AstraZeneca/Oxford Calculations

By Derek Lowe

We have a lot of vaccine news to catch up on! First up is the latest on the possible clotting side effects seen with Oxford/AstraZeneca. This is looking more and more likely to be a rare-but-real effect, and I’m not happy that it’s holding up. I had hoped at first that this was a false signal, but as more data have come in, that possibility is fading. The EMA and others have been coming around to this view, and the question now is what to do about vaccine-induced immune thrombotic thrombocytopenia (VITT). Kai Kupferschmidt and Gretchen Vogel have an excellent overview on this here. As with all public health measures and all drug treatments, there’s a risk/benefit calculation at the heart of this matter. I gather from my email and messages that not everyone is reconciled to those, and while I sympathize – it would be a lot easier if we didn’t have to – I can only reiterate that it’s the only way to go. “You can’t put a price on a human life” is the saying, but while you may well not be able to put a price on yours or on the lives of your loved ones, that statement’s force dissolves when you have to consider millions of people.

Still not convinced? Try an unpleasant thought experiment, then, a sort of vaccination trolley problem: how many people should die of Covid-19 in order to keep one person from dying from a vaccine-induced thrombosis? There are all sorts of calculations imbedded in the answer – coronavirus incidence rates and fatality rates overall and in different cohorts, those same incidence and fatality rates for VITT across those different groups, the estimated effectiveness of the AZ/Oxford vaccine under current conditions in those populations, its availability (and availability versus other vaccine alternatives) and how many people you’re hoping to vaccinate in general with these supplies, the current state of coronavirus spread in the population that you’re looking at, and more. Complicating some of these (or making the comparisons starker, anyway) is the observation that thrombotic events also seem to be more common among people who are infected with coronavirus! You can obviously turn the dials on these numbers to get any answer you want by making enough assumptions. But a pretty solid one number to start with is that fatalities and serious complications from the coronavirus are more likely in older patients, and that tilts the scale towards them getting any vaccine rather than none. With that in mind, German authorities are saying that that AZ/Oxford vaccine should only be used in patients over 60. France puts the cutoff at 55, and the UK at 30 – other countries are all making their own calls, but the EMA itself is not making any recommendations like this (or not yet).

Why wasn’t this picked up in the clinical trials? Numbers, again. The thrombosis risk is still tough to estimate well, but might be roughly around 1 in 100,000. That means that in a clinical trial that doses some 20,000 people, you are simply not going to pick up this signal. No one has ever run a controlled trial (for a vaccine or for anything else) large enough to get good statistics on an incidence rate like that. You’re simply not going to see it, until you get out into a much larger population, which means after approval. That may not be fun to think about, but it’s been the case with every drug that has ever been approved. AstraZeneca and Oxford have made mistakes in the testing and rollout of this vaccine, but this problem isn’t on that list.

And to get all cold and clinical about it again, a one-per-hundred-thousand death rate can mean different things. It would be completely unacceptable for a drug that reduced common cold duration by two days, or one that cut down on the amount of hair growing out of your ears. But if you had a drug that was really effective against glioblastoma, pancreatic cancer, full-blast Huntington’s chorea or other such diseases, patients and families would be beating a path to your door, whooping and waving and shouting for joy and hardly giving that death rate a second glance. After all, nearly everyone with those conditions is going to die sooner rather than later and in relentlessly increasing pain and disability, so why not?

The calculation is nowhere near as stark for the coronavirus, of course, and the calculation is correspondingly harder to make, as detailed above. As we get more data, things will surely adjust. For example, it seems for now that women may be at greater risk than men – if this is true, then you could adjust in that direction. It may be that people with existing antibodies towards the PF4 protein are at greater risk, too, and that would be worth knowing about as well. More important questions that we don’t have answers to yet include whether this is a general adenovirus-vector effect or not. You could imagine that going either way. The Oxford/AZ vaccine uses a chimpanzee-derived adenovirus, whereas J&J uses a less-common human one (and the Gamaleya “Sputnik-V” vaccine from Russia uses that one and the more common adenovirus-5 in its two doses). There could be some protein-binding event that’s unique to the ChAdOx platform, or it could be more or less common across the other adenoviruses. There have been a few scattered reports of what may be VITT symptoms among people who have been vaccinated with the J&J shot, but the situation is far from clear.

And let’s not lose sight of all the ways in which this side effect is bad news. It has directly affected scores of people out of the millions vaccinated, of course. It has also thrown a number of national vaccination campaigns into turmoil, and that’s bad, too. One of the ways that’s happened is through making many people wonder about the safety of vaccination in general, and we certainly didn’t need a real reason for hesitancy thrown onto all the not-so-real ones. The AZ/Oxford vaccine was set to be widely used around the world, with huge numbers of doses being produced, so having this happen with it in particular complicates the timetable of vaccinating the whole planet. No, we didn’t need this at all.

https://blogs.sciencemag.org/pipeline/archives/2021/04/12/az-oxford-calculations