Roche, Abbott COVID Antigen Tests Just So-So in Real-World Data

Two widely used rapid screening tests for COVID-19, which check for SARS-CoV-2 antigens, missed a significant fraction of infected individuals when used in a real-world setting, researchers reported.

Among 2,215 "all-comers" tested at a German diagnostics lab, with 338 ultimately showing positive results with PCR testing, Roche's SD Biosensor and Abbott's Panbio rapid antigen tests showed overall sensitivities of 60.4% and 56.8%, respectively, reported Christoph Wertenauer, of Synlab GmbH in Augsburg and the University of Heidelberg in Mannheim, and colleagues.

"The sensitivities of the two [rapid diagnostic tests] are unsatisfactory. This calls into question whether their widespread use is effective in the ongoing SARS-CoV-2 pandemic," the group wrote in a manuscript posted to the medRxiv preprint server.

They estimated that if 10,000 asymptomatic people were tested in a population with a 5% true positivity rate, these products would yield 124 false negatives. Testing 10,000 asymptomatic individuals from a population with less COVID-19 prevalence -- 0.5% -- would give 34 false negatives, but a higher rate of false positives (18 in all, or more than one-quarter of all positive results).

"Policy and health care providers should account for substantial limitations of [rapid tests] for SARS-CoV-2," Wertenauer and colleagues noted, especially "in asymptomatic persons. Research into alternative approaches to the screening for SARS-CoV-2 should be intensified."

Previous studies have indicated that rapid antigen tests, which can give results in just a few minutes and in some cases are sold over the counter for at-home use and interpretation, are less sensitive than gold-standard PCR assays that check for viral genetic matter. However, most of these studies had smaller patient numbers and/or were conducted in select samples.

In contrast, the Synlab study analyzed results for individuals presenting for routine testing, and sample collection for both the antigen tests and PCR was performed by trained lab workers.

Most of those tested were referred by physicians (32%) or government health agencies (43%), but the other 25% appeared to seek testing on their own initiative.

Not surprisingly, Wertenauer's group found that positivity rates, both with antigen tests and PCR, were higher in those referred by professionals, presumably because of symptoms or exposure. However, over half the patients tested had no symptoms.

Sensitivities for the antigen tests were especially low -- under 40% -- when the individuals had no comorbidities, were asymptomatic (just 23% for the Abbott test and barely 30% for Roche's), or had previously recovered from COVID.

In addition, the antigen tests also did poorly when PCR analysis indicated that the so-called alpha variant (B.1.1.7) was involved, which the researchers could not explain. They noted that earlier lab studies had found no loss of sensitivity with this variant.

Positive and negative predictive values were generally good when estimated for the entire group of 2,215, at well above 90%, but negative predictive value plummeted for those with no symptoms.

Also, as seen in previous studies, sensitivities for the antigen tests were inversely correlated with the PCR cycle threshold values -- indicating that the rapid tests did better when individuals had relatively high viral loads.

Wertenauer and colleagues suggested that the antigen tests were likely to be even less accurate when nonprofessionals or the individuals themselves collect the samples and run the tests.

The study did have limitations, of course, which the authors acknowledged. PCR testing is not perfect, they noted, and cycle threshold values are only loosely correlated with infectivity and may vary substantially from one lab to another.

Comment