Validation of the value of rapid tests

A new study has found that antigen tests are more informative than PCR tests for infection

Welcome to Plugging the Gap (my email newsletter about Covid-19 and its economics). In case you don’t know me, I’m an economist and professor at the University of Toronto. I have written lots of books including, most recently, on Covid-19. You can follow me on twitter (@joshgans) or subscribe to this email newsletter here.

Rapid tests are cheap and they produce fast results which is why many people are advocating their use for population surveillance but critically for clearance for entry into places. Others have criticised their accuracy but, as regular readers, know some of those criticisms are misplaced if your interest is whether someone is infectious rather than just infected. Today’s newsletter reports on a new study that indicates that rapid tests are potentially more valuable than PCR tests for the purpose of finding and isolating infectious people.

The Theory

A month or so ago, I wrote about a new paper of mine that demonstrated that rapid antigen tests may well be more informative than PCR tests if you are interested in determining whether someone is infectious at the moment rather than just infected. The starting point for this idea is captured by this recent diagram from the New York Times.

You can see that while you are infected from exposure onwards (for 21 days or even more), you are only infectious at a later and shorter period (that may overlap with symptoms should you get them).

Now PCR tests are a ‘gold-standard’ for finding the virus and can detect it at very low levels (or even when the virus itself is no longer active). Thus, over the course of your infection you are more likely to receive a positive sign from a PCR test than from the rapid antigen tests. But, as the following diagram shows, this is because the rapid tests only read positive if you have a higher viral load.

What my paper showed was that, if you were interested in whether someone was infectious, a rapid test was more likely to read positive if you are infectious compared with a PCR test that will read positive even if you are not infectious. Thus, if you were looking to isolate people who were infectious, it was possible that a rapid test has around the same false negatives as a PCR test but a lower false-positive rate. That is, for infectiousness, a rapid test was equivalent (rather than worse) than a PCR test in terms of sensitivity but superior in terms of specificity.

The Evidence

A valid criticism of my paper was: how do you know? What was the evidence that the rapid tests were picking up infectious people at about the same rate as PCR tests? What was the evidence that rapid tests have fewer false positives? This was just a theory but the precise conditions and thresholds for these tests surely matter.

I am happy to report that, while I was writing my paper, a study was being conducted to collect the information to test that theory. Here is what it found: “When compared to infectious virus isolation, the sensitivity of antigen-based testing is similar to RT-PCR.” In other words, rapid tests were not less accurate than PCR tests at all if you are interested in infectiousness.

The key table is here:

There was just a single false negative for the antigen test but there were 7 fewer false positives. In other words, the antigen test was slightly less sensitive but more specific than the PCR test.

You can also see from the data where the antigen test had different results from PCR. There was increasing agreement as the viral load was higher (i.e., a lower Ct score).

This study is a good first step but it does not tell us whether a viral load was indicative of infectiousness or not — that is, some threshold or range. In this regard, that still appears to be a judgment call. From what I understand, at viral loads that medical professionals judged to be infectious (e.g, a Ct score of less than 18), there appears to be no difference between rapid antigen tests and PCR tests in catching positive cases.

What did I miss?