Interesting data on accuracy of PCRs - false negatives/positives

[Clariman]

I have recently had as positive PCR test so have been self-isolating. I had one earlier this year. In both cases I was asymptomatic and other tests contradicted the positive PCR, so I have been doing some reading. Earlier in the year the positive PCR was contradicted by a later antibody test which suggested that I had never had Covid-19. The current positive PCR is, so far, contradicted by 5 lateral flow tests: 1 from two days before the PCR, one the day after the PCR and then each day since.

https://www.gavi.org/vaccineswork/false ... ctable-pcr

The above article says that the UK Government's assumption (based on what research or data it doesn't say), is that there are fewer than 5% false positives or 5% false negatives with a PCR. So that means the chance of each of my +ve PCRs being a false positive is less than 1 in 20. My elementary understanding of probability leads me to think that means the chances of both being false positives would be less than 1 in 400. In other words it is HIGHLY likely that at least one of them was positive. Assuming the current one is, I am going to continue doing LFTs to see if they eventually turn positive, given that LFTs detect Covid-19 at a slightly later stage than a PCR. If they don't then I am at a loss to understand what is going on with me and Covid.

See also https://www.gavi.org/vaccineswork/how-l ... 9-be-wrong

Clariman

[9873210]

My elementary understanding of probability leads me to think that means the chances of both being false positives would be less than 1 in 400.

You are assuming independence. This might not be the case.

For example, 5% of the population might have been infected with an unknown virus that always gives a false positive. (Lots of other possibilities for non-independence exist).

[Clariman]

My elementary understanding of probability leads me to think that means the chances of both being false positives would be less than 1 in 400.

You are assuming independence. This might not be the case.

For example, 5% of the population might have been infected with an unknown virus that always gives a false positive. (Lots of other possibilities for non-independence exist).

Do you mean something persistent or transitory? I've also has 6 negative PCRs in 2021, the last one on 22nd November.

[Midsmartin]

My understanding is that lateral flow tests have quite a high false negative rate.

Antibody tests can also have false negatives of there were few antibodies at the time.

A PCR test is extremely sensitive, so could detect traces of virus. Or as you say, perhaps your had two false positives (from contamination of the sample in the lab or whatever).

Unlikely things happen all the time in a large population. So it's all a bit mysterious and I don't think we can solve it easily.

[9873210]

My elementary understanding of probability leads me to think that means the chances of both being false positives would be less than 1 in 400.

You are assuming independence. This might not be the case.

For example, 5% of the population might have been infected with an unknown virus that always gives a false positive. (Lots of other possibilities for non-independence exist).

Do you mean something persistent or transitory? I've also has 6 negative PCRs in 2021, the last one on 22nd November.

It could be either or neither, I'm addressing the math, the biology is a gloss. I can come up with lots of plausible causes for correlation over many time frames and I'm sure a biologist can shoot many of them down.

The statistics could be address experimentally by tracking many people over time. This would be tricky because even if there is some underlying cause for some false positives there are almost certainly many "random" false positives mixed in, so you can't test for one or the other.

Given the population size 1/400 events will show up many times, you could get another three false positives without any need to reach for a non-random explanation.

[Midsmartin]

I can't quickly find a link but I'm pretty sure I heard about research into healthcare workers who failed to develop covid despite exposure and they didn't necessarily have detectable antibodies. You might just have an immune system that can clear covid unusually quickly. If you're lucky.

[Hallucigenia]

Those probabilities only apply for infections above the detection threshold of the respective tests - and the PCR tests are much more sensitive than lateral flows. The most plausible explanation is that you currently have a low level infection which PCR can detect but not lateral flows.

The antibody tests are a bit patchy, particularly if you take them some time after the infection - I had a negative antibody test last year some 9 months after I'm pretty certain I had Covid.

[stockton]

A friend works in a testing lab in on the continent. 2000 PCR tests were delivered at the end of last week; they tested about a dozen on the staff, most of which gave positive results.
After testing with alternative tests the 2000 will be returned to the supplier or put in the bin.

[Clariman]

I've just caught up on some of the Covid Zoe study work. They have identified that about 20% of people who had tested +ve for Covid-19 via a PCR test, later test negative for antibodies, just like I did. This the anti-N test which looks for infection-based antibodies rather than vaccine-based ones.

Out of 8,193 contributors who tested positive, 6,609 (80.67%) had a positive anti-N antibody test result - so they had Anti-N antibodies (1).

It continues...

Looking more closely at the data, we saw that people who had a greater number of symptoms while they were ill with COVID - particularly the ‘classic three’ symptoms of cough, fever and loss of smell (anosmia) - were more likely to have gained antibodies against the virus.

In contrast, people with one or more health issues associated with an underlying condition (known as comorbidities) were less likely to have antibodies after being infected, as well as people who currently smoke (1).

So my asymptomatic Covid-19 PCR test followed by a negative anti-N antibody test is not so unusual. It does, however, beg the question whether all those +ve PCR tests with no symptoms were false positives or not . The article doesn't address that.

From a purely personal note, what do I think of my own circumstances? With <5% false positives from a PCR (2), the balance of probability is that I had Covid earlier this year. The lack of Covid symptoms made it questionable, but 35% of Covid-infected people are asymptomatic according to a recent study, so not that unusual (3). The negative antibody test after my positive PCR combined with the lack of symptoms led me to think I hadn't had Covid, but now I know that 20% of those with a +ve PCR will test negative for antibodies and most of those had fewer symptoms, all these results are beginning to look less contradictory.

So five days after my recent positive PCR and I am still showing no classic covid symptoms. Maybe I am just one of the lucky ones who doesn't get badly hit by Covid. That would still surprise me given my age and other health factors, but I'll take it. Mind you all my lateral flow tests have shoed negative, including today's.

(1) Zoe Study article 9th November 2021 https://covid.joinzoe.com/post/covid-in ... protection
(2) UK Government paper 3rd June 2020 https://assets.publishing.service.gov.u ... atives.pdf
(3) Proceedings of the National Academy of the Sciences USA 24 August 2021 https://www.pnas.org/content/118/34/e2109229118

Any scientists, medics, public health officials who can confirm the validity of these thoughts?

C

[stockton]

A friend works in a testing lab in on the continent. 2000 PCR tests were delivered at the end of last week; they tested about a dozen on the staff, most of which gave positive results.
After testing with alternative tests the 2000 will be returned to the supplier or put in the bin.

After further investigation it looks like I need to correct my account. The unreliable tests were antigen tests, not PCRs.

[Midsmartin]

The results of a PCR test depends on how many cycles of chain reaction it goes through. With each cycle the number of viral DNA strands doubles (or nearly so). Do enough cycles and you can pretty much pick up single molecules from contamination or remnants of previous infection, or not enough cycles and you won't detect anything.

A bit more googling shows this:
https://www.ons.gov.uk/aboutus/transpar ... sitiverate

Where the false positive rate from pcr tests is given as <0.005%. See the answer to Q 7 at the bottom of that link where there are further useful links.

So it's nearly certain that you had covid in you, unless the samples were contaminated in the place you did the test , or in the lab.

[gadjet]

Hi folks

Clariman said "LFTs detect Covid -19 at a slightly later stage than a PCR."
What time-frame is that ?

I have never knowingly had Covid -19, so have sometimes been a bit lax about doing frequent LFTs. I am going out and about a lot more now than I was earlier this year. If I have been in a crowded place eg bus or train, how soon afterwards should I do a LFT to expect an accurate result (either positive or negative) ?

TIA

Sue

[Clariman]

A bit more googling shows this:
https://www.ons.gov.uk/aboutus/transpar ... sitiverate

Where the false positive rate from pcr tests is given as <0.005%. See the answer to Q 7 at the bottom of that link where there are further useful links.

So it's nearly certain that you had covid in you, unless the samples were contaminated in the place you did the test , or in the lab.

Wow! Looks like I have pretty definitely had Covid then! Positive PCR in April and one in December. Unlikely both to have been contaminated?? Thank goodness both were asymptomatic. Does that suggest that I would be asymptomatic if I had Covid again?

[Clariman]

Hi folks

Clariman said "LFTs detect Covid -19 at a slightly later stage than a PCR."
What time-frame is that ?

I have never knowingly had Covid -19, so have sometimes been a bit lax about doing frequent LFTs. I am going out and about a lot more now than I was earlier this year. If I have been in a crowded place eg bus or train, how soon afterwards should I do a LFT to expect an accurate result (either positive or negative) ?

TIA

Sue

I am NOT an expert but my understanding is that PCRs are best between about 2 to 5 days after infection. I think LFTs are when you would be infectious - maybe after 5 days. But I am not an expert and things may change with the new variant.

BTW I think the advice is to do LFTs before going out - to protect others primarily - rather than as a check of oneself after being exposed.

[Midsmartin]

"Does that suggest that I would be asymptomatic if I had Covid again"

It must improve your chances I imagine. But I'm no immunologist.

I believe there is thought to be a link between severity and the quantity of virus you were exposed to. Presumably a high dose gives your body a shorter time to respond as the virus starts to multiply.

Is your blood group O? This seems to be protective
https://www.spring.org.uk/2021/09/blood-typ.php

I think there are other genetic predictors but I forget now.

I also googled this:
https://www.cnet.com/health/medical/cov ... ow-so-far/

Which says that second infections are often less severe, but sometimes they are worse. Not everyone's immune system is the same, or will respond in the same way.
Which says

[DrFfybes]

I posted something like this a while ago, but can't find it.

The PCR, LFT, and Antibody tests are looking for different things.....

A Lateral flow is good (generally) for showing you're 'infectious' - ie. you've got live virus in your airways that has the potential to infect other people. It detects viral proteins in the swab from your airways, and for sufficient quantity you need to become infected and the virus replicate in your system and then be released again into the airways. Generally you will go positive 2 - 5 days post infection.

PCR tests for fragments of the viral genome and is much more sensitive as in theory it can detect at single molecule level. (In reality probably tens).
This means it can pick up early infection - partly as the viral replication cycle also produces bits of RNA in your cells which means that there's more RNA than there is actually viral particles. It is possible to fight off an infection without it entering your system, the non-specific immune system is a first line of defence and can stop a virus entering the body. A lot of people never seem to suffer from colds. They still catch them, they just fight it off before the infection establishes.

People can stay positive to PCR long after they have actually cleared the live virus because the PCR will pick up bits of viral RNA hanging around that are not infectious complete viral particles, but effectively bits of 'dead' virus and viral replication intermediates.


The antibody test lookd for prescence of circulating antibodies in your blood. These are generated when you get a 'proper' infection, although you might have been asymptomatic.


So basically The LFT (Antigen test) looks for viral proteins in your airways, indicative of live virus - if you are LFT positive, you really should keep yourself to yourself.

The PCR looks for prescence of viral RNA - you have been exposed, but might not be infected. If you have a high PCR Cycle threshold (a low amount of detected virus) it's possible to be PCR positive and have no culturable virus - see La Scola et al..

The Antibody test looks to see if you have been infected, AND elicited an immune response. Subsequent exposure/infection should be fought off better, the same as if you'e had mumps/chicken pox etc.

I wonder if Clariman is one of those irritating people who never suffers from colds and flu, as he fights them off before they get into his system.

Paul

[Clariman]

Many thanks for your explanations of the tests, Paul.

I wonder if Clariman is one of those irritating people who never suffers from colds and flu, as he fights them off before they get into his system.

Well I would love to be one of those people, especially given my age, weight and mild asthma - otherwise fit and well. I routinely get offered a flu vaccine because of the asthma but I probably only get round to taking it one year in three. I do not recall having had flu for decades. I've had the occasional nasty cold though.

When I called my GP in May 2020 ref something else and asked about Covid risk I distinctly recall him saying "Oooh, we're seeing a lot of people like you in serious difficulty with Covid - overweight, your age etc., so you should exercise extreme caution". Those highlighted words scared the living daylights out of me and I had been super careful up until about 5 or 6 weeks ago. I am still cautious.

That is why understanding all this is very important to me. If I can have a bit more confidence in my vaccinations and my natural immune system, then maybe I can get out more again .... just in time to catch the new variant

C