13 Jan, 14 tweets, 4 min read
The amazing AZ efficacy against severe cases is a joke!
thelancet.com/journals/lance…
"From 21 days after the first dose, there were ten cases hospitalised for COVID-19, all in the control arm; two were classified as severe COVID-19, including one death."
2 versus 0. I'm speechless!
Just to clarify the amazing AZ severe cases joke, having 0 in the vaccine arm and 2 in the placebo arm means the 95% confidence interval for this efficacy must be from 0% to 100%. They don't bother to compute it, and neither will I. Australia's roll-out is supported by this joke!
Anyone who wants to know why a statistician would laugh out loud at the claim that AZ vaccine stops severe covid19 must read this British Medical Journal article.
bmj.com/content/363/bm…
It's a riot! It contains the same kind of erroneous inference, but just a bit worse. Enjoy!
Computation.
AZ vaccine trial gets 0 hosps vaccine arm, 10 placebo.
Let E=efficacy against hospitalisation, P=Prob(V=0)=(1/(2-E))^10.
E=60%. P=3.46%.
E=65%. P=4.97%.
E=70%. P=7.25%.
So 95% C.I. for E includes 70%.
∴ V=0 hosps consistent with 70% efficacy against general covid19.
A bit more detail.
Let H₀="efficacy of AZ against hospitalisations = 70%".
Let H₁="efficacy of AZ against hospitalisations > 70%".
V = number of vaccine-arm hospitalisations out of 10.
Prob(V=0 | H₀) = 7.25% > 5%.
H₀ can't be rejected at p = 0.05 significance.
H₁ not proven!
H₀="efficacy of AZ against severe covid = 0%".
H₁="efficacy of AZ against severe covid > 0%".
V = number vaccine-arm severe #COVID19 out of 2.
Prob(V=0 | H₀) = 25% > 20%.
H₀ can't be rejected at p = 0.2 significance.
H₁ not proven.
Severe covid efficacy claims not proved!
When I saw the 0 out of 2 basis (in the Lancet paper) for the claim that AZ is completely effective against severe disease, I didn't even laugh out loud.
It gave me a kind of terror to think that Australian so-called medical authorities are intentionally misleading the public.
By the way, I am probably the most pro-vaccination person on this planet.
That's why I really, really, really want one of the 95% effective mRNA #COVID19 vaccines.
They are a miracle of modern science.
I will refuse the almost worthless AZ vaccine which is 6 times less effective.
Here again is the Lancet paper on AZ trial results.
thelancet.com/journals/lance…
It is honest and crystal clear if you know basic statistics.
AstraZeneca have not made deceptive claims.
Only in Australia have I seen "aggressive marketing" of AZ vaccine using deceptive stats arguments.
Anyone who doesn't see the implications of the AstraZeneca article regarding severe disease and hospitalisations should ask a statistics lecturer at a University to interpret it for them.
A 1st year stats student who got this wrong would be put in the stocks and spanked!
#COVID19
Even the ABC is carrying false interpretations.
abc.net.au/news/2021-01-1…
"Professor Pollard argued one of the most significant findings from the clinical trials was that after receiving the first dose of the Oxford vaccine no participant was hospitalised or had severe disease."
For the last couple of days, I've been just a whisker away from giving up on this issue.
After the epic efforts of the last 12 months, Australians now seem resigned to a vaccine which is 6 times less effective than mRNA.
If stats professors don't speak up now, we're doomed.
#COVID19 pandemic requires multiple professions.
Statisticians are not experts in medicine.
Top medical people are not generally experts on statistics.
For medical questions, ask medical experts.
For stats questions, ask stats experts.
I just told someone the Tibetan prayer treatment for covid.
You have 100 participants.
Someone in Tibet prays for 50 of them. Double blind.
Two patients get severe covid, both in the non-prayer arm of the trial.
This proves Tibetan prayer is 100% effective against severe covid.

• • •

Missing some Tweet in this thread? You can try to force a refresh

This Thread may be Removed Anytime!

Twitter may remove this content at anytime! Save it as PDF for later use!

# More from @kennington_u

10 Jan
Strength of vaccines.
If E = efficacy, then S = 1/(1−E) could be called "strength (metric)" of the vaccine.
E.g. E = 95% gives S = 20.
I.e. In situations where unvaccinated group gets 20 infections, same-size vaccinated group gets only 1.
I.e. Vaccine divides infections by 20.
A 95% vaccine is 20 times better protection than placebo.
A 70% vaccine is 3.33 times better protection than placebo.
So a 95% vaccine is 6 times better protection than a 70% vaccine.

If 15 million people get a 70% vaccine, 4.5 million people will have placebo-level protection.
With 95% vaccine, 5% of people will get placebo-level protection.
With 70% vaccine, 30% of people will get placebo-level protection.
So 70% vaccine leaves 6 times as many people unprotected.

It is a mathematical ILLUSION to think 70% is almost as good as 95%.
9 Jan
AZ vaccine 70.4% efficacy for #COVID19 means:
−log₁₀(1−70.4%) = 5.3 dB case reduction.

P/B and M vaccines 95% means:
−log₁₀(1−95%) = 13.0 dB case reduction.

So AZ effectiveness is only 40.6% of P/B and M.
P/B and M vaccines are 2.46 times better.
thelancet.com/journals/lance…
If you compare 70.4% and 95% efficacy without logarithms, the contrast is even more extreme.

(1−70.4%)/(1−95%) = 5.92.

So the Pfizer/BioNTech and Moderna vaccines are 5.92 times better than the AZ vaccine using a linear measure.
In other words, AZ is only 16.9% as effective.
It's beyond my comprehension that Australians are not angry at the plan to give 15 million Australians AZ vaccine which has only 16.9% of the effectiveness of the good #COVID19 vaccines.
This will divide Australians into two classes.
The covid-immune and the covid-vulnerable.