Saliva as an alternative to the naso-oropharyngeal swab for the detection of SARS-CoV-2 by RT-qPCR: a multicenter cross-sectional diagnostic validation study

Study design and ethics review

The protocol for this prospective cross-sectional study was reviewed and approved, with its implementation controlled, by the ManilaMed Ethics Review Committee (MMERC No. 2021-06). The study was performed in accordance with the Declaration of Helsinki and the International Ethical Guidelines for Biomedical Research Involving Human Subjects.


Volunteers considered potentially eligible for inclusion and ultimately invited to participate, through consecutive sampling, were inpatient or outpatient adults receiving the NOS RT-qPCR SARS-CoV-2 test, from May to December 2021, in one of three hospitals in the Philippines: Fe Del Mundo Medical Center (Quezon City, National Capital Region), Dagupan Doctors Villaflor Memorial Hospital (Dagupan City, Pangasinan Province, Luzon) and Ciudad Medical Zamboanga (Zamboanga City, Zamboanga del Sur, Mindanao). Volunteers were included if they are able to give informed consent and independently collect saliva. Volunteers were excluded if they were unable to ensure avoidance of enteral ingestion, gargling with mouthwash, or brushing teeth and smoking for at least 30 minutes prior to supply saliva. Demographic and clinical data such as exposure history and symptoms were obtained from consenting participants.

Test methods

Paired saliva and NOS samples were collected from all volunteers. Samples from each patient were placed in separate labeled tubes (one for NOS and one for saliva), and these tubes were stored for transport in containers distinctly carrying only one or the other type of sample. ‘sample. These were subjected to subsequent laboratory procedures within 24 hours of sample collection, with segregation in terms of personnel and instruments depending on the type of sample. Interpretation of the tests was performed independently by two evaluators who did not have access to clinical information, and the matched sample result from the same study volunteer. The samples were processed and tested at laboratories using harmonized protocols at the aforementioned hospitals that are licensed to operate SARS-CoV-2 RT-qPCR facilities by the Philippine Department of Health.

Salivary RT-qPCR

Prior to NOS sampling, included volunteers were asked to drool at least 1 mL of saliva into a sterile 5 mL tube. The filled sterile tube was then sealed and stored in a chest at room temperature before being transported to the designated laboratory. Each saliva sample was partitioned for conventional Saliva RNA Extraction (SE) and SalivaDirect procedures. The volume allocated for SE was subjected to the conditions prescribed by the manufacturer via the Nextractor NX-48 automated system (Genolution Inc., South Korea). Sample preparation for SalivaDirect was performed according to the method described by Vogels et al.9. Briefly, 2.5 μL (50 mg/mL) of proteinase K was added to 50 μL of saliva in PCR tubes, which were then vortexed at 3200 rpm for 1 min. The samples were then heated at 95°C for 5 min. From these RNA-extracted and SalivaDirect-treated mixtures, 5 μL was used as input in subsequent RT-qPCR processes to amplify SARS-CoV-2 RNA-dependent RNA polymerase (RdRp), envelope (E) and the nucleocapsid (NOT) using the GeneFinder COVID-19 Plus RealAmp Kit (OSANG Healthcare Co., Ltd., South Korea). The kit uses humans RNase P gene template as an internal control, and RdRp, E, NOT and human RNase P RT-qPCR amplified constructs are stained with fluorescein amidite (FAM), Texas Red, 5′-dichloro-dimethoxy-fluorescein/Victoria (JOE/VIC), and Cy5 fluorophores, respectively. An RT-qPCR analysis was deemed valid if (1) the quantification cycle (Cq) readings for RdRp, E, NOT and RNase P were all ≤ 22.00 for designated positive controls and (2) negative control Cq readings for the same genes were all ≥ 40.00 or blank/indeterminate. A sample was considered positive for SARS-CoV-2 if the Cq readings for RNase P and at least either RdRp, E Where NOT were ≤ 40.00, with the characteristic sigmoidal amplification curve noted in all cases. A negative result for SARS-CoV-2 was indicated if RNase P VSq ≤ 40.00 with sigmoid and C amplification curveq readings for RdRp, E and NOT were all ≥ 40.00 or blank/indeterminate. A new sample test was immediately performed if RNase P VSq ≥ 40.00 regardless RdRp, E and NOT VSqvalues. All assay reactions were maintained in the CFX96 RT-qPCR system (Bio-Rad Laboratories, CA, USA).


The conventional RNA extraction-dependent RT-qPCR technique to detect SARS-CoV-2 in NOS was used to assess the utility of saliva-based assays. The volunteers underwent a naso-oropharyngeal swab performed by trained personnel, and the swabs were then placed in sterile tubes containing universal transport medium. The filled tubes were sealed and stored in a cold room at 4–6°C before being transported to the designated laboratory. Manufacturer’s prescribed instructions were followed for sample preparation via the Nextractor NX-48 automated system and RT-qPCR steps to detect the same aforementioned genetic targets (RdRp, NOT, E and human RNase P). Criteria for validity of performing RT-qPCR and positivity of the NOS sample were the same as for saliva samples.

Data analysis

The Buderer techniqueten was used to calculate the minimum sample size. Assuming 10% COVID-19 prevalence during study planning period, target sensitivity and specificity of at least 90% for saliva-based RT-qPCR compared to NOS RT-qPCR , and the level of significance (α) and the maximum acceptable width of the 95% confidence interval (CI) being set at 0.05 and 10%, respectively, gave a minimum sample size of 385.

Descriptive statistics were presented as proportions for categorical variables and median (interquartile range or IQR) for continuous variables. Diagnostic validity estimates (sensitivity, specificity, positive predictive value or PPV, negative predictive value or NPV, and accuracy) and their 95% CIs were calculated using the online calculator MedCalc11, for three scenarios: (1) SalivaDirect RT-qPCR as index test and NOS RT-qPCR as reference test, (2) SE RT-qPCR as index test and NOS RT-qPCR as reference test, and (3) SalivaDirect RT-qPCR, SE RT-qPCR and NOS RT-qPCR as index tests compared separately to a composite reference standard (CRS). In CRS, a participant with at least one NOS or SalivaDirect or SE sample detectable for SARS-CoV-2 was considered positive and a volunteer with all samples without detection of the SARS-CoV-2 gene as negative for the virus.12. Such a paradigm would generate perfect specificity and PPV for all index tests, since there would be no false positive results.

McNemar χ2 was used to statistically compare the sensitivity, specificity, and accuracy of SE RT-qPCR and SalivaDirect RT-qPCR versus NOS RT-qPCR. To compare the PPV and NPV of the two saliva-based index tests, with the swab test as the gold standard, the weighted generalized score test statistical approach developed by Kosinski13 has been used. Cohen’s coefficients κ for concordance between the designated index and benchmark tests were also estimated14. To assess the statistical difference in sensitivity and precision, relative to CRS, between NOS RT-qPCR, SE RT-qPCR and SalivaDirect RT-qPCR, Cochran Q omnibus test with McNemar χ2 test as a post-hoc pairwise comparison technique was implemented. In the case of the three-factor comparison of the resulting NPVs against the CRS, the aforementioned Kosinski approach was carried out in pairs with the Bonferroni correction. An analysis comparing the CqSARS-CoV-2 gene values ​​and RNase P between valid NOS, SalivaDirect and SE samples among volunteers positive in at least one test was also performed via the Skillings-Mack omnibus test15 with the Wilcoxon signed rank test as a post hoc technique for pairwise comparison. Statistical significance was set at p ≤ 0.0500 unless otherwise specified. The rest of the statistical analyzes were performed in Stata 14.2 (StataCorp LLC, College Station, TX, USA).

Comments are closed.