Note: I will be on vacation from 8/8 through 8/22. I will not be writing daily Q&As while away [unless the mood strikes]. Please feel free to keep sending me your questions and know that I’ll answer them when I return, if not sooner.

Question: I just took an antibody test in DC! How much weight should I put on the results?

Answer: Short answer: the validity of the test results depends as there are several important issues at play, described below. In terms of the results, here’s my take:

• If you live in DC, had been sick with COVID-like symptoms earlier in the year, and you get a positive result back from the antibody test, the odds are pretty favorable that it’s a true positive;
• If you had not been sick with COVID-like symptoms earlier in the year and you get a positive result, the validity of the test is murkier;
• If you get a negative result, you can be assured that you do not have B-cell antibodies, but you won’t know anything about T-cell immunity; and
• We still don’t know how protective antibodies are and for how long, and even a positive antibody result is not a definitive positive, so please keep doing all the right public health practices — mask wearing, 6+ feet distance, hand washing, staying home when sick.

Here are the issues affecting result validity:

1. Your own health history: if you had an illness earlier this year that was accompanied by COVID-like symptoms (described in Q&A of 7/6 #Symptoms) then a positive test result is more likely to be a true positive.
2. The sensitivity/specificity of the test: Every diagnostic test has its own sensitivity/specificity (e.g. ability to detect presence/absence of disease; described in Q&A of 4/15). As CDC describes, since population prevalence of SARS-CoV-2 is still low, tests with higher specificity are preferred.
3. The population-level prevalence of the disease: The ability of a diagnostic test to accurately capture true positives/true negatives depends on the overall prevalence of the disease in the population. The lower the population-level prevalence, the higher the false positivity rate. I really like the Q&A of 4/15 on this topic, so if you’re curious check it out! For the US, CDC shares prevalence estimates here) and shared a write up on the first round of prevalence estimates in this paper published three weeks ago in JAMA, which concluded “it is likely that greater than 10 times more SARS-CoV-2 infections occurred than the number of reported COVID-19 cases.”
4. Lasting antibodies: Antibody tests look for the presence of B cells, which are the cells (lymphocytes) that identify invaders and basically mark them for other immune cells to destroy (like military intelligence) (for immune system refresher, see Q&A of 5/9; nice graphic from Science commentary in Figure 1). It seems that for some people who have recovered from COVID, their B cells quickly dissipate and may not be present in antibody tests. But B cells are only part of the immune response. Remember T cells? T-cell immune response could be long-lasting even if B cell response is not (see Q&A of 8/6). So no presence of B cells in the tests does not necessarily mean no presence of some underlying immunity. That said, COVID is still new and we have no idea how long immunity lasts or how protective antibodies are. So far, the fact that we have little evidence of re-infection is quite reassuring (more on that in Q&A of 7/23).

And diving into DC specifically, according to the Washington Post, DC is using 2 antibody tests — DiaSorin Liaison assay or the Abbott Alinity i serology test. The FDA shares sensitivity/specificity estimates here, and I’ve copied the Abbot table here for reference (Table 1). As the FDA estimates, the Abbot test has a positive predictive value (PPV) (i.e. probability that a positive is a real positive) of 84% when prevalence is 5%. Restated — there’s an 84% chance that the positive result is a true positive. If the population prevalence is <5%, the PPV will be much worse. If it is >5%, PPV will be much better. We don’t know what the population prevalence is in DC, but we could make a few guesses. DC has a population of ~700,000 and has 12,518 cases. If DC were finding all cases, the prevalence would be ~1.8%. But we know that we’ve been missing cases. If we extrapolate from CDC’s findings described above — that for every case we found in March-May, we missed 10 cases — then prevalence would be MUCH higher (getting closer to 20%). That is far too high given hospitalizations and the like in DC. Plus, according to the Washington Post, only 6% of antibody tests have been positive, and since there’s likely bias in who is getting an antibody test (people more likely than the general population to have been infected), the population prevalence would presumably be lower than that. All that to say, the FDA’s use of 5% prevalence to estimate positive predictive value is probably a pretty good choice for the DC-area.

Figure 1. Immune Response (from Science)

Table 1. Abbott Sensitivity/Specificity (from FDA)