There are two basic testing questions in COVID-19: how do we test for active infection (i.e. if someone is sick now) and how do we test for whether someone has had the virus in the past. We talk through both in the Testing Explainer. Until now, the primary approach for testing for active infection has been through PCR tests – a nasal swab is used, and a PCR procedure is used to amplify the virus.
Within the last week, the FDA has approved a second approach to detection of active virus, an antigen test. The NYT reports on it here and the FDA has a surprisingly helpful announcement here. Broadly, an antigen test works by looking to detect a particular signature of the virus (technically, a part of the “nucleocapsid”, the protein on the outer layer of the virus).
The big advantage of this type of test is that it is fast. It can work within minutes, even faster than the 15-minute PCR tests. The disadvantage is that it is less sensitive than a PCR test and, in particular, produces many more false negatives. That is: if this test says you have the virus, you very likely do. But if it says you do not, there is a reasonable chance that it’s a false negative.
Our expert contributors had a spirited discussion of the value of this test on our Slack channel. It focused around two questions: (1) How good is this test? (2) Is this solution good in terms of feasibility?
How Good is the Test?
This test is fast: it’s faster than even the fast Abbott test. Having said that, there isn’t a huge difference between (say) 10 and 30 minutes, so it’s not obvious there is a huge gain there.
There was some concern raised about the validation and sensitivity. The test was approved under an emergency authorization with the FDA, which is a fairly low bar at this point. If you read the package insert (fun!) you see that the test was validated with 143 samples (that is, they had 143 people and they tested them with this test, and a reliable PCR, and compared the results).
Of the 59 people who had COVID-19, the test only picked up 47 of them. That’s not a great performance. It’s worse than the fast Abbott machine, and much worse than a standard PCR (although the latter takes a day or two to get results from).
It remains to be seen how this test will perform in real hospital samples (could be better or worse) and of course this is only the first line – later versions may be better. And it is worth saying that some information may be better than none. One can imagine a system where doctors test with this and then also send off a PCR swab for a more comprehensive test. This may simply buy us additional options.
Does this Improve Feasibility?
Let’s imagine this test is similar, perhaps slightly worse, than the current PCR options. Another possible advantage would be to improve testing access. If we had a test that was really easy to perform, even if it was imperfect, it would add a lot of value. We cannot let the perfect be the enemy of the good.
One advantage of this test, in light of this, is it runs on machines which already exist in at least some doctors offices and elsewhere. Estimates suggest there may be 40,000 or so of these machines up and running around the country (office would have to order the particular antigen testing supplies, but not a new machine). This is an immediate increase in testing capacity.
On the flip side, the test requires the same testing supplies that are used for PCR (namely, nasal swabs), so it doesn’t fix that testing supply issue. And, of course, the fact that there is only one company that makes these machines and supplies is…complicated. The FDA has said they expect to approve more antigen tests soon.
Bottom Line:
Progress on testing is hugely important and basically any progress at all should be celebrated. This is a step forward, perhaps not a leap. It will be crucial to think about how to incorporate this testing capacity in a smart way.