNA is quite easy. If you have the ingredients – a big soup of nucleotide bases – you can get your length of RNA that you want, and it will essentially replicate itself. You can do it with relatively low-tech materials and at small scale, and each time you need a new length of RNA (for a different vaccine, perhaps), you can reuse the same facilities. Also, the doses are quite small, so you can make more doses per litre in your bioreactor. 

But you still need a lot of doses. Let’s be optimistic and say that it does protect against infection, and that you need to vaccinate 60% of the population to get herd immunity. (For a discussion of what “herd immunity” is, see here, but essentially, it’s the percentage of the population you need to vaccinate to get the R below 1. So if the R is 2, you need to vaccinate 50% of the population, so it’s twice as difficult for the virus to spread.)

In the UK, you’d need to vaccinate about 42 million people. But we’d need two doses per person, so that’s 84 million doses. And, remember, the vaccine is probably only around 90% effective, so about 10% of them won’t work. You’ll end up needing about 94 million doses to vaccinate the whole country to a herd immunity level. If we actually need 70% coverage, then that number is 110 million.

The UK government has ordered 40 million doses from Pfizer. That’s a big deal – it should be enough to vaccinate about 18 million people successfully. If we concentrated that on healthcare workers and the most vulnerable, then it would significantly reduce the number of deaths. But it won’t stop the spread of the disease itself. 

Of course, when the AstraZeneca vaccine is licensed, we’ll end up with another 30 million in short order and 100 million eventually. So UK herd immunity will probably happen. But even so, it is only a temporary reprieve if the virus is still endemic in the rest of the world. To immunise the entire world to herd immunity levels, you’d need about 10.5 billion doses.

Pfizer thinks it can make about 1.3 billion doses next year. Moderna, who are making another RNA vaccine, say a similar sort of figure. But what’s not clear is whether those two estimates take into account the fact that both companies will need the same raw materials; if both companies are ringing the same suppliers and trying to buy all their nucleotides, then one of them is going to be disappointed. I go into a bit more detail about the problems here.

Suggested reading
When will the Covid-19 vaccine arrive?

By Saloni Dattani

This is why we simply cannot make do with just the Pfizer vaccine. The Oxford one, for instance, is a viral vector vaccine, and won’t need quite the same chemical components as the Moderna and Pfizer ones, so their supply chains won’t interfere with each other so much.

There’s another problem which is that RNA vaccines such as Pfizer’s need to be kept at -80°C for all but the last day or so before they’re administered; Moderna’s needs -20°C. Other types of vaccine just need to be kept at normal fridge temperature, between 2°C and 8°C. This isn’t an insurmountable difficulty, especially in the UK, but it does mean that you couldn’t, say, bring a batch to a GP’s surgery on Friday and administer it on Monday, unless the surgery had a really high-tech freezer.

In the developing world, that will be an even greater challenge, but it is worth noting that the novel Ebola vaccine developed during the 2014-15 outbreak required a -60°C cold chain, and it was rolled out rapidly to hundreds of thousands of people in the middle of a war zone. These cold chains can be established fast.

There’s one more thing that’s worth mentioning. Donald Trump Jr, among others, is floating the possibility that the vaccine announcement has been delayed in order to help Joe Biden get elected. And you might be tempted to believe him – remember that, earlier on, we mentioned that Pfizer were going to release some preliminary results after 32 cases, and again after 62. But they didn’t end up releasing anything until they’d reached 94 – which, as it happened, was just a few days after Donald Trump Sr had lost the US election.

You can understand people’s suspicions. But it’s worth noting that Pfizer couldn’t have known that the results were positive, because they wouldn’t have had access to them. The only people to have access to the unblinded data – that is, the only people who knew which test subjects had the real vaccine and which the placebo – was the DSMB, the independent body supervising the study. And it’s not Pfizer or BioNTech who decided to release the data: it’s the DSMB.

Suggested reading
The trouble with 'Covid denialism'

By Saloni Dattani and Matthew Lesh

It’s not that it’s impossible that this panel of independent scientists were conspiring to help Joe Biden win the election, it just seems unlikely. Gulati thinks the most likely explanation is simply that the DSMB decided that releasing the data early, when the statistical power was still very low, would have reduced public trust in the vaccine. Also, it might have made recruitment for the trial more complicated, if it was obvious that the vaccine outperformed placebo — telling people there’s a 50% chance they’ll be in the placebo arm would not be hugely popular.

To sum up, then: this is a really big deal. It looks like the first vaccine will arrive soon, and that it will work – much better than expected, although with slight question marks over how well it “works” on infection as well as symptoms. And it’s really reassuring that vaccines work at all.

What it doesn’t mean is that we’ll all have the vaccine tomorrow, or even this year, or perhaps even next year. And it certainly doesn’t mean we can stop developing and producing the other vaccines, because each kind of vaccine will have different strengths and weaknesses — some will be easier to manufacture, some will work better in different groups. (There’s a useful Twitter thread on this topic from the CEO of the biotech company GreenLight.)

The best-guess timeline, that most people will probably not see a vaccine until late next year, is still reasonable. And, of course, it would be really nice to see the data from this Pfizer study, rather than just a few bullet-points and self-satisfied quotes in a press release.

Nonetheless, it’s exciting. It’s a light at the end of the tunnel. And finally, it gives a nice straightforward answer to what it is that lockdown is supposed to be buying time for.

Join the discussion


  • December 4, 2020
    The salient fact for me is that the "Useful Idiots" (Stalin's terminology) in bowing to the lobbyists are annuling product liability for the makers of vaccines. Read more

  • November 21, 2020
    I’ve had two respiratory infections since December, one relatively mild, the second more severe but neither half as bad as real flu when for at least a few hours you feel close to death. I’ve had flu twice in my life, first in 1989 when I was half the age I am now. No comparison with this... Read more

  • November 18, 2020
    They need wisdom to rebuild trust! Read more

To get involved in the discussion and stay up to date, join UnHerd.

It's simple, quick and free.

Sign me up