I’ve spent most of the past two months pondering the same question many of you are currently pondering : did I actually have COVID-19 and get over it that one time this past winter that I got really sick with COVID-like symptoms?
In late February and early March I had a cough, fever, and a severe headache that lasted for more than two weeks. Since the day my symptoms started, I’d been entertaining the idea that I might have been exposed to COVID-19 a week earlier at a party in Massachusetts (where the Biogen outbreak later happened).
So on May 4, I got a COVID-19 serology test, both to find out if I’ve ever really been infected and to walk The Prepared’s readers through how to interpret their own antibody test results if and when they get them.
If my blood sample turned out to be positive for antibodies to SARS-COV-2, would that result even be meaningful? How confident could I be in my immunity? Could I still spread the virus to others? Is it safe for me to go back out into the world, or to visit my elderly relatives?
Antibody tests are becoming more widely available
The United States still has a persistent shortage of PCR tests, the ones that go up the nose or down the throat and look for viral RNA to show if a person has COVID-19. Antibody tests, however, look for the antibodies that the human immune system generates to fight COVID-19 to test whether a person has had COVID-19 in the past, and they are relatively abundant.
On March 16, the FDA made it possible for serology test-makers to speed up their development process and get tests out to the public quickly. Since then, people have flocked to antibody testing.
But with tests ranging wildly in price and process, it’s been hard to know which tests worked and what they could really do. Then, last week, the FDA pulled some antibody tests from the market.
“We unfortunately see unscrupulous actors marketing fraudulent test kits and using the pandemic as an opportunity to take advantage of Americans’ anxiety,” the FDA statement read.
Now, 12 antibody tests are approved under an emergency use authorization, and 200 tests are under review.
What it’s like to take an antibody test
I showed up to the Quest laboratory and found the parking lot pleasingly empty. In my mask and gloves, I carefully entered the lab. There was only one other patient in a cavernous waiting room.
I was nervous, even though I knew Quest didn’t test people who thought they currently had COVID-19. And the lab tells people not to come in for testing at all if they’re not in perfect health.
I waited for ten minutes, breathing shallowly, for my name to be called.
I asked the lab technician which COVID-19 antibody test I’d receive. Was it Roche, Abbott, or something else? She didn’t know. The test itself was a blood draw through the arm, drawing a full vacutainer, about a tablespoon of blood. I was out the door in a few minutes.
The technician had told me I’d get my results within five days. In fact, the results arrived overnight, so I saw them on my phone first thing the next morning. I went from blood draw to results in just 15 hours.
The test was an Abbott test, and I was negative.
Unlike a positive COVID-19 antibody test result, a negative is pretty unambiguous. Abbott says its test is over 99% sensitive, so a false negative is very unlikely. Despite my travel history and symptoms, I did not have COVID-19 in March.
When I told my friends about the results, the reactions were mixed. Some people congratulated me on testing negative, while others expressed regret that I wasn’t immune.
For my part, I’m happy that I tested negative. It means that I didn’t spread the disease to anyone else, and, if I can get through the pandemic without getting it, it means I won’t suffer the risks of post-COVID-19 syndrome. But on the other hand, if I do wind up getting it, that could be moot.
But what if I had gotten a positive test? How confident could I be in the results?
What would a positive antibody test result mean?
A positive antibody test would mean it was somewhat likely I’d been infected with COVID-19. That’s because serostudies and antibody tests have a major problem: when it comes to individual tests in areas where the virus has not spread very widely (i.e. the “attack rate” of the virus is low), false positives make up a large percentage of positive antibody tests.
It may seem hard to square the claim above with the fact that the makers of these tests advertise that false positives are unlikely. But there’s some math involved that gives rise to this counterintuitive result. Here’s how it works.
Attack rate in epidemiology is defined as “the proportion of people who become ill with (or who die from) a disease in a population initially free of the disease.” To calculate the attack rate during an outbreak, epidemiologists divide the number of people who were infected by the number of people who were at risk of illness. To be reliable, attack rate estimates have to account for ascertainment.
So for someone testing positive in an area with a low attack rate, a test with a 1% false positive rate could mean different things. Take the following examples:
- An 80% chance of a false positive if the attack rate were 0.25%
- A 50% chance if the attack rate were 1%
- A 25% chance if the attack rate were 3%
- A 10% chance if the attack rate were 9%
- A 4% chance if the attack rate were 20%
- A 1% chance if the attack rate were 50%
These results are calculated using the sensitivity and specificity of the test along with Bayes’ theorem, using the attack rate as a “prior probability” of infection.
So what Bayes’ theorem tells us is that if you live in an area with a 0.25 percent attack rate (e.g. a city of 100,000 people, and 250 of them have been infected), and you get a positive antibody test, it’s still far more likely that your test was wrong than that you’re a truly member of that very exclusive 250-person club in your town.
But if 20 percent of that city’s population (20,000 people) has had COVID-19 and you get a positive result, then yeah, you’re probably in that really large 20,000-person club.
For individuals, calculating the false positive rate is tricky
In my case, while waiting for my results I wanted to do the best I could to sort out if I could trust a positive result or not. To do this, I first had to figure out how common infections were in my area.
So I paired the ascertainment estimate from the New York serostudy with the confirmed case count from my county. I estimated we might have about a 3 percent attack rate; should I view a positive result as a 75% confirmation that I had had COVID-19?
Not necessarily. I also had to consider that I’d been in isolation since the end of February. Did I personally have a lower cumulative risk than the average person in my county, and a lower prior probability?
Or, did I have a higher prior probability because I had been in Massachusetts and spent two weeks with symptoms?
Then, if I came up positive, could I get greater certainty by testing again? That is, are false positives largely the result of idiosyncratic factors, or cross-reaction with non-COVID-19 antibodies that would be consistent by patient?
If it’s the latter, could I evade this problem by getting another test, like the Roche test that advertises an even higher 99.8% specificity?
Basically, it’s a big tangle to figure out if a positive result on any given COVID-19 antibody test is trustworthy.
So I asked myself the following question. If my antibody test came back positive, would I adjust my behavior just based on this test? Probably not. I’d probably continue to stay home, social distance, wear a mask in public, and avoid contact with others.
If I did come up positive, though, there was one thing I’d definitely be able to at least try doing: donate plasma. My doctor informed me that two separate clinical programs in the area had told her clinic they were looking for donors, and that the donated plasma might make it into treatments for current patients. For that alone, getting a positive result would be worth it.
But as I said, I got a negative result, which, given the low attack rate in my area and the quirks of Bayes’ theorem, means I almost certainly did not have COVID-19 that time I got sick and thought I had it.
Should you get a test?
Overall, my experience shows that this kind of testing is pretty painless, and I’m glad to know that I didn’t have COVID-19 in March. In this sense, I can give the system a positive review.
People who live in higher-attack areas like New York will have even more straightforward experiences than I did. Because of the high-attack calculation we did earlier, we know those people can have relatively high confidence in a positive result.
So, should you get tested? I’d say, yes, with conditions. If the cost of the test doesn’t feel meaningful, why not? But make sure you feel comfortable going out in public first.
The test is definitely worth it if you’re either:
- located in a very high-attack area or
- prepared for a positive result to be maddeningly lacking in definitive meaning.
If you have a choice in the kind of test you get, the Roche test is probably the best one. But that choice might not be available to you. I haven’t been able to confirm where and how it’s available. Roche did inform me that Labcorp would be offering the test soon, but I couldn’t confirm whether Labcorp will offer it directly to the public the way Quest has with the Abbott test. And we might get data later which confirms that the Abbott test is as good as the Roche one.
Overall, antibody testing for the general public, despite its many caveats, is a tool we should all welcome in the COVID-19 pandemic.