Salivary SARS-CoV-2 RNA for diagnosis of COVID-19 patients: A systematic review and meta-analysis of diagnostic accuracy

Abstract

Accurate, self-collected, and non-invasive diagnostics are critical to perform mass-screening diagnostic tests for COVID-19. This systematic review with meta-analysis evaluated the accuracy, sensitivity, and specificity of salivary diagnostics for COVID-19 based on SARS-CoV-2 RNA compared with the current reference tests using a nasopharyngeal swab (NPS) and/or oropharyngeal swab (OPS). An electronic search was performed in seven databases to find COVID-19 diagnostic studies simultaneously using saliva and NPS/OPS tests to detect SARS-CoV-2 by RT-PCR. The search resulted in 10,902 records, of which 44 studies were considered eligible. The total sample consisted of 14,043 participants from 21 countries. The accuracy, specificity, and sensitivity for saliva compared with the NPS/OPS was 94.3 % (95 %CI = 92.1;95.9), 96.4 % (95 %CI = 96.1;96.7), and 89.2 % (95 %CI = 85.5;92.0), respectively. Besides, the sensitivity of NPS/OPS was 90.3 % (95 %CI = 86.4;93.2) and saliva was 86.4 % (95 %CI = 82.1;89.8) compared to the combination of saliva and NPS/OPS as the gold standard. Based on low to moderate certainty level these findings suggest a similarity in SARS-CoV-2 RNA detection between NPS/OPS swabs and saliva, and the association of both testing approaches as a reference standard can increase by 3.6 % the SARS-CoV-2 detection compared with NPS/OPS alone. This study supports saliva as an attractive alternative for diagnostic platforms to provide a non-invasive detection of SARS-CoV-2.

Highlights

• •The accuracy of saliva (94.3 %) was similar to the reference standard NPS/OPS.
• •The specificity of saliva was 96.4 % vs. NPS/OPS, which indicates 3.6 % of SARS-CoV-2 detection only in saliva.
• •The sensitivity of saliva was 89.2 % compared to the reference standard NPS/OPS.
• •The sensitivity of NPS/OPS was 90.3 % and saliva was 86.4 % vs. the combination of saliva and NPS/OPS as the gold standard test.
• •Saliva is an attractive alternative sample to perform a non-invasive diagnostic of COVID-19.

1. Introduction

COVID-19 is a worldwide public health issue resulting in unprecedented socio-economic and health impacts [1]. The COVID-19 outbreak triggered by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spread worldwide with at least 6.6 million deaths [2], [3]. A combination of mass screening, mask-wearing, and vaccination are critical measures in public health strategy to contain SARS-CoV-2 [4]. Hence, improved mass testing options can strengthen public health systems by refining the COVID-19 response [5].

Specimens collected from the upper respiratory tract, such as nasopharyngeal swab (NPS) and/or oropharyngeal swab (OPS), are reported as the reference method to SARS-CoV-2 diagnosis using quantitative reverse‐transcriptase polymerase chain reaction (RT‐PCR) [6], [7]. However, the NPS and OPS collections are invasive and uncomfortable for patients. These tests also expose frontline healthcare workers to potential cases of COVID-19, increasing the risk of new infections [7], [8]. Additionally, NPS collection requires trained healthcare professionals to collect samples. Thus, NPS and/or OPS collections present several issues for implementing mass testing for COVID-19 diagnosis [7], [9], which affects the number of tests performed. Alternative diagnostic options must be available, ideally including self-collection tests to improve global testing strategies to contain the rapid spread of novel SARS-CoV-2 variants.

In this context, saliva collection for SARS-CoV-2 detection allows for overcoming some limitations of the NPS and OPS collection [10]. Infectious disease centers worldwide approved salivary tests to detect COVID-19 [11]. Saliva involves a non-invasive, convenient, low-cost, and easy self-collection, not requiring direct contact with healthcare workers [12], [13] and therefore reducing the risk of nosocomial infections. Besides, the collection of saliva is more comfortable for the patient than NPS or OPS method [12], [14]. The stabilization of viruses is more complex in NPS and OPS samples [15], [16] and saliva is easier to collect than NPS and OPS, generating reduced human resources for health systems [11]. Previously, we detailed the potential of salivary diagnosis for COVID‐19, indicating that a non-invasive platform to detect salivary biomarkers could enhance disease detection [17].

Previous systematic reviews with meta-analysis were performed on a limited number of available published studies, including pre-prints, using limited electronic databases, and/or mixing studies with heterogeneous collection methods, including saliva collection associated with fluid from the throat by asking the patient to cough out or sputum [18], [19], [20], [21], [22]. Considering that the number of studies comparing saliva to NPS or NPS/OPS diagnostic increased exponentially with successive waves of COVID-19 caused by novel SARS-CoV-2 variants, it is critical to update the validation of this comparison with more patient cohorts and clinical settings. Hence, the present systematic review and meta-analysis aimed to compare the accuracy, sensitivity, and specificity of salivary diagnostics for COVID-19 based on SARS-CoV-2 RNA with the current reference tests using a nasopharyngeal swab (NPS) and/or oropharyngeal swab (OPS). Besides, considering the detection of SARS-CoV-2 in saliva indicates the presence of COVID-19 infection even if this virus was not detected in NPS/OPS, an additional comparison was performed to evaluate the sensitivity of saliva and NPS/OPS comparing with the combination of saliva and NPS/OPS as a gold standard.

2. Material and methods

2.1. Protocol and registration

The protocol was reported under the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) [23] and registered in the International Prospective Register of Systematic Reviews (PROSPERO) database, under the number CRD [Blinding] (http://www.crd.york.ac.uk/PROSPERO). This systematic review was reported according to the PRISMA [24] and the Joanna Briggs Institute(JBI) Manual [25].

2.2. Inclusion and exclusion criteria for reviewed studies

Diagnostic accuracy studies reporting the number of subjects who tested positive or negative for COVID-19 using simultaneous saliva and NPS (or in combination with OPS) tests to detect SARS-CoV-2 by RT-PCR were selected. Symptomatic and asymptomatic subjects were considered eligible. Studies with subjects declared less than 18 years old were excluded. Moreover, studies were selected with no language restrictions. The exclusion criteria were as follows: (1) review article; (2) guidelines, consensus documents, or expert position papers; (3) protocol studies; (4) case reports; (5) abstract; (6) systematic reviews with or without meta-analysis; (7) non-peer-reviewed articles (pre-prints); (8) studies unrelated to the objective; (9) no paired collection of NPS/OPS and saliva and (10) saliva collection different from spit/drooling method.

2.3. Sources of information and search

The electronic search was conducted in the following electronic databases: MedLine (via PubMed), Scopus, Web of Science, Embase, LILACS, SciELO, and LIVIVO databases up to October 2022. The search strategy was individualized to each database (Table 1).

Table 1.

Strategies for database search.

Database	Search strategy (July 2021)
PubMed https://pubmed.ncbi.nlm.nih.gov/	((“SARS-CoV-2” OR “2019-nCoV” OR “COVID-19” OR “2019 Novel Coronavirus” OR “Coronavirus” OR “COVID-19 Virus Infection”) AND (“Salivary” OR “Saliva”))
Scopus http://www.scopus.com/	(TITLE-ABS-KEY (“sars-cov-2” OR “2019-ncov” OR “covid-19” OR “2019 AND novel AND coronavirus” OR “coronavirus” OR “covid-19 AND virus AND infection”) AND TITLE-ABS-KEY (“salivary” OR “saliva”))
LILACS http://lilacs.bvsalud.org/	((“sars-cov-2” OR “2019-ncov” OR “covid-19” OR “2019 novel coronavirus” OR “coronavirus” OR “covid-19 virus infection”) AND (“salivary” OR “saliva”)) AND (db:("LILACS"))
Web of Science http://apps.webofknowledge.com/	(((“SARS-CoV-2” OR “2019-nCoV” OR “COVID-19” OR “2019 Novel Coronavirus” OR “Coronavirus” OR “COVID-19 Virus Infection”) AND (“Salivary” OR “Saliva”)))
Embase https://www.embase.com	('sars-cov-2′/exp OR 'sars-cov-2' OR '2019-ncov'/exp OR '2019-ncov' OR 'covid-19'/exp OR 'covid-19' OR '2019 novel coronavirus'/exp OR '2019 novel coronavirus' OR 'coronavirus'/exp OR 'coronavirus' OR 'covid-19 virus infection') AND ('salivary' OR 'saliva'/exp OR 'saliva')
SciELO https://www.scielo.org/	*(“SARS-CoV-2” OR “2019-nCoV” OR “COVID-19” OR “2019 Novel Coronavirus” OR “Coronavirus” OR “COVID-19 Virus Infection”) AND (“Salivary” OR “Saliva”)
LIVIVO https://www.livivo.de/	((“SARS-CoV-2” OR “2019-nCoV” OR “COVID-19” OR “2019 Novel Coronavirus” OR “Coronavirus” OR “COVID-19 Virus Infection”) AND (“Salivary” OR “Saliva”))

2.4. Data extraction

Studies retrieved from electronic search were exported to the EndNote Web™ (Thomson Reuters/Toronto-Canada), in which duplicates were removed. Next, two calibrated reviewers analyzed all study titles and abstracts. Besides, experts were consulted to include studies missed by the electronic search. Eligible studies were evaluated to verify whether it fulfilled the eligibility criteria. Non-included studies were registered separately, and the reasons for exclusion were described in Table 1. In case of disagreements between two co-authors, a third one was consulted to reach a consensus.

2.5. Data collection

Two authors independently assessed the full text of all eligible studies to extract relevant data to a structured spreadsheet. Data extracted included a list of authors, year, country, sample size, average age (or age range), and the number of individuals testing positive and negative in each diagnostic test (saliva and NPS/OPS). When available, information on the accuracy, sensitivity, and specificity of saliva tests compared to NPS/OPS was extracted.

2.6. Individual risks of bias

The Joanna Briggs Institute Critical Appraisal Tools for use in JBI Systematic Reviews of diagnostic test studies was used to assess the risk of bias of the selected studies. Two authors independently assessed each domain regarding their potential risk of bias. Each study was evaluated from the questions of the JBI assessment tool [26].

2.7. Statistical analyses

This systematic review with meta-analysis aimed to assess the accuracy, specificity, and sensitivity of the salivary SARS-CoV-2 test compared to the NPS SARS-CoV-2 test (alone or in combination with OPS) as a standard reference. Besides, considering that the detection of SARS-CoV-2 in saliva indicates the presence of COVID-19 infection even if the virus was not detected in NPS/OPS, two additional comparisons were performed to evaluate the sensitivity of saliva and NPS/OPS comparing with the combination of saliva and NPS/OPS as a gold standard.

Accuracy was the first analyzed outcome. Studies providing accuracy that compared salivary tests to NPS (alone or in combination with OPS) were included in the analyzes. We manually calculated the accuracy for studies when the accuracy was not provided, and the absolute number of individuals who tested positive and negative for SARS-CoV-2 in both tests (saliva and NPS/OPS) was described. The diagnostic accuracy was calculated by the number of concordant test results divided by the total number of samples analyzed.

The next outcome was the sensitivity of saliva and NPS/OPS for SARS-CoV-2 detection – calculated by the number of positive results identified in saliva divided by the total number of positive results in the reference standard. Two types of sensitivity measures were analyzed: i) only considering positive results in NPS (alone or in combination with OPS) as the reference standard, and ii) considering a positive result either in saliva or NPS/OPS as the reference standard. This outcome is critical when a unique imperfect reference test can release false negative results [18]. Although NPS is considered a reference standard, this is not fully accepted in the literature [27], [28]. Therefore, we considered both positive results in any test as the gold standard analysis.

The third outcome was the specificity of saliva compared to NPS/OPS as the reference standard. Specificity was calculated by dividing the number of samples identified as negative in the saliva test by the total number of negative tests identified in the reference standard.

All outcomes were calculated as proportions, and proportion meta-analyses were performed to estimate a pooled estimate for each outcome. Estimates were pooled using generalized linear mixed methods using both fixed and random-effects models. The between-study heterogeneity was analyzed using I² and the τ² statistics. I² represents the percentage of variability caused by heterogeneity excluding sampling error, while the τ² refers to the variance between studies. Meta regressions were fitted for each outcome trying to identify sources of heterogeneity. Publication bias was investigated for each outcome using a weighted linear regression with the log of the outcome as the dependent variable and the inverse of the sample size as the independent variable [29]. All analyzes were conducted using R and the meta package. A 5 % significance level was adopted in all analyzes.

2.8. Certainty of evidence

The certainty of the evidence was assessed with the Grading of Recommendation, Assessment, Development, and Evaluation (GRADE) tool [29]. This assessment was based on study design, methodological limitations, inconsistency, indirect evidence, imprecision, and other considerations. The level of certainty among the evidence identified was characterized as high, moderate, low, or very low [30].

2.9. Ethics consideration

This review study is not individual-based and we used peer-reviewed published studies data with ethical approvals.

3. Results

3.1. Study selection

During the first phase of study selection, 10,902 studies were found. After removing duplicates, 2965 studies remained for the analysis of titles and abstracts. After a detailed analysis of titles and abstracts, 115 studies were eligible for full-text analysis. No additional studies were indicated by the expert. After reading the full texts, 1 pre-print, 34 studies unrelated to the topic, 4 did not use NPS or NPS/OPS as the reference standard, 13 used samples from individuals < 18 years old, 3 studies collected no paired of NPS/OPS and saliva, 4 studies collected saliva in different methods, and 12 studies with absent or incomplete data about sensitivity and specificity, remaining 44 studies moved onto qualitative and quantitative analysis (Fig. 1).

Fig. 1.

Flowchart of the search, identification, and selection of eligible studies.

3.2. Study characteristics of eligible studies

All studies were published in 2020–2022 and were conducted in Argentina [31], Belgium [32], Bangladesh [33], Brazil [14], [34], [35], [36], [37], Canada [38], [39], [40], [41], [42], Ethiopia [43], France [44], [45], Greece [46], India [47], Iran [48], Italy [49], Japan [50], [51], [52], [53], Mexico [54], [55], Pakistan [56], Philippines [57], South Korea [58], Thailand [59], Turkey [60], [61], United Arab Emirates [62], United Kingdom [63], United States of America [64], [65], [66], [67], [68], [69], [70], [71], [72], [73]. The total number of participants was 14,043, and the age varied between 18 and 106 years old. Twenty-two studies only included patients who had symptoms of COVID-19 [14], [31], [33], [34], [37], [43], [47], [48], [50], [51], [53], [54], [55], [57], [59], [60], [61], [65], [68], [70], [71], [73], seventeen studies included symptomatic and asymptomatic patients [32], [35], [36], [38], [39], [41], [42], [44], [45], [46], [49], [58], [62], [63], [64], [66], [69], one study included only asymptomatic patients [52], and four studies did not report these data [40], [56], [67], [72]. Other sources of information regarding demographics, characteristics of the population, and characteristics of the samples are available in Table 2.

Table 2.

Characteristics of the population on the eligible studies included.

Author	Country	Population		Symptoms	Diagnosis	RNA extraction kit	PCR-Kit	Target gene for RT-PCR test Swab type	Swab type	Saliva collection
		N	Age
Abdollahiet et al.	Iran	80 (emergency unit’s triage)	56.5 ± 16.8 [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87], [88], [89], [90], [91], [92]	Symptomatic patients	Viral RNA detection with PCR from ONS	Viral Nucleic Acid Extraction kit by RBC Bioscience	2019-nCOV RT-PCR-Fluorescent Probing of Sansure Biotech (Changsha, China)	NR	Oro-Nasopharyngeal	Whole saliva (spit)
Babadyet et al.	United States of America	87 (employees of hospital)	NR	Asymptomatic and symptomatic patients	Viral RNA detection with PCR from NPS	NUCLISENS EasyMag or Chemagic 360	ABI 7500 Fast or the QuantStudio 5	The N1 and N2 genes	Nasopharyngeal	Whole saliva (spit)
Balaskaet et al.	Greece	420 (diagnosing patients and screening healthcare workers)	44.7 ± 13	Symptomatic and asymptomatic	Viral RNA detection with PCR from NPS	NR	Advanta™ Dx SARS-CoV-2 RT-PCR Assay (Fluidigm Corporation, San Francisco, USA) and NeumoDxTM SARS-CoV-2 Assay	The N1, N2, and RP primers and probes (Advanta), and Nsp2 and N genes (Neumo)	Nasopharyngeal	Whole saliva (spit)
Baratet et al.	United States of America	459 (drive-through testing center and hospital)	42 [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87], [88]	Symptomatic patients	Viral RNA detection with PCR from NPS	NucliSENS easyMAG platform (bioMérieux, France)	ABI 7500 fast real-time PCR system (Thermo Fisher Scientific, Waltham, MA)	The N1, N2, and RNase P primers and probes	Nasopharyngeal	Whole saliva (drooling method)
Bergevinet et al.	Canada	773 (hospital patients)	44 [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58]	Symptomatic and asymptomatic	Viral RNA detection with PCR from ONS	Seegene Starlet (Seegene)	Allplex 2019-nCoV Assay (Seegene)	The E, RdRp and N genes	Oro-Nasopharyngeal	Whole saliva (drooling method)
Bhattacharya et al.	India	74 (hospitalized patients)	NR	Mild to‐ moderate symptoms	Viral RNA detection with PCR from NPS	QIAamp Viral RNA Mini Kit	ABI 3500xL Dx Genetic Analyzer	The ORF1 and E genes	Nasopharyngeal	Whole saliva
Borghiet et al.	Italy	192 (hospital)	NR	Asymptomatic and symptomatic patients	Viral RNA detection with PCR from NPS	Direct qPCR	Applied Biosystem 7500 Fast instrument	The N1 and RP genes	Nasopharyngeal	Whole saliva
Callahanet et al.	United States of America	385 (drive-through collection sites)	NR	Symptomatic and asymptomatic	Viral RNA detection with PCR from NPS	NR	Abbott m2000 and Abbott Alinity RealTime SARS-CoV-2 assay	NR	Nasopharyngeal	Whole saliva (spit/drool)
Carrilloet et al.	Philippines	197 (health care workers and outpatient of hospital)	32 [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64]	Symptomatic patients	Viral RNA detection with PCR from NPS	GenAmplify™ Viral RNA Purification Kit (The Manila HealthTek, Inc.)	2019-nCoV Nucleic Acid Diagnostic Kit (Sansure Biotech Inc.) or Real-Time Fluorescent RT-PCR Kit for Detecting SARS-CoV-2 (BGI Genomics Co. Ltd.)	NR	Nasopharyngeal	Whole saliva (snort-spit)
Carrouelet et al.	France	31 (ambulatory patients)	43.0 ± 15.5 [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75]	Symptomatic and asymptomatic	Viral RNA detection with PCR from NPS	NucliSens easyMAG instrument (bioMérieux, France)	Invitrogen SuperscriptTM III Platinum One-Step qRT-PCR system (Invitrogen, France)	The RdRp-IP2 and RdRp- IP4	Nasopharyngeal	Whole saliva (spit)
Caulleyet et al.	Canada	272 (testing center)	NR	Asymptomatic, high-risk and mild symptoms	NR	NR	RT-PCR (NR)	The E gene	Nasopharyngeal	Whole saliva (spit)
de Oliveiraet et al.	Brazil	400 (testing center)	> 18	Symptomatic	Viral RNA detection with PCR from ONS	Chemagic™ 360 instrument (PerkinElmer, USA)	QuantStudio 5™ instrument (ThermoFisher Scientific, USA) and TaqPath™ Covid-19 CE-IVD RT-PCR kit (ThermoFisher Scientific, USA)	The N and RNAseP targets	Oro-Nasopharyngeal	Whole saliva (spit)
Doganet et al.	Turkey	98 (hospitalized patients)	NR	Moderate-severe disease	Viral RNA detection with PCR from NPS and COVID-19 symptoms	Direct Detection Kit (Coyote Bioscience Co., Ltd)	Biorad CFX 96 Real-Time PCR systems	The Orflab and N genes	Nasopharyngeal	Whole saliva (drooling method)
Echavarriaet et al.	Argentina	174 (hospital-emergency)	31 [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50]	Symptomatic	Viral RNA detection with PCR from NPS	Quick‐RNA™ Viral Kit, Zymo Research Corp	CFX 96 Deep Well™ Real Time System (BioRad)	The E gene	Nasopharyngeal	Whole saliva
Genelhoudet et al.	Brazil	229	NR	Asymptomatic and symptomatic patients	Viral RNA detection with PCR from NPS	NR	NR	NR	Nasopharyngeal	Whole saliva
Getachewet et al.	Ethiopia	280	NR	Symptomatic patients	Viral RNA detection with PCR from NPS	RNA Extraction and Purification reagent, by DAAN Gene Co. LTD	RT-PCR Kit fir 2019 n-CoV, by BGI Biotechnology	ORF1ab	Nasopharyngeal	Whole saliva
Ghaniet et al.	Pakistan	48 (patients)	35.9 ± 1.3	NR	Viral RNA detection with PCR from NPS	NR	NR	NR	Nasopharyngeal	Whole saliva (drooling method)
Gonçalveset et al.	Brazil	194	39 ± 13	Asymptomatic and symptomatic patients	Viral RNA detection with PCR from NPS	Maxwell 16 Viral Total Nucleic Acid Purification Kit, by Promega	2019-nCov CDC qPCR Probe Assay, by Integrated DNA Technologies	N1, N2 and RNaseP genes	Nasopharyngeal	Whole saliva (spit)
Güçlüet et al.	Turkey	64 (hospitalized and emergency department)	51.04 ± 17.9	Mild, moderate and advanced symptoms	Viral RNA detection with PCR from ONS and COVID-19 symptoms	EZ1 (Qiagen, Germany) device	RT-PCR SARS-CoV-2 (Primer Design, UK) kit.	NR	Oro-Nasopharyngeal	Whole saliva (drooling method)
Guimarãeset et al.	Brazil	189	39.6 ± 12.3	Symptomatic patients	Viral RNA detection with PCR from NPS	Bio-Gene DNA/RNA Viral Extraction, by Bioclin	SARS-CoV-2 molecular kit, by Bio-Manguinhos	SARS-CoV-2 envelope gene	Nasopharyngeal	Whole saliva (spit)
Hansonet et al.	United States of America	353 (drive-through test center)	35 [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75]	NR	NR	NA	Hologic Aptima SARS-CoV-2 transcription-mediated amplification (TMA) assay (Hologic Inc.)	NA	Nasopharyngeal	Whole saliva (drooling method)
Iwasakiet et al.	Japan	76 (suspicious and diagnosis)	69 [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87], [88], [89], [90], [91], [92], [93], [94], [95], [96], [97]	Mild to moderate disease	NR	QIAamp Viral RNA Mini Kit	One-Step Real-Time RT-PCR Master Mixes and tepOnePlus Real Time PCR System	NR	Nasopharyngeal	Whole saliva (spit)
Jamalet et al.	Canada	91 (inpatients at hospital)	66 [23–106]	NR	NR	NR	Allplex 2019-nCoV Assay (100 T)	The RdRp, E and N genes	Nasopharyngeal	Whole saliva (spit)
Jenkinset et al.	United Kingdom	413	NR	Asymptomatic and symptomatic patients	Viral RNA detection with RT-qPCR from ONS	NR	A triplexed combination of DNA primer and probe oligonucleotides (from the CDC and FDA approved)	The N and e genes	Oro-Nasopharyngeal	Whole saliva (drooling method)
Joet et al.	South Korea	338	39	Asymptomatic and symptomatic patients	Viral RNA detection with RT-PCR from NPS	Real-Prep Viral DNA/RNA kit, by BioSeWoon	Real-Q Direct SARS-CoV-2 Detection Kit, by BioSeWoon	E and RdRp genes	Nasopharyngeal	Whole saliva (drooling method)
Kandelet et al.	Canada	429 (hospital)	42 [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54]	Asymptomatic and symptomatic patients	Viral RNA detection with PCR from NPS	Promega Maxwell HT Viral TNA Kit (NPS)	CFX96 Touch Real-time PCR detection system (NPS) Roche cobas® 6800 analyzer and Roche cobas® SARS-CoV-2 assay (saliva)	The E gene and 5′-UTR (NPS) The E and ORF1 genes (saliva)	Nasopharyngeal	Whole saliva (drooling method)
Landryet et al.	United States of America	124 (Hospital drive through testing)	NR	Symptomatic outpatients suspected	NR	EasyMag	RT-PCR (NR)	The N1, N2 and RNAse P genes	Nasopharyngeal	Whole saliva (spit)
Marxet et al.	United States of America	497	NR	Asymptomatic and symptomatic patients	Viral RNA detection with RT-PCR from NPS	NR	NR	N1 and N2 genes	Nasopharyngeal	Whole saliva (spit)
Maticet et al.	Canada	74 (Inpatients, residents of long-term care facilities, healthcare workers, and outpatients)	51 [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87], [88], [89], [90], [91], [92], [93], [94], [95]	Asymptomatic and symptomatic patients	Viral RNA detection with PCR from NPS	MagNA Pure 96 System	LightMix® ModularDx SARS-CoV (COVID19) E-gene assay and LightCycler® Multiplex RNA Virus Master	The E gene	Nasopharyngeal	Whole saliva (spit)
Mestdaghet et al.	Belgium	2469	31–40	Asymptomatic and symptomatic patients	Viral RNA detection with RT-qPCR from NPS	Total RNA Purification Kit, by Norgen Biotech; and Magnetic bead-based RNA extraction, by University of Liége	iTaq one-step RT-qPCR mastermix, by BioRad; and TaqPath COVID-19 RT-qPCR Combo Kit, by Thermo Fisher	ORF1ab, N, S and MS2 genes	Nasopharyngeal	Whole saliva (spit)
McCormick-Baw et al.	United States of America	155 (hospital)	47.8	Symptomatic patients	Viral RNA detection with PCR from NPS	Direct	Cepheid Xpert Xpress SARS-CoV-2 RT-PCR	The E and N2 genes	Nasopharyngeal	Whole saliva (drooling method)
Miguereset et al.	France	123 (hospitalized and ambulatory patients)	43 (NR)	Asymptomatic and symptomatic patients	Viral RNA detection with PCR from NPS	NR	RT-PCR Panther Fusion™ module	NR	Nasopharyngeal	Whole saliva (spit)
Moreno-Contreraset et al.	Mexico	71 (ambulatory patients)	41 ± 14.4	Symptomatic patients	Viral RNA detection with PCR from ONS	QIAamp viral RNA minikit	StarQ one-step RT-PCR	The gene E and RNase P	Oro-Nasopharyngeal	Whole saliva (spit)
Pasomsubet et al.	Thailand	200 (hospital)	36 [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48]	Symptomatic patients	Viral RNA detection with PCR from ONS	MagDEA® Dx reagents (automated nucleic acid extraction)	CFX96 Real-Time Detection System	The ORF1ab and N gene	Oro-Nasopharyngeal	Whole saliva
Potteret et al.	United States of America	311 (Barnes-Jewish Hospital ED)	NR	Symptomatic patients	Viral RNA detection with PCR from NPS saliva	NUCLISENS-easyMAG	Lyra, SARS-CoV-2 assay, COVID-19 Test (Biofire), cobas 6800 SARS-CoV-2 Test (Roche), or the Xpert Xpress SARS-CoV-2 (Cepheid)	NR	Nasopharyngeal	Whole saliva
Rodriguez-Florez et al.	Mexico	30 (clinical sample)	NR	Symptomatic patients	Viral RNA detection with PCR	QIAamp Viral RNA Mini, Qiagen, MD, United States) or SalivaDirect nucleic acids extraction protocol	SARS-CoV-2 Real-Time PCR kit (Vircell, Granada, Spain).	E gene	Nasopharyngeal	Whole saliva
Sahajpalet et al.	United States of America	240 (healthcare or a community setting)	NR	NR	Viral RNA detection with PCR	Protocol U, Protocol or Saliva All	NR	N gene and ORF1ab	Nasopharyngeal	Whole saliva
Senoket et al.	United Arab Emirates	401 (hospital)	35.5 ± 9.5	Asymptomatic and symptomatic patients	Viral RNA detection with PCR from NPS	Chemagic viral RNA extraction kit on the automated Chemagic™ 360 Nucleic Acid Extractor	NeoPlex COVID-19 kit	The RdRp and N gene targets	Nasopharyngeal	Whole saliva (spit)
Sogbesanet et al.	United States of America	89 (Individuals)	20–83	Symptomatic patients	Viral RNA detection with PCR	NR	Cobas SARS-CoV-2 test	ORF1ab and E gen	Nasopharyngeal	Whole saliva
Uddinet et al.	Bangladesh	596 (Workers hospital)	28–47	Symptomatic patients	Viral RNA detection with PCR	QIAamp viral RNA mini kit (quigen, Germany)	iTaq universal probes one-step RT-qPCR kit	RdRp and N genes	Nasopharyngeal	Whole saliva
Uwaminoet et al.	Japan	196 (hospitalized patients and university staff)	NR	Symptomatic patients	Viral RNA detection with PCR from NPS	NR	2019 Novel Coronavirus Detection Kit (Shimadzu, Kyoto, Japan)	The N1 and N2 primers and probes	Nasopharyngeal	Whole saliva
Vazet et al.	Brazil	155 (healthcare workers and patients at the COVID-19 ward)	40 (33–48.5)	Symptomatic patients – mild to moderate symptoms	Viral RNA detection with PCR from ONS	QIAGEN QIAamp® RNA Mini Kit	RT-PCR - (NR - ONS) and Applied Biosystems 7500 Real Time PCR detector (saliva)	The E and RdRp genes	Oro-Nasopharyngeal	Whole saliva (spit)
Yokotaet et al.	Japan	1924 (contact tracing)	45 [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66]	Asymptomatic patients	Viral RNA detection with PCR from NPS	QIAsymphony DSP Virus/Pathogen kit and QIAamp Viral RNA Mini Kit	One step qRT-PCR was performed using THUNDERBIRD® Probe One-step qRT-PCR Kit and 7500 Real-time PCR Systems	The N2 primers	Nasopharyngeal	Whole saliva
Yokotaet et al.	Japan	42 (Patients)	27–93	Symptomatic patients	Viral RNA detection with PCR	QIAsymphony DSP Virus/Pathogen kit and QIAamp Viral RNA Mini Kit (QIAGEN, Hilden, Germany)	Thunderbird Probe One-step qRT-PCR Kit	N gene	Nasopharyngeal	Whole saliva

Note: NR: Not reported; ONS: Oro-Nasopharyngeal Swab, NPS: Nasopharyngeal Swab.

3.3. Risk of bias within studies

The main shortcoming identified among eligible studies was related to the fact that the index test was interpreted in some studies with knowledge of the reference test results (11 %-5/44). These 5 studies [32], [55], [61], [63], [72] represent 22, 9 % of the total sample size. However, RT-PCR is an analytical test that is independent of a subjective assessment, in this context, it is important to point out that the relevance of knowledge of the reference test to perform the index test (Q4 in table) is more relevant in qualitative and semi-quantitative studies than in quantitative analytical tests, as RT-PCR. Details on the risk of bias for each study are presented in Table 3, suggesting a low risk for other risks of bias.

Table 3.

Risk of bias assessed by the Joanna Briggs Institute Critical Appraisal Tools for use in JBI Systematic Reviews for Diagnostic Test Accuracy Studies.

Authors	Q.1	Q.2	Q.3	Q.4	Q.5	Q.6	Q.7	Q.8	Q.9	Q.10
Abdollahi et al.	√	√	√	U	√	√	U	√	√	√
Babady et al.	√	√	√	U	√	√	U	√	√	√
Balaska et al.	√	√	√	U	√	√	U	√	√	√
Barat et al.	√	√	√	U	√	√	U	√	√	√
Bergevin et al.	√	X	√	U	√	√	U	√	X	√
Bhattacharya et al.	√	√	√	U	√	√	U	√	√	√
Borghi et al.	U	U	√	U	√	√	U	√	√	√
Callahan et al.	√	√	√	U	√	√	U	√	√	√
Carrillo et al.	√	√	√	U	√	√	U	√	√	√
Carrouel et al.	√	X	√	U	√	√	U	√	√	√
Caulley et al.	U	√	√	√	√	√	√	√	√	√
de Oliveira et al.	√	√	√	U	√	√	U	√	√	√
Dogan et al.	√	√	√	U	√	√	U	√	√	√
Echavarria et al.	√	√	√	U	√	√	U	√	√	√
Genelhold et al.	U	√	√	U	√	√	U	√	√	√
Getachew et al.	√	X	√	U	√	√	U	√	√	√
Ghani et al.	√	X	√	U	√	√	U	√	√	√
Gonçalves et al.	√	√	√	U	√	√	U	√	√	√
Güçlü et al.	U	X	√	X	√	√	X	√	X	√
Guimarães et al.	√	√	√	U	√	√	U	√	√	√
Hanson et al.	√	√	√	U	√	√	U	√	√	√
Iwasaki et al.	U	X	√	U	√	√	U	√	√	√
Jamal et al.	√	X	√	U	√	√	U	√	X	√
Jenkins et al.	U	X	U	X	√	√	X	√	√	√
Jo et al.	√	X	U	√	√	√	√	√	√	X
Kandel et al.	√	√	√	√	√	√	√	√	√	√
Landry et al.	√	√	√	U	√	√	U	√	√	√
Marx et al.	√	√	√	√	√	√	√	√	√	X
Matic et al.	√	√	√	U	√	√	U	√	√	√
Mestdagh et al.	√	U	U	X	√	√	√	U	√	X
McCormick-Baw et al.	U	X	√	U	√	√	U	√	√	√
Migueres et al.	U	√	√	U	√	√	U	√	√	√
Moreno-Contreras et al.	√	√	√	U	√	√	U	√	√	√
Pasomsub et al.	√	√	√	√	√	√	√	√	√	√
Potter et al.	√	√	√	√	√	√	√	√	√	√
Rodriguez-Florez et al.	√	√	U	X	√	√	√	√	√	√
Sahajpal et al.	U	X	√	X	U	√	√	U	√	√
Senok et al.	√	√	√	U	√	√	U	√	√	√
Sogbesan et al.	√	√	U	√	√	√	√	U	√	√
Uddin et al.	U	√	U	√	√	√	√	√	√	√
Uwamino et al.	√	√	√	√	√	√	√	√	√	√
Vaz et al.	U	√	√	√	√	√	√	√	√	√
Yokota et al.(a)	X	X	√	U	√	√	U	√	√	√
Yokota et al. (b)	√	√	U	√	√	√	√	√	√	√

Q1. Was a consecutive or random sample of patients enrolled? Q2. Was a case control design avoided? Q3. Did the study avoid inappropriate exclusions? Q4 Were the index test results interpreted without knowledge of the results of the reference standard? Q5. If a threshold was used, was it pre-specified? Q6. Is the reference standard likely to correctly classify the target condition? Q7. Were the reference standard results interpreted without knowledge of the results of the index test? Q8. Was there an appropriate interval between index test and reference standard? Q9. Did all patients receive the same reference standard? Q10. Were all patients included in the analysis? √ – yes; X – no; U – Unclear.

3.4. Synthesis of results and meta-analyses

3.4.1. Accuracy of saliva compared with NPS/OPS as the reference standard for COVID-19 diagnosis

The pooled accuracy for diagnostic tests from forty-four studies clinically and methodologically homogenous using saliva was 94.3 %, which can be considered similar to the reference test (95 % CI = 92.1;95.9) (Fig. 2A). These studies presented a high heterogeneity (I² = 94 %, p < 0.01); however, there was no evidence of a moderating effect (p = 0.532) by the type of reference standard (NPS or NPS in combination with OPS) considered in the comparisons.

Fig. 2.

Accuracy of saliva compared to NPS/OPS as the reference standard for COVID-19 diagnosis with all studies (A), with subgroups of saliva collection methods (B) and the presence of COVID-19 symptoms (C). Events are the accurate results using saliva samples and Total is the sample size number of each eligible study for the meta-analysis.

Due to the high heterogeneity in these studies, an additional analysis was performed to evaluate the effect of spit or whole saliva/drooling method for non-stimulated saliva collection. Besides, we also evaluated the potential impact of the inclusion of symptomatic alone or symptomatic associated with asymptomatic patients for salivary SARS-CoV-2 RNA accuracy in COVID-19 diagnosis. The collection by spit or whole saliva/drooling method (Fig. 2B) presented similar accuracy and the heterogeneity level (spit: 93.2 %, 95 % CI = 89.5;95.6, I² = 93 %, p < 0.01; whole saliva /drooling method: 95 %, 95 % CI = 92.1;96.8, I² = 95 %, p < 0.01) was similar with grouped analysis. However, a higher accuracy was observed in the symptomatic associated with asymptomatic patients for salivary SARS-CoV-2 RNA in COVID-19 diagnosis (96.2 %, 95 % CI = 94.1;97.5) compared to the symptomatic alone patients (92.4 %, 95 % CI = 88.5;95.0). Once again, the high heterogeneity was similar when symptomatic alone (I² =93 %, p < 0.01) or symptomatic and asymptomatic (I² = 88 %, p < 0.01) patients were separately studied (Fig. 2C). In this context, it is noticed that the non-stimulated saliva collection by spit or whole saliva/drooling method and salivary collection in symptomatic alone or symptomatic and asymptomatic patients were not relevant for changes in heterogeneity level.

3.4.2. The specificity and sensitivity of saliva compared with NPS/OPS as the reference standard for COVID-19 diagnosis

The specificity of saliva compared to the reference standard ((NPS alone or in combination with OPS) was 96.4 % (95 %CI = 96.1; 96.7), which indicates 3.6 % of SARS-CoV-2 detection only in saliva (Fig. 3A). Studies included in this analysis presented a high heterogeneity (I² = 91 %, p < 0.01). The same 44 studies previously included, also provided sensitivity to perform this meta-analysis. The sensitivity was 89.2 % (95 % CI = 85.5;92.0). Studies included in this analysis presented a high heterogeneity (I² = 79 %, p < 0.01) (Fig. 3B). The test used as a reference standard (NPS alone or in combination with OPS) was not identified as a relevant moderator (p > 0.05) for sensitivity meta-analyses – thus, all studies were analyzed together.

Fig. 3.

Specificity (A) and sensitivity (B) of saliva compared to NPS/OPS as the reference standard for COVID-19 diagnosis. Events are the accurate results using saliva samples and Total is the sample size number of each eligible study for the meta-analysis.

3.4.3. The sensitivity of NPS/OPS or saliva compared to the combination of saliva and NPS/OPS as the gold standard for COVID-19 diagnosis

Due to the presence of recurrent tests with positive results for SARS-CoV-2 detection saliva with an associated absence of SARS-CoV-2 detection in NPS/OPS samples, we also performed an analysis considering the combination of saliva with NPS/OPS as the gold standard for COVID-19 diagnosis. In these meta-analyses with the same 44 previously reported studies, a positive test detecting SARS-CoV-2 either by saliva or NPS/OPS was considered as a correct diagnosis of COVID-19. Since only positive tests were considered, the only estimate presented in this section is the sensitivity of NPS/OPS or saliva compared to the combination of saliva and NPS/OPS as the gold standard for COVID-19 diagnosis. The pooled sensitivity of NPS/OPS compared to the combination of saliva and NPS/OPS was 90.3 % (95 % CI = 86.4;93.2) (Fig. 4A). Studies included presented a high heterogeneity (I² = 87 %, p < 0.01). Similar compared to the NPS/OPS samples, the sensitivity of saliva was 86.4 % (95 % CI = 82.1;89.8) (Fig. 4B). Studies included in this analysis also presented a high heterogeneity (I² = 86 %, p < 0.01). As described previously, we also accessed non-stimulated saliva collection by spit or whole saliva/drooling method and salivary collection in symptomatic alone or symptomatic and asymptomatic patients for the sensitivity of NPS/OPS or saliva compared to the combination of saliva and NPS/OPS as the gold standard for COVID-19 diagnosis. Both heterogeneity level and accuracy were not affected by unified or single-handed analysis (Supplementary Figs. 1A–B).

Fig. 4.

Sensitivity of NPS/OPS (A) or saliva (B) compared to the combination of saliva and NPS/OPS as the reference standard for COVID-19 diagnosis. Events are the accurate results using saliva samples and Total is the sample size number of each eligible study for the meta-analysis.

3.5. Certainty of evidence

To assess the certainty of the evidence, we considered the sensitivity and specificity as the primary outcome, and the sensitivity estimates of the secondary outcome (a gold standard compared with saliva or NPS/OPS) (Table 4). For the first outcome, sensitivity had moderate certainty of evidence (downgraded due to inconsistency), while specificity had low certainty of evidence (downgraded due to inconsistency and publication bias). For the second outcome, both sensitivities (saliva and NPS/OPS swab) showed moderate certainty of evidence (downgraded by inconsistency).

Table 4.

Grading of Recommendations Assessment, Development, and Evaluation (GRADE) summary of findings table for the outcomes of the systematic review and meta-analysis.

Outcome	Number of studies (Number of patients)	Study design	Factors that may decrease certainty of evidence					Effect per 1.000 patients tested		Test accuracy CoE
			Risk of bias	Indirectness	Inconsistency	Imprecision	Publication bias	Proportion (95 % CI)	pre-test probability of 5 %
Saliva collection compared with nasopharyngeal swab as reference standard to COVID-19
True positives1	44 studies (11,557 patients)	cross-sectional (cohort type accuracy study)	not serious	not serious	seriousa	not serious	none	0.89 (0.85–0.92)	45 (43–46)	⨁⨁⨁ MODERATE
False negatives2									5 (4–7)
True negatives3	44 studies (2587 patients)	cross-sectional (cohort type accuracy study)	not serious	not serious	seriousa	not serious	strongly suspectedb	0.96 (0.96–0.97)	916 (913–919)	⨁⨁ LOW
False positives4									34 (31–37)
Saliva collection only compared with nasopharyngeal swab plus saliva as reference standard to COVID-19
True positives1	44 studies (2896 patients)	cross-sectional (cohort type accuracy study)	not serious	not serious	seriousa	not serious	none	0.90 (0.86–0.93)	45 (43–47)	⨁⨁⨁ MODERATE
False negatives2									5 (3–7)
Nasopharyngeal swab only compared with nasopharyngeal swab plus saliva as reference standard to COVID-19
True positives1	44 studies (2896 patients)	cross-sectional (cohort type accuracy study)	not serious	not serious	seriousa	not serious	none	0.86 (0.82–0.89)	43 (41–45)	⨁⨁⨁ MODERATE
False negatives2									7 (5–9)

GRADE Working Group grades of evidence

High certainty: We are very confident that the true effect lies close to that of the estimate of the effect.

Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect.

Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.

¹ patients with COVID-19; ² patients incorrectly classified as not having COVID-19; ³ patients without covid-19; ⁴ patients incorrectly classified as having covid-19; ^a High (I² > 75 %) and unexplained heterogeneity; ^b Publication bias identified (p < 0.05) by Peters’ test.

4. Discussion

We have performed a systematic review to evaluate the diagnostic accuracy, sensitivity, and specificity of the salivary SARS-CoV-2 RNA test to surrogate a reference standard diagnostic method based on NPS/OPS on COVID-19 diagnosis in adults with ages varied between 18 and 106 years old in several different clinical cohorts. This evaluation was motivated by the potential of saliva for self-collection, which can strongly reduce the risk of COVID-19 transmission in healthcare frontline workers. Besides, saliva collection provides clear advantages over nasopharyngeal swabs as non-invasive, low-cost, and more comfortable for patients.

The choice of the most appropriate clinical specimens for COVID-19 diagnosis remains challenging. In a previous study, saliva was capable to detect COVID-19 infection alone in ∼ 12 % of patients with negative detection (false-negative) using nasopharyngeal swabs [5]. This present systematic review highlights the reliability of saliva specimens as a valuable sample for COVID-19 diagnosis. The application of saliva in RT-PCR showed a pooled sensitivity of 89.2 % and an accuracy of 94.3 % compared to NPS/OPS swabs. Furthermore, the pooled specificity of saliva was 96.4 % compared to NPS/OPS swabs, which indicates 3.6 % of SARS-CoV-2 detection only in saliva. Besides, the sensitivity of NPS/OPS was 90.3 % and saliva was 86.4 % using the combination of saliva and NPS/OPS as the gold standard test. The presence or absence of symptoms, COVID-19 severity, and days of symptom onset may explain differences in sensitivities, however, it provides a suitable spectrum of COVID-19 [74].

The comparison between diagnostic accuracy indicates a heterogeneity that should not be ignored to improve COVID-19 tests. Here, all studies included unstimulated saliva, which avoids dilution of SARS-CoV-2 as could occur in mouth rinsing or stimulated saliva collection [75]. Real-time PCR is considered 100 % specific [76], due to the intrinsic characteristics of this platform [77]. In this context, the present meta-analysis shows similar pooled sensitivity between NPS/OPS swabs and saliva, with just 3.9 % higher sensitivity for the NPS/OPS (90.3 %) compared to saliva (86.4 %) compared to the gold standard combination for COVID-19 diagnosis. This difference can be valid when evaluating an individual diagnostic context, but in the mass screening goal, this difference is diminished given the numerous advantages of saliva [15], [16]. The procedures for sample preservation and DNA extraction were reported in all included studies and presumably did not influence the acquired results. Here, only studies with saliva collection by spit or drooling methods were considered to compare similar conditions, without the presence of sputum [78]. Thus, saliva may be indicated as a diagnostic fluid for detecting SARS-CoV-2 RNA [79].

Considering the SARS-CoV-2 detection in major salivary glands, minor salivary glands, and oral mucosa [80], [81], [82], it should be considered that the presence of SARS-CoV-2 RNA in saliva and negative results in nasopharyngeal samples cannot be classified as false positive but misclassification of currently standard protocols [83]. It is well recognized that updates in COVID-19 diagnostic protocol are crucial, and the current NPS sample reference standard is not perfect. This occurs because NPS sampling method, the stored and transported practice, and RNA extraction protocols need to be optimized [84], [85].

High heterogeneity was observed between the studies when non-stimulated saliva collection by spit or whole saliva/drooling method were combined in the meta-analyses. In the same context, a significant level of heterogeneity was also detected when symptomatic alone or symptomatic and asymptomatic COVID-19 patients were included in this meta-analysis. Therefore, to evaluate the factors that could influence this high heterogeneity, some additional analyses were performed detaching both types of non-stimulated saliva, and the presence or absence of symptoms was performed. However, the heterogeneity level was maintained in these analyzes of the studies was found. In this context, these changes could be related to a wide age range (18–106 years old), different levels of disease severity, or changes in a sample collected from mass test centers and hospitals. The inclusion criteria for the studies included just non-stimulated salivary collection by spit (without sputum or deliberate oropharyngeal fluid collection) or drooling method. However, we performed a meta-analysis with both sample collections spit or drooling separately and it was not effective to change accuracy or heterogeneity level. Furthermore, it is important to point out the notable over-dispersed transmission in COVID-19 [86], heterogeneity in inequities of COVID-19 vaccination [87], heterogeneity in SARS-CoV-2 protein with a large number of amino acid substitutions in genomes sequences worldwide [88], and phenotypic heterogeneity in immunological responses [89]. Finally, it was supported by the presence of fast changes in the level of SARS-CoV-2 RNA detection in both nasal and saliva samples, which can facilitate viral detection when viral genome shedding increases, presenting opposite effects when it reduces. It was associated with heterogeneity in the duration of detectable infectious SARS-CoV-2, SARS-CoV-2 variants, clearance kinetics, and the temporal relationship between SARS-CoV-2 levels [90]. In this heterogeneity context of COVID-19, we believe that the significant heterogeneity level comparing saliva with NPS/OPS samples in this meta-analysis of diagnostic accuracy could be expected.

One of the advantages of saliva-based RT-PCR tests is that they can detect the virus earlier in the infection than NPS tests, and this is particularly important in asymptomatic patients who may not be aware that they are infected [91], [92]. Studies have shown that the viral load in saliva is highest at the time of symptom onset and declines over time, which means that saliva-based tests may be more sensitive than NPS tests in detecting the virus in the early stages of infection [93], [94]. Additionally, in asymptomatic patients, as potential for viral transmission, the virus may be present in the saliva even if it is not present in the nasopharynx, which is the target for NPS tests [95], [96]. In symptomatic patients, both saliva-based and NPS tests are likely to be equally accurate, as the virus is typically present in both the nasopharynx and the saliva in symptomatic patients.

GRADE system found that there is moderate inconsistency and inaccuracy in the results obtained from the meta-analysis, this means that the evidence obtained is moderate and the effect estimate could be different from the real one. More studies are needed with standardization of methodologies to minimize imprecision and inconsistency found in this systematic study.

The evaluation of saliva as a reliable fluid in COVID-19 detection was addressed in previous systematic reviews with meta-analysis. However, these studies were based on a limited number of published studies [21], including non-peer-reviewed articles (pre-prints or grey literature) [18], [19], [97], [98], [99], using limited electronic databases [18], [98], [100], mixing studies with heterogeneous collection [19], [98], [99], [100], [101], OPS test alone as the gold standard [97], [100], excluded studies with low diagnostic power in saliva specimens [102] and did not use the GRADE system to grade the quality of evidence [18], [19], [97], [98], [99], [100], [101], [102]. Although the reliability of saliva as a diagnostic sample has been indicated in previous systematic reviews, the conclusions were less robust due to these mentioned limitations. Despite the limitations in previously systematic reviews, the reported sensitivity and specificity were similar to the present findings. The range of sensitivity and specificity for NPS/OPS was 82–97 % and 97–98 % respectively. Using saliva, the sensitivity range was 82–88 % and the specificity was 92–99 % [18], [19], [85], [86], [87], [88], [89], [90]. In the present meta-analysis, RT-PCR tests in saliva indicate a pooled sensitivity of 89.2 % and an accuracy of 94.3 % compared to NPS/OPS swabs. Besides, 3.6 % of SARS-CoV-2 detection occurred only in saliva without detection in NPS/OPS swabs. Besides, the sensitivity of NPS/OPS was 90.3 % and saliva was 86.4 % using the combination of saliva and NPS/OPS as the gold standard test. Although the sensitivity of salivary SARS-CoV-2 RNA for diagnosis of COVID-19 patients could be considered relatively lower in the saliva compared to NPS/OPS, the advantages of non-invasive salivary diagnostics requiring fewer medical resources and specialized professionals can be acceptable in several conditions.

5. Conclusion

In conclusion, based on low to moderate certainty level, this systematic meta-analysis study supports the promising alternative of saliva for surveillance, screening, and diagnostic platform to provide a non-invasive, low-cost, safe, and more comfortable analysis of SARS-CoV-2 RNA in COVID-19 patients, which can protect healthcare and other frontline workers with self-collection samples. Nevertheless, studies with a more robust experimental design with homogeneous methodologies are needed to confirm our findings and then validate saliva as an alternative sample for the diagnosis of SARS-CoV-2 infection.

Funding

This research was supported by a grant from CAPES - Prevention and Combat of Outbreaks, Endemics, Epidemics, and Pandemics (#23038.014934/2020-59), CNPq (409157/2022-8 and 422205/2021-4), FAPEMIG (#APQ-02872-16, APQ-00476-20, APQ-02148-21), Federal University of Uberlandia and National Institute of Science and Technology in Theranostics and Nanobiotechnology (CNPq #403193/2022-2 e FAPEMIG #APQ-03613-17). Caixeta, D.C. received a fellowship from CNPq. Sabino-Silva, R received a fellowship from FAU/ImunoScan/UFU, CNPq Productivity Fellowship (306050/2021-8), and PrInt CAPES/UFU.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

Supplementary material.Sensitivity of saliva compared to the combination of saliva and NPS/OPS as the reference standard for COVID-19 diagnosis with subgroups of saliva collection methods (A) and the presence of COVID-19 symptoms (B). Events are the accurate results using saliva samples and Total is the sample size number of each eligible study for the meta-analysis.