February 9, 2021

It’s amazing how things become old hat. The “British variant” which kept us locked down over Christmas  was the terrifying newcomer only a few weeks ago. Now I feel almost nostalgic for it.

The “British variant” had many mutations, but its key one affected its ability to spread: it was about 50% better at jumping from person to person than the original Covid. More recently, different strains have arisen, with that same mutation — but also another one, which appears to make it harder for vaccine-primed immune systems to fight against. One of these arose in South Africa, hence the talk of the “South African variant”, but the same extra mutation has arisen twice, independently, in the British strain too. The idea of the vaccines not working, after all this time waiting for them to save us, is genuinely chilling.

So: what’s the reality? Do the vaccines work against the new variants? This week more bad news came in: a study from South Africa finding that the Oxford/AstraZeneca vaccine “fails to protect” against mild and moderate cases of the mutant strain first detected in that country. 

It’s worrying, of course, but there are complicating factors. I think there are some reasons to be less worried than the headline suggests.

First: this is an extremely uncertain finding. As the virologist Müge Çevik points out, the study was small: it had 1,700 participants, which sounds reasonable enough, but your study’s statistical power comes from the number of cases, not the number of people in the study. And there were only 42 cases of coronavirus in the control and the treatment arm put together.

Of those 42, 19 had been given the real vaccine, and 23 had the placebo. The study itself isn’t available, but screenshots taken from an online briefing say that it is believed to be 22% effective. But because of the small numbers, the uncertainty around them is huge: the 95% uncertainty interval stretches from 58% to -49% efficacy.

So, while it is very likely that the vaccine is less effective against the South African strain than against the more old-school varieties, it is still likely to give some protection against mild to moderate disease, and it is very possible that it gives much better protection than this study suggests.

An aside

I wanted to have a somewhat tangential, but serious, grumble at this stage. It is ridiculous that we only learnt about this study via a Financial Times story (and, later, a Times story). It is even more ridiculous that the only reason we have any of the data at all is because a couple of enterprising journalists took screengrabs from the briefing. 

There are all sorts of problems with scientific publishing. One of them is that peer review, as well as being flawed in many ways, is slow: it takes time for (unpaid) scientists to review a paper before it can be published in a journal. But the rise of the internet has meant that it is possible to avoid that problem, by publishing on preprint servers, and letting the peer review process happen in its own time. In the Covid era, that speed is vital; days or even hours might matter, might save hundreds of lives, if some discovery is made about whether some drug or intervention works or doesn’t work.

There have been tens of thousands of Covid-related papers published on the biomedical preprint server MedRxiv in the last 12 months; some of them have been absolute garbage (the XKCD comic reminds us that it probably makes sense to simply refer to them as “PDFs”, rather than “unpublished studies”), but at least it is possible for other scientists, and science journalists, to look at the data. 

So why on Earth wasn’t the South African study made available on preprint? It’s not that the paper hasn’t been written: the FT story was based on a pre-peer-review, i.e. preprint, version, and the slides were clearly written from a written-up paper. It’s just baffling to me. The Oxford vaccine is likely to save millions of lives; its scientific team should rightly be lauded as heroes, as should everyone involved in its manufacture. But communicating science is part of science, and this has been a big failure. I’ve asked the Oxford press team to see if they can help me understand it; if they respond with something interesting, I’ll add it in here.

Back to the point 

As I was saying. My main reason for worrying slightly less than the headlines suggest is because the headlines say that it doesn’t protect very well against mild to moderate disease. The study only looked at younger people — the median age was 31 — so there were no severe cases at all, in either the treatment or the control groups.

Intuitively, you might think that if a vaccine doesn’t protect against mild disease, it won’t have a chance against severe disease. But that’s not how it works. “Most severe disease is a consequence of someone’s immune system not dealing with the level of viral load,” says Stephen Evans, a professor of pharmacoepidemiology at the London School of Hygiene and Tropical Medicine. “If you drive the viral load down in someone they’re less likely to get severe disease.” It may not be enough to stop them from developing a PCR-detectable, or even a symptomatic, case of the disease, but it might stop them from ending up in hospital.

Or at least that’s the theory, according to our understanding of virology. But it’s not actually been easy to show in the case of Covid because severe disease is, mercifully, rare. Even in the main Lancet study into the safety and efficacy of the Ox/AZ vaccine, there were only 11 severe cases, including one death, among the 11,600 participants in the study. Ten of them, including the death, were in the control group, suggesting a greater-than-90% efficacy against severe disease; but the uncertainty interval (determined using an odds-ratio calculator) is huge, from 99% to 22%. Pfizer’s results are similar; Moderna’s are a bit more robust, but still hard to draw firm conclusions from.

Evans had an idea, though, which I thought would be interesting: what if you aggregate the results from the main vaccine studies? It would obviously have many problems — the various studies have slightly different endpoints; they looked at different populations; and of course they are different vaccines with different effects. And still the sample size is small: across the Pfizer, Moderna, AstraZeneca and Novavax trials, there were only 52 cases of severe disease or death, of which only two were in vaccinated people. (Notably, 29 of them came in the Moderna placebo group, so that vaccine’s results are overrepresented.)

But it would, nonetheless, be somewhat reassuring to find that the vaccines are, on average, more protective against severe disease than milder disease. I did that, and found that they are: 96% effective against severe disease, compared to 87% against mild or moderate. 

(Confidence intervals 99% to 84% for severe, 90% to 83% for mild/moderate; full data here. Prof Evans hopes to write a rather more professional version of this up as a short scientific paper in the future.. He tells me that the comparison between mild and severe disease is a bit more complicated than this, because the data aren’t entirely independent, but I think it’s OK for a crude comparison.)

I don’t want to make too much of this — the fact that those confidence intervals overlap mean we can’t be too sure that the effect is real — but it seems likely that Covid vaccines are better at keeping people out of hospital than they are at stopping them getting the disease at all. So even if the Oxford vaccine isn’t brilliant at stopping you getting the South Africa variant, I’d expect it to be somewhat better at stopping you dying of it.

This doesn’t mean that the South African data (or what we can see of it) isn’t bad news. The 22% efficacy finding is still unnerving; and even if it is better at preventing severe disease, there’s still plenty of room for “better” to mean “still not great”. And it is probably less effective at stopping transmission, so the disease can carry on spreading, and mutating, and eventually becoming even better at evading vaccine-primed immune systems. In separate bad news, the South Africa strain was found to be less susceptible to antibodies produced in response to the Pfizer vaccine, which doesn’t mean it won’t work – there’s more to the immune system than antibodies – but isn’t a great sign.

But on the other side of the ledger, we now have several approved vaccines that can easily be repurposed for newer variants – Sarah Gilbert, of the Oxford study, was saying on Andrew Marr’s show that they are already preparing a new version, just as they started preparing this one back in January last year. This time it won’t take so long to get approval – there are protocols in place for the flu vaccine, which changes its target slightly each year to keep up with new variants, and doesn’t need full-on Phase I, II, III trials each time. The mRNA vaccines are even easier to make new versions of, and new vaccines are going to keep on coming for the next couple of years. And we have all the factories up and running already. The vaccination process might have to go on a bit longer, but this isn’t a disaster.

Of course, by then there’ll be new new variants. The South African one will be a distant, nostalgic memory, which we look back on with inappropriate fondness. But the same facts will apply.