The importance of getting a good antibody test - the maths and science behind it

[chomalungma]

Lab tests aren't perfect. There will always be false positives and false negatives.

Getting the test right and using it in the right people is so important.
I used to be a Clinical Biochemist (a long time ago) until life got in the way.

The key words are specificity and sensitivity.

Sensitivity - say it is 99%. That means that if 100 people have the disease, 99 of them would test positive

Specificity - say it is 99%. That means that if 100 people did not have the disease, then 99 of them would test negative.

This is really important when you screen people.

Imagine 10% of the population you screen have the antibody - so they have been exposed.

10,000 people tested. 1000 have the antibody, 9000 don't.

People with the antibody - 99% sensitive

990 positive (true positive)
10 negative (false negative)

People with no antibody

99% specificity
9801 negative (true negative)
99 positive (false positive)

So out of 1089 positive tests, 990 are actually positive.

So if you have a postive test, you have a 90% chance of having antibodies.

That's assuming 10% of the population have had the disease and the test has 99% sensitity and specificity

Well I think it's interesting

I do too! Although I have dyscalculia so my maths is ropey but I try

The key is:

1. How many people in the population have the disease?
2. How good are your tests?

If you know those answers, you can then plug them in and see what comes up

Interesting post thank you.

I’ve never understood though- are the false negatives and positives down to the test or the testee? In other words, if you repeated the test with a fresh sample (maybe taken at a different time?) would you expect to see the same results or not?

So if the sensitivity and specificity are too low for purpose, could you improve the accuracy of the test by doing two tests? If you got two positives or two negatives then you would be more certain, and if they differed then a third test would clarify the position.

Or is it something about the testee and their antibodies/samples that causes false results?

A whole new question

Repeat a test and you might get a different result on the same person

Let's say your sodium level is 136 .

I could do it 10 times - and the result could vary between 130 - 140 - because the result may vary every time.

It's complicated - the result may change even if you had the same sample and did 10 tests on it, and the result may change if you took the same person and did many tests on them

I could repeat the same test and get a different result.

’ve never understood though- are the false negatives and positives down to the test or the testee? In other words, if you repeated the test with a fresh sample (maybe taken at a different time?) would you expect to see the same results or not

Lab tests are funny things. You are always going to get variation when you do a test. I could do the same test on the same sample and get a range of results - machine tests tend to have less variation than human tests because machines just do these things better.

And if you took a sample one day, and then took another sample from the same person the next day, you could also get a different result.

1 of the tests I have seen works this way.

You have coloured particles coated in the Covid-19 antigen. This is the protein on Covid -19 that the body recognises as foreign.

The particles are mixed with the sample from the patient. Any human antibodies (Ig G) will bind to the particles.

The particles then move up the membrane.

On a line in the membrane, there are antibodies to human Ig G. These then bind to the human Ig G antibodies that have bound to the coated particles.

So basically you have a sandwich. On the line, there are anti - Ig G bound to Ig G antibodies to Covid -19 - which are bound to the coated particles.

If there are no Ig G antibodies, then there is no sandwich.

Then basically a line appears - to show if it positive.

It's the same principle as a pregnancy test

www.assaygenie.com/covid-19-15-min-rapid-poc-test

It doesn't tell you the level of Ig G antibodies you have.

This particular test seems reasonably sensitive - but they have only done it on limited samples and only compared it to PCR - and not to the Gold standard

And a bit on PCR

www.assaygenie.com/2019-novel-coronavirus-2019-ncov-triplex-rt-qpcr-detection-kit/

The enzyme used is Taq Polymerase.

Remember in biology how you were told that proteins generally work best at 37 C. Not this one. It comes from Thermus Aquaticus - a species of bacteria found in deep sea vents and can tolerate high temperatures.
en.wikipedia.org/wiki/Thermus_aquaticus

So this enables the amplification of DNA to be very fast because it takes place at a high temperature.

(I am really missing science at the moment)

This is really interesting- thank you!

Do you think we'll be getting the antigen test anytime soon? Or whichever the one is that tells you you've already had it. Why is that more difficult? Is it because once you've had it, it's a more sensitive test needed to pick that up?

This is really interesting. Do you think the media are being irresponsible politicising the testing issues? Is it likely that behind the scenes there are complications and they're trying to get things like the quality of the test right?

Why is that more difficul

It takes time to evaluate it. It's basically a pregnancy test - and you then have to step up manufacturing of it.

They need to get it right - because with this test, you are looking for a line to see if you have antibodies. Chances are that you won't get many false positives - but you could get quite a lot of people coming back as 'negative' who do have antbodies - and we all know what looking for a 'line' means when you look at the MN threads on pregnancy tests.

(I do have to point out that I left this field for complex life reasons many years ago - but I am kind of regretting it now and would love to get back into it)

Is it likely that behind the scenes there are complications and they're trying to get things like the quality of the test right

Tests need to be good. They need good quality control and they have to know how accurate they are - which needs lots of evaluation and trust in the manufacturers.

They should really be compared to the gold standard. You need to know if someone has had Covid-19 by using lab culture (not PCR) and then you do evaluation of this test - and you need enough patients - probably at different time intervals.

I would also throw in analysis with more complex antibody tests that give you an actual level of antibody so you can evaluate the sensitivity of it. It can be hard to see a line

So what sort of timescale do you think.we might be looking at here in the uk?

Sorry, to be clear, time til a widespread antibody test being available.

Honestly - I have no idea. I would guess a lot of manufacturers are working really hard to get a reliable test

This scientist explains it quite well - especially on the anti-body tests.There is a time lag from having had the virus to being able to detect antibodies - and this must be what they are working hard to evaluate

This is useful

www.sciencemediacentre.org/expert-comment-on-different-types-of-testing-for-covid-19/

They keep talking about an antigen test for Covid-19.
But it seems that this 'antigen' test is actually talking about detecting the RNA - which is not an antigen

A bit more on the tests

The reagents are running low - as well as the equipment needed to do it

www.theguardian.com/world/2020/apr/03/uk-ministers-struggle-to-keep-promise-of-100k-coronavirus-tests-by-end-of-april