A few responses:

1. It isn't just about deaths (from TTS or COVID) as both can cause long term health sequelae - eg stroke from CVST, extensive intestinal resection after splanchnic vein thrombosis, clotting after COVID. There are also the benefits that are harder to quantify (eg protection of close contacts, contributing to the community, reduced anxiety about getting COVID). The other endpoints (hospitalisation, ICU) also have limitations - people are hospitalised with COVID for reasons other than severity (eg where a nursing home is evacuated), and few people >80 are admitted to ICU due to advance care directives.

2. As I noted, all the three data sources (EMA, UK/MHRA and Australian data) have their limitations - EMA don't use a case definition (if an adverse event report mentioned "platelet" it would included, even if it wasn't TTS), under- and over-ascertainment possible in both EMA and UK, and relatively smaller numbers in Australian dataset particularly <50 years.

3. Mortality estimates in the cited risk benefit document were based on last year (as indicated in the footnote - Australian first wave and Victorian second wave) where no people in the 20s died. Obviously this may be different with delta (as might vaccine effectiveness and age specific incidence/severity). But the largest source of uncertainty is the risk of COVID over the period of time between now (when AZ is available) and whenever Pfizer/Moderna/Novavax is widely available to all (Sept-Nov).

Op COVID SHIELD National COVID Vaccine Campaign Plan (health.gov.au)

4. Worth re-emphasising that we carefully used the word "prefer" and are trying to respect patient autonomy, recognising that how those <40 perceive risk and benefit is a personal decision. I saw a tweet from a younger person saying that there were happy to have AZ as she had been on the pill for acne since she was 15 and had done illicit drugs off a toilet seat (I'm paraphrasing). That's not to say that all young people would concur, but I suppose that puts a 1:20000-40000 risk of TTS into perspective for that person (although if I were her GP I might be considering a few non-judgmental words about risk taking and life choices).

Allen

I have worked through some of this.  Using the EEA data the incidence in 20-29  is about 2.5 times the Australian TGA rate.  So you can take the mortality rate that the TGA found of 3% and that gives 1.6 deaths per million. Similarly for 30-49 it is about double, giving 1.1 deaths per million. Then, you look at the NEJM paper and you see a death rate of 22%.  The EEA paper has 52/269=19%. So have they only found more severe patients? Probably, yes. The UK adverse events reporting is known to be very poor.  Most of the cases in the EEA actually come from the UK, but are not the subjects in the NEJM paper which is for a later time period.  An interesting question is why the EEA has information on the age distribution of vaccination while the NEJM article doesn't? Maybe BREXIT mess. Anyway, if I apply a death rate of 19% to the EEA paper I get 10 deaths per million in 20-29 and 8 deaths per million in 30-49. 

Compared to the mortality in the Victoria wave from last year, they look good. Now the difficult part. In Sydney most cases are in South Western Sydney and vaccination is a good idea. If someone is in one of the more affluent areas, the cumulative case rate is 1 in 1,000 or less, so assuming that doesn't differ too much across ages, the fatality rate is 2 in a million, using 2 deaths per 1,000 cases in 20-29 from Ian Marschner's paper. Much better to follow the social distancing rules rather than be vaccinated.

I feel that this really needs another look at the TGA and ATAGI data updated until now, and broken down by all the age groups under 50, and including deaths. I believe there has already been one death in the 30-39 group so it would be worthwhile.  By October these age groups will probably be opened to Pfizer or Moderna anyway.

As there are more under 60s currently being vaccinated with AstraZeneca and more data becomes available, the risk estimates (of severe TTS and death) will become more precise and less prone to bias from the early younger vaccinated group. My question is, are the people currently giving informed consent to take AstraZeneca, being informed by data that is appropriate to their age group?  Especially the under 50s. Even if they are given a range rather than a point estimate like "Risk of death is 1 in a million".

There may be newer data by now, but there’s quite a detailed EMA assessment report from 24 March:

https://www.ema.europa.eu/en/documents/prac-recommendation/signal-assessment-report-embolic-thrombotic-events-smq-covid-19-vaccine-chadox1-s-recombinant-covid_en.pdf

There is an analysis by age on p.19, actually looks like quite a strong association.

I have to comment on the mortality risk from the AstraZeneca vaccine. Across all age groups there have been 6 deaths from approximately 6 first doses, so an incidence of 1 in a million, which is what is being reported in the media.

However, if you look at the 30-49 age group,  who are currently being urged to receive AZ instead of waiting for Pfizer, a different picture emerges

According to the latest TGA and ATAGI reports, the incidence of TTS in <50s is 3.4 in 100,000, or 1 in 30,000. In the 30-49 year old age group, there have been 8 Tier 1 (serious) clots and 4 deaths (1/2 in 30-39, 3/6 in 40-49). So among those who contract TTS in this age group, there's a death rate of 50%, leading to an overall death rate of 1 in 60,000 first AZ doses, or 1.7 TTS deaths per 100,000. This is higher than the expected deaths in this age group from COVID-19 in a low or medium risk scenario in this risk-benefit document. It's only when the scenario changes to high risk (3,544 cases per 100,000 in 16 week period) that there is a clear benefit from the vaccine. Is this where NSW or the other states with outbreaks are at the moment?

There isn't a gender breakdown of the TTS incidence, but I expect the mortality rate would be even higher in women, since they have had 3/4 deaths in this age group (5/6 deaths in all age groups have been women).

Admittedly there have only been a small number of deaths and I haven't calculated confidence intervals, but this incidence rate is 16.7 times higher than 1 in one million.

Conversely, the risk of death in the >60s, who are recommended to receive AZ, is much less than 1 in one million.