Thursday, May 21, 2020

Fancy some serology? (2)

I thought I would just take a break from trying to simplify the Protons electron transport chain as regards ultra low fat diets and talk about sensitivity and specificity of serology tests for a break.

People may have noticed I'm quite keen on serology and am rather less than enthusiastic about PCR for test, track and trace in a situation where the SARS-CoV-2 virus is present throughout the country, as it is here in the UK. Stupid is as stupid does.

However, serology is not quite as straight forward as I might like either.

There are a number of serology tests coming on to the market, and many have a 100% sensitivity and 98% specificity. It difficult to express how phenomenally accurate these test are. If I submit a blood sample for some routine analysis I accept that 95% for these sorts of accuracy assessments is pretty good, we're dealing with biological systems, there is room for grey zones.

So presently serology has a 100% sensitivity. That means it will always pick up seropositive people. If you have antibodies, this test will find them. Always. Getting a negative needs some thought.

This is addressed by specificity. A 98% specificity means that a negative on the test will be correct 98 times out of 100. If the test says you don't have antibodies, it is also most likely correct, a one in fifty error rate there.

It is difficult to over emphasise quite how good these values for sensitivity and specificity are for a lab test. They are very, very good.

At detecting antibodies.

If antibodies are found in a healthy person it is, with a test this good, pretty well certain that they have been exposed to the disease and, in the absence of illness, that they are immune. Or at least they were at the time of exposure.

Sadly human immune systems can be recalcitrant in cooperating with serology.

The Royal College of Pathologists short presentation on serology is now up on Youtube

The COVID 19 pandemic: testing – serological diagnostics for COVID 19

and here is a screenshot from just after 18 minutes in:

The dotted red line is the lower limit of the serology assay used. All of the patients have had known, absolutely certain, clinical disease. If you use a serology test which is 100% sensitive and 98% specific, you will pick up everyone over the red dashed line. A negative result will be correct 98% of the time.  That is what a highly sensitive, highly specific test does. To put that in a more visually clear image here is another screenshot:

Again, below the red dotted line you will be classified as seronegative, you are seronegative. That does not mean that you have not been exposed. It doesn't matter how good your test is. The test cannot see below the red line, above the red line the test is phenomenal. This not a problem with the test, the test is not for exposure/recovery from the disease. It is just for antibodies above a certain level. This is the limit of serology testing, it is undermined by the ability to recover from this infection without seroconverting. It happens, it's on the graphs. It's not the test's fault.

To a large extent recovery without seroconversion suggest that the innate immune system is at work (or you are simply unable to become infected) and that would fit nicely with the reports suggesting that hyperglycaemia over 10mmol/l is bad news and hyperglycaemia below 10mmol/l gives a slightly better outlook in severely ill patients. Hyperglycaemia is a good way of suppressing the innate immune system.

Passthecream put an interesting link in the comments of the last serology post which suggests the innate immune system is also adaptive, it remembers, no antibodies needed...

Adaptation in the Innate Immune System and Heterologous Innate Immunity

Sadly, at our rudimentary level of understanding of the immune system, we are in no position to assess whether a given person might be seronegative but still immune.

Having said all of this, it's worth remembering that being seropositive without having being ill suggests you are immune. It's interesting to see the WHO position on this. The WHO currently suggests that there is no evidence that having antibodies confers immunity.

That is interesting and absolutely, currently, technically correct. Thus far seropositive people have never been challenge-tested with virulent virus, so there can be no evidence that antibodies are protective (try getting that one through ethics committee review!). It is theoretically possible that a person could have been exposed to virulent virus, have never been ill, have developed antibodies, and yet is still be susceptible to the virus. You can imagine that this might be the case.

Well, actually, I can't.

So, if you are the head of an ITU in a UK district general hospital in London and you find you are one of those lucky people who are solidly IgG seropositive without ever having been ill, what would you do as regards PPE for yourself?

As the WHO says, there is no evidence that being seropositive is protective, as yet. But for antibodies produced in vivo, by someone who was never unwell, for these antibodies not to be protective would have to be a first of a kind as regards immunology (vaccine induced seropositivity is a whole different ball game).

I love this guy:

COVID-19: ICU care, long-term effects and immunity with Dr Richard Breeze

(Hat tip to Unknown for the link and no, Breeze didn't use any PPE while treating the large wave of COVID-19 patients which passed through Lewisham District General Hospital's upgraded ITU)

Knowledge over protocol. He also strikes me as the sort of person who might look at a patient on a ventilator who was developing barotrauma because "protocol" suggest "Xml/kg" as the "correct" tidal volume setting and who might reach over and reduce (gasp) the tidal volume setting. Just my guess. Or avoid intubated ventilation if at all possible (which was what they did).

You have to contrast this with the hospital managers who discharged SARS-CoV-2 positive patients in to unprotected nursing homes because "it's protocol".

I get the impression that good medics (and there are some excellent ones out there) don't seem to be the sort of people that become the politico-medics who guide the government...


Aside: I just can't get over Dr Breeze working without PPE. Sort of thing I might have done under the circumstances. I can't believe it was allowed nowadays!


Sondra Rose said...

I appreciate you putting this all together with such coherence!

Gyan said...

How does WHO define immunity?
Is it not developing symptoms of the disease?

I have never understood why it is not possible for a person to develop antibodies but maybe in inadequate amounts or maybe deploy these antibodies in some way imperfect and thus be susceptible to the infection again?

Why it is assumed or is it known that the above scenario is just not possible?

Peter said...

You're welcome Sondra.

Gyan, the WHO is very careful what it says. Having antibodies per se is not invariably protective, but having those antibodies because you have recovered from an acute viral infection (whether you showed signs or not) is usually so. In the current pandemic no one, other than Dr Breeze, has proved that those antibodies are protective...


Passthecream said...

There are a couple of preliminary reports along these lines atm, encouraging iff they prove to be accurate:

"we detected SARS-CoV-2−reactive CD4+ T cells in ∼40-60% of unexposed individuals, suggesting cross-reactive T cell recognition between circulating ‘common cold’ coronaviruses and SARS-CoV-2."

Davidp said...

Thanks, but I'm confused by this bit. You say: "If you have antibodies, this test will find them. Always." And also: "If the test says you don't have antibodies, it is also most likely correct, a one in fifty error rate there."

But if the test says you don't have antibodies, and it is wrong, that means you _do_ have antibodies. But by the first statement if you do have antibodies the test will find them always, so it won't say negative. So something seems a bit askew in this explanation. I guess if the test always finds antibodies when they are there, then the remaining error it can make is a false positive, saying you have antibodies when you really don't.

Eric said...

A chief concern is that coronaviruses do not tend to trigger long-lasting immunity. About a quarter of common colds are caused by human coronaviruses, but the immune response fades so rapidly that people can become reinfected the next year.

Researchers at Oxford University recently analysed blood from recovered Covid-19 patients and found that levels of IgG antibodies – those responsible for longer-lasting immunity – rose steeply in the first month of infection but then began to fall again.


Last week, scientists at Rockefeller University in New York found that most people who recovered from Covid-19 without going into hospital did not make many killer antibodies against the virus.

Peter said...

Nice Pass,

Hi Davidp, Making a test have 100% sensitivity is easy. No need to do any lab work, just declare all samples positive. You will miss no one. Of course your false positive rate will be very high, unknowably so... This is why you need specificity. While 100% specificity would be nice absolutely no test can do this. It's a continuous trade off between avoiding false negatives and false positives. The test is remarkably good.

Eric, it's a coronavirus. No one would expect anything else. If you get a year's immunity you'll be doing well. The vaccine will be pretty useless too. As always, try not to be diabetic is the best plan. Note I've dropped the "try not to be elderly" component of my standard advice.