In the Reggio Emilia cohort, 60.6% of positive SARS-CoV-2 cases diagnosed before 22 March achieved viral clearance, measured as first negative swab, by 22 April 2020. Median time to viral clearance was found to be 30 days from diagnosis and 36 days from symptom onset, with a trend that increased with increasing age and that was slightly longer in hospitalised patients, suggesting that clearance was slower in the more severe cases.
About one fifth (21.3%) of viral clearances in the follow-up period were not confirmed by the second swab, suggesting that there was a high rate of false negatives in this population.
The percentage of confirmed viral tests increased significantly as the interval between the first positive swab or symptom onset and the first negative follow-up swab increased. This result confirms the predictions of a model built based on the results of a number of reports on clearance.19
It must be noted that the endpoint of viral clearance can only be observed at the moment of testing, a negative swab does not tell us when clearance actually occurred, meaning that we only have a terminus ante quem. The longer the interval between tests, the greater the overestimation of time to clearance.
In this study it was not possible to assess the sensitivity of RT-PCR. Nevertheless, we considered the occurrence of a positive test after a negative one as a proxy of a false negative result, even if we could not exclude that some negative tests followed by a negative confirmation test might have been false negative results as well. However, repeated tests for SARS-CoV2 RT-PCR has been considered an acceptable reference standard in previous studies and systematic reviews.25 26
The testing protocol in this study was consistent with those recommended by ECDC, with a longer interval only for the second retesting due to the healthcare system overload yet reflecting real-world practice, including outpatient data.
As this was a population-based study, clinical information was not available of all included subjects. We therefore considered access to the ED and hospitalisation as proxy of disease severity. Even if this could limit the accuracy of disease severity assessment, the distribution of deaths in the three groups confirmed a strong association between hospitalisation and the probability of dying of COVID-19 in our cohort. Moreover, defining groups by healthcare service use was more appropriate to support public health decision making since this easily available information could be used to organise testing schedules.
The median time to viral clearance observed in our cohort was longer than that reported by two cohort studies of hospitalised patients in Wuhan, China, both of which had a follow-up of about 1 month. The first, which involved 191 subjects, reports a median of 20 days in survivors (IQR 17.0–24.0), with a maximum of 37 days from symptom onset. This study does not report any difference between patients undergoing antiviral therapy with lopinavir/ritonavir. However, longer intervals were observed in patients with more severe disease, as our results also suggest.27 As the inclusion criterion in this study was hospital discharge between 29 December 2019, and 31 January 2020, however, it is not clear whether patients with longer disease duration may have been excluded.
The second study reports a median of 23 days (IQR 18–32 days) between symptom onset and viral clearance. However, the median was calculated only for those patients who had had two consecutive negative swabs during follow-up (120/168, 71.4%). This way of estimating the median time may lead to an underestimation if the actual number of cohort subjects truly followed up is not taken into consideration. Further, the study reports that 86.7% of the 120 included subjects achieved viral clearance within 37 days of follow-up but that 10 subjects (8.3%) were still positive by day 40.28 This study also observes an increase in time to viral clearance with greater disease severity, with increasing age and in the absence of antiviral therapy.28
A recent case report states that viral shedding was detected up to 49 days from symptom onset.29
As Atkinson and Petersen discuss, to be able to use these results to make public health decisions, it must be remembered that RT-PCR can identify even fragments of the virus, meaning that subjects who do not have any active replication and are thus not infectious will nevertheless test positive.18
A number of studies have assessed the viral load in SARS-CoV-2-positive subjects in various biological matrices, reporting consistent results. These results describe a period of high viral load in the respiratory airways and, presumably, high transmissibility, starting about 3 days after symptom onset,25 with a peak in viral load identified between the day before and 4 days after symptom onset and a decrease in load starting from day 8 after symptom onset.30–33
Various studies have, however, detected a viral load 20–28 days from symptom onset,30–34 even when the virus itself was at times undetectable in the same period, reporting fluctuating results when the viral load approched the limit of detection of diagnostic systems.31 33
The half-life of up to 3 months of respiratory epithelial cells and the detectability of genetic materials from a live virus or even from fragments of dead virus by RT-PCR should be also considered to understand the inconsistency in negative results over a prolonged period.35 After a phase of active viral replication, estimated in 8 days from symptom onset,33 the persistence of dead virus fragments at concentrations close to the limit of detection could explain the unconfirmed negative test rate in the first weeks after clinical recovery.35
Some authors have shown that late positive samples have low viral load and scarce ability to infect cells in vitro,33 35 suggesting a low, if any, potential for generating new infections. Based on this, the WHO changed the recommendations to discontinue transmission-based precautions for COVID-19 patients.36 However, virological and epidemiological evidence on the risk of transmission during the convalescent phase characterised by positive RT-PCR is weak, and current serological data have not provided any additional insight.37 Furthermore, current epidemiological evidence of transmission has been influenced by how quarantine has been managed thus far.
The results concerning the differences in viral load in terms of disease severity are partially discordant.30–32 34 One study on 3497 samples of different biological matrices from 96 patients admitted to the hospital in Zhejiang, China, found different distributions of viral load in moderate and in severe cases, with a median time to viral clearance on samples taken from respiratory airways of 14 and 21 days, respectively, and a peak in the second week after symptom onset in patients with moderate disease and in the third week in patients with more severe disease. The authors also report longer viral persistence in patients over age 60 and in males.34 The median time to viral clearance in this study is shorter than that which we observed in our cohort, however, the inclusion criteria were also different, with only cases reaching a negative swab included, and testing was done much more frequently than in our cohort.
Other studies, instead, do not report any differences in viral load between symptomatic and asymptomatic subjects.4 32 38 Further, from the study on the entire population of the municipality of Vò Euganeo, 43.2% of the subjects who tested positive to SARS-CoV-2 were asymptomatic, and from the reconstruction of the chain of disease transmission, two of the eight new cases observed during follow-up had had contact only with asymptomatic subjects.38
These results have important implications for policies of tracing and isolation: they suggest the possibility that asymptomatic and pre-symptomatic subjects are as infectious as symptomatic subjects are, although perhaps for not as long.4 30–32 38
Our data indicate that testing at 14 days from diagnosis, as many regional surveillance protocols recommend, will result in most cases still being positive. So that at least half of these tests are negative, testing should be done after more than 4 weeks once patients are symptom-free. What’s more, given the high probability a priori of viral persistence, negative tests 3 weeks from diagnosis have a high probability of being false negatives.
Second, our data suggest that recommendation for tailored surveillance based on age, sex and disease severity of each patient is not warranted, since median times are quite similar even in very different patients, and personalised time for retesting would not increase surveillance efficiency more than would an overall delay in start of testing.
A third important implication of our results for practice concerns the management of isolating and monitoring paucisymptomatic suspected COVID-19 subjects who have not been tested due either to our health services’ difficulty in performing the test at home during the most impactful phase of the epidemic or because not enough tests were available. At the moment, paucisymptomatic subjects receive the recommendation to self-isolate during the symptomatic phase, but there are no clear indications on what to do once symptoms have disappeared. If these subjects have indeed been infected with SARS-2-CoV, all the evidence suggests that viral clearance even in them will not be achieved rapidly. To avoid generating secondary cases, either the isolation period should be longer (over 30 days from symptom onset) or at least one follow-up test should be done before ceasing isolation.
Finally, our results point out that almost all asymptomatic COVID-19 patients and a large proportion of symptomatic patients who will be eligible to discontinue transmission-based precautions (including isolation) according to the most recent WHO recommendations of 27 May 202036 will test positive for SARS-Cov-2 on RT-PCR when released. Since there is still uncertainty regarding whether these same patients are infectious, our results have relevant public health implications.