[ExI] ;covid test

Stuart LaForge avant at sollegro.com
Sun Oct 17 17:21:32 UTC 2021

Quoting BillW:

> Report today:  Thousands of BRits may be false negatives.
> Well, duh, if you don't have a perfect correlation and make a cutoff you
> are going to have false negatives and false positives.  The question is:
> where do you set the cutoff?  I assume that the worst thing is a false
> negative.  For a false positive you test again.  For false negatives?
> What?  Anyone know?
> billw

Yeah. . . Test again in 5 days; false negatives are largely due to  
timing and the way viral infection works and not a test flaw. So for  
example, the day someone is in infected, they will always test as a  
false negative, that is to say that on day 1, the false negative rate  
(FNR) is 100%. The FNR declines over time as the virus infects more  
cells and starts to become detectable in nose swabs. On day 5 of  
infection when symptoms usually start, the FNR is still 67% and that  
is still  lot of false negatives. The test reaches minimum FNR at 20%  
on day 8, 3 days after the onset of symptoms. Then the false negatives  
rise again so on day 21, the FNR is back to 67%


Overall, in this pooled analysis, false negative RT-PCRs were least  
common 3 days after symptom onset. The rate of false-negative RT-PCRs  
was highest the day of infection, were lowest 8 days after infection,  
and then began to rise again.

Study population:

Pooled analysis of a mix of inpatients and outpatients with SARS-CoV-2  
infection in 7 studies of RT-PCR performance in the upper respiratory  
tract by time since symptom onset or exposure (n = 1330 respiratory  
Primary endpoint:

Estimation of false-negative rates by day since infection.
Key findings:

Over the 4 days between infection (day 1) to the typical time of  
symptom onset (day 5), the probability of a false-negative result in  
an infected person decreased from 100% on day 1 (95% CI, 100%-100%) to  
67% (CI, 27% - 94%) on day 4.
On the day of symptom onset (day 5), the median false-negative rate  
was 38% (CI, 18% to 65%).
On day 8 the median false-negative rate decreased to 20% (CI, 12%-  
30%), and then began to increase again (21% [CI 13%-31%] on day 9).
On day 21 the median false negative rate was 66% (CI, 54% -77%).

Quoting Darin Sunley:

> If false positives are, and always have been massively pervasive, it means
> the entire narrative around "asymptomatic transmission" [You know, that
> thing that has never been a meaningful component of any upper respiratory
> tract pandemic in the history of mankind, and for which we turned all of
> western civilization into a dictatorship run by public health
> professionals? That thing?]  has been massively flawed, right from the
> beginning.
Yeah. For any test, it turns out that the false positive rate (FPR)  
goes up when the incidence of infection goes down. So in the case of  
the diagnostic RT-PCR tests with 95% sensitivity and 95% specificity,  
if 10% of your population is infected , then only about 16% of your  
positive test results are uninfected false positives. If only 1% of  
your population is infected, however, then about 84% of the positives  
will be false. Diagnostic testing is a numbers game and at the tail  
end of an epidemic most of the positives will be false. That does make  
the whole notion of asymptomatic transmission, kind of suspect. But  
sensationalism sells better than math.


"In summary, we have provided additional evidence that false positive  
SARS-CoV-2 PCR test results do occur in the clinical setting and are  
especially a problem in a low prevalence screening situation where the  
prior probability of a positive test is low. Although it is  
acknowledged that resource limitations may constrain the amount of  
retesting performed, we posit that the human and economic costs of  
considering all positive results to be definitive evidence of  
infection warrant an evaluation for the possibility that the result is  
falsely positive in an asymptomatic individual without known exposure  
to an actively infected person."

Stuart LaForge

More information about the extropy-chat mailing list