Substantial evidence already exists on the non-inferiority of saliva relative to NOS as a specimen for conventional RNA extraction-dependent RT-qPCR detection of SARS-CoV-2 [7, 8]. Compared to NOS as a testing sample, the use of saliva has several advantages. In this alternative setup, compared to the number of staff that must be trained for adequate-quality NOS collection one patient at a time, fewer personnel needing less technical training are needed to facilitate simultaneous self-collection of saliva by multiple patients. This potentially translates to a less aggregate risk of SARS-CoV-2 transmission between patients and personnel (due to avoidance of close-contact interactions secondary to swabbing, and fewer personnel contact with materials that could not be sanitized beforehand such as patient samples) and reduced demand for personal protective equipment (PPE). In terms of logistics, storage of samples before transport to the laboratory is less complicated (with the use of plain sterile vials rather than a combination of tubes, ice-cooled chest boxes, and proprietary swabs and transport reagents). Patients are also expected to be more amenable to such sample collection for SARS-CoV-2 testing than invasive and uncomfortable swabbing. It is also important to note that variation of several patient and personnel factors related to naso-oropharyngeal swabbing introduces iatrogenic preanalytical variability that may adversely affect testing results [14, 15]. Furthermore, these findings may be useful in light of recent reports suggesting changing patterns of viral shedding, tissue tropism, and transmission mechanics [16–18], in favor of oral cavity and saliva, depending on the emerging SARS-CoV-2 variant such as Omicron [19].
Nevertheless, there remains fewer endeavors on attempting to streamline these laboratory processes without significant decline in analytical and/or diagnostic performance [9, 20]. This is particularly important in the geographic and sociopolitical contexts of developing countries. Such settings are associated with challenges in addressing demands for COVID-19 pandemic response, with meager resources, such as sustaining testing coverage that is adequate, timely and accurate [21]. In the Philippines, a previous undertaking [22] adapted the RNA extraction-free technique, known as covidSHIELD, of the University of Illinois at Urbana-Champaign [20] for saliva RT-qPCR detection of SARS-CoV-2. This has led to the first regulatory approval of the technology in the country through Sansure Biotech (Hunan, China)-based molecular biology facilities of the Philippine Red Cross [23]. The present study, on the other hand, sought to evaluate the SalivaDirect procedure[9] through the harmonized laboratory system of a private hospital network, which utilizes GeneFinder reagents and Bio-Rad RT-qPCR platforms. While it may appear that covidSHIELD involves theoretically cheaper reagents for saliva processing (Tween 20 and Tris-Borate-Ethylenediaminetetraacetic acid [TBE] buffer) compared to SalivaDirect’s Proteinase K, an advantage of the latter is that its heat-inactivation step is only 5 minutes or one-sixth of the time required by the other. On top of the aforementioned benefits of the testing regime shift, from swab-based to saliva-based, streamlined pre-RT-qPCR saliva processing through heat inactivation and Proteinase K addition further reduces (1) the resource cost to release SARS-CoV-2 RNA copies from capsids compared to the use of proprietary reagents, and (2) the risk of laboratory personnel exposure to viable SARS-CoV-2 units in patient samples (as the virus is known to be denatured at a fraction of time and thermal energy applied during the prescribed inactivation step [24]).
The results of the present study add to the growing body of evidence that indicate favorable comparability of saliva and NOS as specimens for SARS-CoV-2 nucleic acid-based detection. The head-on comparisons between SE RT-qPCR and NOS RT-qPCR and between SalivaDirect RT-qPCR and NOS RT-qPCR reveal statistically similar diagnostic accuracy parameter estimates. Furthermore, our post-hoc analyses showed that among volunteers whose NOS, SE and SalivaDirect samples were also positive for SARS-CoV-2, the mean (SD) Cq values for all three SARS-CoV-2 gene targets (RdRp, N and E) and the human RNase P internal control are not significantly different between the SE and SalivaDirect specimens. As the Cq for a genomic target is inversely proportional to the initial amount of target gene copies in the post-processed RT-qPCR mixture, these findings imply that the presumed increase in heterogeneity of heat-inactivated and Proteinase K-treated saliva, compared to saliva that was processed conventionally, may not have adversely affected the analytical, diagnostic and quantitative performance of the RT-qPCR procedure common and downstream to both tests. However, compared to counterpart values for NOS, the SARS-CoV-2 viral load and the human cell content appear to be lower and higher, respectively, in both saliva specimen types. Higher RdRp, N and E Cq readings for NOS suggest that either (1) the viral load in the nasopharyngeal area is higher than in the oral cavity or (2) the liquid volume of saliva led to a “dilution” of the corresponding SARS-CoV-2 content. The significant discrepancy in human RNase P Cq values may be due to the higher human cell content in saliva than in the nasopharyngeal swab scrapings.
Because volunteers having a positive saliva test result and a negative swab test result will be considered false positives in head-on comparison (with NOS RT-qPCR as the reference standard), the estimated specificity and PPV would have appeared to be lower than what could be expected in an RT-qPCR platform. Assuming that optimal quality control procedures were observed from test sample collection to processing, RT-qPCR specificity and PPV are theoretically 100% (making it impossible to assign a sample as a false positive) because the primers designed to detect the genetic material of a pathogen is highly and molecularly specific to its genomic sequence [2]. It is thus more likely that saliva-positive but swab-negative volunteers represent truly-infected cases missed by swab testing, rather than false positives, and vice versa. Arguably, a false sense of security may befall such patients if they were tested only by NOS RT-qPCR. We hypothesize that this subset of the population may be a greater clinical or public health risk than their swab-positive but saliva-negative counterparts, considering that SARS-CoV-2, in terms of anatomy and biomechanics, is more likely to be released from the host to the environment (through aerosolization and droplet formation) in saliva than in the nasopharynx [25]. With this consideration of perfect specificity and PPV in mind, we then analytically constructed a CRS wherein the criterion for a positive case is at least either a SARS-CoV-2-positive NOS, SD or SalivaDirect sample (virtually ensuring specificity and PPV of 100% since no volunteer can be considered false positive), and that for a negative case is SARS-CoV-2-negative NOS, SD and SalivaDirect samples. With this benchmark, the diagnostic performance of NOS RT-qPCR, SE RT-qPCR and SalivaDirect RT-qPCR was separately assessed. All tests performed statistically similar to one another in terms of sensitivity, NPV, accuracy and agreement in this scenario, again suggesting that both saliva as a sample and the streamlined technique to handle this specimen are non-inferior, at the very least, to the conventional test in epidemiologically detecting SARS-CoV-2. Another remarkable finding that is apparent in this construction is that all tests, even NOS RT-qPCR which is the considered reference standard, was not able to detect SARS-CoV-2 in some volunteers whose other samples turned positive. While this can be due to iatrogenic factors (e.g., patient-specific and personnel-specific aspects concerning naso-oropharyngeal swabbing quality), potential pathophysiological and fluid-tissue localization mechanisms that are yet to be elucidated in SARS-CoV-2 infections may also be at play.
We recognize several limitations of this study. As a cross-sectional design was implemented, follow-up through clinical assessment and laboratory testing for at least one time point after initial study contact was not performed. Such a procedure could demonstrate viral kinetics in the specimens through the preclinical, clinical, and convalescent or mortality phases of the infection. This study was also insufficiently powered to determine the impact of sample collection timing (from the history of COVID-19 exposure and/or onset of relevant symptoms) on test results. RT-qPCR of NOS samples that underwent RNA extraction-free processing was also not performed; this would have provided direct information on the difference in categorical and quantitative (Cq) results due to the change in pre-RT-qPCR processing technique for swab specimens. We also suggest that head-on comparison of the covidSHIELD and the SalivaDirect techniques be conducted to generate primary evidence on the relative performance and utility of these two major streamlining approaches for saliva samples. Nevertheless, considering the results of our study, we recommend the implementation of SalivaDirect for its potential to speed up testing turnaround time and to reduce laboratory costs.