Covid 19 testing

lulukyriacou
Posts: 22
Joined: Tue Jul 28, 2020 5:37 pm

Re: Covid 19 testing

Post by lulukyriacou »

Guy, thank you for your excellent and informative answers much appreciated.

guy153
Posts: 48
Joined: Wed Apr 22, 2020 5:29 pm

Re: Covid 19 testing

Post by guy153 »

Have looked a bit more into this claim that you might be more likely to get a false positive if you are infected with another coronavirus that isn't SARS2.

It seems this mostly comes from a controversial off-guardian article called something like "COVID19 PCR Tests are Scientifically Meaningless". That article is full of distorted and misleading claims, one of which is that the test may confuse SARS2 with other coronaviruses and that the WHO seem to be encouraging this.

What seems to have happened is that a German lab developed a SARS2 test before SARS2 had even been sequenced, by looking for bits of SARS1 that they had in common (which presumably they found by guessing). The test found bits of the "E" gene that SARS1 and SARS2 share. But the sequences they were looking for are not found in any other known human coronaviruses.

Newer tests also look for other parts of the genome (most tests look for a few). So the question arose what conclusion should you draw if you match on 'E' but not on anything else? Either you have SARS1 (or some other unknown related virus) or, which is far more likely, the test just didn't find enough of that other gene to pass the threshold. The WHO concluded that if you match only on E then you might as well call that a match.

If you're using the test to decide whether to quarantine or trace contacts during an epidemic (such as happened in Iceland) it makes sense to treat an 'E' gene only match as a positive. You will not match on 'E' if you have any other cold (any more than the normal false positive rate). The bits they're looking for are only shared between SARS1 and SARS2 out of known coronaviruses. You probably don't have SARS1 but if you do, you should quarantine yourself anyway, since it's like SARS2 only worse. On this occasion the WHO are giving good advice.

The test is highly specific and will not match on other common cold viruses. It's not 100% specific because nothing is. Contamination and human error are the most likely causes of false positives. The false positive rate is low, nobody knows exactly how low-- certainly less than 5% but more than 0%. This just means that when prevalence is down to less than that you need to treat the results with a lot of caution. The low sensitivity is generally considered a bigger problem, but this matters when prevalence is high, which in the UK it isn't. But it will be a problem in Australia for example where the epidemic is just starting to kick off.

MRG
Posts: 7
Joined: Thu Jul 16, 2020 3:16 pm

Re: Covid 19 testing

Post by MRG »

Interesting article yesterday by Prof. Carl Heneghan:

https://www.cebm.net/covid-19/covid-cas ... heres-why/

The more you test the more you find. But are the results significant?

guy153
Posts: 48
Joined: Wed Apr 22, 2020 5:29 pm

Re: Covid 19 testing

Post by guy153 »

Yes, very good article. He's put his finger on the real problem, which is very real, impossible to argue with, and doesn't rely on outlandish claims about the test being "scientifically meanginless" or anything.

"The potential for false-positives (those people without the disease who test positive) to drive the increase in community (Pillar 2) cases is substantial, particularly because the accuracy of the test and the detection of viable viruses within a community setting is unclear."

This is exactly the problem.

MRG
Posts: 7
Joined: Thu Jul 16, 2020 3:16 pm

Re: Covid 19 testing

Post by MRG »

What I can't get my head around is why Pillar 2 testing gives far more positive results per 100,000 tests when the tests are done on a Wednesday, and fewest positive results when done on a Sunday or Monday??

There's definitely a weekly pattern; why should far more positive test results occur mid-week compared to the beginning?

MRG
Posts: 7
Joined: Thu Jul 16, 2020 3:16 pm

Re: Covid 19 testing

Post by MRG »

Another thing about the final graph in Carl Heneghan's article

https://www.cebm.net/covid-19/covid-cas ... heres-why/

the difference between the Pillar 1 and 2 trendlines is roughly 500.

500 in 100,000 is 0.5%.

Presumably the Pillar 1 tests in hospital are more accurate.

So I wonder how close this 0.5% is to the false positive rate of the commercial Pillar 2 tests - specificity of 99.5% anyone?

guy153
Posts: 48
Joined: Wed Apr 22, 2020 5:29 pm

Re: Covid 19 testing

Post by guy153 »

MRG wrote: Mon Aug 03, 2020 11:55 am Another thing about the final graph in Carl Heneghan's article

https://www.cebm.net/covid-19/covid-cas ... heres-why/

the difference between the Pillar 1 and 2 trendlines is roughly 500.

500 in 100,000 is 0.5%.

Presumably the Pillar 1 tests in hospital are more accurate.

So I wonder how close this 0.5% is to the false positive rate of the commercial Pillar 2 tests - specificity of 99.5% anyone?
They're also testing different populations. Pillar 1 is people in hospitals, Pillar 2 is people who turn up to carparks. So you would expect Pillar 2 to be higher.

I wouldn't be surprised if the Pillar 2 test kits are less accurate as well, but we can't tell this from this data.

guy153
Posts: 48
Joined: Wed Apr 22, 2020 5:29 pm

Re: Covid 19 testing

Post by guy153 »

Sorry, no you're right, Pillar 1 should be higher since that's in healthcare settings.

Pillar 2 are finding about 0.5% of the population infected. But the ONS Covid-19 Infection Survey of households (supposedly randomly sampled) are finding 0.05%.

You might expect some selection bias for Pillar 2-- you only bother to go to a carpark to get tested if you think you might test positive. But a factor of 10?

I think a more likely explanation is what you suggested: the Pillar 2 tests, who are using a different commercial test kit, have a specificity of about 99.5% and the ones the ONS are using are more like 99.95%.

Interesting that this also gives you a factor of 10 difference in the IFR. If the government believe their Pillar 2 testing is representative then there are lot of cases out there, which means the IFR (taking deaths and prevalence at face value from official stats and Pillar 2) works out rather less than seasonal flu... so why are we even bothering?

I suspect that the truth is that Pillar 2 test data is basically junk, the ONS data is quite good, and that the true number of active cases in the general population is around 0.05%. We can't estimate IFR without a better breakdown of how many hospitalizations and deaths are coming from care homes compared to the population outside.

TomK
Posts: 4
Joined: Wed Aug 05, 2020 12:34 am

Re: Covid 19 testing

Post by TomK »

guy153 wrote: Mon Aug 03, 2020 8:52 am Have looked a bit more into this claim that you might be more likely to get a false positive if you are infected with another coronavirus that isn't SARS2.

It seems this mostly comes from a controversial off-guardian article called something like "COVID19 PCR Tests are Scientifically Meaningless". That article is full of distorted and misleading claims, one of which is that the test may confuse SARS2 with other coronaviruses and that the WHO seem to be encouraging this.

What seems to have happened is that a German lab developed a SARS2 test before SARS2 had even been sequenced, by looking for bits of SARS1 that they had in common (which presumably they found by guessing). The test found bits of the "E" gene that SARS1 and SARS2 share. But the sequences they were looking for are not found in any other known human coronaviruses.

Newer tests also look for other parts of the genome (most tests look for a few). So the question arose what conclusion should you draw if you match on 'E' but not on anything else? Either you have SARS1 (or some other unknown related virus) or, which is far more likely, the test just didn't find enough of that other gene to pass the threshold. The WHO concluded that if you match only on E then you might as well call that a match.

If you're using the test to decide whether to quarantine or trace contacts during an epidemic (such as happened in Iceland) it makes sense to treat an 'E' gene only match as a positive. You will not match on 'E' if you have any other cold (any more than the normal false positive rate). The bits they're looking for are only shared between SARS1 and SARS2 out of known coronaviruses. You probably don't have SARS1 but if you do, you should quarantine yourself anyway, since it's like SARS2 only worse. On this occasion the WHO are giving good advice.

The test is highly specific and will not match on other common cold viruses. It's not 100% specific because nothing is. Contamination and human error are the most likely causes of false positives. The false positive rate is low, nobody knows exactly how low-- certainly less than 5% but more than 0%. This just means that when prevalence is down to less than that you need to treat the results with a lot of caution. The low sensitivity is generally considered a bigger problem, but this matters when prevalence is high, which in the UK it isn't. But it will be a problem in Australia for example where the epidemic is just starting to kick off.
The following is an extract from Page 2 (Intended Use) of the 'CDC 2019 Novel Coronavirus (nCoV) Real-Time RT-PCR Diagnostic Panel Instruction - Instructions for Use' (revision 13/7/2020):

"Results are for the identification of 2019-nCoV RNA. The 2019-nCoV RNA is generally detectable in upper and lower respiratory specimens during infection. Positive results are indicative of active infection with 2019-nCoV but do not rule out bacterial infection or co-infection with other viruses. The agent detected may not be the definite cause of disease. Laboratories within the United States and its territories are required to report all positive results to the appropriate public health authorities."

https://www.fda.gov/media/134922/download

My interpretation of the above extract is that any positive test results can be considered inaccurate unless a further thorough diagnosis is undertaken of each case. I wonder how often this is actually done, as there seems to be a genuine inclination to inflate numbers..

Granted, the test kit instructions are for the US; however I'm assuming that as the CDC is a primary source of all things COVID-19, that other countries are likely to be using the same or very similar testing protocol. If anyone can provide divergent sources/info, I would appreciate it.

guy153
Posts: 48
Joined: Wed Apr 22, 2020 5:29 pm

Re: Covid 19 testing

Post by guy153 »

TomK wrote: Wed Aug 05, 2020 1:07 am The following is an extract from Page 2 (Intended Use) of the 'CDC 2019 Novel Coronavirus (nCoV) Real-Time RT-PCR Diagnostic Panel Instruction - Instructions for Use' (revision 13/7/2020):

"Results are for the identification of 2019-nCoV RNA. The 2019-nCoV RNA is generally detectable in upper and lower respiratory specimens during infection. Positive results are indicative of active infection with 2019-nCoV but do not rule out bacterial infection or co-infection with other viruses. The agent detected may not be the definite cause of disease. Laboratories within the United States and its territories are required to report all positive results to the appropriate public health authorities."

https://www.fda.gov/media/134922/download

My interpretation of the above extract is that any positive test results can be considered inaccurate unless a further thorough diagnosis is undertaken of each case. I wonder how often this is actually done, as there seems to be a genuine inclination to inflate numbers..
It sounds on first reading like they're saying bacterial infection or co-infection with other viruses might cause the positive test. But they aren't, and it won't-- it's just that if you have a SARS2 infection you might also have some other infection, and it might be the other one that's making you ill.

This is good advice, especially at the height of the epidemic, when the prior probability of having a SARS2 infection is high. At that point if you test positive you probably do have a SARS2 infection, but you might also have a bacterial pneumonia and the doctor should make sure so she can decide on the best treatment.

Now that the epidemic is over, there's a similar problem: if you test positive for Covid in the UK now in "Pillar 2" you only have about a 6% chance of actually having Covid (it's more likely a false positive). It's still more likely, even given the positive test, that your symptoms are caused by a different virus or bacterium. The doctor needs to be careful not to assume it's Covid when you might need antibiotics instead.

Post Reply