The Covid-19 Files (5): Test for Antibodies NOW

A test for antibodies to SARS-CoV-2 does not have to be extremely reliable at the individual level in order to provide valuable information about the incidence of Covid-19 in a population. Oh, and I had a letter published in New Scientist!

A few days ago I decided my New Scientist (NS) magazine (dated 21st March) had been left in direct sunlight long enough to destroy any SARS-CoV-2 virus particles on the outside. Nevertheless, I deployed a bomb-disposal robot to remove the no doubt biodegradable plastic wrapper, then destroyed the wrapper with a flame-thrower. And then destroyed the robot with a small thermonuclear device.

I wasn’t disappointed with my magazine. There’s a useful article by Michael Marshall explaining how the virus works (paywall, but the article appears to be reproduced here, not sure what that’s all about – you can even get the pictures by downloading the PDF, which is odd for a paid subscription magazine, but let’s not digress too much).

I leafed through the rest of the NS to see if there was anything else to read, and then, well, you know how your own name jumps out of a page of text? Mine did, on the Letters page. I’d moreorless forgotten that I’d written to them on 2nd March:

Editor’s pick – Flaws and a ray of hope in pandemic policy (3)

From Tim Joslin, London, UK

You report that, in some countries, many new covid-19 cases can’t be traced to their source of infection. A test exists for the virus itself, but is it also possible to deploy one for antibodies to the virus?

Such a test would help detect transmission chains by revealing those who have recovered from covid-19. It would also allow its morbidity and mortality rates to be more accurately determined, simply by randomly sampling the population of an infected area.

The editor writes:

We have since reported online that many labs are trying to develop tests for the antibodies (6 March, newscienti.st/NS-tests).

I’m not really sure why this letter was published, since the issue of a test for antibodies to SARS-CoV-2 had become a topic of discussion since I wrote. Maybe they should try to reduce the time between publication of articles and correspondence related to them to less than 3 weeks.

Needless to say, though. the UK is yet to deploy an antibody test, though other countries seem to be doing so.

At the same time, as I have previously pointed out, we seem to be deluding ourselves into believing the Covid-19 mortality (and serious illness) rate is lower than the global scientific consensus suggests. The estimate (pdf) of 0.5 – 1%, depending on demographics (mortality is much worse in older people) and the availability of healthcare (so we really don’t want the NHS to be overwhelmed), seems to me to be fairly robust, e.g. for South Korea where a great deal of testing and contact-tracing has been carried out. Why it’s not believed in the UK probably has more to do with a fixation in parts of the epidemiology community with achieving “herd immunity”.

Let’s have an intermezzo:

“When the truth offends, we lie and lie until we can no longer remember it is even there. But it is still there. Every lie we tell incurs a debt to the truth. Sooner or later, that debt is paid.”

Chernobyl TV series, quoted by Marina Hyde in the Guardian, 21/3/20.

Since it seems some key decision-makers are not going to be happy until we determine the prevalence of Covid-19 in the UK – including mild and asymptomatic cases – we’d better get on and sample the population for antibodies to the disease,

There is complete confusion as to when antibody tests will become available in the UK. At yesterday’s (Wednesday 25th March) Downing Street press conference, Boris was flanked, as on previous occasions, by his two fall-guys, Chief Medical Officer Chris Whitty and Chief Scientist Sir Patrick Vallance.

From what was said at the press conference, it seems the tests are still being validated (yes, apparently the UK has bought 3.5 million tests without knowing whether or not they actually work), so there is indeed uncertainty about when you’ll be able to order them from Amazon. One of the two scientists burbled about how dangerous it would be to release tests that gave false positives – that is, an incorrect indication that you have had Covid-19, and would therefore have become immune and could safely return to work and a normal existence – or false negatives, an incorrect indication you had not had Covid-19, when you in fact had, which could cause you to subsequently self-isolate to avoid infection or take other precautions unnecessarily.

But individual diagnoses are only one of the reasons to deploy antibody tests.

The key question we need to answer is the level of infection in the UK. That would tell us whether there is an “iceberg” of unsymptomatic cases or whether we have to believe the evidence from the rest of the world about the proportion of serious cases and hence mortality to expect. At yesterday’s (Wednesday 25th March) press conference the PM cited the Oxford modelling study, that incredibly irresponsibly suggested (or at least led the media to suggest) that half the UK population may already be infected with the virus. If true, that would imply a huge number of asymptomatic cases. Problem solved. Some believe or want to believe that is the case.

Here’s my point: why don’t we sample the population using the antibody tests we have even if they’re inaccurate?

I have one proviso: if the tests show positive for antibodies to other viruses (or anything else) as well as SARS-CoV-2, they’re probably not a lot of use.

But if, for example, the tests are insufficiently sensitive, we would still obtain useful information. In fact, we could calibrate the tests by using them on a sample of people who we already know have recovered from Covid-19. If we found the tests miss a proportion of these known cases we could apply a correction to the results of the random population sample.

Conversely, if the tests give spurious false positives we could also apply a correction. We’d need to use the tests on a population we were fairly sure had not been exposed to the virus (e.g. individuals self-isolating to avoid infection, in communities with no known cases) to determine the correction factor.

A sample of, say, 10,000 would be sufficient to provide regional Covid-19 incidence data as well as an accurate national estimate.

Some of the hundreds of thousands of NHS volunteers could be used to conduct the tests. If the results aren’t reliable for individuals it could be explained to them that the test is only indicative (this would hardly be new – it’s true of many diagnostic tests), or the results not even supplied.

I’m sure the government has already planned such an exercise (perhaps someone could check they’re good to go at one of the Downing Street press conferences).

My point is simply that we don’t necessarily need a 100% reliable antibody test to determine the incidence of Covid-19 in the population.


By way of a postscript, it has occurred to me that it must be possible that some of the aymptomatic “cases” aren’t cases at all. Maybe the existing test for the presence of the virus is giving some false positives, perhaps due to contamination of family members of Covid-19 cases. Maybe the virus can be present without you actually getting infected. I mention this just for completeness. I doubt it’s likely except in a few cases. (And, hot dog news, it turns out the Hong Kong Pomeranian was infected, so even that wasn’t a false positive).