US health authorities are calling for a pause in the use of the Johnson & Johnson Covid-19 vaccine, after reports of extremely rare blood clotting cases. The Food and Drug Administration (FDA) said it was acting "out of an abundance of caution".

It said six cases of severe blood clotting had been detected in more than 6.8 million doses of the vaccine.

BBC 13th April 2021

John Hopkins reports of Cerebral Venous Sinus Thrombosis that :

CVST is a rare form of stroke. It affects about 5 people in 1 million each year.

Hopkins Medicine Org

[Thus, during the whole of 2021, we would expect there to be 1,663 cases of CVST in the entire population (332.5 million) of the USA. On average, that is 31 cases per week.]

If the expectation of CVST is that 5 people in a population of one million will be affected, is the incidence in vaccinated persons any higher ?

As mentioned in comment, the statistic is very similar regarding the EU's experiences regarding Astra Zeneca : 72 cases in 80 million vaccinations.

It is interesting that in both these instances the occurrence of CVST is very similar, about one in a million.

  • 1
    How do you determine if something is overcautious? Also it says 6 cases have been detected but does it discuss how many possible cases might have been missed? It shouldn't be hard to imanage that not every case gets detected. – Joe W Apr 13 at 14:45
  • 3
    All 6 cases were in women between 6 and 13 days after vaccination. So 6 women in a 1 week period relative to the vaccination date. – DavePhD Apr 13 at 15:20
  • 1
    If the incidence rate is truly no higher than the "background" rate, we would expect to see the same rate among people who received the other COVID vaccines - or any vaccine, for that matter. Presumably that is not occurring, since we've heard nothing about it. – Mark Apr 13 at 15:30
  • 2
    @DJClayworth I think that if these cases had resulted in death, they would have said so. If everyone in the US were to get the vaccine and get blood clots at this rate, ~300 people would get blood clots, versus 1000x DEATHS. I don't think that "risking hundreds of thousands of deaths to prevent a few hundred blood clots is a perfectly reasonable thing to do" is an opinion deserving of any respect. And yes, there are other vaccines, but if people on average take another week to get a vaccine, the cost/benefit is still overwhelmingly in favor of not blocking J&J. – Acccumulation Apr 15 at 3:10
  • 1
    @Acccumulation one of the cases has resulted in death so far. The risk the vaccine may be worth it for one age group, but not another. – DavePhD Apr 15 at 15:07

Yes, the incidences of CVST following vaccination with the J&J vaccine and the AstraZeneca vaccine are higher than the background incidence of CVST.

For the Johnson and Johnson vaccine: Between March 19 and April 12, 6 cases of CVST with thrombocytopenia were reported to the Vaccine Adverse Event Reporting System (VAERS),1 a database maintained by the CDC and FDA. These 6 cases, all in women under 50, were what triggered the joint CDC and FDA recommendation on April 13 to pause vaccinations with the J&J vaccine.

An April 14 CDC presentation (a day after the federally-recommended pause) stated that, among women 20-50 years old, the observed CVST cases exceeded the expected background rate by "3-fold or greater." (Slide 24)

At the time, 6.86 million J&J vaccine doses had been administered, 1.4 million of which had been in women 20-50 years old. This is 0.87 cases of CVST with thrombocytopenia per million J&J vaccine doses administered, or 4.3 cases per million J&J vaccine doses administered in women 20-50.

Since the initial 6 reports, the CDC has confirmed 14 cases of CVST with thrombocytopenia as of April 25 (very recent April 30 article). By April 25, 8.1 million J&J vaccine doses had been administered. This is 1.7 cases of CVST with thrombocytopenia per million J&J vaccine doses administered. (The statistics for women 20-50 alone are not available.)

Going into the mathematics:

A Johns Hopkins page states CVST "affects about 5 people in 1 million each year." This means in 2 weeks, CVST affects 0.19 people per million. Out of 8.1 million doses administered, we would expect to see 1.55 CVST cases. The 14 cases reported to VAERS and confirmed by the CDC is close to 9 times the background rate of CVST.

For the AstraZeneca vaccine: On April 7, the European Medicines Agency (EMA)2 wrote:

As of 4 April 2021, a total of 169 cases of CVST and 53 cases of splanchnic vein thrombosis were reported to EudraVigilance. Around 34 million people had been vaccinated in the EEA and UK by this date.

This is somewhere between 2.5 to 5 cases of CVST per million AstraZeneca doses.3 (I am unsure where the question's figure of 72 cases in 80 million vaccinations came from.)

Going into the mathematics:

Using the same rate of 5 CVST cases per million people each year, we would expect to see anywhere between 6.54 and 13.08 cases out of 34 million people vaccinated (depending on how many had received one or two doses). The 169 cases reported to EudraVigilance exceeds both figures and is between 13 to 25.8 times the background rate of CVST.

1 The caveat with all VAERS reporting is reports do not mean a vaccine caused an adverse event.

2 You can think of them as the EU equivalent of the US FDA.

3 Unlike the J&J vaccine, the AstraZeneca vaccine requires 2 doses. It's unclear how many doses of the AstraZeneca vaccine had been administered by April 4, only that 34 million people had been vaccinated.

  • Very surprised nobody has answered yet since the numbers are out there. Let me know any suggestions! I have not written an answer in awhile. – Barry Harrison May 1 at 7:11
  • If the second sentence ("However, regulators have reaffirmed both vaccines' benefits outweigh the risks.") should be deleted because the question doesn't ask about that, let me know too! – Barry Harrison May 1 at 7:37
  • 3
    "both vaccines' benefits outweigh the risks" may be relevant for a government deciding on a "death minimization" strategy, but is not as relevant for an individual making a choice about different vaccines. – pipe May 1 at 11:12
  • It would be instructive to explain the established causal link further or illustrate this bland authority statement of 'benefit/risk still OK' with two numbers: absolute risk increase and relative risk increase (like for the vaccines vs illness: 60%–95% RR but only 0.7%–1.2% AR) – LangLаngС May 1 at 14:16
  • @pipe Thanks, I deleted that. – Barry Harrison May 1 at 14:42

107 to 130 cases of potentially deadly blood clots occurred each year per 100,000 Caucasian individuals from 1985 through 2002. This translates into about 1 to 3 cases per 1,000 people.


Over a 13 day period, then, we would expect a background rate of about 200 cases among 6.8 million people.

DavePhD gives a rate of only 3 people, which if we treat as a Poisson process with lambda = 3, then p(x>=6) = 8%. That's well above what is usually the cutoff for statistical significance. And we have to consider all the cherry-picking that is going on: the choice to look at blood clots, the choice to look at a particular type of blood clots, the choice to look at a specific time frame after the shot, the choice to look at a particular vaccine, the choice to look at a particular demographic, etc. This is like saying that among Floridians who ate green M&Ms, the number of people who stubbed their left pinky toe between 26 and 37 days later was 3 greater than the background rate (I'm not saying the Bayesian prior for "green M&Ms cause stubbed toes" is comparable to "vaccines cause blood clots", just that the evidence is comparable).

  • 3
    Your link is to 'deadly blood clots' which is mainly a figure covering DVT and other general thrombosis. The vaccine issue is exclusively a matter of CVST, a much narrower focus. – Nigel J Apr 15 at 5:10
  • 3
    Why should we trust your analysis here? – Oddthinking Apr 15 at 7:12
  • @NigelJ That's exactly my point in saying the data is cherry-picked. – Acccumulation Apr 15 at 14:11
  • 3
    "if we treat as a Poisson process" Why would we do that? Please provide a reference showing this is the appropriate model to use for this problem. As always, it isn't the arithmetic that is hard, it is working out whether the mathematical process is appropriate in the first place. – Oddthinking Apr 16 at 2:25
  • 1
    As to whether Poisson distributions are "basic", the cut-off we normally use is high-school level, and I genuinely can't remember whether I was introduced in Poisson in high school or uni. – Oddthinking Apr 17 at 1:39

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .