Interview with Dr. Cassandra Kelly, director of Emerging Threats at FIND.
Why is FIND assessing the performance of COVID-19 tests?
How many brands of COVID-19 tests are around?
Cassandra Kelly: Within weeks have seen over 500 tests. But the total number is likely over 1000. There are hundreds under development.
What is the role of tests in the current phase of the pandemics?
Cassandra Kelly: As countries emerge from lockdown, it’s critical to find new infections, trace the contacts, and limit how many other people become infected. Several countries are using rapid diagnostic tests (RDTs) for these purposes. [RDTs are quick tests, some of them looking like a pregnancy test, that aim at detecting either active infections through antigens or past infections through antibodies].
RDTs raise doubts: Spain and the UK bought lots that did not work, for example.
Cassandra Kelly: RDTs need to have as high sensitivity as possible, because you don’t want to miss cases: low sensitivity would yield a lot of false negatives that would end up in another surge of the disease. In the current phase, when cases start to drop and prevalence becomes low, also specificity has to be high. We want to make sure that a positive is actually a positive. Otherwise, we could burden the healthcare system by treating false positives.
Are there good enough rapid tests available?
Cassandra Kelly: In general, specificity is OK in quick tests. But not sensitivity. In antibody tests, it can range from less than 10% up to 80% to 90%. Moreover, the sensitivity of antibody tests depends a lot on the day. They have the lowest performance in the early days of infection, when the organism produces few antibodies. Then performance increases, between 8 to 14 up to 20 to 25 days after infection. And then it goes down again. In antigen tests, the sensitivity ranges from 16 up to 93-94%. We are not seeing anything coming up to 98-99%, which is where we need to go for a reliable test. In order to get a high sensitivity, by now you need to resort to nucleic acid [PCR tests] or ELISA test [another test that requires laboratory processing].
But why not resort directly to PCR, then?
Cassandra Kelly: PCR is the optimal test for active infections, the most sensitive and specific. The challenge is that it requires a laboratory infrastructure and time. In the case of a new outbreak, these testing strategies would face a new bottleneck. In the case of a low prevalence scenario, like the one we are in, quick tests yield an enormous number of negatives, and many of them may be false negatives because of low sensitivity. So you would need to funnel anyway a lot of cases to PCR.
So is there a reasonable way to use quick tests?
Cassandra Kelly: For active case finding, antigens tests should be used to triage potentially active infections. Antibodies test should only be used much later in the course of infection, and always in conjunction with an antigen test or a PCR test. For public health measures, like contact tracing, antigens tests could be used to screen whether there is somebody actively infected, although asymptomatic. Antibody tests should be used for seroprevalence: large screenings of the population to understand how many people may have an immune response. From an epidemiological perspective, they may be useful to understand the risk we have of a new wave and how big it may be.
What wrong uses of tests have you seen around?
Cassandra Kelly: Antibody tests are being used too early. Quick tests are also being used to understand whether somebody is infected. If I get a negative, I assume I don’t have COVID-19 but maybe just an allergy instead. That misuse is extremely impactful for sustaining active transmission of the virus. It’s critical that it does not happen. The other mistake is related to immunity, for example when tests are used to decide whether a person previously exposed can go back to work. While their body has launched an immune response, we still don’t know whether they have sufficient immunity, and how long it will last.
What advice would you give to government and companies regarding quick tests?
Cassandra Kelly: Right now, I would say: be cautious about what you are buying. Don’t just look at the manufacturers’ claims about performance. Make sure you see independently generated data. Also: decide the type of test based in what questions you are trying to answer. Third: you can use low performing tests if you have a strong PCR and ELISA infrastructure to complement them. But if you don’t have this infrastructure, then it’s better not to use tests, until better ones appear. Instead, apply social distancing and hygiene. Apps that ask a couple of questions to users to diagnose them are not a good alternative, because they are even less accurate then tests.