Simple Saliva Sample Collection for the Detection of SARS-CoV-2 Variants Compared With Nasopharyngeal Swab Sample

Participants were prospectively recruited from June 2021 to May 2022 at a hospital in Seoul, Republic of Korea, among patients who were hospitalized in the COVID-19 treatment center or who visited the hospital because of COVID-19 symptoms or after contact with a confirmed COVID-19 case. Patients aged 13 years or older who voluntarily agreed to participate were enrolled in this study. Clinical information was obtained from each participant. NP and saliva samples were collected from each participant at the same time and were subsequently tested for SARS-CoV-2 using rRT-PCR. The leftover samples were stored at −70°C and later tested for SARS-CoV-2 variants.

This study was approved by the Institutional Review Board of Eunpyeong St. Mary's Hospital, Seoul, Republic of Korea (approval Nos. PC21DDST0051, PC21DISI0153, PC21DDST0078, PC22DDDT0044, PC22DDDT0045). Informed consent for sample collection was obtained from all participants after the nature of the procedures had been fully explained. The institutional review board waived the requirement for consent for the variant testing because it used leftover samples.

NP samples were placed in universal transport medium (UTM; AB Transport medium, AB Medical, Gwangju, Republic of Korea) by skilled medical staff. UTM was also used for the collection of saliva samples. Food and drink were restricted for 30 minutes before saliva collection. A specific collection kit was prepared and used to quantitatively collect saliva into UTM using the drooling technique. In 2021, saliva was collected into a plastic container for quantification and then transferred to UTM. In 2022, saliva was collected directly into UTM using a graduated UTM container and a plastic funnel for saliva transfer. Using these methods, 200 to 300 μL of quantified saliva was collected per sample. The NP and saliva samples were vortexed for 10 seconds before aliquoting, and viral RNA was extracted using 250 μL per sample. RNA extraction was performed using Real-Prep Viral DNA/RNA kit (BioSewoom, Seoul, Republic of Korea) and the Real-Prep DNA/RNA Extractor (Hanwool TPC Co, Bucheon-si, Republic of Korea) according to the manufacturer's instructions. PCR amplification of SARS-CoV-2 was performed using Real-Q Direct SARS-CoV-2 Detection Kit (BioSewoom) and the Applied Biosystems 7500 Real-Time PCR System (Thermo Fisher Scientific, Waltham, Massachusetts). This method has a 95% limit of detection of 7.87 copies/μL for the E gene and 8.32 copies/μL for the RdRp gene, determined by probit analysis.²¹  The E gene and RdRp gene of SARS-CoV-2 were amplified for 40 cycles, and cycle threshold (Ct) values of 38 or lower were considered positive for the 2 RNA targets. The human RNase P gene was amplified as an internal control (IC), and Ct values of 35 or lower were considered successfully amplified.

NP and saliva samples that were confirmed positive by rRT-PCR were analyzed for SARS-CoV-2 variants. Samples stored at −70°C were thawed at room temperature and RNA was extracted from 200 μL of each sample using the Real-Prep Viral DNA/RNA kit and the Real-Prep DNA/RNA Extractor, according to the manufacturer's instructions. Variant analysis was processed using the Illumina COVID Seq assay RUO (Illumina, San Diego, California) and the Illumina MiSeq sequencing system (Illumina) according to the manufacturer's protocols.²²  Fastq files obtained from the Miseq instrument were analyzed for SARS-CoV-2 variants using Illumina DRAGEN COVID Lineage version 3.5.4 in the Illumina BaseSpace environment.

Statistical analysis with figure production was performed using MedCalc software (version 20.111; MedCalc Software, Ostend, Belgium) or GraphPad Prism (version 9.4.0; GraphPad Software, San Diego, California). The Kolmogorov-Smirnov test was used to assess the normality of the distribution. The rRT-PCR results of the NP and saliva samples were compared using the Mann-Whitney U test and Bland-Altman analysis. The rRT-PCR results were compared according to variant using ANOVA or the Kruskal-Wallis test. The Cohen κ coefficient was calculated to evaluate intersample reliability for agreement between the NP and saliva samples; κ values of less than 0, 0 to 0.20, 0.21 to 0.40, 0.41 to 0.60, 0.61 to 0.80, and 0.81 to 1 indicated no agreement, slight agreement, fair agreement, moderate agreement, substantial agreement, and almost perfect agreement, respectively. P values <.05 were regarded as statistically significant.

During the study period, NP and saliva samples were collected from 338 patients (180 females and 158 males; median age, 39 years; interquartile range, 31–57 years) and tested using rRT-PCR. Of the 338 participants, 106 were confirmed positive by the test results of both or either of the NP and saliva samples. Of the participants with positive test results, 59 were recruited from June to August 2021 and 47 were recruited from April to June 2022. A total of 100 and 101 participants tested positive by rRT-PCR using NP and saliva samples, respectively (Table 1). The saliva samples showed a positive percent agreement (PPA) of 95.0% (95% CI, 88.7%–98.4%), a negative percent agreement of 97.9% (95% CI, 95.2%–99.3%), and an overall percent agreement of 96.8% (95% CI, 94.3%–98.4%) with the NP sample results. Of the 106 participants with positive results on one or both specimens, 7 were asymptomatic. SARS-CoV-2 was detected in the NP samples of 98 of 99 symptomatic participants and 2 of 7 asymptomatic participants, and in the saliva samples of 96 of 99 symptomatic participants and 5 of 7 asymptomatic participants. The NP and saliva sample results were discrepant in 11 participants (Table 2), of whom 7 were asymptomatic. Six participants had a negative NP sample result and a positive saliva sample result. The median rRT-PCR Ct values using NP and saliva samples were 16.5 versus 22.6 (P < .001) for the E gene, 16.2 versus 22.4 (P < .001) for the RdRp gene, and 23.6 versus 21.2 (P < .001) for the IC, indicating higher viral loads in the NP samples than in the saliva samples. The significantly lower Ct value of the IC in saliva than in NP samples suggested a higher concentration of human cells in saliva.

The median time between symptom onset and testing among the 99 participants with symptomatic SARS-CoV-2 infection was 3 days (interquartile range, 3–5 days). The rRT-PCR Ct values of the RdRp gene are shown according to the time since symptom onset in Figure 1. Although the Ct values were lower in the NP samples than in the saliva samples, the rise in the Ct values of the NP samples according to the time since symptom onset was more pronounced than that of the saliva samples.

The Ct values of 95 samples that tested positive by rRT-PCR in both the NP and saliva samples were compared using Bland-Altman analysis. Most of the Ct values for saliva were within 1.96 SD of the NP Ct values, with a mean difference of 5.3 (Figure 2, A). In participants with a low Ct value in the NP sample, there tended to be a large difference between the Ct value of the NP and saliva samples, but the difference became smaller as the Ct value of NP samples increased.

Samples from 99 of the 106 participants with confirmed infection were tested for SARS-CoV-2 variants. Six samples in which the target amplicon did not amplify and 1 sample that was insufficient were excluded from the analysis.

Among the positive participants recruited between June and August 2021, 57 sets of samples were analyzed for variants, which consisted of 41 Delta, 11 B.1.619.1, 3 B.1.620, and 2 Alpha variants. Among the positive participants recruited between April and June 2022, 42 sets of samples were analyzed for variants, all of which were confirmed to be the Omicron variant (Table 3), consisting of 3 BA.1.1, 14 BA.2, and 25 BA.2.3. The rRT-PCR Ct values of the RdRp gene in the NP samples showed a significant difference between the Omicron and B.1.619.1 variants (P = .045), but no significant differences were observed between the Ct values in the other variants. The rRT-PCR Ct values of the RdRp gene in saliva samples showed no significant differences between variants. For each variant, rRT-PCR using saliva samples showed a PPA of 90% or higher compared with the NP sample results and almost perfect agreement according to the κ values. The Bland-Altman analysis by variant revealed 1 saliva sample result differing from the NP sample result by more than 1.96 SD for the Delta variant (Figure 2, B), 2 saliva sample results differing from the NP sample result by more than 1.96 SD for the Omicron variant (Figure 2, C), and 1 saliva sample result differing from the NP sample result by more than 1.96 SD for the other variants (Figure 2, D). Similar to the overall trend for all positive sets of samples, in participants with a low Ct value in the NP sample, there tended to be a large difference between the Ct value of the NP and saliva samples for each variant, but the difference became smaller as the Ct value of NP samples increased.

In this prospective study, saliva was collected into a UTM container with no specific pretreatment and tested by rRT-PCR for SARS-CoV-2 in the same way as the NP sample. SARS-CoV-2 gene detection using saliva showed a favorable PPA and negative percent agreement compared with the standard method using NP samples. In addition, 5 asymptomatic participants with negative NP sample results had SARS-CoV-2 detected in their saliva sample. Moreover, the results of variant analysis suggested that saliva can be used as an effective sample type for detecting variants, including the Omicron variant.

Saliva collection is a noninvasive procedure that does not need to be performed by skilled medical staff, enabling self-collection without a visit to the hospital. Moreover, it can minimize the exposure of medical staff to SARS-CoV-2 infection, induces no pain, and has fewer side effects than NP sample collection.^4,5  Saliva is a complex biomixture of secretions, fluid, and transudate in various proportions.¹²  Collection methods for posterior oropharyngeal or deep throat samples have also been used to collect saliva samples for SARS-CoV-2 PCR testing.^4,23,24  In this study, a direct collection method was used, using a specifically manufactured plastic UTM container. Saliva collection was carried out by the participants themselves using the provided kit under the observation of the medical staff who collected NP samples. Two types of assisting devices were used for saliva quantification, with the method using a plastic container being easier for quantification. In this study, the drooling technique was used to collect saliva, with no precollection preparation such as forceful coughing and no collection of posterior oropharyngeal saliva. However, saliva sample collection was more time-consuming than NP sample collection because many of the SARS-CoV-2–positive participants experienced a dry mouth and thirst. In spite of using the drooling technique, saliva samples submitted by participants occasionally appeared mucoid or bubbly. However, RNA was extracted from the saliva sample collected in UTM after vortexing for 10 seconds with no specific pretreatment, in the same way as the NP samples, and no samples showed inhibition of rRT-PCR amplification.

Bland-Altman plots were used to compare the rRT-PCR results of the NP and saliva samples. Although, in general, lower Ct values were observed with the NP samples than with the saliva samples, the difference between the NP and saliva samples was smaller in participants with higher Ct values in the NP sample. This suggests that saliva could be a useful sample type in patients with a low viral load, as the difference in viral load between the NP and saliva samples was less. Although the viral load can be more severely underestimated in saliva than in NP samples, the detection rate of SARS-CoV-2 by rRT-PCR was similar between the saliva and NP samples, with minimal impact on the diagnosis. That the Ct values increased with time since the onset of symptoms in SARS-CoV-2–positive participants and the degree of increase was lower in saliva than in NP samples suggests that saliva could be used for virus detection as time passes after the onset of symptoms. SARS-CoV-2 has been shown to be more frequently detected in saliva than in NP samples after the diagnosis of COVID-19.²⁵  Moreover, the difference in PPA of rRT-PCR results tends to decrease with increasing time since symptom onset,²⁶  which is more marked in asymptomatic patients. Likewise, among 11 participants with a discrepancy between NP and saliva sample results, 7 participants were asymptomatic, of whom 5 were confirmed positive only in the saliva sample and 2 were confirmed positive only in the NP sample. Taken together, these results suggest that saliva may be useful as an ancillary sample to the NP sample in asymptomatic patients with a low viral load. Moreover, as saliva and NP samples showed a similar SARS-CoV-2 detection rate in participants with a moderate or high viral load, saliva could be used as an alternative to NP samples.

In this study, various SARS-CoV-2 variants were identified using amplicon-based next-generation sequencing. Among the samples collected in 2021, the most common variant was the Delta variant, followed by the B.1.619.1 variant. Starting in April 2021, an increase in the B.1.619.1 and B.1.620 variants was reported in South Korea.²⁷  Regarding the viral load of NP samples by variant, the Omicron variant showed a tendency toward lower Ct values, but no significant differences were observed with the other variants except B.1.619.1. The Ct values of RdRp gene in saliva showed no significant differences between variants. As was observed for all the variants combined, each variant exhibited a similar pattern of difference in the Ct values between NP and saliva samples. The Bland-Altman plot showed a tendency of the difference in the Ct values between the 2 types of samples to decrease as the Ct value increased. In a previous study, the Delta variant showed a significantly lower Ct value than the Alpha, Gamma, and historical clade 20G variants, especially in presymptomatic or asymptomatic patients.¹⁵  These differences in viral load between virus variants indicate the high infectivity of the Delta variant.¹⁷  In another study, the Delta variant showed a higher viral load in both saliva and NP samples than the Omicron variant and induced more severe clinical symptoms.²⁸  Conversely, Cornette et al¹⁸  reported no significant difference in the viral load between patients infected with the Omicron and Delta variants, although saliva was more sensitive than NP samples. In NP samples, the Omicron variant showed no significant difference in viral load compared with the Delta variant in both symptomatic and asymptomatic patients.¹⁶  We confirmed that the viral load in saliva samples did not differ markedly from that of standard NP samples and identified the trend in the difference in viral load between the 2 types of samples.

This study has some limitations. First, this study did not include children, in whom obtaining saliva samples is a useful option. A recent study reported that in children, saliva samples have a lower utility than NP samples because of a low viral load, and that the younger the patient, the larger the difference between the 2 sample types.²⁹  In contrast, another study found that saliva was an adequate alternative to NP samples in children.³⁰  This study recruited only patients aged 13 years or older, and although 17 participants were aged 19 years or younger, only two 19-year-olds and one 18-year-old were confirmed positive by rRT-PCR. As a consequence, no sufficient validation was performed in the young age group including children. Second, no follow-up testing was performed. Although symptoms and clinical information were recorded during history taking, a subsequent rRT-PCR and a follow-up of clinical features were not performed. In individuals with SARS-CoV-2 infection, the presence of symptoms can affect the viral load^12,16 : the higher the viral load, the greater the likelihood of viral spread³¹  and the greater the disease severity.³²  Thus, further studies are warranted to determine the association between the time of symptom disappearance and clinical features using saliva samples. Third, in this study, food and drink were restricted for 30 minutes before saliva collection, and the history of food and drink intake, smoking, and medications was not investigated. Thus, it is impossible to determine whether these factors affected saliva samples. According to one study, eating, drinking, and smoking before testing had no impact on SARS-CoV-2 positivity rate.¹⁸  Studies should be conducted on substances that can interfere with viral RNA detection or that can enhance the specificity of saliva samples.

In summary, this study confirmed that saliva is a useful sample type for SARS-CoV-2 detection using rRT-PCR testing, and that saliva and NP samples can be collected at the same time. Various SARS-CoV-2 variants were effectively detected in saliva samples, and saliva samples were useful in asymptomatic participants and participants who were tested several days after symptom onset, especially when used with NP samples. Considering the advantages of saliva collection, saliva could be useful as a sample type for SARS-CoV-2 diagnosis.