Abstract

A systematic review and meta-analysis were conducted to investigate the SARS-CoV-2 viral load in human saliva and compared it with the loads in oropharyngeal swabs, nasopharyngeal swabs, and sputum. In addition, the salivary viral loads of symptomatic and asymptomatic COVID-19 patients were compared. Searches were conducted using four electronic databases: PubMed, Embase, Scopus, and Web of Science, for studies published on SARS-CoV-2 loads expressed by CT values or copies/mL RNA. Three reviewers evaluated the included studies to confirm eligibility and assessed the risk of bias. A total of 37 studies were included. Mean CT values in saliva ranged from 21.5 to 39.6 and mean copies/mL RNA ranged from 1.91 × 101 to 6.98 × 1011. Meta-analysis revealed no significant differences in SARS-CoV-2 load in saliva compared to oropharyngeal swabs, nasopharyngeal swabs, and sputum. In addition, no significant differences were observed in the salivary viral load of symptomatic and asymptomatic COVID-19 patients. We conclude that saliva specimen can be used as an alternative for SARS-CoV-2 detection in oropharyngeal swabs, nasopharyngeal swabs, and sputum.

1. Introduction

Coronavirus disease 2019 (COVID-19), caused by SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2), was confirmed as an outbreak reported in Wuhan, China, in December 2019 [1]. Already by March 11th, 2020, it was declared as a global pandemic, indicating the contagiousness and related fast spreading of the virus. By March 16th, 2022, the virus had globally infected over 462 million people with approximately 6 million deaths [2]. To date, these numbers are still increasing.

Most individuals who become infected show mild to moderate flu-like symptoms and recover without hospitalization. Clinical symptoms of COVID-19 are diverse ranging from mild to severe including fever, dry cough, smell- and taste-loss, dyspnea, muscle pain, headache, and respiratory tract infection. In most severe cases, it may lead to lung failure, hospitalization, and death [3]. However, it has been shown that 24% of the population infected with SARS-CoV-2 remained asymptomatic [4, 5]. Several risk factors relate to interindividual differences in sensitivity to COVID-19 including age (fatality rate of patients in the age group 70–80 years is 8% higher than the age groups below [6, 7], gender (higher mortality in males) [8, 9], genetic factors, and underlying comorbidities (cardiovascular diseases, diabetes mellitus, hypertension, chronic kidney disease, and chronic lung diseases) [6]. Differences in viral load kinetics in various body fluids may play a role as well [1015].

The main human-to-human transmission of SARS-CoV-2 occurs via inhalation of aerosols, generated through coughing, sneezing, or direct contact with mucous membranes of the eyes, mouth, and nose [3, 1625]. The receptor-binding domain (RBD) of the coronavirus spike (S) glycoprotein, located on the surface of the viral envelope, mediates viral entry into host cells by binding to the ACE2 (angiotensin-converting enzyme 2) receptor. The binding of the S-protein to ACE2 is subsequently primed by a host cell protease, TMPRSS2 (transmembrane protease, serine 2), which facilitates cell entry [2022]. High expressions of ACE2 and TMPRSS2 are found in the epithelial cells and human acinar granular cells of the salivary glands [2226]. In line, the salivary glands may serve as a reservoir of the virus facilitating viral replication and shedding of infectious particles into saliva. The viral load profile of SARS-CoV-2 in saliva seems to peak during the first week of symptoms onset [27]. However, the virus may still be detected in low amounts such as approximately ∼2 log10 copies/mL after 20–30 days in saliva, despite the range of salivary antiviral molecules which potentially contribute to counteract the viral load and transmission [1, 13, 14, 2730].

The collection of respiratory tract secretions such as nasopharyngeal swabs (NPS), oropharyngeal swabs (OPS), and sputum followed by detection of viral genome with RT-PCR has become the gold standard for SARS-CoV-2 screening and diagnosis. However, collection of these matrices has a series of drawbacks regarding discomfort of patients, risk of exposure to healthcare workers, need for specific instruments, and limiting self-collection [31]. In turn, saliva has been regarded to be an attractive matrix for sampling compared to NPS and OPS collection because it offers benefits such as noninvasive and quick and easy sampling with minimum risk of exposure to healthcare workers and decreasing the need of personal protective equipment [1115, 3234].

Based on the abovementioned, we hypothesized that SARS-CoV-2 screening and diagnostics in saliva is a good alternative for NPS, OPS, and sputum. It appears, so far, that studies have investigated the detection of SARS-CoV-2 viral loads in saliva specimens indicated in measures of sensitivity and specificity. However, until now, no studies with meta-analysis have compared the SARS-CoV-2 viral load in saliva to other biofluids expressed in CT values and copies/mL RNA. Therefore, the aim of this systematic review was first to address the SARS-CoV-2 load (expressed in cycle threshold (Ct)-value or copies/mL RNA) in human saliva, and secondly, to compare the viral load in saliva with OPS, NPS, and sputum. Furthermore, the SARS-CoV-2 load in saliva samples of symptomatic and asymptomatic COVID-19 patients was compared. A meta-analysis was conducted to systematically compare the viral load data from different studies.

2. Materials and Methods

2.1. Protocol Registration

This review was registered in PROSPERO International Registration of Systematic Reviews (CRD42021245877) (https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=245877) and written using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) approach, see Table 1 [35].

2.2. Search Strategy and Data Sources

Advanced literature search strategy was applied using four electronic databases including PubMed, Embase, Scopus, and Web of Science. The search strategy was conducted using the combinations of the following key words: (COVID-19 (title/abstract)) OR (coronavirus (title/abstract)) OR (SARS-CoV-2 (title/abstract)) OR (2019-ncov (title/abstract)) AND (saliva (title/abstract)) OR (saliv (title/abstract)) OR (salivary (title/abstract)) OR (oral (title/abstract)) OR (mouth (title/abstract)) OR (oropharynx (title/abstract)) AND (viral load (title/abstract)). A manual search was conducted in order to include other relevant articles. The search strategy was performed monthly up until April 2021.

2.3. Inclusion and Exclusion Criteria

Inclusion criteria included original published scientific articles in English that reported on SARS-CoV-2 load inhuman saliva until April 2021.

Eligibility criteria were conducted using the PICO guidelines [35]:

2.3.1. Population/Patients (P)

Humans, individuals, determined with SARS-CoV-2 load in saliva (all ages).

2.3.2. Intervention/Exposure (I)

SARS-CoV-2 load detected using RT-PCR.

2.3.3. Comparison (C)

SARS-CoV-2 load in OPS and/or NPS and/or sputum, if available.

2.3.4. Outcome (O)

The difference of SARS-CoV-2 load in saliva compared to NPS, OPS, and/or sputum (expressed in CT values or copies/mL RNA).

Research on the SARS-CoV-2 load was first addressed for saliva. Then, a comparison was made in the viral load in saliva with OPS, NPS, and sputum.

Studies that did not report the viral load in saliva and OPS, NPS, and/or sputum in humans were excluded. Animal studies, reviews, opinion articles, letters to the editor, and case reports were excluded.

2.4. Selection Process

One author (MF) performed the initial literature search. Subsequently, three authors (MF, FB, and ML) examined the titles and abstracts of all identified records. Studies were chosen based on the inclusion and exclusion criteria. A single author (MF) extracted the data from the included articles, which again was verified by the authors FB and ML. Disagreements were resolved by discussion.

2.5. Data Collection Process

For the included studies, the following parameters were extracted: author(s); year of publication; SARS-CoV-2 viral load in saliva; OPS, NPS, and/or sputum (expressed in CT value or copies/mL RNA); methods to detect viral load; saliva sampling; total cohort size; percentage of SARS-CoV-2 positive saliva; days of symptom onset; and salivary viral load in symptomatic and asymptomatic COVID-19 patients, if available. If information was missing, corresponding authors were contacted to complete the data.

Firstly, the SARS-CoV-2 load (expressed in CT value or copies/mL RNA) in saliva was obtained, and secondly, the viral load in saliva was compared to OPS, NPS, or sputum. Finally, the difference in salivary viral load of symptomatic and asymptomatic COVID-19 patients was obtained.

2.6. Risk of Bias in Individual Studies

The potential risk of bias in the included studies was assessed using the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies developed by NIH (National Heart, Lung, and Blood Institute) [36]. Three authors performed the quality assessment independently. Based on the number of “Yes” answers, a rating of good (9–11), fair (5–8), or poor (≤4) was allocated to the individual study. This tool includes 14 questions which were answered by (Yes/No/Not applicable/Not reported/Cannot be determined), see Table 2. Differences in quality rating were discussed by all reviewers (MF, FB, and ML) to reach a consensus.

2.7. Data Synthesis

Data on SARS-CoV-2 salivary load were summarized and compared with SARS-CoV-2 load in OPS, NPS, and/or sputum. When ≥3 comparable studies were available, a meta-analysis was conducted using Review Manager (RevMan version 5.4, the Cochrane Collaboration, 2020), where appropriate, the mean (of viral CT value and viral copies/mL RNA) and standard deviations (SD) were derived. If the mean and SD were not reported, then they were derived from the sample size, median, interquartile range (IQR), and minimum and maximum values using an online calculator at https://www.math.hkbu.edu.hk/~tongt/papers/median2mean.html.Random-effects. A model in RevMan 5.4 was selected to measure the standard mean difference for continuous outcome data with 95% confidence interval (CI). Forest plots were conducted to visualize characteristics of the selected studies; the standard mean difference of viral load in saliva was compared to OPS, NPS, and sputum and the heterogeneity between the studies (I2). A random effects model was applied for moderate heterogeneity (>30%), otherwise the fixed effects model was applied. The overall mean was obtained. value <0.05 was considered as statistically significant.

3. Results

3.1. Study Selection

A total of 712 articles were retrieved through database search (Figure 1). After duplicate removal, 259 articles were screened by the title and abstract and 147 articles were included for full-text reading after which 111 were excluded. Finally, a total of 37 papers were included. Three additional articles were included by manual search.

3.2. Study Characteristics

A total of 21 of the 37 selected studies reported the viral load as a mean or median CT value (Tables 35), while 16 studies reported the viral load in copies/mL RNA (Tables 69). Ten articles reported the viral load solely in saliva and 21 articles reported it in saliva compared with OPS, NPS, and/or sputum. The remaining six studies reported the viral load in OPS [1, 49, 50, 59, 60] and sputum combined with saliva [7]. Five of the 31 studies that reported on salivary viral load collected unstimulated whole saliva (UWS) by drooling: the saliva was collected at the bottom of the mouth and then relieved into the collection device [12, 31, 3739]. Other studies reported saliva collection methods including spitting (three studies) [13, 57, 58], self-collection (eight studies) [11, 14, 33, 4244, 47, 48], funnel (one study) [32], gargling (one study) [10], saliva stimulated by rubbing outside of the cheeks and then spitting (one study) [15], by coughing (two studies) [41, 54], and by collecting naso-oropharyngeal saliva (two studies) [45, 46]. One study purchased saliva from COVID-19 patients [51]. Seven studies did not report the saliva collection method; however, these studies were included because the viral loads were reported in all cases.

In 24 studies, the viral load dynamics of different respiratory tract samples was evaluated at the early phase of infection (first week), while in five studies, it was assessed in the second week of the infection. The remaining eight studies did not report the days of symptom onset. Furthermore, five studies included the viral load of saliva in symptomatic and asymptomatic COVID-19 patients; in four studies, the mean viral load was reported as CT value.

3.3. SARS-CoV-2 Load in Saliva

The mean SARS-CoV-2 load in saliva derived from 22 studies included 916 patients in total and showed mean CT values ranging from: 21.5 to 39.6 (Tables 3, 4, 6, and 7). Eleven studies included a total of 216 patients with a mean range of 1.91 × 101 to 5.69 × 1011 copies/mL RNA (Tables 6 and 7).

3.4. SARS-CoV-2 Load in Saliva Compared with NPS

A total of 13 studies were included for comparison of the standard mean difference in CT values of saliva and NPS (Figure 2). No significant differences were found in the mean viral load between saliva (overall mean: 26.4) and NPS (overall mean: 26.9 (). However, there was considerable heterogeneity between these studies (; I2 = 93%; 95% CI: −0.36–0.64), demonstrating that these data should be interpreted with caution but might be considered as a trend. Five studies compared the standard mean difference of the viral load given in copies/mL RNA in saliva and NPS (Figure 3). No significant differences were found in the mean viral load between saliva (overall mean: 1.80 × 1022) and NPS (overall mean: 2.78 × 1020) (), and moderate heterogeneity was observed across the studies (; I2 = 63%; 95% CI: −0.47–0.59).

3.5. SARS-CoV-2 Load in Saliva Compared with OPS

Four studies were included for comparison of the standard mean difference in CT values of saliva and OPS (Figure 4). No significant differences were found in the mean viral load between saliva (overall mean: 28.8) and OPS (overall mean: 30.5) (). Moderate heterogeneity was found between the studies (; I2 = 36%; 95% CI: −0.88–0.13).

3.6. SARS-CoV-2 Load in Saliva Compared with Sputum

Data from four published studies were selected to compare the mean CT values of saliva with sputum (Figure 5). No significant differences () and no heterogeneity was found in the mean viral load between saliva (overall mean: 29.3) and sputum (overall mean: 28.8) (; I2 = 0%; 95% CI: −0.65–0.50), demonstrating that these data are homogenous.

3.7. SARS-CoV-2 Load in Saliva of Symptomatic and Asymptomatic COVID-19 Patients

A meta-analysis was conducted to explore the standard mean difference of SARS-CoV-2 load in saliva of symptomatic and asymptomatic COVID-19 patients. Data from four published studies were selected to compare the mean CT value of saliva in symptomatic and asymptomatic patients (Figure 6). Results indicate that no significant differences were found in the mean viral load between symptomatic (overall mean: 26.06) and asymptomatic patients (overall mean: 25.7) (). However, a substantial heterogeneity was obtained between these studies (; I2 = 66%; 95% CI: −0.63–0.37).

3.7.1. Risk of Bias Assessment

Overall, 32 studies had a fair risk of bias (Table 2). Three studies were deemed to have a low risk of bias and one study had a high risk of bias. The overall rating in the quality of the studies was fair.

4. Discussion

Meta-analysis of 37 included articles revealed that the viral load of SARS-CoV-2 in saliva was comparable to that in NPS, OPS, and/or sputum. Data also disclosed that the viral load in saliva of symptomatic and asymptomatic patients were not significantly different.

Similarities in the viral load of saliva and NPS corresponded to values reported by others [50, 61, 62]. It was shown that saliva has comparable sensitivity to NPS for the detection of SARS-CoV-2 by RT-PCR. Some studies demonstrated higher viral load in saliva compared to NPS [37, 48, 6365]. In contrast, others showed a lower viral load in saliva; analysis of these values, however, revealed no statistically significant differences [45]. Though, interestingly, it has also been reported that the viral load in saliva peaks earlier, i.e., the first week after infection, and declines less rapidly compared to NPS, suggesting a higher postinfection window of opportunity in saliva for screening and diagnostic purposes [66]. It is thought that the higher viral load and longevity of the virus in saliva may be due to a higher level of ACE2 receptors at various sites in the oral cavity (gingiva, shed epithelial cells in saliva, mucosa, tongue, hard and soft palate, and salivary glands) compared to the nasopharynx [17, 19, 2125]. Saliva has also been shown to be sensitive enough to detect the majority of viable infections compared to NPS with potential higher likelihood of viral transmission [66].

A considerable heterogeneity was obtained in the meta-analysis of viral load in saliva compared with NPS, which could be explained by the sample size of the studies. To exemplify, the study of Yee et al. (2021) and Teo et al. (2021) had the largest sample sizes: n = 127 and n = 209, respectively, whereas the sample sizes of other studies varied between 2 and 41. Furthermore, differences in saliva collecting methods may contribute to the heterogeneity. For example, the study of Yee et al. (2021) used a different method for saliva collection compared to the other studies. Furthermore, the authors described that saliva was first stimulated by gently rubbing the outside of the cheeks and subsequently by spitting without interference of coughed-up saliva. Potentially, this method could have stimulated minor salivary glands and parotid glands, secreting predominantly serous saliva potentially loaded with SARS-CoV-2 particles. The saliva sampling methods of the other 11 studies were diverse: six studies reported self-collection [14, 4244, 47, 48], one study used the so-called drooling method [12], two studies were instructed to collect naso-oropharyngeal saliva [45, 46] and subsequently were asked to spit repeatedly in a sterile cup [45], one study reported coughed-up saliva from the throat [10] while two studies did not report the collection method at all [34, 40]. Currently, there is a lack of a globally accepted and standardized saliva collection protocol for SARS-CoV-2 analysis. However, despite the different saliva collection methods, PCR primers, and conditions, the study set-ups are not likely to have a major influence on the viral loads [67, 68]. The passive drooling technique is generally recommended as standard for saliva collection [6971]. It is stated that this method provides the greatest sensitivity and allows collecting whole saliva excluding mucous secretions from the oropharynx and sputum [37]. It is an easy and safe technique that can be done with relative simple instructions. As this study revealed that the viral load is comparable in all sample types, we recommend the use of sampling unstimulated saliva, unless other techniques are preferred, e.g., for sake of efficiency, logistic reasons, or standardization. To exemplify, for patients that are intubated and are not able to drool, it is suggested to pipette the saliva sample [70]. Another explanation for the heterogeneity could be that the viral load in saliva changed by food intake and by the circadian rhythm. Wyllie et al. (2020) and Hung et al. (2020) found the highest viral load of 61.5% in the morning, compared to before lunch 23.1%, 3PM, before dinner 7.7%, and at bedtime 0%. Exact times of sampling, however, were not reported. The relative high viral load in the morning may be due to overnight fasting and decreased salivary flow rate during sleep [72]. Consequently, it is, therefore, suggested to refrain from consumption of food and drinks in the morning prior to saliva collection [73]. The same study showed that the salivary flow rate increased after food consumption, which may dilute or wash out the viral RNA [28, 74, 75]. Another factor causing heterogeneity might be the dilution of saliva samples after collection in viral transport medium (VTM). In line, some studies showed that collecting undiluted unstimulated saliva is preferable since the sensitivity and viral detection rates were higher than diluted unstimulated saliva. This processing method also showed no RNA degradation [10, 15, 33]. Most studies were found to have a fair risk of bias, largely due to not providing sample size calculation and power description, as well as not adjusting for potential confounding variables that might impact the outcome such as age and gender.

Meta-analysis from this study is in line with previous studies and demonstrated that no significant differences were found in the viral load of saliva compared to sputum [43, 7679]. The viral load of sputum showed greater variation than saliva [78, 80, 81]. This could partly be related to the fact that the thick mucus from sputum hampers the viral RNA extraction [82]. It has also been observed that many patients are unable to produce enough sputum and coughs, making it an unsuitable method leading to decreased test sensitivity [77, 83].

We found that the viral load in saliva was comparable to OPS as indicated by CT values. This finding is in line with other studies [10, 84]. In contrast, however, Moreno-Contreras et al. (2020) found that saliva had a significantly higher viral load compared to OPS, whereas OPS and NPS combined (NPS + OPS) were shown to have a comparable viral load with saliva, suggesting that saliva is a good alternative sampling matrix for NPS + OPS. The reason for the difference between OPS and saliva viral load is unclear, but it is tempting to hypothesize that OPS was not sampled correctly due to the risks associated with this process. A total of 73.1% of NPS positive cases were negative in OPS [85], rendering it a less reliable specimen, as also reported by Khiabani et al. (2021).

Meta-analysis from the current study showed that the mean SARS-CoV-2 loads in saliva of symptomatic and asymptomatic COVID-19 patients are comparable as revealed by CT values and also shown by other authors [8688]. Similar viral loads have been also found in other fluids (NPS, OPS, and sputum) [89, 90]. A possible explanation for their comparable viral load could be the shedding of SARS-CoV-2 viral RNA originating from fragmented/degraded genomes of dead viral particles within the oral epithelial cells which has been shed into the saliva of asymptomatic individuals. It has been reported that a high amount of viral RNA does not necessarily mean greater infectivity [89, 91, 92].

It has to be noted that in due course of the current study, new variants of SARS-CoV-2 emerged. Studies on the so-called Omicron variant (B.1.1.529) reported that the viral shedding rate is higher in saliva than in nasal samples [9395]. It is shown that the salivary Omicron load peaks 1-2 days earlier than the nasal swabs detected by RT-PCR [93]. Marais et al. also concluded that saliva swabs performed better than midturbinate samples up to day 5 postinfection with positive percent agreement (PPA) of 96%. Individuals in the cohort study from Adamson et al. showed to develop symptoms within 2 days after first positive saliva PCR test [93]. Even more, faster and more efficient infection rates have been found for the Omicron variant in the human bronchus compared to the previous SARS-CoV-2 variant, leading to symptoms such as loss of smell and taste which are, therefore, better detected in saliva compared to NPS [93, 94, 96]. Saliva antigen tests and RT-PCR, however, showed a declined sensitivity in Omicron infections after day 5 postinfection with an overall PPA (of RT-PCR) of 96% to approximately 50% [95]. Several studies conclude that saliva swabs are a promising alternative to NPS and midturbinate samples, especially early in infection [9395]. It is, therefore, advised to use saliva samples as a diagnostic matrix for detecting the Omicron variant, instead of the currently used NPS. Many previous studies have also shown that the diagnostic performance of saliva tests has been successful in other viral infections, i.e., HIV [9799]. More research is needed to reveal the diagnostic accuracy of saliva, especially in late-stage of infection, for identifying the Omicron and possibly future variants of concern.

5. Limitations

Some data of the viral load (in CT values or copies/mL RNA), SD, and/or IQR were not available and, therefore, could not be included in the meta-analysis. Secondly, the fact that only four studies reported the CT value and SD of saliva from symptomatic and asymptomatic patients, provided only a small basis for comparison. Thirdly, in some studies, the methods of saliva collection were not reported in detail or at all. Also, saliva characteristics such as viscosity may have influenced the SARS-CoV-2 detection. UWS has usually a mucous consistency, whereas stimulated saliva is relatively serous produced [100].

6. Conclusion

This systematic review revealed that SARS-CoV-2 load in saliva is comparable to OPS, NPS, and sputum. Saliva specimen can therefore be used as alternative for SARS-CoV-2 detection since it is noninvasive, convenient, safe, and therefore ideal for mass screening. In addition, it was found that the SARS-CoV-2 loads in saliva of asymptomatic and symptomatic COVID-19 patients were not significantly different.

Data Availability

The data used to support the findings of this study are available within the article. This is a review based on published data.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

We thank the authors of included studies for sharing their datasets for meta-analysis. We also thank Zainab Assy, Henk Brand, Wendy Kaman, and Toon Ligtenberg for their helpful discussion. This research was financially supported by the institution of the authors.