By Grant Jacobs 15/03/2021

Some nations are pausing their AstraZeneca COVID-19 vaccine campaigns over concerns that it may be linked to blood clots. International media is is reporting it widely. I’m seeing concerned people asking questions on social media. Evidence so far indicates this is unlikely to be a real issue, but one that is being checked out of caution.

One thing it illustrates is how helpful it is to have baseline data when rolling out major health efforts like these vaccination campaigns.

Baseline statistics can head off potential ‘scares’ by telling us what are real concerns, and what are not. If after a treatment the same or fewer people have the ill-health event as those without the treatment, it’s unlikely the treatment is causing it. In New Zealand, a research group lead by Helen Petousis-Harris is working on baseline data for NZ’s vaccination effort.

In the case of these blood clots, early reports say that the number of cases found are fewer than what would be expected in the population-at-large.[2]

This is not the vaccine we are rolling out in NZ. The vaccine we are currently using is the BioNTech/Pfizer vaccine.


So, what happened?

Several nations have paused using the Oxford/AstraZeneca vaccine because of reports of a small number of cases of blood clots in the days after vaccination.

Austria reported in total four cases of thrombosis (blood clots), the same batch of 1 million doses. One of these people died. Similarly a small number of cases have occurred in other countries.

People get blood clots for a wide range of reasons. That people got a blood clot does not mean “it’s the vaccine that done it”.

In fact data reports suggests the number of people with clots who have been vaccinated is actually lower than number of people of comparable age, etc., who have not been vaccinated.

In any event, several nations have temporily suspended use of the vaccine. These include Denmark, Norway, Iceland, as well as Estonia, Lithuania, Luxembourg, and Latvia. More recent additions include Ireland, Romania, Bulgaria, Thailand, and the Netherlands.

Some of the concern is about the particular batch of the vaccine, rather than the vaccine in general. This is also being checked.

The European Medicines Agency (EMA) has released findings showing there are no indications these events are caused by the vaccine.

They find 30 cases have been reported amongst the nearly 5 million people in Europe who have had the AstraZeneca vaccine. This rate is not higher than for people not given the vaccine.[1]

Vaccine programs are being interrupted. For example, in the Netherlands they have cancelled 43,000 appointments at short notice.

You also have to weigh in the small risk of blood clots against the risk that in pausing the vaccine rollout some people will suffer or die from COVID-19. This is particularly true as most of the people currently being vaccinated are those more vulnerable to COVID-19.

When you think how this is disrupting a very important public health effort, it emphasises how useful it is to have baseline data in advance.

Update from AstraZeneca

In a recent release, AstraZeneca notes –

A careful review of all available safety data of more than 17 million people vaccinated in the European Union (EU) and UK with COVID-19 Vaccine AstraZeneca has shown no evidence of an increased risk of pulmonary embolism, deep vein thrombosis (DVT) or thrombocytopenia, in any defined age group, gender, batch or in any particular country.

So far across the EU and UK, there have been 15 events of DVT and 22 events of pulmonary embolism reported among those given the vaccine, based on the number of cases the Company has received as of 8 March. This is much lower than would be expected to occur naturally in a general population of this size and is similar across other licensed COVID-19 vaccines. The monthly safety report will be made public on the European Medicines Agency website in the following week, in line with exceptional transparency measures for COVID-19.

These results suggest it is unlikely to be a vaccine-related health risk. They quote Ann Taylor, their Chief Medical Officer, as saying: “Around 17 million people in the EU and UK have now received our vaccine, and the number of cases of blood clots reported in this group is lower than the hundreds of cases that would be expected among the general population.”[1]

It raises a question, though: what if these baseline rates of blood clots, and related conditions, were known before the vaccine campaign, so that they could be looked up as any possible concerns arose?

Reflects vaccine age groups?

Scientists writing for a briefing from the Australian Science Media Centre, and in other media, have suggested that these clots may be showing up because early vaccination efforts focus on people who are more likely to get these clots without a vaccine (the elderly, those with chronic disease).

Thrombosis is linked to many conditions. It’s possible that part of the issue is simply that in focusing initial vaccination efforts on more vulnerable people, we’re also focusing it on people more likely to have health problems.

(It could also be related to previous infections, including COVID-19; see Updates.)

If ill health events occur shortly after vaccination they might unfairly be attributed to the vaccine unless we have a good idea what the rate of these health events usually are. That’s what baseline statistics are for.


There’s no easy way to tell if one thing—a vaccine, a drug, some surgical operation—caused an effect simply from that some patients had ill effects afterwards. (Or a positive effect, for that matter.)

An old saying in science is that “correlation is not causation”. Just because some event happened after the first, does not in itself mean the first event caused the second.

One way to partially resolve this is to compare how often the second event occurs before and after the treatment.

If it’s the same, most likely the treatment isn’t changing that outcome.

You’d like to have an estimate of the number of cases in time above which ought to flag a warning. The greater the frequency of the ill-health event above the baseline, the more concerned you should be.

It’s a bit more that just simple counts. We considering them, we also need to put them in context of things that might influence the outcome like the person’s age, the current season (some things are seasonal), what other conditions a person has. It’s not easy to do all of these, of course.

Time frames matter, too: if, say, the vaccine program has only been running for one month, but your baseline statistics are for a year, you need to take that into account, bearing in mind any seasonal effects. Similarly, you may want to take into account the number of days after the treatment.

Sorting out what the baseline statistics are is a big job!

In New Zealand, Helen Petousis-Harris’ team is working on this. She writes over at Diplomatic Immunity here on Sciblogs.

Hopefully this work will head off any unnecessary dramas in NZ’s vaccine rollout!

Other articles in Code for life


Update, 16 March pm: An agency of the German Federal Ministry of Health, the Paul-Ehrlich-Institut, has published a press release on March 15th saying their scientists, “now see a striking accumulation of a special form of very rare cerebral vein thrombosis (sinus vein thrombosis) in connection with a deficiency of blood platelets (thrombocytopenia) and bleeding in temporal proximity to vaccinations with the COVID-19 vaccine AstraZeneca.”

Annoyingly—for a scientist or science writer, at least—this press release doesn’t give the number of cases, nor say what age, gender, etc. these patients are, making it impossible to judge this or interpret it further. The EMA has given no further details about a specific type of blood clot.

The article above talks about baseline statistics. One possibility is that these are not elderly patients, but younger adults, and that these blood clots are a consequence of exposure to COVID-19. If that’s true, then the ‘true’ baseline for comparison is not the general population, but younger people exposed to COVID-19. Frontline workers are also a focus of early vaccination efforts, and they have the highest rate of exposure to COVID-19.

These clots can occur more often after other infections. There you would have to know if they have (recently) had some other infection.

Update, 16th March am: Spain, France and Germany now also have paused the use of the AstraZeneca vaccine.


This batch of vaccine was delivered to: Austria, Bulgaria, Cyprus, Denmark, Estonia, France, Greece, Iceland, Ireland, Latvia, Lithuania, Luxemburg, Malta, the Netherlands, Poland, Spain, Sweden. (Source: EMA report, 10-March-2021.)

An aspect I haven’t covered is clues coming from cases such as blood tests. I’m not trying analyse the likely cause, but introduce one factor that is useful, baseline data.

I’m a bit surprised at this blowing into a longer and larger media event. When I tweeted about this on March 12th, it was clear then that it was unlikely to be genuine health concern.

Cited footnotes

  1. On the face of it, this would be read as the vaccine reducing the number of clots. Better to think of it as “not more”, esp. as it‘ll almost certainly be an undercount.
  2. Professor Sarah Cowley tweeted that, “The annual incidence of venous thromboembolism is approximately 2 in 1000 of the general population & the annual incidence of diagnosed pulmonary embolism in the UK has been reported as 7–8 per 10,000 people.” Similarly David Spiegelhalter writing at The Guardian uses 1 person per 1,000 per year. Even allowing for that the vaccination is only part of the year, it’s easy to see from simple arithmetic that the rate of clots in vaccinated people (<40 in ~5 million) is not higher than those without the vaccine. Spiegelhalter’s back-of-the-envelope estimate comes in at more than 100 a week. (I’ve written ‘<40’ here because the exact numbers are changing as reports come in.)


“A diagram showing terms to do with letter height and positioning on the baseline.”

This is not the use of ‘baseline’ used in the article, but there’s a slight analogy: think of graphs; we’re looking for the parts of the data (letters) standing above the baseline (red line).

(Cropped from original. Author: Max Naylor. Public domain. Source: Wikimedia Commons.)