October 21, 2020

If there’s a finish line in sight for the grim marathon that has been Covid-19, it’s a vaccine. It will be – we all hope – the thing that signals a return to some sort of normality, eventually. And there have been some promising headlines recently: notably, the Oxford-AstraZeneca vaccine has started production, with the UK deputy chief medical officer saying a rollout could start before Christmas, while Pfizer says that it hopes to file for FDA emergency use approval for its own vaccine within the next few weeks.

I don’t want to downplay these developments. They’re important, and if one or both of those is authorised, then we could start seeing people being vaccinated relatively soon. But I did want to talk about some of the difficulties and complexities involved. It is perfectly possible that one vaccine is already in production, and that another will soon be approved for emergency use, and that still, you, ordinary British person1 will not see either for over a year; and that most of the world will not see one for another year after that.

Like what you’re reading? Get the free UnHerd daily email

Already registered? Sign in

First, let’s talk about the different kinds of vaccine. All vaccines work on one basic principle, which is presenting the immune system with something (an “antigen”) that looks like the thing you’re vaccinating against, usually a virus. But there are different ways to make that antigen.

There are, depending on how you want to break it down, essentially four different kinds. The traditional method – which goes back to Louis Pasteur – involves getting hold of the actual virus you’re vaccinating against, killing or weakening it somehow, and using that as an antigen. These are known as “killed virus vaccine” and “live attenuated virus vaccines”, but I’ll refer to them both as traditional vaccines.

The second involves using a part of the virus, a “subunit”, as the antigen. In the case of coronaviruses, the part involved is the spike protein, the lumpy bits outside the body of the virus. In modern subunit vaccines, you can use a genetically modified virus to infect yeast cells and get them to produce the protein, so it is faster.

The third and fourth, viral vector vaccines and RNA vaccines, also both use the protein spike as the antigen, but instead of producing it in a factory, they recruit the body to do that for them. Viral vector vaccines use a genetically modified virus to infect the body’s cells and make them produce the spike protein, just as the actual virus co-opts the body’s cells to force them to make copies of itself.

RNA vaccines do much the same, except instead of using a genetically modified virus, they simply use tiny lengths of RNA, which – once snuck inside a cell membrane, in a tiny blob of fat called a lipid nanoparticle – commandeers the cell’s machinery and starts it making the spike protein.

All of these have different strengths and weaknesses. Traditional vaccines are reliable; they’re a mature technology and tend to give good immune responses. But they are slow to make – they need to be grown in insect cells or hens’ eggs, so a batch can take a couple of weeks. The exact process for each one is subtly different, so you can’t easily repurpose a factory. And it takes a lot of infrastructure – a lot of steel in the ground – to make a factory.

Subunit vaccines are pretty well field-tested now; they’ve been around for a few decades. They’re somewhat faster than the traditional ones to make, and the factory hardware is more interchangeable because you can get your yeast to produce any protein you like, but they still need large vats.

Viral vector vaccines, like traditional ones, require the growth of actual virus particles in some biological substrate, such as hens’ eggs or mammalian cells. But unlike traditional vaccines, you don’t need to make a different virus each time – you can just grow the same virus (in the case of the Oxford-AstraZeneca vaccine, a member of a family of viruses called adenoviruses) for each vaccine, and paste in the genetic code for the particular protein you want it to build. They are, however, very much untried – as far as I know, only one has been used in humans before, for protection against Ebola.

RNA vaccines are much faster to produce – it’s “a chemical rather than a biological process”, according to Al Edwards, a professor in biomedical technology at the University of Reading. They also require far smaller doses, so you can get many more out of, say, a 200-litre bioreactor, and, like viral vector vaccines, you can paste in the RNA sequence for whatever protein you need: you can use the same equipment for a vaccine for any given virus. It can also be very small scale, with relatively low-tech equipment; Imperial College is planning to make a small demonstration plant in rural Uganda. But RNA vaccines have literally never been used before, and, inconveniently, while all the other kinds of vaccine can happily be kept at normal fridge temperature (2°C to 8°C), the lipid nanoparticles that are used to get RNA vaccines into the machinery of the cell are very unstable, and need to be kept at -80°C.

There are about 200 vaccines currently in development for the SARS-Cov2 virus. All of the above kinds of vaccines are represented. However, the two we mentioned at the top, the Pfizer and Oxford/AstraZeneca vaccines, are the two most advanced – both are in Phase III trials (that is, large trials to show that they are not merely safe in humans, but actually work). They are an RNA vaccine and a viral vector vaccine, respectively. Moderna, Johnson & Johnson, and Novavax also have vaccines in Phase III trials, but are not, I think, quite so far along.

As a bit of an aside, something strange has happened with the Oxford/AstraZeneca vaccine. Two patients got ill. The first was with MS, and the vaccine study declared on its consent form that it was unrelated, so the trial continued after a brief pause. The second was transverse myelitis, an inflammation of the sheath around the spinal column; again, the trial paused, and, while it restarted in the UK, Brazil and South Africa, in the US, it did not.

suggested reading
The paranoid style in vaccine science

By Tom Chivers

That is extremely unusual, according to Stephen Evans, a professor of pharmacoepidemiology at the London School of Hygiene and Tropical Medicine. Usually, drug trials are monitored by independent groups called Data and Safety Monitoring Boards (DSMBs). If a patient gets ill, these boards look at the situation – often, as in the AstraZeneca case, pausing the trial to do so – and decide whether the illness was caused by the vaccine, and, if so, whether the risk outweighs the benefit. If it does, they might recommend to the regulator that they end the trial.

But normally you only have one DSMB for the whole trial. In this case, there are apparently two – one for the UK, Brazil and South African parts of the trial, and one for the US. And while the UK one decided it was safe to carry on with the trial after a brief pause, the US one has not. The US Food and Drug Administration (FDA) is now investigating it and not allowing it to restart. “It’s a very strange thing to have two different DSMBs for the same vaccine,” says Evans. “I’ve not known it to happen before.”

Andrey Zarur, the CEO of the RNA biotech company GreenLight – who have their own RNA vaccine in animal trials, and for whom, full disclosure, I’ve been doing some writing over the past few weeks – thinks that the UK regulators have been very lenient. “AstraZeneca have had very favourable treatment in the UK,” he says. “They say the myelitis is unrelated, but the FDA says BS.” It’s a fairly rare condition, and he thinks it’s likely that it’s connected; he thinks “you may get partial approval” in the UK, but it won’t be approved in the US, at least not soon. Evans is more hopeful, and thinks that the FDA is being “super-cautious”. He notes that a previous vaccine candidate developed by the Oxford team for the MERS virus also saw a serious adverse event. That one was also declared unrelated, but Evans speculates that the coincidence might be enough to make the FDA more wary.

Either way, AstraZeneca are confident enough that they’ll get approval – in the UK, at least – to start production; the UK government has ordered 30 million, with an option for 70 million more. And if Pfizer does get enough data to apply for emergency use approval, they’ll start manufacturing too, at least at small scale.

So let’s talk about manufacturing. All the vaccines currently under development look like they’ll need two doses each. It’s one thing making 10,000 or 20,000 doses for a clinical trial – in order to get the world to herd immunity, even assuming that the vaccines all bestow complete immunity after the booster shot, we are going to need to vaccinate more than four billion people, twice each. Eight billion doses is an awful lot. A normal bioreactor for traditional vaccines might be 200 litres and able to make a million or so doses in a period of weeks; you’ll need a lot of them.

You can’t simply use a bigger vat, says Edwards. It’s like making a cake: if you can make a cake 20cm across and 10cm deep in your oven, that doesn’t mean you can make a cake 20 metres across and 10 metres deep in a much bigger oven. “It would be burnt on the outside and soggy in the middle.” Making live viral vaccines is a biological process, with lots of steps that depend on precise temperatures and chemistries. “The bigger your batch, the more you have to change your system of production,” he says, “so if you’re making it in a big batch, you have to check you’re still making the same thing. One disaster is that if you make it a different way, it can have a different effectiveness or even not work.” You can make lots of smaller reactors, but obviously that is expensive and difficult.

That’s true for traditional virus vaccines, but it’s also true for the modern viral vector ones, which similarly have to brew up a virus. “Making an adenovirus particle has never been scaled,” says Zarur. The only medical treatments which use similar technologies, gene therapies, have only been used on “single-digit thousands” of patients, he says. AstraZeneca says that it can create two billion doses by the end of next year, but Zarur is sceptical: “there just aren’t enough reactors in the world” as it stands.

RNA vaccines can be made more easily, and because they need much smaller doses, you can get more doses out of the same-size reactor. But Zarur points out other problems. Moderna and Pfizer, both developing RNA vaccines, have each said they can produce around a billion doses by the end of next year. But, Zarur says, that’s an estimate based on how much of their raw materials they can get hold of. “When Moderna says they’re going to make a billion doses by the end of 2021,” he says, “it’s about calling on suppliers and saying how much nucleotide can you make.” And the suppliers of nucleotides might say, we can build enough to make a billion doses. “But those are the same suppliers that Pfizer is calling! Those materials, the nucleotides and the enzymes, are in very short supply.” RNA vaccines, he says, have enormous upsides, but there just isn’t a supply chain in place for them yet.

It is, Zarur says, “a twist of fate” that “the vaccines that are in the lead are also the ones with the least developed supply chain”. The Coalition for Epidemic Preparedness Initiatives (CEPI), the global body established to prepare vaccines for new viruses, estimates that even putting all the pharma companies together, globally there is only capacity to make between two and four billion doses by the end of 2021; even in the most optimistic scenario, we will only have about half of what we need.

(It’s worth noting here that the Oxford-AstraZeneca vaccine trial will be declared a success if there are half as many infections among people given the real vaccine as among those given the control — so if there were 100 in the control arm and 50 in the vaccine, that would be enough. This means it may be that “successful” vaccines only prevent 50% of infections. If that were the case, we’d need to vaccinate a lot more than half the world.) 

Manufacturing the vaccine isn’t the end of the problem. “After you’ve made billions of doses,” says Edwards, “you have the opposite problem: we’ve got factories making tonnes of these vaccines, now you’ve got to get them into little tubes, stick them in a fridge, get them all around the world and put them in needles and stick them in someone’s arm.”

It is a huge logistical challenge. Recently, I spoke to Toby Peters, a professor of cold-chain economy at the University of Birmingham, and Nilay Shah, the head of the chemical engineering department at Imperial College London, about the difficulties of mass rollout of vaccines. At the moment, we vaccinate only a subset of the population at a time: small children, vulnerable people, pregnant women. Now, though, we need to vaccinate everyone, fast. To get a sense of how much the system needs to be scaled up, at the moment, about 50 million vaccine doses – for measles, polio, things like that – are distributed every month in India. There are nine different vaccines, so that’s about 6 million people. But, he reckons, we’d need to get that up to about 200 million people a month. The vaccines need to get from the factory to the clinic to the patient without spending more than a day or so at room temperature.

This “cold chain” delivery is a significant enough challenge if you’re working in a city, but to get to people living in rural parts of India, or sub-Saharan Africa, you will need infrastructure. You’ll need trained people to deliver it, especially if you’re putting several doses in a single vial, which will save money but will mean the healthcare workers need to know how to safely take the right dose out while keeping the remainder sterile and secure. And the vaccine needs to reach a majority of the population. So, Shah points out, you need to get it to them – it needs to be walking distance at most for everyone. We can’t expect billions of people to travel long distances; in underdeveloped areas, that might mean mobile vaccination stations in trucks.

Peters points out that millions of tonnes of food are transported chilled around the world, and that that cold chain is extremely effective and has huge capacity, so it’s not impossible; but it will require a major effort, not least to find out how many fridges and things are available in the developing world, and second to design a cold-chain system, separate from the existing vaccine cold chain. Even in the UK, this is going to be a non-trivial problem. “One thing that bothers me slightly,” says Edwards, “is that a mass vaccination programme is as difficult as a mass testing program, and that” – this last with admirable understatement – “hasn’t always run as smoothly as we hoped.”

Again, RNA vaccines can be made at a smaller scale, so more locally. But then if you want to keep them for more than 24 hours or so, you need to keep them at -80°C. “You can do that in London or New York,” says Zarur, “but how are you going to get it to Laos or Sao Paolo? And unless you vaccinate those populations, you’re wasting your time.” He is hopeful that the second generation of RNA vaccines, including GreenLight’s, will be made in small, almost popup factories anywhere you can get some cheap chemical ingredients and an electrical supply, and then in the 24 hours you have before it degrades, you can get it to where it needs to be. “You couldn’t get it from Cambridge, Massachusetts to France,” he says. “But you could get it from Mexico City to Chiapas.” But that won’t be true of the first-generation ones, which will require more complex and expensive ingredients.

None of this is to say that it’s not important that vaccines are starting to roll off the shelf. It is. Even if only a few million doses are built at first – even if, as Zarur says, “it’s enough for the football players, the Trumps, and the healthcare workers” – that’s a big deal. As he points out, in the first wave, “healthcare workers were dying in the hallways”, and the shortage of medical professionals caused by illness was part of the reason it killed so many others – people weren’t getting treated properly. And as we vaccinate the most at-risk populations, the susceptible population will be reduced, so the virus will find it harder to infect people, and the death toll will come down, and slowly the situation will improve for all of us.

But it absolutely will not be: we get a vaccine in December and we’re out of this by February. The light is real, but the tunnel it’s at the end of is still pretty long.

  1. I’d say “you and I”, but since I’m on the Oxford trial, I’ll get the real vaccine if/when it’s declared efficacious, if I haven’t had it already.