COVID-19: Vaccine and testing procedures Megathread Part 2 [Mod Warning - Post #1]

plodder

antiskeptic wrote: »

I understand the group is large and within the realms possible, identical. But it alters not the problem.

The problem is that the numbers involved in calculating efficacy are tiny. 9 out of 24,000 in the vaccinated group and 132 (iirc) from the placebo group.

Now you can decide there is statistical significance between 9 and 132 and decide no statistical significance between 9 and 12.

But that's not the issue. The issue is whether you would get a significant spread of numbers were you to monitor a number of 24,000 sized groups. Take the vaccinated group with 9 cases. Would you get 9/24000 if you examined 100 x 24000 vaccinated groups? Obviously not. So what would the spread be? 4 to 20? 3 to 144? 7 to 231? And would that spread follow a normally distributed curve whereby the significant mean is in the middle?

That you cannot practically obtain an answer is neither here nor there. The issue is that 9 and 132 can be complete darts at a board numbers.

I think the numbers belie the significance of it. It's much easier to see in a timeline graph like this one, and in fact the vast majority of vaccine cases occurred during the period immediately after vaccination (as expected).



The low overall infection rate is only of secondary interest. There has been speculation that people likely to sign up for a vaccine trial are more careful in general about getting infected.

schmoo2k

antiskeptic wrote: »

I understand the group is large and within the realms possible, identical. But it alters not the problem.

The problem is that the numbers involved in calculating efficacy are tiny. 9 out of 24,000 in the vaccinated group and 132 (iirc) from the placebo group.

Now you can decide there is statistical significance between 9 and 132 and decide no statistical significance between 9 and 12.

But that's not the issue. The issue is whether you would get a significant spread of numbers were you to monitor a number of 24,000 sized groups. Take the vaccinated group with 9 cases. Would you get 9/24000 if you examined 100 x 24000 vaccinated groups? Obviously not. So what would the spread be? 4 to 20? 3 to 144? 7 to 231? And would that spread follow a normally distributed curve whereby the significant mean is in the middle?

That you cannot practically obtain an answer is neither here nor there. The issue is that 9 and 132 can be complete darts at a board numbers.






As ought be clear, the size of the test group is irrelevant when the numbers you base your decision on are tiny. The question is what spread can be expected when you compare multiple groups of 24,000

If you knew high blood pressure caused heart attacks and were developing a drug to lower blood pressure you would take a large sample of people with high blood pressure. You would give half the drug and half a placebo. And then see how many went on to have heart attacks.

All the participants would be exposed to the risk and you have a significant result.

But you can't expose all of the participants of a Covid vaccine trial to Covid - that would be unethical. And so you don't know how many of the trial subjects are exposed to Covid. It will be a small fraction of your trial sample in any case.

In the case of the blood pressure drug you wouldn't expect to get significant variation in 100 x 24000 sized groups: all participants would have high blood pressure. Not so with this vaccine trial. Circumstances on the ground will mean there will be a range in the numbers contracting Covid if you could monitor100 groups x 24000.

Like I say, given the comparatively tiny numbers involved (9 and 132 from 20000), that range/spread could swallow up whole, all your statistical significance obtained from your mere two data points (9 and 132)

Then they wouldn't have seen similar results in the different phases of the trial ? If they had it would have caused them to go back to the start (and this often happens).

chimneypot20

chimneypot20 wrote: »

Question: with the availability of a vaccine, I'm aware that it's not yet known if the vaccine prevents one from transmitting the vaccine. Supposing it is the case that the vaccine does NOT prevent transmission, how does the vaccine help end the pandemic? Or is it the case that the vaccine just prevents serious illness and this pandemic becomes endemic?

bump

Geuze

chimneypot20 wrote: »

Question: with the availability of a vaccine, I'm aware that it's not yet known if the vaccine prevents one from transmitting the vaccine. Supposing it is the case that the vaccine does NOT prevent transmission, how does the vaccine help end the pandemic? Or is it the case that the vaccine just prevents serious illness and this pandemic becomes endemic?

It will reduce the demand/pressure on hosps and ICU, I assume.

But I see your point, masks and physical distancing must remain.

funnydoggy

Geuze wrote: »

It will reduce the demand/pressure on hosps and ICU, I assume.

But I see your point, masks and physical distancing must remain.

Until how long? The first measure grand, but physical distancing is impossible to remain in place forever. People will either ignore it, or businesses left right and center will close because they can't stay afloat.


I think (hope) once the most vulnerable are vaccinated and our hospital/ICU admissions tank, restrictions will follow suit.

odyssey06

Geuze wrote: »

It will reduce the demand/pressure on hosps and ICU, I assume.

But I see your point, masks and physical distancing must remain.

This is a good article in the NY Times:

The coronavirus vaccines are injected deep into the muscles and stimulate the immune system to produce antibodies. This appears to be enough protection to keep the vaccinated person from getting ill. Some of those antibodies will circulate in the blood to the nasal mucosa and stand guard there, but it’s not clear how much of the antibody pool can be mobilized, or how quickly. If the answer is not much, then viruses could bloom in the nose — and be sneezed or breathed out to infect others.
“It’s a race: It depends whether the virus can replicate faster, or the immune system can control it faster,” said Marion Pepper, an immunologist at the University of Washington in Seattle.

https://www.nytimes.com/2020/12/08/health/covid-vaccine-mask.html

plodder

chimneypot20 wrote: »

bump

I think it is hoped that the vaccines will reduce transmission too, and there is some data that suggests it (from Moderna). I'm not an expert, but I've been surprised more hasn't been said about that. There was a concern that to prevent transmission would require a nasal spray type vaccine, which targets the bug directly at the point where it spreads from, but these don't exist yet.

So, maybe there is an abundance of caution around the Moderna data (which was a 67% reduction in virus presence in swabs taken just before the second vaccine shot). Also, obvious concern that a message getting out that the vaccine prevents tranmission will be counter productive in the short-term.

mandrake04

opinionated3 wrote: »

Apologies if this has been posted before. My understanding was that as long as the vulnerable, elderly and front line staff were vaccinated that society in general could reopen. In simple terms, is this now not the case?

Most of the experts seem to be touting mid/late summer before getting back to normal, sense would tell you it’s nots going be an instant silver bullet.

I was listening to Prof Paddy Mallon earlier who is pretty grounded and he says to administer a vaccine effectively you need have a healthy health system to begin with. If your health system is under pressure from a Tsunami of cases it’s going eat into your resources to administer the vaccine. Also there’s risk that in vaccinating so many people in such a short time you don’t actually end up creating an environment that the virus can spread and you end up giving the patient the virus at same time as the vaccine the disease would probably take hold before the vaccine could provide protection.

wes

funnydoggy wrote: »

I think (hope) once the most vulnerable are vaccinated and our hospital/ICU admissions tank, restrictions will follow suit.

I think some restrictions will remain until we achieve herd immunity via the vaccine. I think we can go down to a level 2 or 1 before that however.

funnydoggy

wes wrote: »

I think some restrictions will remain until we achieve herd immunity via the vaccine. I think we can go down to a level 2 or 1 before that however.

It makes sense alright! Hope so

astrofool

antiskeptic wrote: »

I understand the group is large and within the realms possible, identical. But it alters not the problem.

I think the problem lies with your misunderstanding of data despite it being shown clearly to you time and again, which makes me believe you're an anti-vaxxer pretending to be a sceptic.

The vaccine has proven to be safe.
The vaccine has proven to be effective.

Likely by the time you can take it, millions more people will have got the vaccine, further proving it's safe and effective.

Will you take it then?

What are you looking for in the numbers that would change your mind?

Again, you seem to be doubling down on misunderstood figures by you, do you now understand what they mean and how they are measured?

What are you trying to prove by repeating the same wrong information over and over again?

stephenjmcd

A very steady stream of vaccines to arrive. First vacination 30th Dec

https://twitter.com/Philip_Ryan/status/1341352425739923457?s=19

nommm

https://twitter.com/philip_ryan/status/1341352425739923457?s=21

40k /week doesn’t inspire me that the end is near

ACitizenErased

nommm wrote: »

https://twitter.com/philip_ryan/status/1341352425739923457?s=21

40k /week doesn’t inspire me that the end is near

I’m assuming that is Pfizer only. Technically no other vaccine “exists” in the EU scheme of things.

funnydoggy

stephenjmcd wrote: »

A very steady stream of vaccines to arrive. First vacination 30th Dec

https://twitter.com/Philip_Ryan/status/1341352425739923457?s=19

I'm suddenly thinking of the big trucks carrying all that gold in Die Hard With a Vengeance :pac:

mandrake04

An interesting one, Australia could possibly vaccinate border, quarantine, airline and airport staff first.

https://www.theguardian.com/commentisfree/2020/dec/22/a-covid-vaccine-will-soon-arrive-in-australia-heres-how-we-should-deploy-it


I also heard the health minister Greg Hunt talking about vaccine approval, he says the TGA is not going be giving EUA but rather go for full approval and that they are monitoring the use of Pfizer in US and UK first. There’s also 30 million doses of AZ/Oxford sitting in the CSL plant in Melbourne ready to go but they don’t want to rush the approval.

antiskeptic

antiskeptic wrote: »

I understand the group is large and within the realms possible, identical. But it alters not the problem.

The problem is that the numbers involved in calculating efficacy are tiny. 9 out of 24,000 in the vaccinated group and 132 (iirc) from the placebo group.

Now you can decide there is statistical significance between 9 and 132 and decide no statistical significance between 9 and 12.

But that's not the issue. The issue is whether you would get a significant spread of numbers were you to monitor a number of 24,000 sized groups. Take the vaccinated group with 9 cases. Would you get 9/24000 if you examined 100 x 24000 vaccinated groups? Obviously not. So what would the spread be? 4 to 20? 3 to 144? 7 to 231? And would that spread follow a normally distributed curve whereby the significant mean is in the middle?

That you cannot practically obtain an answer is neither here nor there. The issue is that 9 and 132 can be complete darts at a board numbers.






As ought be clear, the size of the test group is irrelevant when the numbers you base your decision on are tiny. The question is what spread can be expected when you compare multiple groups of 24,000

If you knew high blood pressure caused heart attacks and were developing a drug to lower blood pressure you would take a large sample of people with high blood pressure. You would give half the drug and half a placebo. And then see how many went on to have heart attacks.

All the participants would be exposed to the risk and you have a significant result.

But you can't expose all of the participants of a Covid vaccine trial to Covid - that would be unethical. And so you don't know how many of the trial subjects are exposed to Covid. It will be a small fraction of your trial sample in any case.

In the case of the blood pressure drug you wouldn't expect to get significant variation in 100 x 24000 sized groups: all participants would have high blood pressure. Not so with this vaccine trial. Circumstances on the ground will mean there will be a range in the numbers contracting Covid if you could monitor100 groups x 24000.

Like I say, given the comparatively tiny numbers involved (9 and 132 from 20000), that range/spread could swallow up whole, all your statistical significance obtained from your mere two data points (9 and 132)

You are trying hard, but your understanding of statistical significance is non existent. There are not 2 data points. Each event is a data point. And for the data with 9 events in the vaccine group you produce a 95% confidence interval that the true rate of infection in the vaccine group will be between 2.9% and 11.7% - eg in 19 of 20 trails the maximum rate of infection you would expect would be 11.7%.

If 95% confidence was not enough and you wanted 99 the range would be 2.2 to 13.6% - in 99 trial out of 100 the data would fall in the range based on 9 infections detected in the vaccine group

If 99% wasn't enough at 99.9 the range would be 1.6% to 16% - in 999 trials out of 1000 the data would fall in that range

Sky King

funnydoggy wrote: »

I'm suddenly thinking of the big trucks carrying all that gold in Die Hard With a Vengeance :pac:

I was thinking of the convoy of illuminated trucks in the coca cola christmas ad

funnydoggy

Sky King wrote: »

I was thinking of the convoy of illuminated trucks in the coca cola christmas ad

https://twitter.com/laoneill111/status/1330216372840886272

is_that_so

nommm wrote: »

https://twitter.com/philip_ryan/status/1341352425739923457?s=21

40k /week doesn’t inspire me that the end is near

It will take a few months for all of them to ramp up supplies. By the end of January we may have 3 approved but from March onwards is where they expect a whole lot more.

namloc1980

nommm wrote: »

https://twitter.com/philip_ryan/status/1341352425739923457?s=21

40k /week doesn’t inspire me that the end is near

At 40k a week it would take over 3 years to get to 70% of the population vaccinated. But you'd hope that the other vaccines will come on stream in the first few months of 2021 and numbers would seriously ramp up.

Sky King

Does the 40k include the soon-to-be-approved vaccines I wonder? Hopefully not....

Van.Bosch

stephenjmcd wrote: »

A very steady stream of vaccines to arrive. First vacination 30th Dec

https://twitter.com/Philip_Ryan/status/1341352425739923457?s=19

Why the 30th if others are starting 27th?

antiskeptic

plodder wrote: »

I think the numbers belie the significance of it. It's much easier to see in a timeline graph like this one, and in fact the vast majority of vaccine cases occurred during the period immediately after vaccination (as expected).



The low overall infection rate is only of secondary interest. There has been speculation that people likely to sign up for a vaccine trial are more careful in general about getting infected.

I don't see how the graph addresses the question. Narrow it down to just the placebo group for simplicities sake. The graph shows a steady cumulative increase of Covid cases in the unvaccinated group, as you would expect. You end up after a period of time with a total. 132 or whatever

My question is: what would the range of numbers be if you observed 100 groups x 24,000.

You would naturally see a cumulative increase in all groups over time. The cumulative total wouldn't be 132 each time. That much is certain. So what would it be?

What if it was 9? Wouldn't the graph look just like the vaccinated group graph?

The low numbers do count. If makes the possible range of totals obtainable in like groups (say 100 groups of 24,000 unvaccinated) a more relevant issue.

antiskeptic

[Deleted User] wrote: »

You are trying hard, but your understanding of statistical significance is non existent. There are not 2 data points. Each event is a data point. And for the data with 9 events in the vaccine group you produce a 95% confidence interval that the true rate of infection in the vaccine group will be between 2.9% and 11.7% - eg in 19 of 20 trails the maximum rate of infection you would expect would be 11.7%.

If 95% confidence was not enough and you wanted 99 the range would be 2.2 to 13.6% - in 99 trial out of 100 the data would fall in the range based on 9 infections detected in the vaccine group

If 99% wasn't enough at 99.9 the range would be 1.6% to 16% - in 999 trials out of 1000 the data would fall in that range

Thanks for the lesson but my comment (and ignorance on the precise workings of confidence intervals) isn't the central issue.

You might address the central issue:

1 group of 24000 placebo'd produced 132 (or some such) Covid cases.

You would agree that if we ran 100 placebo groups of 24000 we wouldn't get 132 Covid cases in each group. We would get a range of results.

Q.1 What is that range?

Q.2 How is that range arrived at (assuming we can't practically run an experiment of 100 groups of 24000 plaebo'd people).

antiskeptic

antiskeptic wrote: »

I don't see how the graph addresses the question. Narrow it down to just the placebo group for simplicities sake. The graph shows a steady cumulative increase of Covid cases in the unvaccinated group, as you would expect. You end up after a period of time with a total. 132 or whatever

My question is: what would the range of numbers be if you observed 100 groups x 24,000.

You would naturally see a cumulative increase in all groups over time. The cumulative total wouldn't be 132 each time. That much is certain. So what would it be?

What if it was 9? Wouldn't the graph look just like the vaccinated group graph?

The low numbers do count. If makes the possible range of totals obtainable in like groups (say 100 groups of 24,000 unvaccinated) a more relevant issue.

Answered.

Deleted User wrote: »

You are trying hard, but your understanding of statistical significance is non existent. There are not 2 data points. Each event is a data point. And for the data with 9 events in the vaccine group you produce a 95% confidence interval that the true rate of infection in the vaccine group will be between 2.9% and 11.7% - eg in 19 of 20 trails the maximum rate of infection you would expect would be 11.7%.

If 95% confidence was not enough and you wanted 99 the range would be 2.2 to 13.6% - in 99 trial out of 100 the data would fall in the range based on 9 infections detected in the vaccine group

If 99% wasn't enough at 99.9 the range would be 1.6% to 16% - in 999 trials out of 1000 the data would fall in that range

hmmm

Sky King wrote: »

Does the 40k include the soon-to-be-approved vaccines I wonder? Hopefully not....

Just doing some back of the envelope calcs - suggests for the EU has a whole that represents about 200 million doses a year, so that is 1 vaccine.

Hopefully we have 4 vaccines in the new year. If we got similar from each we'll be humming along at 160,000 vaccinated a week which wouldn't be long making a difference.

antiskeptic

antiskeptic wrote: »

Thanks for the lesson but my comment (and ignorance on the precise workings of confidence intervals) isn't the central issue.

You might address the central issue:

1 group of 24000 placebo'd produced 132 (or some such) Covid cases.

You would agree that if we ran 100 placebo groups of 24000 we wouldn't get 132 Covid cases in each group. We would get a range of results.

Q.1 What is that range?

Q.2 How is that range arrived at (assuming we can't practically run an experiment of 100 groups of 24000 plaebo'd people).

Statistics -you can go at it from two sides - observe the number of events you get and then get a confidence statement over a range as I have shown or complete multiple studies and use the standard deviation to calculate a distribution. And its not the number of patients in the trial that gives the confidence in efficacy, just the ratio of events. The more patients, the quicker you reach the endpoint. The overall number is important for safety. Either way a result that produces 9 events in 141 occurring in the vaccine population would occur about 1 time in 1,000,000,000,000 by chance if the vaccine was ineffective.

Gael23

Looks like were going to be getting about 20,000 vaccines a month unless more come along soon

Stheno

Gael23 wrote: »

Looks like were going to be getting about 20,000 vaccines a month unless more come along soon

Its 40000 doses a week