What is the false positive rate for SARS-CoV-2 tests?

[Jeff_Lichtman]

I’m sorry if this has already been answered somewhere in this forum. I did a search and didn’t find any actual numbers (although there’s been a lot of discussion of false positives and negatives).

I want to know the false positive rate for tests that detect whether someone is currently infected with SARS-CoV-2. I don’t mean the antibody test, but the test (or tests) for the infection itself. If there’s more than one test, it would be good to get numbers for all of them. Citations would be helpful.

The reason I’m asking is that I’m engaged in a discussion with a non-Doper about the meaning of a positive test. She seems to think that a positive test for the virus means it’s 100% certain you have the disease. We know that’s not true, and that if the infection rate is low enough, even a small false positive rate means a positive test isn’t a reliable indicator that someone is infected.

Thanks for any help you can give.

[DSeid]

For false positives I can find this.

Review of external quality assessments revealed false positive rates of 0-16.7%, with an interquartile range of 0.8-4.0%. … <snip> … the likeliest source of these false positives is sample contamination or human error. … <snip> … We suspect that under-appreciation of the potential impact of false positives in SARSCoV-2 testing has two sources. First, because PCR-based diagnostic protocols are effectively designed to eliminate false positives due to cross-reactivity, some individuals assume that false positives have been eliminated from the test process. In practice,
however, minute levels of contamination, which are extremely challenging to control, can produce false positives in PCR-based tests, and the potential for human error in sample handling or records management can probably never be entirely eliminated.

I was surprised that it is that high.
This is the most recent review of false negatives I am aware of.

Researchers at the Cleveland Clinic tested 239 specimens known to contain the coronavirus using five of the most commonly used coronavirus tests, including the Abbott ID NOW. The ID NOW has generated widespread excitement because it can produce results in less than 15 minutes.

But the ID NOW only detected the virus in 85.2% of the samples, meaning it had a false-negative rate of 14.8 percent, according to Dr. Gary Procop, who heads COVID-19 testing at the Cleveland Clinic and led the study. …

… The Cleveland Clinic plans to rely on other tests that performed better in the study, Procop said. That includes the test developed by the Centers for Disease Control and Prevention, which detected 100% of positive samples. Another test, made by Roche, detected 96.5% of positive samples. The fifth test in the study, made by Cepheid, detected 98.2% of infected samples, Procop said. The Cepheid test produces results in less than an hour. …

And while you are not asking about it, the Roche antibody test is said to be very good. Claim at least of “a specificity greater than 99.8% and sensitivity of 100% (14 Days post-PCR confirmation)” I am not aware of independent verification. Others not so good. Really need to know what product was used in any study that comes out.

Obviously you already understand how Bayes theorem and priors apply.

[Melbourne]

@Jeff_Lichtman

I want to know the false positive rate for tests that detect whether someone is currently infected with SARS-CoV-2. I don’t mean the antibody test, but the test (or tests) for the infection itself. If there’s more than one test, it would be good to get numbers for all of them.

I think there are two kinds of tests: PCR and Antigen. PCR – DNA amplification – is what has been used so far, and the test itself seems to have zero false positive: either you have DNA that can be amplified, or you don’t. It takes 24 hours because it takes that much amplification to get a recognizable result.

Antigen tests aren’t antibody tests. They use something like antibodies to test for the presence of antigens. Antigens are those parts of the virus that are antigenic, that is, those parts of the virus that are detected by antibodies These tests are new, and fast, but aren’t amplifying, so they depend on you having enough virus in you so that they can see the result.

Cross-immunity can be a problem with antigen tests. You may test positive for small pox, but actually only have cow pox. I would imagine that new COVID19 tests are tested for cross-reactivity with other corona viruses, but I haven’t seen any numbers.

Not an expert. No citations. Maybe there is some other kind of virus test I don’t know about.

[Corry_El]

This article gives a summary of manufacturer claimed results for various antibody tests, not tests for active infections. As other answers noted those are pretty different things.

Covid-19 antibody tests face a very specific problem

There is no general answer AFAIK where independent testers have scientifically verified each of those results. There are reports of very poor accuracy for non-brand name antibody tests not on that list, and some reports of lower accuracy than that table for a few of the entries. However, that includes effects from improper use of the test, contamination etc. Obviously the manufacturers are not claiming the tests produce the stated results even if the facility administering the test, or a lab, screws it up. Then again maybe some some tests are easier to screw up. IOW it’s not necessarily a simple answer.

Also, sticking with antibody tests, we have to consider that very high specificity is needed for false positives to be a small % of all positives, and that also depends on what % of a give population is actually positive.

So to take the lowest claimed specificity in that table 90.6% specificity, that looks not so low but would generate basically useless results in a population where only 1% really had antibodies. Taking the specificity simplistically, it would mean 9.4% of the 99% who didn’t have antibodies would test positive, 9.3% test positive w/o antibodies, 1% really have them (not quite all of whom would test positive).

However even assuming that the several tests claiming 100% specificity really only have 98% specificity, in a population with 10% actual positives (It might be higher than that where I live, I tested positive on a test claiming 100% specificity) it would be around 1.8% of population false positive, 10% who really have the antibodies, perhaps 9% have them and test positive (if the sensitivity were 90%). Those are probably quite useful results, especially if you were say tracking the general trend of the spread of exposure, rather taking an individual viewpoint of ‘I have to be absolutely sure either way’.

Also keeping in mind that every day medical tests generate wrong results not rarely, but it isn’t Issue 1 or politicized like the COVID response has inevitably become. My sister in law got a false positive result for serious cancer (in a different developed country than the US, it’s not a one country issue). Scary, though all’s well in that case when you find out it’s wrong.