Worldwide, molecular detection of SARS-CoV-2 is primarily focused on swab-based sampling of the nasopharyngeal, nasal, and/or oropharyngeal cavities, but as the pandemic has progressed there has been increasing interest in the role of saliva as a sample type.[[1]]

Saliva holds several advantages over swab based techniques: it can be self-collected, which may reduce pressure on healthcare providers in the community, and it is less invasive, which may be desirable for people who require frequent testing.[[2]] Early on in the pandemic, where swabs and extraction reagents were scarce, an ability to bypass these requirements was also an advantage though this has largely been overcome by increased manufacturing capability and wider use of integrated machines combining extraction and amplification.[[3]]

Several systematic reviews have now been published, comparing molecular detection of SARS-CoV-2 from saliva and swab-based techniques, and these have shown that the sensitivity is similar,[[1,4,5]] though the studies are heterogenous in nature and several have indicated poorer performance and technical issues depending on methodology and use scenario.[[6,7]]

However, only a limited number of diagnostic companies have sought authorisation for their swab based molecular tests for testing saliva,[[8]] which prompts the question as to whether commonly used swab based nucleic acid amplification tests (NAATs) can be repurposed for saliva.

We, therefore, sought to compare detection of SARS-CoV-2 from paired nasopharyngeal swabs and saliva in order to understand the performance of saliva using NAATs in common use for testing swabs in New Zealand.

Methods

LabPLUS, Auckland Hospital

One hundred and seventy-three paired saliva and nasopharyngeal samples were available from 150 individuals at a managed isolation facility (MIF) for international arrivals in 2020 and 23 patients hospitalised with COVID-19 during the community outbreak in Auckland in August and September 2021.

Nucleic acids were extracted from 200μL of saliva or 200μL of viral transport medium (VTM) containing the nasopharyngeal swab on the MagNA Pure 96 with the MagNA Pure 96 DNA and Viral NA Small Volume extraction kit (Roche Diagnostic, Germany), prior to testing on the LightCycler 480 instrument (Roche), using the in-house E gene assay according to previously described method.[[9]]

For the hospitalised patient samples, 200µL of saliva or VTM (nasopharyngeal swab) were extracted on the KingFisher™ Flex using NucleoMag (Machery-Nagel) extraction kits prior to testing on the ABI 7500 (Thermofisher, US) instrument using the TaqPath COVID-19 Combi kit (Thermofisher).

Southern Community Laboratories, Dunedin

Thirteen paired saliva and nasopharyngeal samples were available from crew on board a quarantined ship. 400µL of saliva or VTM (nasopharyngeal swab) were extracted on the KingFisher™ Flex using the MagMAX™ Viral/Pathogen II Nucleic Acid Isolation Kit prior to testing with the TaqPath™ COVID-19 Combo assay on QuantStudio™ 7500 Real Time PCR instrument.

Southern Community Laboratories, Wellington

Paired saliva and nasopharyngeal samples from two maritime crew members and eight persons in the Wellington managed isolation and quarantine facility (MIQ) were included in this study. 200µL of each sample was extracted on the MagNA Pure 96 using the MagNA Pure 96 DNA and Viral NA Small Volume Kit (Roche). RT-PCR for SARS-CoV-2 was performed on the maritime samples using the LightMix SARS-CoV-2 (E-gene) assay (Roche), and on the MIQ samples using the Perkin Elmer SARS-CoV-2 QRT-PCR assay according to the manufacturer's instructions.

Comparisons

Analysis of the data set was performed retrospectively on samples which had been gathered independently and tested at the study sites across New Zealand as part of validation studies. To be eligible for inclusion, both sample types were required to have been collected from an individual on the same day.

Delta (∆) Ct (the difference between cycle threshold values) was calculated for nasopharyngeal swabs and saliva based on a single gene. Chi-squared tests were used to compare proportions. Fisher Exact tests were used to compare performance by laboratory and assay.

Ethics

The MIF samples tested at Auckland Hospital were taken as part of an Institute of Environmental Science and Research (ESR) excretion of SARS-CoV-2 in saliva and faeces study for which ethical approval for the study was obtained from the Health and Disability Ethics Committee (ethics reference: 20/NTB/216/AM01). Individuals at Auckland Hospital, and the other sites provided, informed verbal consent to partake in assay validation by providing additional samples; this approach was endorsed by the New Zealand Ministry of Health.

Results

One hundred and ninety-six paired nasopharyngeal and saliva samples from unique individuals were tested, with 46 (23%) positive from either sample type, 43 (93% of total positive) from nasopharyngeal swab, and 42 (91% of total positive) from saliva, indicating no significant difference between performance of the two sample types (p=0.69). In three instances, saliva samples were positive where nasopharyngeal swabs were negative, and in four instances the nasopharyngeal swabs were positive and the saliva samples were negative. Comparing the assays, there was no statistically significant difference between the performance of samples at any laboratory or assay combination (p=0.069); see Table 1 for performance at specific laboratory and assay combinations.

The positive percentage agreement of saliva compared with nasopharyngeal swab was 91%, (95%CI 81.2–95.5%) and negative percentage agreement was 98%, (95%CI 95.4–99.4%), with a kappa of 0.90 (95%CI 0.77–0.96%).

In comparison the percentage agreement of nasopharyngeal swabs compared with saliva was 92.9% (95%CI 83.2–97.7%) and negative percentage agreement was 97.4% (95%CI 94.8–98.7%) with a kappa of 0.90% (0.77–0.96%).

The average difference in Ct between saliva and nasopharyngeal swabs across the sample set was 0.22 cycles, indicating excellent concordance; however, the difference between nasopharyngeal swabs and saliva collected from the same individual was quite variable with up to 19 cycles difference between the sample types (see Table 1); 20/46 (43%) of positive saliva samples had lower Cts (implying higher viral loads) compared with the nasopharyngeal swab, and 26/46 (57%) had lower Cts in the nasopharyngeal swab (p=0.21).

View Table 1.

Discussion

We compared the performance of nasopharyngeal swabs with saliva as sample types for the molecular detection of SARS-CoV-2 at three laboratories, across five test combinations and found that the sensitivity was similar between both sample types. It is notable that a non-statistically significant lower overall detection rate was seen for saliva compared with NPS in our dataset (91% versus 93%). Both of these observations are consistent with findings of several meta-analyses.[[1,4,5]] Overall, these findings support that saliva is an appropriate sample type for detection of SARS-CoV-2 by NAAT for diagnosis and surveillance in New Zealand.

We found that saliva and nasopharyngeal swabs detected positives where the other tested negative, and vice versa, and that the cycle threshold, an imperfect marker of viral load for semi-quantitative assays, was quite variable between the two sample types. These findings were not significantly skewed towards one sample type: there was a wide range of difference between cycles thresholds of between -16.67 to 19.07 cycles between saliva and NPS across the dataset, and whilst there was a predominance of higher viral loads in NPS for 57% of positives, the viral load was higher in saliva in the other 43%, with an net overall difference of only 0.22 cycles between sample types. This variability in viral load is most likely explained by differential viral dynamics in the anatomical spaces which change over time,[[10]] is a limitation of either sample type and indicates neither sample type was significantly less sensitive than the other.

Shedding of viral RNA via the nasopharynx can occur for some time following infection, with a median shedding time of 19 days reported in a large population-based cohort in Canada,[[11]] and as the prevalence of past infections rises, detection of non-viable RNA can potentially lead to unnecessary isolation of individuals. There is some evidence that saliva shedding is highest in the first week of illness and drops thereafter, indicating it may better reflect infectiousness compared with a nasopharyngeal PCR,[[12]] which may be of benefit in the pivot from an elimination strategy to living with endemic COVID-19.

Over the New Zealand Delta outbreak, saliva had a role in the public health response in New Zealand, particularly as an option for regular surveillance testing of international border workers or essential workers crossing regional boundaries,[[13,14]] and in some settings for healthcare workers caring for COVID-19 infected patients. It is not quite clear what role saliva NAAT testing will have in the New Zealand public health response in the future. During the New Zealand Omicron outbreak, general surveillance testing has reduced and rapid antigen testing (RAT) has largely replaced saliva NAAT testing, and has greatly reduced the volume of laboratory NAATs. RAT has similar advantages to saliva NAAT for frequent asymptomatic testing,[[15,16]] with additional advantages such as decentralisation of testing, self-testing, lower cost, and ability to provide results in as little as 10 minutes.[[17]] Judicious use of this modality can reduce reliance on laboratories and save resources for diagnostic testing for public health and healthcare purposes.

Saliva appears attractive as one of the possible alternative sample types for individuals where swab-based testing is not tolerated. Given that New Zealand is now several months divorced from its elimination approach, it is now less critical to identify every case of COVID-19 in the community, and providing more tolerable options to nasopharyngeal swabs is desirable for people presenting for testing. In addition to saliva, swabbing of the throat, mid-turbinates or anterior nares alone and self-swabbing could all be offered as alternative sample types where logistical and operational considerations allow.

Saliva NAAT testing is not universally available in New Zealand laboratories, due to largely practical reasons: saliva is a heterogeneous matrix and in common with others,[[4]] we have found that it requires different pre-processing steps compared with swab-based testing to support high-throughput workflows. These steps may reduce the efficiency of existing workflows (particularly for integrated testing platforms), impacting on turnaround times for those individuals tested. Secondly, because diagnostic companies have not sought authorisation for their products to be used for saliva, the requirement for additional validation of the sample type is difficult for smaller laboratories to achieve, particularly when under constraints of responding to high NAAT test demand.

The purpose of the study was to compare detection of SARS-CoV-2 from paired nasopharyngeal swabs and saliva in order to understand the performance of saliva using NAATs in common use for testing swabs in New Zealand. There were substantial challenges obtaining this data over the course of the pandemic due to low infection rates in New Zealand, logistical challenges, and inter-agency co-operation. Therefore, this study is imperfect due to its retrospective nature and the small numbers tested at sites outside Auckland, but it nevertheless provides valuable aggregate information on the performance of different sample types for the detection of SARS-CoV-2. Whilst Pitman et al.[[18]] performed an analysis on imported patient samples from the United States, local data, such as ours, obtained in the New Zealand setting, is essential to help inform national testing strategies. The smaller numbers tested at each location limits our ability to comment on individual assay performance; however, whilst outside the scope of this study, this work has been performed separately and is required before any assay combination is used to test saliva for SARS-CoV-2 by NAAT. It is also important to note there were quite large differences in positivity amongst the participating laboratories, reflecting the different sources of samples. This study was restricted to those samples which were simultaneously paired, in order to best assess analytical and sample factors. This restricted the numbers of samples we were able to include in the study but ensured that the patients’ infection status was the same at the time of sampling. Further ongoing studies would be useful to assess the impact of vaccination status, variants, and time from onset of infection on detection in different anatomical spaces.

In summary, we found that saliva is an equivalent sample type to nasopharyngeal swabs for the detection of SARS-CoV-2 in our laboratories using multiple assay combinations and is suitable for use as a diagnostic and surveillance test.