This is the first point where 300 odd people were excluded from the analysis because of protocol deviations. Worryingly, this number was much higher in the treatment group. While I hope that blinding would have prevented any motivated exclusion, it's still quite concerning.
His second main point (which is actually really interesting), is that there were 3500 cases of symptoms consistent with Covid-19 that weren't revealed by a PCR test. I'm not sure what to think of this (and note that this number swamps the actual numbers of confirmed cases by about 10x), except that it matches with anecdotal evidence I've seen around negative tests in the presence of symptoms.
>Again, note the juxtaposition. These 477 patients who reported COVID-19 symptoms but never received nasal swabs were across the entire trial of well over 40,000 subjects, but the way it’s phrased, not put into context of the size of the entire trial, makes this sound like a huge number
477 untested patients is absolutely a huge number in the context of this trial: only 170 symptomatic infection were confirmed over the entire trial (the only endpoint). Even if the missed positives were evenly distributed between the two arms of the trial, the relative efficacy would be skewed by the additional cases.
For example, if we assume all of the untested people were actually positive (not unreasonable, they were symptomatic), the reported efficacy would drop by ~30%: instead of 8 confirmed cases in the vax arm vs 162 in the control arm (95.2%), it would become 218 vs. 429 (66.3%). Not a trivial difference.
Actually, if you look at the data Pfizer give to the FDA (the stuff that was initially released), there were more 'suspected by not confirmed' cases of COVID-19 in the control group than the test group.
Pfizer did not actively test each participant. It only selectively tested some patients and not others, based on symptoms. They provided no explanation of 'suspected but unconfirmed' cases. If they suspected them, perhaps they should have tested them, but they didn't.
They have a good point about missing data, but they're glossing over the fact the fatality rate goes up drastically if the medical system fails.
I 100% agree that we should be running antibody tests to see how many people have caught COVID-19 and recovered. All of our testing until recently was based on RNA tests, which may only be positive during a relatively short window. (I saw a case study claiming that even some hospitalized patients are testing negative inside 7 days.) We need to know how many people catch this and beat it quickly.
That said, even in populations that have been RNA-tested early and extensively (such as the Diamond Princess and South Korea), the number of completely asymptomatic cases is less than 50%. Using the most optimistic data, I personally have less than a 0.5% chance of dying assuming I get all the medical care I need.
And that's the problem. This virus hospitalizes about 20% of identified cases. They require some supplemental oxygen, and maybe an IV. With good care, probably less than 1% die.
So there are really two key fatality rates:
1. How many people die if they get all the care they need, and
2. How many people die if 30+% of the population catches this at the same time?
Even if we're overestimating (1) by a factor of 10, that's still enough to make (2) catastrophic. What happened in Wuhan and Lombardy can happen here, and there's absolutely no reason that it couldn't get 10x worse. Even if we're overestimating the disease.
So let's start testing aggressively for antibodies. Until we get that number, I'm all for extreme caution.
In VA we've done 30,645 tests and have 3,645 confirmed cases. That tells me that 88% of people tested were negative. Either they're spreading a whole lot of "potentially" exposed people or there are a lot of people with symptoms who actually have something else.
It must have taken a special effort to write all those words without once mentioning the false positive rate of the PCR test (~2.3%), which (unlike earlier in the pandemic) is now in the ballpark of reported case numbers: https://www.medrxiv.org/content/10.1101/2020.04.26.20080911v...
We can't tell anything from the article.
* We don't know how many people were tested.
* We don't know how many people tested positive.
* We don't know the false positive rate for the test.
We need to be really careful about the claims of asymptomatic infection.
The problem with this concept is it becomes impossible to detect false positives. In fact it makes the test the ground truth rather than actual observable clinical sickness. Any bug in the test thus creates a pseudo-epidemic:
If you look at that paper you can see it admits that other research studies found no asymptomatic infection at all. Others on the other hand said they saw it. So there's no real consensus that this phenomenon actually exists in this case, and the asymptomatic people didn't seem to transmit the virus at all (so why are we all wearing masks now then?).
The numbers involved are tiny. Their conclusions are based on the antibody test returning positive without symptoms for just 6 people. The amount of antibodies found in the asymptomatic cases was significantly lower than for other cases, leading to the question of whether they were picking up noise at the edge of the test's capabilities and whether they truly understand how the test works (given that there weren't many SARS-CoV-1 cases, there'd be limited cases on which to test or calibrate it).
Thank for pulling out that false positive statistic. It is incredibly irritating when an article, even a press release type one like this, makes it a point to give an exact number for the true positive / false negative rate and then fails to answer the obvious other half of the question. It made me sneer "oh yeah? Well, a magic 8 ball that only ever says 'yes you have covid you're gonna die' would catch that remaining 1.5%; why aren't you doing that?"
But as usual, the fault is in the summary, not the research.
Let's assume the "real" share of infectious people in the population is 10/100k (That's twice as much as the most recent reported 7-day incidence (5.1/100k) of new infections for Leipzig, the area where the study was performed.)
Further let's assume a PCR Test has a 30% FNR, and a 10% FPR (numbers completely made up, I don't have a source).
Then out of 1500 People who have tested negative, we'd expect 0.05 to have the virus.
Ah, ok. I think this was the article I originally read which implies they only used 2,500 covid positive patients. But rereading it, maybe it’s just unclear reporting.
“Across around 2,500 captured cough recordings of people confirmed to have COVID-19, the AI correctly identified 97.1 percent of them – and 100 percent of the asymptomatic cases.”
"- 360,304 patients with suspected SARS-CoV-2 infection
- Of these 360,304
cases, 26,815 were confirmed by a positive PCR test. The rest had negative
PCR tests.
- In the set of patients with case definition, 1,292 received HCQ (at least 2 grams
per month)
- The proportion of HCQ chronic treatment was higher in negative patients
(0.36% vs. 0.29%, P = 0.04)
- We were able to show that patients taking HCQ have had reduced odds of
SARS-CoV-2 infection"
No joke, that is the basis for their claim. Just like "compared to those that we suspected could have tested positive, it turned out that among those that weren't positive there were more people proportionally receiving HCQ." I don't understand how that can actually prove something. One can surely find a lot of things that are proportionally more present in those tested negative, but without any meaning.
These “cases” are inaccurate as SARS-COV-2 virus detection does not mean people have the disease COVID-19. From the NYT: “In three sets of testing data that include cycle thresholds, compiled by officials in Massachusetts, New York and Nevada, up to 90 percent of people testing positive carried barely any virus, a review by The Times found.”
> As we learned that more and more infections were effectively asymptomatic, early numbers on what was happening are hard to take at face value.
According to [1], it's 40.5% of the confirmed population. It's indeed a high rate, but those cases are also taken into consideration wrt to mortality rate as far as I am aware. It also highlights just how transmissible this virus is, allowing it to eventually infect the vulnerable population.
We also learned that covid was a component in a very large portion of excess deaths. Yes, somebody may have died of cancer, but covid was also present. It's possible that the patient sans covid would not have died.
Plus, there is the inverse, long covid and its symptoms lingering and weakening the population making it susceptible to subsequent diseases.
The problem however, in all of this, is that any trust to certain institutions in the US is now gone, which will only cause more harm.
> In this meta-analysis of the percentage of asymptomatic SARS-CoV-2 infections among populations tested for and with confirmed COVID-19, the pooled percentage of asymptomatic infections was 0.25% among the tested population and 40.50% among the confirmed population.
https://www.fda.gov/media/144245/download#page=18
This is the first point where 300 odd people were excluded from the analysis because of protocol deviations. Worryingly, this number was much higher in the treatment group. While I hope that blinding would have prevented any motivated exclusion, it's still quite concerning.
His second main point (which is actually really interesting), is that there were 3500 cases of symptoms consistent with Covid-19 that weren't revealed by a PCR test. I'm not sure what to think of this (and note that this number swamps the actual numbers of confirmed cases by about 10x), except that it matches with anecdotal evidence I've seen around negative tests in the presence of symptoms.
The details of the second point can be found here: https://www.fda.gov/media/144245/download#page=42
As a meta-point, the world would be a better place if more people clicked on linked and attempted to interrogate the soruces of things.
reply