Sensitivity of rapid antigen testing and RT-PCR performed on nasopharyngeal swabs versus saliva samples in COVID-19 hospitalized patients: results of a prospective comparative trial (RESTART).

Objectives: Saliva sampling could serve as an alternative non-invasive sample for SARS-CoV-2 diagnosis while rapid antigen testing (RAT) might help to mitigate the shortage of reagents sporadically encountered with RT-PCR. Thus, in the RESTART study we compared antigen and RT-PCR testing methods on nasopharyngeal (NP) swabs and salivary samples. Methods: We conducted a prospective observational study among COVID-19 hospitalized patients between 10th December 2020 and 1st February 2021. Paired saliva and NP samples were investigated by RT-PCR (Cobas 6800, Roche-Switzerland) and by two rapid antigen tests: One Step Immunoassay Exdia(R) COVID-19 Ag (Precision Biosensor, Korea) and Standard Q(R) COVID-19 Rapid Antigen Test (Roche-Switzerland). Results: A total of 58 paired NP-saliva specimens were collected. Thirty-two of 58 (55%) patients were hospitalized in the intensive care unit and the median duration of symptoms was 11 days (IQR 5-19). NP and salivary RT-PCR exhibited sensitivity of 98% and 69% respectively whereas the specificity of these RT-PCRs assays were of 100%. NP RAT exhibited much lower diagnostic performances with sensitivities of 35% and 41% for the Standard Q(R) and Exdia(R) assays respectively, when a wet-swab approach was used (i.e. when the swab was diluted in the viral transport medium (VTM) before testing). The sensitivity of the dry-swab approach was slightly better (47%). These antigen tests exhibited very low sensitivity (4 and 8%) when applied to salivary swabs. Conclusions: Nasopharyngeal RT-PCR is the most accurate test for COVID-19 diagnosis in hospitalized patients. RT-PCR on salivary samples may be used when nasopharyngeal swabs are contraindicated. RAT are not appropriate for hospitalized patients.

Introduction 72 additional manipulation step using VTM tubes was performed for an in vitro evaluation of a 122 possible dilution effect of VTM (see section "In vitro testing of dilution effect" below). 123 All samples were taken by two specialists in infectious diseases (AK and GC) or a member of 124 a paramedic team trained in NP swabs collection (coming from a team performing all COVID-125 19 samples in our hospital). NP was performed according to the recent CDC and WHO 126 guidelines (19, 20) and saliva sampling protocol was based on previously published data 127 adapted for hospitalized patients (supplementary material, S1) (5). Each nasopharyngeal 128 swab was performed on a different naris, which was randomly chosen for each sample. 129 130

Quantitative SARS-CoV-2 PCR and RAT 131
RT-PCR was performed in our microbiology laboratory using the automated Cobas 6800 ® 132 system (Roche-Switzerland) (18). In order to quantify the viral load (VL) based on the 133 number of cycles threshold (Ct) obtained with the molecular platform, we used the following 134 equation, derived from RNA quantification: VL = (10^((Ct -40.856)/ -3.697))*100. Details on 135 methods used to derive this equation were described elsewhere (21). The analytical limit of 136 detection was determined to be at 1000 copies per ml. For graphical representation 137 purposes, NP or salivary samples with undetectable VL are represented in graphs with VL 138 determined to be at 500 copies/ml. 139 NP swabs and saliva samples were used to assess RAT performances. NP swabs were either 140 directly suspended in the buffer solution ("Dry approach"), or initially suspended in 3 ml of 141 VTM ("Wet" approach). One hundred and fifty µl of the sample were subsequently mixed 142 with the buffer solution. Saliva samples were only treated with the "Wet" approach. Reading 143 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

In vitro testing of dilution effect 150
In order to test in vitro a possible dilution effect generated by the use of VTM tubes instead 151 of direct testing, we simulated the two sampling scenarios: the "Wet" scenario versus the 152 "Dry" swab scenario (supplementary material, S2_Figure 1). A SARS-CoV-2 positive clinical 153 NP sample was diluted 7 times and the series was used as internal reference for the limit of 154 detection. Each initial sample of this dilution series represented a "Dry" swab. Then, two 155 clean swabs were soaked in each one of the dry samples and then suspended in two further 156 VTM tubes, thus simulating two series of "Wet" samples.

Patients' characteristics 187
All patients with confirmed SARS-CoV-2 infection by NP RT-PCR and admitted in ICU or 188 Internal Medicine ward during the study period were screened for eligibility criteria. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Diagnostic performance of RT-PCR (NP versus saliva) and rapid antigen testing 199
The diagnostic performance of RT-PCR and RAT for diagnosis of SARS-CoV-2 infection is 200 shown on Table 2. A positive NP RT-PCR or salivary RT-PCR was selected as the reference 201 standard comparator. 202 NP and salivary RT-PCR exhibited an overall sensitivity of 98% and 69% respectively whereas 203 the specificity of both assays was of 100%. Noteworthy, the sensitivity of salivary PCR 204 increased to 81 % (95% CI: 59-88) in patients presented with a duration of symptoms of less 205 than 10 days. VL (copies/ml) in NP swabs was significantly higher than that detected on 206 salivary specimens for up to 20 days after illness onset (Figure 2). Median VL value in 207 positive NP swabs with negative paired saliva specimens was 3700 copies/ml (IQR, 2900-208 9675). Median duration of illness for those patients was 15 days (IQR, 9-21). ICU patients 209 had higher VL compared to patients hospitalized in internal medicine ward. Pair testing 210 results are shown in Figure 3. An analysis of the agreement between the two specimens (NP 211 versus saliva) revealed a fair agreement with a kappa coefficient of 0.37 (95% CI 0.16-0.59; 212 p=0.001) and an overall proportion of agreement of 72% (proportion of positive agreement 213 80% and proportion of negative agreement 53%) (supplementary material, S3_Figure 2). 214 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101/2021.04.09.21255105 doi: medRxiv preprint approach was used (Table 2). Interestingly, the sensitivity of the dry-swab approach was 217 slightly better [sensitivity 47% (95% CI: 35-62) for Standard Q ® assay). These antigen tests 218 exhibited very low sensitivity of 8% and 4% for Exdia ® and Standard Q ® assays, respectively, 219 when applied to salivary swabs. Figure 4 shows RAT results according to illness duration and 220 VL. All RAT performed better in high VL or early in the course of the disease (especially if VL 221 >10 6 copies/ml and illness duration < 10 days) (supplementary material, S4_Figure 3). The present study sought to evaluate the role of alternative and non-invasive methods for 235 diagnosis of SARS-CoV-2 infection in moderately and critically ill-hospitalized patients. 236 Moreover, to our knowledge this is the first study to evaluate the impact of VTM on RAT 237 results for SARS-CoV-2 diagnosis. 238 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 15, 2021. ; https://doi.org/10.1101/2021.04.09.21255105 doi: medRxiv preprint ease of collection and comfort for repetitive testing as well as highly reliable results (2-6, 9-240 12). Nonetheless, only few studies evaluated the diagnostic performance of saliva as 241 compared to NP swab in hospitalized patients (3, 11-13), with results being conflicting. Our 242 study evaluated hospitalized patients with moderate to very severe disease. Most of them 243 were still symptomatic upon sampling (84%) and had a wide distribution of VL, ranging from 244 10 3 to 10 9 copies/ml. Overall, NP RT-PCR was more sensitive in diagnosis of SARS-CoV-2 245 infection than RT-PCR performed in saliva sample. VL detected in NP swab was higher (1-2 246 log copies/ml) than in saliva specimens. It is noteworthy however, that sensitivity of salivary 247 RT-PCR increased considerably (from 69% to 81%) when considering patients presenting 248 early in the course of the disease (<10 days). Progressive decrease in VL over time (14) might 249 explain loss of sensitivity of the salivary RT-PCR, which is more pronounced when testing 250 patients with more than 10 days of symptoms. In fact, patients in our study with a positive 251 NP swab and negative saliva sample had low VL (median value of 3700 copies/ml) and a 252 median duration of symptoms of 15 days. This raises the question of the infectivity of those 253 patients and suggest that patients not detected by salivary RT-PCR are those who have low 254 infectivity potential and are late presenters in the course of the disease. Therefore, the 255 results of our study, along with previous published data (3, 9, 11-13), suggest that saliva 256 specimens could be a fair non-invasive alternative to NP swabs (if NP swab is contraindicated 257 for example), especially for those presenting early in the course of illness. 258 A second goal of our study was to evaluate the diagnostic performance of RAT among 259 When we tested the diagnostic performance of RAT for patients presenting with less than 10 265 days of illness, sensitivity remained very low. RAT yielded high diagnostic performances only 266 for patients with high VL (≥10 6 copies/ml) (supplementary material, S4_Figure 3). The much 267 lower diagnostic performances of RAT in saliva might be explained by the lower VL in saliva 268 as compared to NP swabs or eventually the presence of mucosal secretory immunoglobulins 269 targeting SARS-CoV-2 antigens and thus competing with RAT for the same target (14, 23). 270 Finally, our study evaluated the role of VTM in SARS-CoV-2 diagnosis by RAT. While both 271 "Wet" and "Dry" procedures are recommended by the manufacturer, previous data suggest 272 that VTM can influence diagnostic performances of RT-PCR and RAT for SARS-CoV-2 273 detection (15, 24). This is the first study to our knowledge that evaluates "head to head" a 274 "Wet" versus "Dry" RAT procedure for SARS-CoV-2 detection. Our in vitro evaluation showed 275 higher Ct levels for the "Wet" series of both antigen tests used, suggesting a dilution effect 276 when the swab is immerged in the VTM. Our "head to head" clinical comparison confirmed 277 the in vitro experimentation, showing that the "Dry" NP swab performed slightly better than 278 the "Wet" one (sensitivity increasing from 35% to 47%), likely due to a decreased dilution of 279 the sample. While RAT can not be recommended for hospitalized patients, the observed 280 difference between "Wet" and "Dry" swabs should be taken into account when performing 281 RAT for SARS-CoV-2 diagnosis in an outpatient setting. Still, when RAT are just used as a 282 supplementary triaging step at the hospital entry to fasten isolation of highly contagious 283 subjects, the use of a wet swab approach is also acceptable since the lower sensitivity is 284 compensated by the fact that in high pandemic period, performing a single sampling for both 285 RAT and PCR is likely a good and effective option (14). 286 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. On the other hand, this study has a few limitations as well. Its monocentric nature and 292 limited sample size require our results to be confirmed by larger prospective trials. Our 293 patients were initially diagnosed with NP swab RT-PCR that might have induced a bias 294 towards subsequent NP swabs being more often positive versus other samples. Hospitalized 295 patients may have altered saliva production or composition (25) that could influence saliva 296 based diagnostic strategies or even explain the differences observed in salivary RT-PCR 297 performances among severely and mildly ill COVID-19 patients. We chose to use a validated 298 and easy to use non-invasive saliva collection procedure (5). It is possible that other methods 299 of saliva collection (such as throat washing for example) would have improved diagnostic 300 yield and should therefore be tested in other comparative trials. 301 In conclusion, NP swab RT-PCR was the most sensitive method to diagnose SARS-CoV-2 302 infection in moderately to critically ill hospitalized patients. Salivary RT-PCR could be used as 6.
Large parallel screen of saliva and nasopharyngeal swabs in a test center setting 358 proofs utility of saliva as alternate specimen for SARS-CoV-2 detection by RT-PCR. [ . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.