What's the current state of US coronavirus antigen and antibody testing?

July 24, 2020 10:37 AM

In particular (almost said "specific"--accidental immunology humor?), what's happening with the accuracy of tests? In April, we were seeing antigen tests yielding up to 30% false negatives but almost no false positives. In May, that seemed to be flipped for antibody tests (lots of false negatives, almost no false positives), particularly the Quest Diagnostics test. But as July ends, what improvements have been made to these tests, if any?

Also, I realize this study was published a few days ago. It's a small sample and seems like it might be motivated by the author's own test development goals, but it's part of what made me ask the question.

posted by yellowcandy to Science & Nature (6 answers total)

False negative and false positive rates depend on how prevalent the condition is in the population being tested. If a disease is extremely uncommon in the tested population, any positive result is likely to be a false positive. A positive result becomes less likely to be a false positive as disease prevalence increases. You can see this explained more mathematically here.

Note that this is tested population, not overall population - so restricting tests to people who are extremely likely to have Covid will greatly reduce the number of false positives, whereas large-scale screening of asymptomatic people will have a much higher number of false positives, assuming the same test is being used in both scenarios.

Aside from that, another huge factor is that different tests (and testing procedures, personnel, etc) may have different accuracy.
posted by randomnity at 11:13 AM on July 24, 2020 [2 favorites]

Response by poster: (Not babysitting, but I want to make it clear that I'm not looking for information on how biostats or stats in general work; I'm more interested in the tests themselves.)
posted by yellowcandy at 11:25 AM on July 24, 2020 [1 favorite]

There was a review paper of some recent ELISA antibody tests, which I can't find. If I remember it correctly, it reported that such tests yielded false positives from the presence of immunoglobulins for common coronaviruses (such as HKU1 and others) as well as adenoviruses and other respiratory viruses, which could cause false signal.

An antibody test may also not work if administered too early in the course of infection. After infection, it is still unknown how long once-infected people have detectable levels of IgA, IgG and IgM proteins specific for novel coronavirus antigens, combined or individual measurements of which are used for assessing a positive/negative result. IgM tests seem to have low specificity, generally. Different assays may not work as well over time, for those same reasons.

It seems that antibody tests may be less useful as a diagnostic for individuals, and they may be more useful for assessing infected rates for populations.

Antigen tests are cheap and scalable but have their own problems. You need a sample containing enough virus particles to get a positive reading, and adding that sample dilutes antibodies involved in signaling a reaction. The antibodies in antigen tests try to bind to the N and S proteins of the novel coronavirus. However, other coronaviruses with similar-enough proteins could bind and cause a reaction. Antibodies can also react with each other, which yields false positives. Improvements in antigen-based assays will likely try to improve upon the technical aspects of the antibodies involved, to limit these issues.

It might also help to search for and read the protocols for these tests, which some companies make available. This may provide more specifics on what is being tested, and how tests change and improve over time.
posted by They sucked his brains out! at 12:59 PM on July 24, 2020 [1 favorite]

Quest says they are using this Abbott test and the EUROIMMUN Anti-SARS-CoV-2 ELISA assay. Loads of details at the links - I'm not sure where you got the idea that these tests produced "lots of false negatives."

Be aware that immunity can persist even after antibodies fade (link to posts by Dr. Florian Krammer).
posted by exogenous at 1:53 PM on July 24, 2020

I think this is an area of active research.

Another thing to think about, in addition to antibody tests, as well as the difference between specificity and sensitivity, is the purpse for which you are testing. You may be interested in this thread on rapid tests.
posted by latkes at 4:56 PM on July 24, 2020

An article today in the NY Times reveals that major commercial antibody tests are not directed against the spike protein but instead "look for antibodies to a protein called the nucleocapsid, or N, that is bound up with the virus’s genetic material." Antibodies to the nucleocapsid seem to wane more quickly than the neutralizing antibodies against the part of the virus that binds to human cells. “God, I did not realize that — that’s crazy,” said Angela Rasmussen, a virologist at Columbia University in New York. “It’s kind of puzzling to design a test that’s not looking for what’s thought to be the major antigen.”

Why? "The N protein is plentiful in the blood, and testing for antibodies to it produces a swifter, brighter signal than testing for antibodies to the spike protein. Because antibody tests are used to detect past infection, however, manufacturers are not required to prove that the antibodies their tests seek are those that actually confer protection against the virus." (emphasis added)
posted by exogenous at 12:21 PM on July 26, 2020