Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

CNN has more numbers:

> In Moderna's trial, 15,000 study participants were given a placebo, which is a shot of saline that has no effect. Over several months, 90 of them developed Covid-19, with 11 developing severe forms of the disease.

> Another 15,000 participants were given the vaccine, and only five of them developed Covid-19. None of the five became severely ill.

https://edition.cnn.com/2020/11/16/health/moderna-vaccine-re...



Important caveat to keep in mind:

> The analysis was based on the first 95 to develop Covid-19 symptoms.

https://www.bbc.com/news/health-54902908


151 out of 30,000 subjects is considered statistically significant. They will reach that number around Thanksgiving.


I was thinking about this since I read the article. The number you state seems to be 0.5% of total subjects. To my understanding, this is not how "p < 0.005" is employed.

If we take the control group, we have P(Corona)=90/15000. The likelyhood of getting as an extreme result in the vaccine group is then P(0 cases) + ... + P(5 cases) = [math and statistics] = the actual p-value.

How did you determine the statistical significance in your post?


What's the caveat?


The caveat is that “dczot” doesn’t know what a caveat is.


What seems to be vague... They didn't contract Covid, but we're they exposed to it in a lab environment, or just set off on their way to be evaluated x months later?


It is fine for it to be vague. It is implied because medical ethics and international humanitarian law preclude exposing subjects to a deadly virus, let alone for testing purposes, let alone in a blind study.

Additionally, you can apply common sense that if 15,000 people were exposed to the virus with no protection, more than 90 of them would become infected.


They are planning to do just that in the UK https://www.nature.com/articles/d41586-020-02821-4


Opening paragraph "first-of-its-kind" "if it receives final regulatory and ethical approval"

The Nazis did it, and US military does similar kind of shit (I reserve the word science) too.


>They didn't contract Covid, but we're they exposed to it in a lab environment, or just set off on their way to be evaluated x months later?

This is what is referred to as a 'challenge trial'[0]. It is something the UK is apparently working on but afaik, neither of the current vaccine candidates have used any form of challenge trial. This is one of the difficulties with vaccine testing...if I were to vaccinate everyone in the world against small pox, how do I prove it's effectiveness?

[0] https://www.the-scientist.com/news-opinion/a-challenge-trial...

Edit: for some fun, I figured I should link to the 'COVID challenge trial volunteers advocacy organization'. Yes, an advocacy group for people who want to be infected with COVID. https://1daysooner.org/


Deliberately exposing people ("challenge trials") is not allowed.

It could have saved a huge number of lives, but that's not how the system works.


Challenge trials also don't provide as much useful information.

Nobody is getting covid on purpose, in a lab. Being protective or unprotective against a lab dose isn't the same as being protective or unprotective against a real-world dose.


Lab doses can be set to mimic the “real-world” dose. Not to mention that the hope is that the protection is effective against even higher doses.


We don't really know what the "real-world" dose is.


I think the first goal of these challenge studies is to figure this out. Give slowly increasing doses and see what the threshold is.


I don't think anyone has really proposed this. Challenge trials are much more about getting quick answers to efficacy of treatment -- ethically, it's at least somewhat defensible to expose a volunteer if you also have medicine/vaccine for them, even if you're not sure how well it will work. It would be wildly unethical to give the virus to folks just to see how much virus it takes to get sick.


> I don't think anyone has really proposed this.

They seem to talk about it right near the top of major news stories, so I guess they are proposing it:

https://www.bbc.com/news/health-54612293

You might also get useful information about how people's immune systems react, when you give them too little of the virus to get infected. It's not only about figuring out the dose for the main trial. It also could have been done 6 months ago...


"It would be wildly unethical to give the virus to folks just to see how much virus it takes to get sick."

Depends. If there are volunteers for it, (which I can imagine to be the case) - and the volunteers are fully aware of the risk - then it might be ethical to let them proceed and save millions of other people.


Ethics are a bit more nuanced than "consent makes it OK". Suicide is consented to as well.


Thats why some people accept the right of other people over their own life.


And of course:

1. You have to have a control group, so half of test subjects would be exposed to the virus without having received the vaccine.

2. The highest-priority recipients for a vaccine are members of high-risk groups; it'd be pointless to run a challenge trial full of low-risk individuals.

So a challenge trial wouldn't just mean giving the virus to a few thousand vaccinated macho 20-year-olds. It'd mean giving the virus to unvaccinated 80-year-old care home residents.


> it'd be pointless to run a challenge trial full of low-risk individuals.

This is binary all-or-nothing thinking. It considers 90% the same as 0%, since both are not 100%.

Testing on healthy young people you learn how a normal immune system reacts to a vaccine candidate. High risk groups mostly have similar immune systems.

Even if somehow you could only vaccinate everyone under 60, that would do enormous good stopping the spread.


Another way of making the point: we did COVID challenge trials on monkeys first, because their immune systems are a decent model for that 80-year-old human's. Well, a 30-year-old human is an even better model of an 80-year-old human. A challenge trial on young people wouldn't prove everything, but it would give a high-information signal quickly.


> This is binary all-or-nothing thinking. It considers 90% the same as 0%, since both are not 100%.

Thinking is not just "binary or not"; there are degrees. If exposing that group of people would be 90% pointless, it's not much less out of the question than if it were 100% pointless.


I like your recursive "degrees of binarity" argument :)

You're right about the principle, of course. I just think it would be "10% pointless", and thus well worth doing.


In all phase3 trials large groups of people, as similar to each other in all variables as possible, will be either given a placebo or the actual vaccine and then go back to their normal lives.

The trials have a set "Covid-19 cases" mark where the stop to evaluate. The Pfizer vaccine has 164 contractions of Covid-19 as their mark to conclude the research. This research seems to aim for 151.

The Moderna vaccine, which is based on similar mRNA technology as BioNTech’s, is expected to be assessed by the FDA on a final analysis of 151 Covid cases among trial participants who will be followed on average for more than two months.

If both groups are similar enough you can then say how effective the vaccine is in preventing it.

This, by the way, is why most phase 3 trials take place in the US and/or Brazil (among other places): the more Covid-19 is around the faster you can get to the set number of contractions and conclude the phase 3.


The latter. Exposing them willingly would be unethical since they could develop a severe form and possibly die.


I don't think they have started yet, but such trials are planned in the UK.

There's no easy blanket "unethical" ruling from on high. Obviously this isn't a risk-free thing to volunteer for, but it isn't risk free for nurses to go to work either, and thankfully, they still do.


Also, you'd possibly be giving an unrealistic dose in the lab. If you give someone 1000x the viral load they'd ever get in the real world, the vaccine might not work. Or might work but send the immune system into a crazy overdrive cytokine storm.


How likely is this due to random chance?


Well, given that 90/15000 = 0.6% in the control group developed the disease, you can consider the vaccine group as a Bernouili trial of n = 15000 with probability p = 0.6%. Then the probability of observing 5 or fewer cases is 3.4*10^-32, from the tail probability of the binomial distribution.

https://en.wikipedia.org/wiki/Bernoulli_trial

Of course, that's assuming that five guys from the vaccine group didn't get infected at the same after-ski party, or any funny business that violates statistical independence ...


Naive question: If the null hypothesis is that there is no difference, wouldn't that imply 95/30,000, p = 0.0031, putting the probability of observing less than 5 cases at a much more reasonable 1*10^(-14).


I think Fisher's exact test [1] is most commonly used in these types of trials. But the P-value roland (parent comment) provided also make sense to me. For the Fisher's test R says...

                 Covid    NoCovid
                 _____    _______
    Vaccine        5       1495  
    NoVaccine     90       1410  

    9.0e-22  one tailed
    4.5e-22  two tailed
[1] https://en.wikipedia.org/wiki/Fisher%27s_exact_test


Shouldn't the NoCovid values be 14995 and 14910?


Ah yeah good catch (2.7e-21 for 2-tailed, mutatis mutandis)


No, I think you must estimate p only using the cohort where people were not treated, otherwise you will underestimate the population fraction. In order to test if the null hypothesis is true, we can't assume that's its true when constructing the test.


A 90 to 5 by random chance?


Not at all likely




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: