Hacker News Reader

Comment by jjk166 | Hacker News Reader

6 cases in 7 million doses over 3 months is exactly the rate that CVST would normally be expected at. While that doesn't mean we shouldn't investigate, it does mean there is no reason to be alarmed at this time.

> Cerebral venous sinus thrombosis is rare, with an estimated 3-4 cases per million annual incidence in adults. While it may occur in all age groups, it is most common in the third decade. 75% are female. [0]

[0] https://en.wikipedia.org/wiki/Cerebral_venous_sinus_thrombos....

View on HN

Over the past 2 weeks, the reported rate of these events associated with the AZ vaccine has gone from 1 in a million, to 1 in 30,000 [0].

[0] https://www.nejm.org/doi/full/10.1056/NEJMoa2104882

The source you mention does not explicitly give a one in 30.000 figure, and it is not something that can be safely concluded from the data in the paper. The authors state that at the time 132,686 people had received their first AZ shot in Norway, of which five had developed thrombosis.

This is definitely not enough data to conclude a 1 in 30,000 figure. And neither does the cited paper conclude this.

5 developed the disease 7 or 8 days after the vaccination. Not sure what you're getting at...sampling error? It's a binomial distribution, right? Back of the fag packet maths, we can easily reject 1 in a million, and put a reasonable upper bound below 1 in 100,000.

The clots associated with the AZ vaccine aren't CVST.

Though there have been SVTs and DVTs the vast majority have been CVSTs with thrombocytopenia:

https://www.theguardian.com/world/2021/apr/13/astrazeneca-bl...

There's the question of timing too. It's not clear if the number includes all CVST incidences or only those ocurring in a time window after vaccination. (Update: according to NYT these cases are 1-3 weeks after vaccination, so the background rate is too low to matter here). If there is a strong correlation in timing you can not simply use frequency data to rule out causation.

Of course the cases are shortly after vaccination, the data is symptoms reported shortly after vaccination which might be side effects. You're not going to have anyone who got a clot before the vaccine or long afterwards in the dataset, but those cases still did/will exist. There is no evidence yet of a correlation in timing.

The background rate you quoted are expected to be spread out independent of vaccination. If you pick a two-week window after vaccination you should only expect 1/6 of the cases that are expected in 3 month, which is the number you calculated.

But we're not observing people for 3 months and picking an arbitrary two week window. People do not get added to the population until they have taken the vaccine, and they are all observed for different lengths of time. We did not hypothesize a 2 week period beforehand and compare how many landed within to how many were outside that range, the 2 weeks is just a circle drawn around the datapoints after the fact.

Here is the chain of logic:

1. First assume independence, that CVST is unrelated to vaccination.

2. Take any two weeks, you can calculate the expected cases using population background rate because we assume independence.

3. The number is around 1 using Wikipedia data.

4. The observed case number 6 greatly exceeds the expected number based on independence assumption.

5. We conclude that with high probability that our independence assumption is wrong, i.e. there's correlation with high confidence.

Your logic breaks down at step 2. As a counter example, what are the odds that in the other 7 weeks you would have zero cases? Of course the answer is that this isn't how the data works.

You are not sampling a random, normally distributed event over a fixed interval, you are sampling the spacing between two different events where one of them is systemically linked to how you define the population and the interval.

> All six cases were in women aged between 18 and 48, with symptoms appearing six to 13 days after vaccination.

All the observed cases occurred shortly after vaccination, so I’m not sure 3 months is the right divisor to use in your comparison.

You are only looking at people a short time after vaccination. Anyone in the population who had a clot prior to being vaccinated or who will get a clot later this year have not been counted.

That is exactly the point. Using the background rate you quoted from Wikipedia within a two week window you are expected to see around one case. So most of the cases can not be explained by the background.

> While that doesn't mean we shouldn't investigate, it does mean there is no reason to be alarmed at this time.

Except to the general public this overcautious behavior and resulting media hype raises vaccine suspicions, and will decrease public trust in J&J (and even vaccines in general) as a result. It’s already happening.

Public trust in J&J has been low ever since they got caught putting asbestos in baby powder

https://www.reuters.com/investigates/special-report/johnsona...

People need to accept that this is a real effect. Denying it isn't helping anyone. It's happening, it's rare, but it's happening.

If it's statistically significant, the evidence showing so will emerge quickly and we can decide at that time how the increase in risk compares to that of remaining unvaccinated.

Otherwise, there isn't a whole lot of difference between "not real" and "real but so rare as not to matter."

Reply on news.ycombinator.com