As those of you who have visited our offices during the COVID-19 pandemic will know, we have relied on rapid antigen testing as both a screening and diagnostic tool since they became available. There are a number of advantages to rapid antigen testing – but the main one is in the name – they’re rapid. With results in 10-15 minutes, rapid antigen testing was a cornerstone in our ability to reopen our office safely and confidently.
The major concern with antigen testing, unlike PCR which gets sent out to a lab and can take days to return, is reliability. Because antigen tests look for antigen proteins on the surface of the virus instead of traces of the virus’s RNA, they simply are less accurate. But how much less accurate? And, equally importantly, if used as a measure of one’s infectiousness (i.e. ability to transmit the virus), do we need a test to be as remarkably sensitive as PCR?
We have performed hundreds of antigen tests in the preceding months, and for a subset (a total of 144) of these we collected and sent concurrent PCR samples. Of the 144 that we collected, 2 had positive antigen results which were both subsequently confirmed by PCR. But, there were 2 additional samples that tested negative on the antigen test but later returned positive from PCR.
Two primary characteristics describe the clinical utility of a diagnostic test. The first is sensitivity – which is the probability that a test result will be positive when the disease is present (true positive rate). The second is specificity – which is the probability that a test result will be negative when the disease is not present (true negative rate). In our sample, the sensitivity of the rapid testing is poor, only 50% (error range: 6.8% to 93.2%) but the specificity is 100% (error range: 97.4% to 100.0%).
In a low prevalence population with a 2% infection rate, the rapid antigen test functions quite well. A positive test result would be expected to be a true positive 100% of the time. A negative test result would be expected to be a true negative result 99% of the time. Overall accuracy: 99%
However, in a high prevalence population, such as we are in now, with a 16.6% infection rate, this testing modality becomes significantly less accurate but not useless. A positive test remains 100% predictive but a negative test likelihood drops to 90.9%. Overall accuracy: 91.7%.
Our primary concern moving forward would be to substantially decrease the number of missed infections. For example, in our experience we have missed two times as many infections as led to believe. On the other hand, we successfully identified 140 of 142 negative infections. To quote Dr. Meagan Fitzpatrick writing in the Lancet (link: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)32635-0/fulltext):
“Should rapid testing become a viable, trusted screening strategy for control of COVID-19, performance characteristics should be well understood and screening strategies should be designed with test imperfections clearly in mind.”