The FDA just approved the first antibody test for the Coronavirus, made by Cellex. The test has a significant false positive rate though (see here). Now for a lot of tests, false positives can be reduced by retesting positive cases, as the probability of two false positives for the same person is usually low. But this tweet by epidemiologist Zachary Binney raises the possibility that retesting may not help in this case:

Running the test 2x & only telling someone they're immune if they get 2 +s might help reduce false +s, depending on the error source. If it's anything systemic - say it's detecting antibodies from a similar virus that don't grant immunity to COVID-19 - no good.

My question is, is anything known about the sources of error for Cellex’s antibody test, and whether it’s anything systemic along the lines that Binney outlined?

  • We require a link to the claim that this has a high false positive test rate. Otherwise it's just a rumor and not suited for this site. – Graham Chiu Apr 4 at 3:08
  • @GrahamChiu No, it is suitable for this site because we lowered the standards of evidence for COVID-19 questions. However, I do agree that Keshav should provide a source for his false positives statement. – Carey Gregory Apr 4 at 4:22
  • @CareyGregory See this chart by Binney: twitter.com/zbinney_nflinj/status/1245789676264460288?s=21 I think it’s calculated using either the sensitivity or the specificity of the test (which are 93.8% and 95.6% respectively). – Keshav Srinivasan Apr 4 at 7:27
  • You need to edit your question and add that link. – Carey Gregory Apr 4 at 22:03
  • 1
    @CareyGregory OK, I added it. – Keshav Srinivasan Apr 5 at 1:38

You can see all the tests being produced here http://www.centerforhealthsecurity.org/resources/COVID-19/Serology-based-tests-for-COVID-19.html

Many of the tests share similar sensitivities and specificities.

The Cellex test looks at both IgG and IgM antibodies.

RDT, lateral flow assay, which detects IgM and IgG to the nucelocapside protein of SARS-CoV-2. The sensitivity is 93.8% and specificity is 95.6%, when tested at 2 Chinese hospitals in a total of 128 COVID19 positive patients, and 250 COVID19 negative patients (as detected by RT-qPCR).

So they are approximately picking up 4 per 100 as positive that don't have PCR evidence of infection. This could mean that the test is actually more accurate because some of these false positives did actually have the disease in the past, or, it is detecting an antibody against another coronavirus like SARS or MERS. They have already controlled for the seasonal coronavirus.

False positives can occur due to cross-reactivity with antibodies from previous infections, such as from other coronaviruses.

This may be less of a problem in the USA population and also if you have a history compatible with infection the more likely the test is a true positive.

It will be more interesting if they can get validity data from a USA population.

Edit: there is a new paper from nature medicine suggesting that the sars-cov-2 might have been circulating for some time, even years before it jumped to become more lethal. This might explain the false positive rates.



| improve this answer | |
  • there is a new paper from nature medicine suggesting that the sars-cov-2 might have been circulating for some time, even years before it jumped to become more lethal. This might explain the false positive rates. my words from the beginning! But this is the first time that I hear something similar. Never in discussions or in news or in talk shows with experts this assumption or possibility has been mentioned. It would have been very, very sensationally accidentally if cause Zero found in Wuhan would have been cause zero of the world. – Albrecht Hügli Apr 8 at 11:32
  • What is the 95% CI for 11 cases (=95.6%) in a sample of 250 (the specificity test)? I think the lower end is something like a specificity of 93%, or a false positive rate of 7%, which is unusable for determining whether there is, say, a prevalence of 1% or 3%. – Peter - Reinstate Monica May 16 at 5:16
  • Andrew Gelman has a pretty detailed discussion of the specificity problem on his blog. – Peter - Reinstate Monica May 16 at 5:19
  • @AlbrechtHügli Why would it be sensational? It's rather likely that it is some Chinese place, given the population density, customs and proximity to animals. The role of the laboratory there is also still open (and the case may indeed never be closed, but that is a sensationally unlikely coincidence, conspiracy theories aside). – Peter - Reinstate Monica May 16 at 5:20
  • @AlbrechtHügli There is apparently 50 years of evolution between the bat coronavirus and sars-cov-2. There's no surprise it has to have been somewhere in the wild mutating over that period. – Graham Chiu May 17 at 2:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.