SARS-CoV-2 kinetics and shedding in bodily
fluids
Respiratory shedding
Key points
• SARS-CoV-2 replication in the throat and diagnostic use of upper respiratory tract sampling in the early infection days (regardless of the severity of symptoms)
• Viral RNA loads peak within the first infection days in the upper and later in the lower respiratory tract; need for standardisation of viral load testing and reporting
• SARS-CoV-2 RNA detectable from respiratory samples up to 6 weeks in mild and 8 weeks in severe cases, and beyond symptoms resolution
• Insufficient systematic comparisons between all respiratory samples’ types; higher viral loads in sputum than swabs, nasopharyngeal swabs more sensitive than oropharyngeal swabs
• CT findings could sometimes precede viral RNA detection in the upper respiratory tract and full clinical presentation should always be evaluated
We reviewed 262 studies with respiratory sampling for ≥31957 COVID-19 cases, including data on nasopharyngeal/midturbinate/nasal (NP) swabs (≥13286 cases), oropharyngeal/throat (OP) swabs (≥7301), combined NP and OP swabs (≥1493), sputum (≥409), bronchoalveolar lavage fluid (≥21), and other respiratory specimens (endotracheal aspirate, bronchoalveolar swab, ≥21) (Supplementary Dataset) [1, 9-255][256-268].
Lower respiratory tract (LRT) bronchoalveolar lavage (BAL) sampling allowed the initial and subsequent virus culture of SARS-CoV-2 [1, 11, 13]. Almost all BAL specimens described in peer-reviewed literature had detectable viral RNA regardless sampling timing, disease severity or co-morbidities and were useful for ultimate confirmation of difficult cases [11, 44, 71, 77, 78]. Virus isolation was successful from NP swabs 2 dps in 2 mild cases (using Vero E6 cells) [120]; NP and OP swabs 4 dps (using Vero CCL-81 cells) [76] in a mild case [63], and also at 4 dps from NP swab and nasopharyngeal aspirate in another mild case (using Vero E6 cells) [98]. SARS-CoV-2 was isolated from NP swabs of 2 severe cases 1 and 10 dps [135]. In a German study of 9 mild COVID-19 cases, SARS-CoV-2 was isolated up to 8 dps from both upper respiratory tract (URT) swabs (16%) and sputum samples (83%) with viral loads (VL) >106 copies/ml. Furthermore, the authors detected viral subgenomic messenger RNAs (sgRNA) which led them to conclude there was ongoing viral replication in the throat up to 5 dps. Sequencing data also showed the continuous presence of two genotypes of SARS-CoV-2 differing by a single mutation in the throat and lungs samples of a patient [131]. La Scola and colleagues cultured 174 NP swabs and 9 sputum samples testing positive via PCR (from 155 patients total) and succeeded with virus isolation from 129 samples (124 with observable cytopathic effect on Vero E6 cells). They observed a strong correlation between Ct values and virus isolation: 100% isolation rates from samples with Ct 13-17 decreasing to 12% at Ct=33 and no isolation from samples with Ct≥ 34 [211]. An Indian study was successful in isolating SARS-CoV-2 (using Vero CCL-81 cells) from respiratory samples in 9 of 12 samples with Ct values ranging 16-25.1 [217].
Despite the growing amount of literature, only two studies documented NP swabs, OP swabs, and sputum, collected sequentially in the same 16 [172] and 49 [240] patients respectively, while 2 studies described an upper respiratory specimen paired with sputum for a total of 11 cases [39, 131]. Nevertheless, we aggregated the data from 32 observational studies with available infection timeline and respiratory sampling for 216 cases [9, 18, 23, 25, 29, 34, 37, 39, 42-45, 58, 59, 63, 72, 85, 98, 111, 116, 120, 131, 134, 158, 160, 171, 172, 176, 187, 197, 206] to provide a summary (Figure 1, Figure 2, Table, and Supplementary Table). SARS-CoV-2 RNA was detected in NP and OP swabs from symptoms onset up to 42 dps in mild cases [158] and 50 dps in severe ones [172]. Sputum yielded positive results up to 27 dps in mild cases [131], but 55 dps in severely ill [172] (Figure 1 and Figure 2). Several studies reported SARS-CoV-2 RNA detection from the URT for a median period of 10-20 days [53, 59, 79, 99, 105, 187, 191] with a prolonged one seen in severe cases [99, 105, 191]. SARS-CoV-2 RNA was detectable in URT samples well beyond waning of respiratory symptoms [37, 46, 107, 131]. SARS-CoV-2 remained detectable in OP swabs ≥2 weeks, (nine mild cases) and in sputum >3 weeks (six mild cases) despite symptoms resolution [131]. Chen and colleagues described recurring SARS-CoV-2 RNA positivity in OP swabs of a patient until 30 dps (VL 4.56x102 copies/mL), well after pneumonia resolution and hospital discharge [37]. Prolonged viral RNA detection in OP swabs (range 5-30 dps for 22 cases) was also reported for mild cases regardless of symptoms, incl. >3 weeks in 8 cases [199]. They did not study if infectious virus could be detected. Sun and colleagues aggregated data on 175 NP swabs, 88 OP swabs, and 62 sputum samples, and estimated a median/95th percentile time until loss of detection as follows: NP 22.7/46.3 dps, OP 15.6/32.8 dps, and sputum 20/43.7 dps for 43 mild cases vs. NP 33.5/49.4 dps, OP 33.9/53.9 dps, and sputum 30.9/44.7 dps for six severe cases [240].
A Nanchang study (21 mild, ten severe cases) observed clearance in NP swabs within 10 dps in 90% of the mild cases compared to continuous RNA detection >10 dps in all severe cases [99]. Feng and colleagues detected SARS-CoV-2 RNA in NP swabs of 24 mild cases for 16±7 days compared to 22±4 days in eight ICU patients [105]. An aggregation of retrospective observations from 191 hospitalised adults in Wuhan showed a median duration of URT viral RNA detection of 20 (IQR 16-23) dps with continuous detection until death in non-survivors and ranges 8-37 dps in survivors [79]. Another study including 66 COVID-19 cases found a median of 9.5 (6-11) dps until the first negative results in OP swabs [53]. A total of five patients presented with mild symptoms and detectable SARS-CoV-2 RNA 4-15 days following last negative OP swab and previous hospital discharge [104], though the COVID-19 reactivation as reported by the authors could have been prolonged disease course too. Another Wuhan-based study reported re-hospitalisation with mild symptoms for 11 cases within a median of 16 (range 6-27) days, of which four has viral RNA detected in OP swabs [254].
Duration of shedding may be related to patient’s general health condition: in a Wuhan-based study 27 out 56 mild cases had prolonged SARS-CoV-2 RNA detection >24 days in NP/OP swabs, associated with old age and comorbidities. The proportion of positive respiratory samples decreased from 89% to 66%, 32%, 5%, and 0% in weeks 2-6 since symptoms onset [187]. Xu and colleagues summarised respiratory samples data from 113 patients and observed a median RNA detection for 17 (IQR 13-22) dps. Prolonged detection ≥15 dps was associated with males, old age, hypertension, severe illness upon admission, invasive mechanical ventilation, and corticosteroid treatment [150]. A Wuhan-based study on 41 discharged severe cases reported SARS-CoV-2 RNA detection in OP swabs for a median of 31 (IQR 24-40, range 18-48) dps and no significant difference between male and female, nor between cases <65 and ≥65 years [186].
In the studies that provide quantitative results (n = 26), [9, 10, 25-27, 34, 37, 39, 40, 59, 63, 99, 111, 119, 120, 122, 131, 158, 160, 161, 170, 172, 202, 206, 220, 248] the highest VL in URT specimens were reported in the early days of the disease [34, 39, 59, 63, 99, 120, 131, 170, 172], also before development of respiratory symptoms [39, 119, 170]. A study of nine mild cases reported SARS-CoV-2 RNA detection from all NP and OP swabs in the first 5 dps with (average VL 6.76x105 copies/swab and maximum 7.11x108 copies/swab), whereas detection rate in subsequent swabs was only 40% reaching up to 28dps (average VL 3.44x105 copies/swab) [131]. Among four patients (2 mild, 2 severe) viral loads in NP swabs ranged from 7.4 log10 copies/1000 cells (mild case 2 dps) and 7.1 log10 copies/1000 cells (severe case 6 dps) to negative 9-14 dps, while for a critically ill case they were in the range 6.7-4.4 log10 copies/1000 cells and persisted 24 dps until death [120].
He and colleagues reported high VL for 414 OP swabs from 94 patients in the early days of infection and gradual decrease until about 21 dps with no difference when stratified by sex, age, or disease severity [170]. However, VL in the sputum of 22 mild cases reached a peak in week 2 since symptoms onset and were significantly lower than those of 74 severe cases [191]. In a paediatric dialysis unit with 12 COVID-19 cases (6 asymptomatic, 6 mild), the viral loads were significantly higher for the patients with symptoms [212]. Kimball and colleagues reported no significant difference between the Ct values in NP swabs collected from 10 symptomatic, 10 presymptomatic and 3 asymptomatic cases in a nursing facility [119]. Among 61 healthcare workers, VL of self-collected NP+OP swabs were significantly lower for 56 asymptomatic vs. five symptomatic cases [248]. A study on 12 cases (9 mild, 3 severe) showed that Ct values from respiratory samples were correlated with disease severity scores like ARDS index PaO2/FiO2 ratio and lung injury Murray score, as well with biochemical indicators like albumin levels and lymphocytes and neutrophils percentages, and concluded viral loads could be as a COVID-19 severity predictor [17]. In a cohort of 92 cases (51 mild, 11 mild turning severe, 30 severe) low Ct values (high VL) in sputum were correlated with severe COVID-19 and risk of progression to severe disease [202].
Comparing respiratory specimens, higher VL were reported for sputum than respiratory swabs [9, 40, 122, 131, 172]. A study using RT-PCR and droplet digital PCR found both significantly higher positive rates and average VL in sputum (66% and 17429±6920 copies/test) compared to OP swabs (37% and 2552±1965 copies/test) and NP swabs (16% and 651±501 copies/test) [122]. For 16 critically ill patients sputum and endotracheal aspirate samples all had detectable SARS-CoV-2 RNA at levels significantly higher than NP and OP swabs with positivity rates of 81% and 63% respectively [172]. In nine mild cases up to 5 dps, the maximum SARS-CoV-2 VL in sputum (2.35x109 copies/mL) was higher compared to respiratory swabs (7.11x108 copies/swab). However, examining paired sputum and swab samples 2-4 dps in seven patients showed higher virus concentration in swabs (two cases), sputum (two cases), and similar virus concentrations in both for the remaining five cases [131]. In a cohort of 52 patients, the positivity rates of sputum samples (77%) were higher than OP swabs (44%) with a significant difference when comparing cases with positive sputum sample and negative OP swab (40%) vs. those with negative sputum and positive OP swab (8%) [180].
Choosing the most appropriate respiratory specimen depends on timing in the infection course. However, well-documented studies comparing all respiratory and other sample types collected in a known timeline are limited. A preprint study in Guangdong on 866 respiratory samples from 213 symptomatic cases (37 severe), showed that apart from BAL, sputum is the most sensitive sample type for COVID-19 laboratory diagnosis, followed by NP swabs [26]. In the period 0-7 dps, the highest positivity rate was observed in sputum (severe cases 89%, mild cases 82%), followed by NP swabs (73%, 72%) and OP swabs (60%, 61%). In the period 8-14 dps the same order in positivity rates was observed: sputum (severe cases 83%, mild cases 74%), NP swabs (72%, 54%), and OP swabs (50%, 30%), as for ≥ 15 dps: sputum (61%, 43%), NP swabs (50%, 55%), and OP swabs (37%, 11%). Sputum samples 0-7 dps also yielded the lowest median Ct values (25 in severe cases, 28.5 in mild cases). BAL samples 8-14 dps were positive for SARS-CoV-2 RNA in 12 severe cases and negative for three mild cases. Beyond 15 dps the study showed 79% positivity in BAL in severe cases [26]. Further peer-reviewed data was also in favour of sputum, or NP swabs [71, 77]. A retrospective study of 4880 cases in Wuhan found 38% positive rate for 4818 NP and OP swabs compared to 49% for 57 sputum specimens and 80% for 5 BAL [71]. Another Chinese study including 205 COVID-19 cases yielded overall 1070 samples with the following positivity rates: 93% for 15 BAL samples, 72% for 104 sputum specimens, 63% for 8 NP swabs, 46% for 13 fibrobronchoscope brush biopsy samples, 32% for 398 OP swabs, 29% for 153 feces samples, 1% for 107 blood samples [77]. Only BAL specimen sequencing, and BAL and sputum samples PCR could confirm SARS-CoV-2 co-infection with influenza A in an ICU patient, whereas repeated NP swabs were negative [78]. Although sputum might seem like a sample of choice, it was described that only a third of 1099 COVID-19 patients had a productive cough [50], suggesting that in practice NP swabs would be preferable in most cases. Higher sensitivity for NP swabs in comparison to OP swabs was observed as well in other studies and case reports [34, 52]. A study with sequential sampling in 18 patients (72 NP and 72 OP swabs) showed higher VL in the nose than the throat [34].
SARS-CoV-2 shedding potential in asymptomatic and pre-symptomatic individuals needs to be elucidated [74, 94, 101, 119, 139, 144, 155, 156, 160, 167, 170, 224, 227, 269, 270]. Multiple studies worldwide reported SARS-CoV-2 RNA detection in respiratory samples from cases with epidemiological link and no (189 adults, 29 children) [9, 31, 34, 36, 61, 70, 74, 94, 116, 119, 123, 139, 144, 155, 160, 199, 205, 214, 224, 227, 248] or mild/non-specific symptoms (23 adults, 1 child) [9, 12, 15, 18, 20, 24, 25, 28, 30, 39, 61, 74, 116, 227]. One described SARS-CoV-2 RNA detection in OP swabs for 17 days in an otherwise asymptomatic patient [70]. In a series of testing campaigns in Iceland that led to SARS-CoV-2 RNA detection in NP+OP swabs of 1321 cases total, those reporting any symptoms ranged between 46% (random population screening) and 94% (targeted screening) [167]. At this stage, it is unclear whether SARS-CoV-2 affects the upper or the lower respiratory tract first, or maintains independent replication in both sites. Thus, choosing between NP and OP swabs, or sputum as sampling strategy should be done with the purpose (general population screening or confirmation of suspected cases) and potential infection timeline in mind. URT sampling would be preferable in the early infection days, especially in asymptomatic or mildly symptomatic suspected cases, whereas lower respiratory sampling provides more reliable confirmation in advanced COVID-19 with lungs involvement. Cases with an epidemiologic link, radiologic findings, and an initial negative result should be monitored further by PCR and evaluated in conjunction with their clinical presentation [10, 22, 35, 38, 41, 55, 56]. The discrepancy between URT and LRT test results has triggered a discussion about the lack of sensitivity of PCR testing. In a study among 1014 patients in Wuhan 59% (95%CI, 56%-62%) of OP swabs were positive whereas 88% (95%CI, 86%-90%) had chest CT findings within a median of one day, consistent with an earlier resolution of viral replication in URT than in LRT samples. Furthermore, 308 patients (75%) had negative PCR results in conjunction with radiologic findings and 14 out 15 cases with CT findings tested positive on a follow-up PCR within a mean of 5 days, a finding that may be more difficult to explain [41]. Another study that aggregated data on PCR results and CT imaging showed all 167 patients had a positive OP swab by the end of their hospitalisation [22]. A multicentre study of 80 COVID-19 cases (no critically ill) reported the following positivity rates in repeated OP and/or NP swabs collection until confirmation: 51%, 38%, and 11% upon the 1st-3rd test PCR test [56]. Another Hubei-based study reported SARS-CoV-2 RNA detection in OP swabs of 74%, 12% and 14% of 91 cases upon the 1st-3rd test and no significant difference between the 30 severe and 61 mild cases included [221]. A New York-based study on 5700 COVID-19 cases (incl. 1281 in ICU) reported detection upon 1st NP swab test in 97% (n=5517) of cases compared to 3% (n=183) with repeated tests [193].