PCR Test Caveats After Discharge Following Covid-19

To the Editor—We read with interest the paper published by Yuan et al [1]. The authors suggest recrudescence of 24 cases of coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The authors did not report whether the viral nucleic acid recovered was viable or replication-competent SARS-CoV-2. Nor do they present sequence data that could support the assumption that this was full-length virus. From South Korea, press reports described 91 patients testing positive again after having recovered from COVID-19. The findings by Yuan et al and the press reports from South Korea may have large implications for public health measures (eg, quarantine procedures). However, we question their findings and suggest that they may be explained—at least in part—by assay sensitivity and specificity.

Without a gold standard for true COVID-19 status, a positive reverse transcriptase–polymerase chain reaction (RT-PCR) has become a proxy for the presence of disease. This creates the problem of confusing test sensitivity (conditional probability of a positive test given true presence of disease) with the positive-predictive value (PPV) (conditional probability of true presence of disease given a positive test). While sensitivity is inherent to the test, the PPV further depends on disease prevalence. This may have affected the interpretation of the study by Yuan et al.

With 9812 RT-PCR tests per million, South Korea has been leading the world in testing their population [2, 3]. In an estimation of COVID-19 prevalence, we multiplied the number of new cases on 11 April 2020 [4] with a disease duration of 14 days and multiplied this with a dark figure estimate from the Danish health authorities of a factor of 70, although a (not yet peer-reviewed) study from Austria estimated a dark figure factor of 9) [5]. Prevalences were 0.004% (China), 0.06% (South Korea), 0.5% (Japan), 10.5% (United States), 3.2% (Denmark), and 7.5% (United Kingdom).

Assuming that RT-PCR has a sensitivity of 99.0% and specificity of 99.99%, the PPV of a positive test would be 99.99% in a population with a true COVID-19 prevalence of 50% (eg, hospitalized patients). However, PPVs vary as a function of prevalence and would be 29.8% (China), 85% (South Korea), 98% (Japan), and above 99.7 for the United States, United Kingdom, and Denmark. Thus, 1 out of 6 with a positive RT-PCR in South Korea and 2 out of 3 in China might not have COVID-19. Although our calculations come with certain assumptions, they illustrate that the low PPV in the situation with broad testing may, in itself, cause problems with interpretation in the absence of a confirmatory test. The problem of false positives is amplified by extended testing in low-prevalence populations.

In conclusion, a positive RT-PCR test is not the same as true COVID-19 nor does a positive RT-PCR test during follow-up imply infectiousness of the individuals unless viral replication competency has been proven. The issue of false-positive tests may affect the public’s adherence to and trust in quarantine measures.

Note

Potential conflicts of interest. T. B. reports unrestricted research grants from Pfizer, Novo Nordisk Foundation, Simonsen Foundation, GlaxoSmithKline (GSK), Gilead, and Lundbeck Foundation and personal fees from GSK, Pfizer, Boehringer Ingelheim, Gilead, and Merck Sharp and Dohme, outside the submitted work. H. F. L. reports personal fees from LEO Pharma, Sanofi, and Astra Zeneca, outside the submitted work. S. S. reports no conflict of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.