Comparison of Mid-turbinate Nasal Swabs, Saliva, and Nasopharyngeal Swabs for SARS-CoV-2 Reverse Transcription–Polymerase Chain Reaction Testing in Pediatric Outpatients

This study consists of a prospective cohort of children recruited at the time of clinically indicated SARS-CoV-2 testing. Enrollment into this study occurred from August 2020 to January 2021 at 2 clinical testing sites including 1 “drive-through” testing site and 1 emergency department (ED). Institutional review board approval was obtained.

Samples were obtained from children at the time of a clinically indicated NP sample collection for the standard-of-care test. Participants were either seeking care in the ED or were referred to the COVID-19 drive-through collection site by their clinical provider. Inclusion criteria were children/young adults up to 21 years of age with the ability and willingness to provide study samples, including self-expectorated saliva. Participants could be symptomatic or asymptomatic, and detailed symptom information was obtained at the time of sample collection. Participants were asked if they had experienced the following in the 10 days before the test: fever, chills, congestion/rhinorrhea, cough, headache, sore throat, fatigue, arthralgias, myalgias, photophobia, vomiting, nausea, diarrhea, abdominal pain, loss of sense of taste or smell, shortness of breath, or any other symptoms. Participants were also asked when their symptoms began. Consent, and assent when appropriate, was obtained before specimen collection.

Samples for this study were collected by a pediatric nurse in the drive-through location and by a pediatric nurse, research coordinator, or respiratory therapist in the ED. All personnel collecting samples were trained in the specific study protocol. The saliva and mid-turbinate nasal samples were collected just before collection of the NP swab, which served as the comparator sample. Saliva samples were collected by providing a 30- or 50-mL sterile conical tube to the participants, asking them to allow spit to collect in their mouths, and then spitting into the sterile tube until at least 2 to 3 mL of fluid was obtained. No additional instruction was provided to the participants, such as rinsing the mouth or not eating within a certain timeframe before collection. The saliva tube was immediately capped and stored at 4°C. A mid-turbinate nasal swab (Pur-Flock Ultra swab, Puritan Medical Products, Guilford, Maine) was collected by the health care practitioner, who inserted the swab just until resistance was met, and swirled gently for 15 seconds in the first naris and then repeated with the same swab in the second naris. The swab was immediately placed into 3 mL of sterile saline in 3.6-mL cryovials (Thermo Fisher, Waltham, Massachusetts), capped, and stored at 4°C. For the standard-of-care sample, the health care practitioner inserted an NP swab (Pur-Flock Ultra) into the nasopharynx and it remained in place for 5 seconds. The NP swab was removed and placed into a transport tube containing 3 mL of sterile saline (Tube Media, Cardinal Health 200 LLC), capped, and immediately stored at 4°C. All samples were transported at 4°C the day of collection by a dedicated courier. The NP sample for the standard of care was transported to the Children's Healthcare of Atlanta at Egleston clinical laboratory for testing the day it was collected. Saliva and mid-turbinate samples were transported to the Emory Center for Clinical and Translational Discovery biorepository and stored overnight at 4°C. Once the positive NP results were available, remaining saliva and nasal swab media were transferred to the clinical laboratory and tested for SARS-CoV-2 the next day. A similar number of patients with a negative SARS-CoV-2 test finding that was obtained on the same day as the positive samples were identified as negative control patients. The day following collection and testing of the NP swab sample, the paired saliva and mid-turbinate swab samples from the patients negative for SARS-CoV-2 were submitted to the clinical laboratory for testing, along with the samples from the patients who were positive for SARS-CoV-2. Stability of SARS-CoV-2 RNA in samples stored overnight at 4°C had been validated in our laboratory for all sample types used in this study (data not shown).

NP and mid-turbinate nasal swab samples were tested for SARS-CoV-2 by reverse transcription–polymerase chain reaction (RT-PCR) on the Panther Fusion instrument (Hologic, Marlborough, Massachusetts). Briefly, 0.5 mL of saline from NP or mid-turbinate nasal specimens was added to the Specimen Lysis Tube and run on the Panther Fusion instrument per standard procedure for the RT-PCR analysis.

Saliva samples were tested according to the following procedure, derived from review of the literature and initial experiments in our laboratory.⁴  A 0.4-mL aliquot of the liquid portion of the saliva was transferred to a 15-mL Falcon tube prefilled with 0.4 mL of sterile DNase- and RNase-free saline (G-Biosciences, St. Louis, Missouri). This 1:1 mixture was briefly vortexed, and 0.5 mL was added to a Specimen Lysis Tube and run on the Panther Fusion instrument. The cycle threshold (Ct) cutoff for a positive result was 40 for all sample types. Tests with invalid results were repeated with the same lysis tube on the Panther Fusion. If the repeated test was invalid, the samples for that patient were excluded from analysis.

Descriptive statistics of demographic and clinical data were reported as medians with 25th to 75th percentiles (interquartile range) for continuous data and frequencies (percentages) for categorical data. Between-group differences between NP positive and negative results were compared by using Student t tests for continuous data and χ² tests or Fisher exact tests for categorical data. If continuous data were nonnormally distributed, nonparametric tests were used (ie, Wilcoxon rank-sum tests). Concordance between the gold standard NP swab result and saliva or mid-turbinate nasal swab result was estimated by using Cohen κ coefficient with exact 95% CIs. Sensitivity and specificity values were also reported with exact 95% CIs. Cochran-Armitage trend tests were used to test for trends in percent agreement across increasing days since reported symptom onset. A generalized linear 1-way analysis of variance adjusted for unequal variance and Tukey pairwise comparisons were run to test for differences of NP Ct values across days since reported symptom onset. Statistical significance was set at the .05 level. All significance tests were 2-sided. Statistical analysis was performed with SAS 9.4 (Cary, North Carolina).

Positive and negative sample sets were tested from 154 children. Twelve saliva samples gave invalid results after 2 attempts, and all samples from these patients were excluded from further analysis. Samples from the remaining 142 children constituted the final data set. Of 142 children in the study, 73 had positive SARS-CoV-2 NP results and 69 had negative SARS-CoV-2 NP results.

Study subjects ranged in age from 5 to 19 years (Table 1). There were 75 males and 67 females. Children were divided into 3 groups on the basis of presence or absence of symptoms and the timing between symptom onset and sample collection.

One hundred fourteen children (80%) reported symptom onset in the past 10 days. Children were further divided into groups representing duration from onset of symptoms for analysis. Ninety-two children (65%) had symptom onset ≤5 days before sample collection, 22 (15%) had symptom onset >5 days and ≤10 days before sample collection, and 28 (20%) had symptom onset >10 days before sample collection or did not report symptoms. Of these 28, 20 (14% of the total) did not report symptoms. The asymptomatic children were most commonly referred for testing before an aerosol-generating surgical or endoscopic procedure, or for a history of SARS-CoV-2 exposure. There was no significant difference in the demographics of children who were PCR positive or negative for SARS-CoV-2 when using NP swab samples. There was also no difference in the demographics for participants with positive or negative SARS-CoV-2 test results in the 3 groups when assessed by the time of symptom onset.

There was 88% positive agreement and 99% negative agreement between saliva and NP results, with a 93% overall agreement (Table 2). There was 86% positive agreement and 100% negative agreement between the mid-turbinate nasal swab and the NP swab results, with an overall agreement of 93% (Table 3). Sample agreement was then assessed on the basis of length of time from symptom onset. In samples collected ≤5 days from symptom onset and >5 to ≤10 days from symptom onset, the positive agreement for saliva compared to the NP swab sample increased to 92% and 90%, respectively (Table 4). Positive agreement for the mid-turbinate nasal swab compared to the NP swab sample increased to 90% and 100% for these 2 groups, respectively (Table 4). Negative agreement for both saliva and mid-turbinate nasal swab samples compared to the NP swab sample was 100% in children with symptom onset ≤10 days previously. The overall agreement dropped to 82% for both saliva and mid-turbinate nasal swab samples compared to NP swab samples in participants with symptom onset >10 days before collection or in participants without symptoms.

Twenty-six saliva samples (18%) initially gave invalid results, but each result was valid after repeating the test. All of the tests for the NP and mid-turbinate samples were valid on the initial run. Ct values for all positive NP swab samples had a mean of 25.2. The mean NP Ct value for children with symptom onset ≤5 days before collection was 23.4; for children with symptom onset >5 to ≤10 days before collection, 25.9; and for children with symptom onset >10 days before collection or reporting no symptoms, 30.5 (P < .001) (Table 5). Of 14 patients who tested positive when using the NP swab sample but negative for the mid-turbinate nasal swab sample and/or the saliva sample, only 2 had an NP Ct value of 30 or less (Table 6). Ct values from the mid-turbinate nasal samples were more closely correlated to the Ct values from the NP sample than from the saliva sample (Figure, A and B). A single negative NP swab and mid-turbinate nasal sample had a positive corresponding saliva sample, with a Ct of 38.

To our knowledge, our study is the first to provide comparative SARS-CoV-2 RT-PCR data on contemporaneously collected saliva, mid-turbinate swab, and NP swab samples in children. Both mid-turbinate nasal swabs and saliva are easier specimens to obtain than NP swabs and cause less discomfort to the child, making them an attractive alternative for nonmedical setting testing such as schools, daycare facilities, and at home. However, given the number of invalid results observed for the saliva samples in our study and the need for repeated testing with this specimen type, mid-turbinate samples may be more reliable unless preprocessing of saliva is performed.^5,6 

Our findings reinforce those of Sahni et al,⁷  who also performed a prospective analysis of NP and mid-turbinate samples collected in children. The overall sensitivity reported for their mid-turbinate to NP swab was 82.5%, similar to our percent positive agreement (PPA) of 86%. When sample collection occurred within 5 days of symptoms, they reported a sensitivity of 89.5% for the mid-turbinate compared to the NP swab and we reported a PPA of 90%, increasing to 100% when the sample was collected >5 to ≤10 days from symptom onset, but with smaller numbers of subjects. Both studies reported 100% correlation for negative samples, and so our overall correlation varied between 95% and 100% for samples collected within 10 days of symptom onset. Of note is 1 saliva sample that was positive for SARS-CoV-2 with correlating negative mid-turbinate and NP samples. The Ct for this positive result was close to the limit of detection.

Both our study and the study by Sahni et al⁷  identified decreasing sensitivity as the length of time between symptom onset and test collection increased. Sahni et al⁷  reported that when the sample was taken up to 14 days following onset of a patient's symptoms, the sensitivity of the mid-turbinate to the NP swab was 82.1%. This differs from our sensitivity value for patients with symptom duration longer than 10 days from specimen collection, or with no symptoms in the past 14 days. In that group of patients, our PPA for the mid-turbinate to NP swab was 67%.

While our respective studies had overlap, there were differences. Sahni et al⁷  ascertained their positive samples from patients who reported symptoms, whereas we had a mix of symptomatic and asymptomatic children. The inclusion of asymptomatic children in our group could account for the difference in sensitivity for the mid-turbinate compared to the NP swab.

We also collected saliva, which adds a third dimension to our study, and tested the saliva by using a dilution protocol published previously. The inclusion of saliva collection also ensured that our patient population was older, given the need for cooperation of the child to collect saliva. Eighteen percent of saliva specimens included in the study gave an invalid result—requiring a repeated test to obtain a valid result—compared to no invalid tests from NP or mid-turbinate swab samples. Despite this variability, saliva showed an overall PPA of 88% compared to the NP swab sample, similar to 86% for the mid-turbinate swab, and the decrease in PPA when compared to the NP swab was seen with increasing duration of symptoms in relation to the time of sample collection.

Strengths of this study include the prospectively designed methods and conduct of the study. All specimens were tested within 24 hours of collection; and all saliva, mid-turbinate nasal swabs, and NP swabs were collected contemporaneously and tested on a single platform. The strength of our study can also be a limitation, as there are many different ways to collect samples, and there are multiple testing platforms; and moreover, our population was limited to pediatric outpatients, instead of including inpatients as other studies have. Other limitations include the fact that only 1 testing platform was used, and so generalization of our results to other testing platforms would need validation. Additionally, the testing platform we used was validated for mid-turbinate swabs by the manufacturer, but was not validated for saliva. While we used a methodology for saliva testing that was published, as well as tested in our laboratory, we still obtained invalid results. As has been reported, saliva often benefits from additional processing before testing.⁶  Finally, our sample size of 142 did not allow for some detailed analysis, such as separating asymptomatic children from children with symptom duration longer than 10 days. During the study, there was only 1 variant circulating in Georgia, so this possible limitation did not exist.

In conclusion, the message of our study is simple. Both saliva and mid-turbinate nasal swabs have a 95% or greater agreement with NP swab collections in children with symptoms consistent with SARS-CoV-2 infection for up to 10 days before testing. Given the clinical benefit of less invasive sample types in children, we suggest that either a mid-turbinate swab or saliva could be considered for use in clinical testing for SARS-CoV-2, with the caveat that an invalid result may be more likely with saliva than with either the NP or mid-turbinate swab sample. Additional considerations include ensuring that the test is validated for saliva and/or a mid-turbinate swab and that other respiratory pathogens, which require NP swab collections, are not needed in the same test.