The study population and testing methods have been described (1). Persons were recruited at a free, appointment-based, community antigen testing site in Winnebago County, Wisconsin, USA. Approximately 30 minutes after providing an initial nasal swab specimen for antigen testing, 2 additional self-collected specimens were collected under Centers of Disease Control and Prevention (CDC) staff supervision from the anterior nares simultaneously in an alternating fashion to maximize uniformity.

Of 2 simultaneous swab specimens, we used 1 specimen for rapid antigen testing by the Abbott BinaxNOW SARS-CoV-2 Antigen Card Test, a point-of-care lateral flow test with results available within 15 minutes of specimen collection. We placed the other specimen in viral transport medium and transported it on ice to the Marshfield Clinical Research Institute laboratory (Marshfield, Wisconsin, USA) for RT-PCR testing. Specimens with a cycle threshold (C_t) value <37 for at least 2 of 3 SARS-CoV-2 gene targets (open reading frame 1ab, spike gene, and nucleocapsid gene) were considered positive, according to the instructions of the manufacturer (TaqPath COVID-19 Combo Kit; Thermo Fisher Scientific, https://www.thermofisher.com).

We attempted viral culture at a CDC laboratory for all participants testing positive by RT-PCR or antigen test by using Vero-CCL81 cells, which were inoculated with clinical specimens, and observed daily for 7 days (8). All cultures that had a visible cytopathic effect were used for RNA extraction and SARS-CoV-2 RT-PCR confirmation. Any specimen that showed a cytopathic effect, was positive by RT-PCR, and had a C_t >2 lower than that for the original clinical specimen was considered culture positive.

We collected symptoms at time of specimen collection, symptom onset date, and exposure history by using paper questionnaires and entered data into REDCap database version 11.0.3 (https://www.vumc.org/dbmi/redcap). Participants reporting >1 of 15 symptoms at the time of specimen collection were considered symptomatic. Possible symptoms were fever, rigors, nasal congestion, sore throat, shortness of breath, headache, diarrhea, loss of taste, loss of smell, chills, muscle aches, fatigue, cough, nausea, and abdominal pain.

We define known exposure as being within 6 feet of a person who tested positive for SARS-CoV-2 within the last 14 days for >15 minutes over a 24-hour period. We analyzed data by using SAS version 9.4 (https://www.sas.com). We made comparisons by using Kruskal‒Wallis test for continuous variables or χ² tests for categorical variables; statistical significance was defined as α<0.05. This analysis was reviewed by CDC and was conducted consistent with applicable federal law and CDC policy.

During November 16–December 15, 2020, we collected 2,112 specimen pairs that had valid results for PCR and antigen tests; most (56.3%) participants were symptomatic (age range 5–95 years, median 42 years). Of 2,112 specimen pairs, 334 (15.8%) were positive by RT-PCR, 269 (12.7%) were positive by antigen test, and 200 (9.5%) had recoverable virus (culture positive). Of the 200 culture positive specimen pairs that had a positive RT-PCR result, 191 (95.5%) had a positive antigen test result. Positive predictive value (PPV) of antigen test for culture positivity (191/269, 71.0%) (Table 1) was higher than PPV for RT-PCR (200/334, 59.9%). Virus was successfully isolated from 191 (71.5%) of 267 specimen pairs with concordant positive antigen/RT-PCR results, 9 (13.4%) of 67 specimen pairs with positive RT-PCR and negative antigen test results, and 0 of 2 specimen pairs with positive antigen and negative RT-PCR test results.

Figure

Figure. Box plots of C_tvalues among participants with recoverable virus who had concordant positive SARS-CoV-2 RT-PCR and antigen test results (A) compared with those who had positive RT-PCR and...

All participants who had culture-positive specimens and false-negative antigen tests were symptomatic (7/9; 77.8%) or had a known exposure in the past 14 days (5/9; 55.6%). Among culture-positive symptomatic participants, those who had false-negative antigen and concordant positive antigen/RT-PCR results were tested a similar number of days after symptom onset (median 2 days vs. 3 days) (Table 2). The 2 persons who had recoverable virus and false-negative antigen test results and who were asymptomatic at the time of testing had known exposures the day before testing. For those who had recoverable virus, nucleocapsid gene C_t values were significantly lower in those with concordant positive results (median 19.1, interquartile range 17.1–21.3) than those who had false-negative antigen test results (median 26.6, interquartile range 25.6–31.0) (p<0.0001) (Table 2; Figure).

Consistent with previous studies assessing the relationship between antigen tests, RT-PCR, and ability to culture virus (9–11), we found that SARS-CoV-2 was more likely to be recovered among specimen pairs for which antigen test and RT-PCR results were positive than among pairs in which antigen test results were negative and RT-PCR results were positive. Although some studies have shown similar PPV for viral culture when comparing RT-PCR and antigen tests (12), we found higher PPV for the antigen test than for RT-PCR (13), suggesting that antigen test positivity might be a better marker of infectiousness than a positive RT-PCR result. However, a small but nontrivial proportion of samples that had negative antigen and positive RT-PCR results had recoverable virus, suggesting that antigen tests are misclassifying some infectious persons as SARS-CoV-2 negative. This finding, consistent with those of similar studies (6–11), suggests that lower sensitivity of antigen tests when compared with RT-PCR cannot be attributed exclusively to lingering positive RT-PCR results for persons who are no longer infectious.

Symptoms on the day of testing for most infectious persons who had false-negative antigen test results suggests that CDC’s current antigen testing guidance, which recommends confirmatory RT-PCR testing after negative antigen test results for symptomatic persons in community settings (14), is appropriate. Both asymptomatic participants who had false-negative antigen test results and recoverable virus had exposures within the previous 48 hours. Therefore, all participants who had false-negative antigen test results were unlikely to infect others if following CDC guidance because they would have been advised to quarantine because of exposure (asymptomatic close contacts) or while awaiting confirmatory RT-PCR results (symptomatic persons).

One limitation of this study was that although recoverable virus is indicative of infectiousness, lack of ability to isolate virus does not necessarily imply lack of infectiousness (15). Symptom status was only measured at the time of testing. Because we did not attempt virus isolation on antigen-negative and RT-PCR‒negative specimens, PPV was the only reported measure of agreement between antigen test, RT-PCR, and recoverable virus in culture. Because RT-PCR testing was not performed with calibrators, we are not able to report values in copies/milliliter. Finally, this investigation assessed only the BinaxNOW antigen testing platform.

This study suggests that antigen test positivity is more predictive of infectiousness than RT-PCR test positivity. However, false-negative antigen test results can be obtained for infectious persons, especially among those with symptoms, supporting CDC recommendations to follow negative antigen testing among symptomatic persons with RT-PCR confirmatory testing within 48 hours (14).