Sage Journals: Discover world-class research

Abstract

By using a meta-analytical approach, this study aimed to analyse the diagnostic capacity of protein-based biomarkers in saliva for the differential diagnosis of oral potentially malignant disorders (OPMDs) and oral squamous cell carcinoma (OSCC) from healthy individuals as control group (HCG).

Articles on protein-based biomarkers in saliva, which provided quantitative expression in individuals with clinical and histopathological diagnosis of OPMD or oral leukoplakia (OL) were considered eligible. Searches were conducted in eight electronic databases. The methodological quality was assessed using the Quality Assessment of Diagnostic Studies tool (QUADAS-2). Functional analysis was also performed. Meta-analyses were performed using the OpenMeta tool (Analyst).

Meta-analysis was possible for 4 of the 11 biomarkers studied. Only the carcinoembryonic antigen (CEA) and the soluble fragment of cytokeratin 19 (CYFRA21) were significant for the OSCC/OPMD subgroup, both with a very low heterogeneity. CEA had an OE $=$ 25.854 (CI95%: 13.215–38.492, $p<$ 0.001, I2 $=$ 0) and CYFRA21 had an OE $=$ 9.317 (CI95%: 9.014–9.619, $p<$ 0.001, I2 $=$ 0). For the OPMD/HCG subgroup, only CYFRA21 was significant, with an OE $=$ 3.679 (CI95%: 0.663–6.696, $p=$ 0.017) although with high heterogeneity (I2 $=$ 91.24).

The CEA and CYFRA21 markers proved very useful when differentiating OSCC from OPMD. The CYFRA21 was the only protein that was capable of distinguishing between OPMD and healthy controls.

Keywords

Systematic review meta-analysis protein-based biomarkers saliva oral leukoplakia oral potentially malignant lesions

1. Introduction

The oral squamous cell carcinoma (OSCC) represents more than 95% of all malignant neoplasms of the mouth [1, 2] and it has a high prevalence and morbidity, with 354,900 new cases and 177,400 deaths registered per year worldwide [3]. Despite advances in the understanding and treatment protocols for different malignant tumours, the five-year survival rate for patients with OSCC does not exceed 50–60% overall [1, 4]. As a result, prevention and early diagnosis are considered to be highly necessary in order to improve the prognosis and survival rates.

The oral potentially malignant disorders (OPMD) are clinical conditions with a variable malignant transformation capacity. This group of lesions is comprised of leukoplakia, erythroplakia, erythroleukoplakia, lichen planus, nicotinic stomatitis, actinic cheilitis, amongst others [2, 5, 6]. The OL is the main lesion within the OPMD group, with an annual global average of malignant transformation rate of between 1 and 2% [2, 5, 7], nonetheless, these estimations may be higher, depending on the geographical region observed and the level of dysplasia, therefore meaning that it is possible to achieve rates that are equal to or greater than 39% [2]. The OL is clinically defined as a “white plaque with a questionable risk of malignancy once all other similar clinical lesions that carry no risk have been excluded” [2, 8]. This lesion may present microscopic aspects, which can vary from hyperkeratosis, with or without various levels of dysplasia (low or high), to carcinoma in situ [2, 4, 8]. Therefore, a biopsy is considered to be strictly necessary in order to establish a definitive diagnosis.

Currently, the diagnoses based on these clinical and histopathological criteria represent a problem given that they are poorly reproducible and have high inter-observer variability, therefore impeding a real standardisation in the early detection and prevention of the development of OSCC [2, 9, 10, 11]. As a result, one of the most important challenges in the diagnosis of OL is to predict which of the lesions will progress to OSCC.

Saliva is a complex biological matrix, the composition of which can vary depending on multiple factors [12, 13, 14]. The idea of using saliva as a method for the diagnosis and control of different diseases (periodontitis, cancer, recurrent aphthous stomatitis, OPMD, etc.) [13, 14] has proved to be of great interest, not only due to the varying molecular content, the low invasiveness of the procedure, and the fact that it is easy to collect, but also given the low cost that is involved [13]. The advances that have been made into the study of saliva have meant that it has been possible to establish its usefulness as a source of biomarkers that is comparable to blood, cerebrospinal fluid, pleural fluid and urine [13, 14, 15, 16], therefore allowing for the early diagnosis and screening of different diseases through “liquid biopsies” [14, 15, 17].

Despite innovations in the field of protein identification that involve several highly sensitive procedures (such as mass spectrometry techniques) [15, 16, 18, 19, 20], the ELISA method is still one of the most widely used, given its wide dissemination, reproducibility and low cost in comparison with the new omics techniques. [21, 22, 23, 24]. As a consequence, the search for protein markers in saliva offers further possibilities for understanding the structural interrelationships of OPMD and how these might determine progression to oral carcinoma.

The main goal was to perform a systematic review and meta-analysis with the markers that met the established requirements. The secondary goals were 1) To obtain information on the protein-based salivary markers described in the literature for OPMD, 2) To study the expression differences compared to healthy controls and malignant OSCC lesions.

2. Methods

A systematic review of the literature was conducted by a multidisciplinary team between the 1 ${}^{\text{st}}$ of December 2019 and the 31 ${}^{\text{st}}$ of January 2020. The review was written according to the Cochrane Handbook for Systematic Reviews of Interventions version 5.1.0 [25] and the PRISMA guide schecklist [26, 27] for systematic reviews (Supplementary 1).

2.1 PICO question

The study was organised based on the following PICO question: Patient, index test, comparison, outcome. P: patients diagnosed with an OPMD or OL. I: protein biomarker in saliva. C: patients with OSCC and the control group of healthy individuals (HCG). O: expression differences between OPMD and OSCC-HCG. Based on this the following scientific question was developed: Are there any protein-based markers in saliva that make it possible to quantitatively obtain diagnostic differences between healthy patients, OPMD and OSCC?

2.2 Criteria for considering studies for this review

2.2.1 Types of accuracy study

Studies of protein-based molecular markers (index tests) in saliva that revealed quantitative expression results related to clinically and histopathologically diagnosed diseases (reference standard). Therefore, the following were included 1) Human studies and controls, in the English language, which considered the OL or the OPMD and that included cases of OL as the reference standard; 2) Studies that included protein and saliva-based markers exclusively; 3) Studies that used ELISA and/or equivalent methods of analysis with quantitative expression determination (Bromocresol green method, Immunoturbidometry and LUMINEX). The exclusion criteria included: 1) Patients with explicit diagnosis, systemic disease, or a disorder that interfered with the conditions of interest; 2) Studies that only evaluated blood or tissue markers; 3) Studies in which the biomarker was not validated and was only discovered in label-free analysis; 4) Studies that only tried to validate a technique for diagnosis; 5) Betel quid chewing population as the only risk factor; 6) Studies that did not define the biomarker; 7) Studies that investigated changes in cytokines in saliva pre/post-treatment and that did not distinguish a potential biomarker; 8) Studies that did not define or include groups with the conditions of interest; 9) Studies that only included the OSCC or OSCC/HCG group. 10) Studies that quantified protein expression in units of measurement other than weight/capacity (i.e. mg/dl, ng/dl or pg./dl), such as Sequence coverage, mass/load ratio (m/z), Ion score, Fold-change o Mascot score. 11) Studies in which the marker was expressed under a minimum breakpoint.

2.2.2 Participants

The participants included in the study were those with clear and validated diagnoses, both clinically and histopathologically. Studies on experimental animals or in vitro models were excluded.

2.2.3 Control and target conditions

The patients included had been diagnosed according to the WHO Classification for Head and Neck Tumours. OPMD and OL were considered as the target condition and OSCC and HCG were considered as the healthy control group, both positive and negative according to the compared subgroups. The HCG group included patients without systemic diseases and who were not treated with immunosuppressive medication.

2.2.4 Reference standard

The reference standard for the diagnosis of OPMD, OL and OSCC was defined based on clinical and histopathological parameters (dysplasia and malignancy). Studies that did not detail any reference standard for the diagnosis of the described pathologies were not considered eligible for inclusion in this review. Given the great diversity of studies, clear inclusive diagnostic criteria were established. For the OSCC group: clinically diagnosed patients who had been verified through a histopathological study, regardless of the degree of differentiation, and excluding cases of warty carcinomas were included. For the OPMD group: cases of oral lichen planus, erythroplakia, oral submucous fibrosis and OL were included. We considered as OPMD any studies that did not mention a clear clinical diagnosis, but that did mention a non-malignant lesion with some degree of dysplasia [28, 29]. When only OL was found within the OPMD group, this had to be: white, non-detachable, homogeneous and non-homogeneous (mottled or warty) lesions with hyperkeratosis, acanthosis and/or histological squamous hyperplasia, with any degree of dysplasia (except carcinoma in situ, which was excluded).

2.2.5 Index test(s)

Any single protein-based biomarker detected in saliva, which was analysed from a quantitative point of view in order to differentiate between controls and targets. Studies analysing multiple biomarker panels were excluded because of the impossibility of analysis.

2.3 Search methods for the identification and selection of studies

2.3.1 Information sources and search strategy

Electronic searches in MEDLINE were performed using PubMed, EMBASE, OVID, Web of Science, Scopus, Cochrane Library, ClinicalTrials and the WHO regional bibliographic databases (AIM, IMEMR, IMSEAR, WPRIM) and the Conference Citation Index databases. Following the recommendations established by the Cochrane Systematic Reviews of Diagnostic Test Accuracy Group [30], the search strategy used to identify precision studies included three sets of terms related to the target condition (OPMD or OL), the index tests (protein-based biomarkers), and the type of oral sample (saliva) (Supplementary 2).

2.4 Data collection and analysis

The search was conducted by two observers (MPS and EA). Any ongoing discrepancies were resolved through discussion with two different reviewers (XMM and CMCP). The reasons for excluding studies were detailed. During the first phase, the title and abstract of the retrieved articles were read, and any studies that complied with the inclusion criteria, as well as those that did not provide sufficient data for a clear decision to be made were judged in full text. Subsequently, the entire content of all of the studies that were considered eligible was reviewed in their entirety in a second-round, and a final decision was made as to whether they were to be included or not.

2.5 Assessment of methodological quality

Two reviewers (MPS and EA) assessed the quality of the included studies separately using the revised Quality Assessment of Diagnostic Studies tool (QUADAS-2) [31]. This tool consists of four domains: 1) patient selection, 2) index test, 3) reference test, 4) flow and time. In addition, each domain was evaluated in terms of its risk of bias, and the first three domains were also evaluated in terms of their applicability.

2.6 Qualitative analysis

The unit of analysis was the expression of each protein biomarker in saliva, supplemented by the author and the year of publication. In most of the studies, the quantitative expression or level of a quantifiable protein was reported. Accordingly, these results were interpreted according to the presence or absence of disease (OSCC, OPMD or OL) based on a numerical measurement that was classified using a specific threshold (predetermined or not). The expression estimates were shown as integer values and their standard deviation for each biomarker and subgroup. The weighted mean difference (WM) between the case group and the control group was obtained for each model, giving the overall expression (OE) and the 95% confidence interval (CI95%). We distributed the meta-analysis in two subgroups, the first OSCC (standard reference)/ OPMD and the second OPMD (standard reference)/HCG, including the isolated OL reports within OPMD group.

2.7 Quantitative analysis

A meta-analysis was performed when the number of data for a biomarker was at least two papers and when the units of measurement of the marker were the same. The analysis strategy was to include all of the expression studies, regardless of their threshold value. The statistical analysis of the results was performed using the OpenMeta tool (Analyst).

2.8 Functional networks, and pathway mapping

Analyses of biological pathways within the differential protein markers were performed using the PANTHER enrichment analysis tool [32], only for interactions with high confidence (0.7). The Reactome Biological Pathway enrichment analysis of proteins of interest was performed using the STRING tool [33]. In both cases, we used proteins that presented statistically significant differences in our study, and in some cases, we performed node enrichment to be able to obtain more protein interactions. Finally, we used a systematic bioinformatics analysis including Database for Annotation, Visualisation and Integrated Discovery (DAVID) and gene ontology enrichment [34].

2.9 Assessment of heterogeneity

Firstly, the threshold effect of expression was evaluated graphically by looking at the forest plot and secondly, the threshold effect was evaluated statistically using the $I^{2}$ test. The $p$ values $>$ 0.1 and an index $I^{2}<$ 50% indicated a low heterogeneity among the studies, and, as such, the fixed-effect model was performed. On the other hand, $p$ values $<$ 0.1 and an index $I^{2}>$ 50% indicated considerable heterogeneity. All of the bilateral differences with a p-value of less than or equal to 0.05 were considered as significant [35].

3. Results

3.1 Study selection

The overall combined result of the references obtained from the aforementioned databases was 279 citations. After removing the duplicates, we obtained 217 unique citations. The biomarkers described in the literature for OPMD included: carcinoembryonic antigen (CEA), C-reactive protein (CRP), soluble fragment of cytokeratin 19 (CYFRA21), human epidermal growth factor receptor 2 (Her-2/neu; erbB-2), interleukin-1 $\alpha$ (IL-1 $\alpha$ ), interleukin-1B (IL-1B), interleukin-6 (IL-6), interleukin-8 (IL-8), Naa10p, Resistin (RETN), and Tumour necrosis factor-alpha (TNF- $\alpha)$ . Out of these papers, only eight [36, 37, 38, 39, 40, 41, 42, 43] fulfilled all of the eligibility criteria. The flowchart of research and data processing is shown in Fig. 1.

3.2 Characteristics of protein-based studies in saliva

Table 1 summarises the characteristics of the eight selected studies, disaggregated by biomarkers. The

Table 1
Descriptive data from the studies included in the meta-analysis

First author	Year	Country	Study subgroup	Proteomic biomarker	N	Age	Male (%)	Smokers (%)	Mean expression	Method	Dysplasia
Zheng	2018	China	OSCC	CEA	112	63.20	55.35	–	146.02	ELISA ng/ml	–
			OPMD	CEA	30	61.43	53.33	–	116.72	ELISA ng/ml	–
			HCG	CEA	60	60.57	61.66	–	77.34	ELISA ng/ml	–
Airoldi	1986	Italy	OPMD	CEA	17	52	82.40	82	58.52	ELISA ng/mL	–
			OSCC	CEA	11	66	81.80	81	80.72	ELISA ng/mL	–
			HCG	CEA	19	58	70	42	46.31	ELISA ng/mL	–
Awasthi	2017	India	OSCC	CYFRA 21	30	49.60	93.30	–	17.50	ELISA ng/ml	–
			OPMD	CYFRA 21	9	34.20	88.90	–	5.9	ELISA ng/ml	–
			HCG	CYFRA 21	25	48.10	88	–	3.9	ELISA ng/ml	–
Rajkumar	2015	India	OSCC	CYFRA 21	100	55.50	68	14	17.46	ELISA ng/ml	–
			OPMD	CYFRA 21	100	55.5	58	14	8.15	ELISA ng/ml	–
			HCG	CYFRA 21	100	43	65	0	3.06	ELISA ng/ml	–
Rhodus	2005	USA	OSCC	IL-6	13	59.46	76.92	61.53	0.0882	ELISA pg/ml	–
			OPMD	IL-6	13	56.76	30.76	53.84	0.0708	ELISA pg/ml	Moderate (38.46%) Severe (61.54)
			HCG	IL-6	13	58.70	53.84	0	0.0014	ELISA pg/ml	–
Sharma	2011	India	OL	IL-6	20	50	100	35	0.4140	ELISA pg/ml	Mild (40%) Moderate (35%) Severe (25%)
			HCG	IL-6	20	35	100	0	0.0172	ELISA pg/ml	–
Rhodus	2005	USA	OSCC	IL-8	13	59.46	76.92	61.53	3154	ELISA pg/ml	–
			OPMD	IL-8	13	56.76	30.76	53.84	1918	ELISA pg/ml	Moderate (38.46%) Severe (61.54)
			HCG	IL-8	13	58.70	53.84	0	1580	ELISA pg/ml	–
Gleber-Netto	2016	Taiwan	OSCC	IL8	60	51.40	95	83.30	283.75	ELISA pg/ml	–
			OPMD	IL8	60	50.80	95	96.70	140.35	ELISA pg/ml	–
			HCG	IL8	60	50.50	91.7	81.70	127.79	ELISA pg/ml	–
Punyani	2012	India	OSCC	IL8	25	53.20	64	–	1718.610	ELISA pg/mL	–
			OPMD	IL8	25	32.16	76	–	299.513	ELISA pg/mL	Mild (20 %)
			HCG	IL8	25	45.12	76	–	210.096	ELISA pg/mL	–

Note: CEA: carcinoembryonic antigen, CYFRA 21: soluble fragment of cytokeratin 19, IL-6: interleukin-6, IL-8: interleukin-8.

Table 2

Flow chart describing the management of data from the search, data processing and selection of articles for meta-analysis.

Table 3

Overall expression values

Biomarker	Study subgroup	OE	CI 95% ( $p$ )	Heterogeneity ( $I^{2}$ , $p$ )
CEA	OSCC/OPMD	25.854	13.215–38.492 ( $p<$ 0.001)*	Null ( $I^{2}=$ 0, $p=$ 0.582)
	OPMD/HCG	25.394	$-$ 1.221–52.008 ( $p=$ 0.061)	High ( $I^{2}=$ 79.91, $p=$ 0.026)
CYFRA21	OSCC/OPMD	9.317	9.014–9.619 ( $p<$ 0.001)*	Null ( $I^{2}=$ 0, $p=$ 0.428)
	OPMD/HCG	3.679	0.663–6.696 ( $p=$ 0.017)*	High ( $I^{2}=$ 91.24, $p<$ 0.001)
IL-6	OPMD/HCG	233.561	$-$ 88.265–555.387 ( $p=$ 0.155)	Extreme ( $I^{2}=$ 99.89, $p<$ 0.001)
IL-8	OSCC/OPMD	912.622	$-$ 108.248–1933.492 ( $p=$ 0.08)	Extreme ( $I^{2}=$ 97.69, $p<$ 0.001
	OPMD/HCG	45.967	$-$ 24.473–116.407 ( $p=$ 0.201)	Low ( $I^{2}=$ 39.30, $p=$ 0.193)

Note: Overall expression values (OE) with a 95% confidence interval (CI95%) and the heterogeneity ( $I^{2}$ ) for each marker in the different analysis subgroups. * Stands for statistically significant differences.

Figure 1.

Forest plot with the weighted mean differences, overall expression (OE), 95%.

Figure 2.

Forest plot with the weighted mean differences, overall expression (OE), 95% confidence interval and heterogeneity for each protein biomarker in saliva for the OPMD/HCG subgroup analysis A) CEA. B) CYFRA21. C) IL8. D) IL6.

eight selected studies were conducted in five different countries: China [36], Italy [43], India [37, 39, 40, 41], USA [42] and Taiwan [38]. A total of 986 patients were included who were distributed in the following groups: HCG ( $N=$ 335), OL ( $N=$ 20), OPMD ( $N=$ 267) and OSCC ( $N=$ 364). The average age of the HCG was 49.9 years, with an average age of 50 years in the OL group, 49.6 years in the OPMD group, and 55.9 years in the OSCC group, respectively. The analysed markers were: CEA (GO:0006915, apoptotic process), CYFRA21 (GO:0006955, immune response) IL-6 (GO:0006955, immune response) and IL-8 (GO:0006955, immune response). With regards to the level of dysplasia, this aspect of OPMD was only described in the studies by Rhodus et al. [42] (Moderate (38.46%), Severe (61.53)); Punyani et al. [40] (Mild (20%)) and Sharma et al. [41] (Mild (40%), Moderate (35%), Severe (25%)). No information was provided with regards to the size or clinical appearance of the OL (homogeneous non-homogeneous), or the treatment modality or follow-up (oncological events or recurrences).

3.3 Quality assessment of protein-based studies in saliva

With regards to the methodological quality of the studies (Supplementary 3), after applying the QUADAS-2 scale regarding the risk of bias, it was determined that this was very high for patient selection (100%) and very low (0%) for Flow and Timing and Reference Standard. With respect to concerns regarding the applicability, the methodological risk was generally very low, especially for the patient selection and index test fields, although the study by Zheng et al. (1 study of 8, 12.5%) had an uncertain evaluation of the reference standard, given that it did not define the number of cases for each type of diagnosis included in the OPMD group.

3.4 Synthesis of the qualitative and quantitative analyses of four salivary biomarkers

In terms of the different markers and study subgroups, Table 3 shows all the values of OE and $I^{2}$ . Only two markers were significant for the OSCC/ OPMD subgroup, CEA and CYFRA21, both with very low heterogeneity. CEA had an OE $=$ 25.854 (CI95%: 13.215–38.492, $p<$ 0.001, $I^{2}=$ 0) and CYFRA21 an OE $=$ 9.317 (CI95%: 9.014–9.619, $p<$ 0.001, $I^{2}=$ 0). Both markers included two studies respectively (CEA: Zheng et al and Airoldi et al and CYFRA21: Awatashi et al and Rajkumar et al) (Fig. 1A and 2B). IL-8 was not a predictable marker for the OSCC/OPMD subgroup (Fig. 1C). For the OPMD/HCG subgroup, only CYFRA21 was considered as such with an OE $=$ 3.679 (CI95%: 0.663–6.696, $p=$ 0.017) although it had a high heterogeneity ( $I^{2}=$ 91.24) (Table 3, Fig. 2A). CEA presented differences between OPMD and HCG approaching significance ( $p=$ 0.061) but with high heterogeneity; additionally, IL-6 and IL-8 showed no differences and presented extreme ( $I^{2}=$ 99.89) and low ( $I^{2}=$ 39.30) heterogeneity, respectively (Fig. 2B–D).

Once these proteins had been obtained from the eight selected references, we performed interaction studies to determine the possible relationship between them. To do this the STRING program was used, and as can be seen in Fig. 3, there is a direct relationship between IL-6 and IL-8 (CXCL-8), forming a small cluster of pro-inflammatory cytokines with two proteins, the CXC chemokine receptor type 2 (CXCR2) and the IL-6 receptor (IL6R), which are potent chemotherapeutics targets and neutrophil activators (Supplementary 4).

Given the relationship between the cytokines, which showed a well-defined cluster, several studies were performed to determine the position of these proteins in the immune system’s response to the inflammatory process. Using the DAVID program, we were able to observe how these proteins are found in the TNF signalling pathway activated by macrophages (Supplementary 5). The study, which used databases such as PHANTER provided us with a global view of the biological processes in which the selected proteins are involved. According to Supplementary 6, proteins are involved in signalling processes, the immune system, regulation, localisation, locomotion, etc. A new protein-protein interaction analysis was performed, in which it was observed that the selected biomarkers presented a theorical direct molecular relationships between IL-8 (or CXCL-8), IL-6 and CYFRA21 (KRT19). The data gathered from STRING indicated that KRT19 is involved in the organisation of myofibrils alongside KRT8, while CEA plays a very important role in cell adhesion and intracellular signalling processes (Fig. 3, Supplementarys 7 and 8). Only CYFRA21 can be considered as a good marker for the OPMD/HCG subgroup. By asking the program to indicate more possible nodes/interactions related to this protein, we determined a strong cluster composed of cytoskeleton proteins and Mesothelin, a membrane-binding protein that plays a very important role in the cell adhesion process (Supplementary 7 and Supplementary 9).

4. Discussion

The hypothesis of this study proposed the idea that there are protein-based biomarkers in saliva whose quantitative expression in OPMD is different to that of healthy control groups and malignant lesions, therefore making it possible for reference points of expression to be established. Although we found studies in which 11 different biomarkers had been analysed, it was only possible to perform a meta-analysis on eight studies with four biomarkers.

Figure 3.

Protein interaction network of the four protein markers selected using the STRING databases.

CEA is a transmembrane glycoprotein that has multiple functions. In OSCC of the tongue, its overexpression correlates with neutrophil infiltration and both are associated with poor prognosis [44]. In our meta-analysis, CEA presented differences in expression between OSCC and OPMD ( $p<$ 0.001). The value of the expression in the Chinese population [36] and the Italian population [43] was 29.3 (CI95%: 11.6–46.9) and 22.2 (CI95%: 4.1–40.3), respectively, with an OE of 25.85 (CI95%: 13.2–38.4). However, CEA did not present a significant difference in expression when comparing OPMD to HCG in the considered studies. The distribution of men and women in the population in Zheng et al.’s study [36] was similar, although the number of cases of OSCC was almost four times higher than the number of cases of OPMD (Table 1), which also included the diagnoses of oral lichen planus, OL and erythema, without specifying the number of cases that corresponded to each diagnosis (risk of bias). The salivary levels of CEA in the OSCC group showed a significant association with the tumour stage ( $p=$ 0.018) and lymph node metastasis ( $p=$ 0.035) [36]. On the other hand, more than 70% of the subjects in Airoldi et al’s study [43] were male, identifying that 70.6% of the OPMD corresponded to OL ( $n=$ 12). In this study, the CEA was significantly greater in OSCC than in OPMD ( $p<$ 0.025) and HCG ( $<$ 0.001).

Historically, CYFRA21 has been researched as a biomarker for non-small cell lung carcinoma and it was subsequently evaluated in other malignant [45] and potentially malignant disorders [46]. It may also be useful as a biomarker in the head and neck region [47]. The OE for CYFRA21 was highly significant in OSCC compared to OPMD ( $p<$ 0.001), and the latter group also showed significance when compared to HCG ( $p=$ 0.017). In the considered studies, populations of the same origin were studied [37, 39]. The OPMD group exhibited an OE that was 3.6 (CI95%: 0.6–6.6) times greater than HCG, despite this, there was a wide heterogeneity in the estimation of the included studies, for Awasthi et al. [37] with 2 (CI95%: 0.2–3.7) and Rajkumar et al. [39] with 5 (CI95%: 4.9–5.2). The Indian population in the study by Awasthi et al. [37] was comprised of more than 88% male individuals. This study evaluated a quantity of OPMD that was equivalent to 1/3 of OSCC cases, specifying that the OPMD group was comprised of 5 OL, 2 submucous fibrosis, 1 lichen planus and 1 squamous papilloma. CYFRA21 exhibited a positive association with the tumour size, the clinical tumour stage and the lymph node status. Furthermore, analysis of the ROC curve showed a sensitivity of 90% and a specificity of 97%, with a minimum cut-off value of 8.7 ng/mL for the detection of OSCC. The study by Rajkumar et al. [39] included an OPMD group that consisted of 50 cases of OL and 50 cases of oral submucous fibrosis, with males representing 60% of the cases and a reported rate of 14% of smokers for both OSCC and OL. The analysis of the ROC curve showed a specificity and sensitivity of 75%, with a cut-off limit of $>$ 10.4 ng/mL to differentiate OSCC from OPMD; additionally, it showed a specificity of 95% and a sensitivity of 83.6%, with a cut-off limit of $>$ 7.91 ng/mL for the detection of cases of OSCC in contrast to HCG.

In studies that were performed to determine the interaction between CEA and CYFRA21, the results demonstrated that both present a strong association (Fig. 3) and that they are directly related to a whole cluster of keratins that are involved in the cellular cytoskeleton, one of the key factors in oral carcinogenesis [48]. All this was in line with studies in which it has been determined that both proteins can be used as diagnostic markers in other epithelial tumours such as lung cancer [49] and head and neck cancer [50], therefore concurring that these two proteins are differentially expressed in healthy and potentially malignant tissues.

The IL-6 salivary mRNA was previously identified as a promising biomarker for OSCC [51], however, studies into the role of protein expression in saliva or OPMD are limited. There was no significant difference in the IL-6 cytokine salivary OE between the OPMD group and HCG. The anticipated means of expression for the studies included in the meta-analysis were considerably different, the highest being 397.8 (CI95%: 381.3–414.3) in the study by Sharma et al. [41], compared to 69.4 (CI95%: 56.1–82.6) in the study by Rhodus et al. [42], which resulted in a value of extreme heterogeneity. In the study by Rhodus et al, no clinical diagnoses were indicated in the OPMD group, only different histological degrees of dysplasia [42]. The concentration of IL-6 was significantly higher in OSCC and OPMD when compared to HCG ( $p<$ 0.001). On the other hand, in Sharma’s study, which was conducted in India, the means of salivary expression were almost 10 times higher than those observed in the US population, and it also included patients who presented periodontitis together with OL, which could have an impact on the amount of IL-6 measured. The OL group showed a significantly higher mean IL-6 compared to the HCG ( $p<$ 0.001). The levels of IL-6 in patients who smoked and had a higher degree of dysplasia were always higher.

Changes in the expression of pro-inflammatory cytokines may play a role in the malignant transformation in OSCC and are likely to be used as diagnostic markers in the serum of patients, due to their excessive production by tumour cells [14], and, likewise, they could also act as possible pharmacological targets. IL-8 plays an important role in cancer invasion, angiogenesis and metastasis [52]. In this meta-analysis, we found no significant difference for IL-8 between OSCC and OPMD, nor between OPMD and HCG. Likewise, we did not find any relationship between these pro-inflammatory cytokines and the previously described markers (CEA and CYFRA21), however, as can be seen in Fig. 3, the pro-inflammatory cytokines would form a strong signalling cluster by themselves, through the functional analysis that has been performed with the DAVID program, corresponding to the immune response and inflammation signalling pathway (Supplementary 4). The studies included in the meta-analysis showed a very high value of heterogeneity in all of the subgroups (Figs 1C and 3C). The study by Gleber-Netto et al. [38] covered an almost exclusively male population in Taiwan, with a percentage of smokers that exceeded 80% in all groups. The corresponding clinical diagnoses were not reported in the OPMD group, only those lesions with different degrees of dysplasia were included. On the other hand, the study conducted by Punyani et al. [40] in India reported that the OPMD group was comprised of 13 oral submucous fibrosis and 12 OL, and that 5 of these had mild dysplasia. The identification of IL-8 was significantly higher in OSCC compared to OPMD ( $p<$ 0.0001) and HCG ( $p<$ 0.0001). The OPMD group presented higher levels of IL-8 compared to HCG, however, this difference was not significant ( $p=$ 0.738). In the study by Rhodus et al. [42], the clinical diagnoses were not determined in the OPMD group, only the existence of lesions with different degrees of dysplasia was mentioned. The concentration of IL-8 was significantly higher in the OSCC group compared to HCG ( $p<$ 0.001), and it was also significantly higher in the OPMD group compared to HCG ( $p<$ 0.05), although the statistical significance between OSCC and OPMD was not noted.

The main constraints of this study included the lack of clinical-pathological information on the lesions that are mentioned as dysplastic in the OPMD group, and likewise, the degree of dysplasia has not been characterised. Additionally, there were considerable discrepancies between the number of analysed cases of OL and OPMD, compared with the cases of OSCC. Very few studies have analysed pure OL groups, independent of other OPMD, which therefore makes it difficult for specific conclusions to be drawn. Another aspect that must be taken into consideration is that the gender and age of the individuals in the meta-analysis was quite heterogeneous, without there being a strictly matched selection in terms of gender or similar age intervals between all of their study groups. Moreover, not all of the studies reported the presence of risk factors such as smoking (a key factor in oral carcinogenesis) and others included habits that are typical of the geographical region under study, such as chewing betel nut, therefore meaning that extrapolation to a global level was not possible. In terms of the strengths of this study, it is important to highlight the clear and strict inclusion criteria that made it possible to limit analysis biases. Furthermore, even though there were different techniques such as proteomic techniques that allowed for massive quantitative analyses of the literature, this systematic review only included studies in which the quantitative values were representative, reproducible and comparable.

5. Conclusion

This systematic review and meta-analysis provides a summary of the quantitative expression of biomarkers based on salivary proteins. Based on the results, we can conclude that it is quantitatively possible to discriminate the OPMD compared to malignant lesions and healthy controls. The CEA and CYFRA21 markers proved to be very useful when differentiating OSCC from OPMD, however, the usefulness of exclusively differentiating the group of leukoplakia from OSCC does not seem totally clear. Besides, the CYFRA21 marker proved to be the only protein that was capable of distinguishing between OPMD and healthy controls. The definition of these protein-based salivary markers as a molecular signature, which identifies potentially malignant lesions, would allow for new, less invasive protocols to be established for early diagnosis and screening, as well as making it possible to determine therapeutic targets that would improve the prognosis of patients with potentially malignant lesions and those at risk.

Author contributions

Esteban Arroyo: Conception, Interpretation and analysis of data, Preparation of the manuscript, Revision for important intellectual content.

Sergio Piñeiro Donís: Conception, Interpretation and analysis of data, Preparation of the manuscript, Revision for important intellectual content.

Cintia Chamorro: Interpretation and analysis of data, Preparation of the manuscript.

Monica Oliveira Alves: Interpretation and analysis of data, Preparation of the manuscript.

Xabier Marichalar Endia: Interpretation and analysis of data, Preparation of the manuscript.

Darcy Fernandes: Interpretation and analysis of data.

Alejandro Lorenzo Pouso: Interpretation and analysis of data, Preparation of the manuscript, Supervision.

Andreia Bufalino: Conception, Supervision.

Susana Bravo López: Conception, Preparation of the manuscript, Supervision.

Pilar Gandara Vila: Conception, Supervision.

Mario Pérez Sayáns: Conception, Interpretation and analysis of data, Preparation of the manuscript, Revision for important intellectual content, Supervision.

Sources of support in the form of grants

Esteban Arroyo received funding for this research from Programa Estudantes-Convênio de Pós-Graduação – PEC-PG, CAPES/CNPq, Brazil (88881.154490/2017-01); Andreia Bufalino received funding for this research from SÃ£o Paulo Research Foundation (FAPESP) (grant #2017/01438-0) and the National Council for Scientific and Technological Development CNPq (423945/2016-5); Darcy Fernandes received funding for this research from São Paulo Research Foundation (FAPESP) (grant #2017/01798-7). This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES).

Supplementary data

The supplementary files are available to download from http://dx.doi.org/10.3233/CBM-203043.

sj-pdf-1-cbm-10.3233_CBM-203043.pdf - Supplemental material

Supplemental material, sj-pdf-1-cbm-10.3233_CBM-203043.pdf

References

Scully

and Bagan

, Oral squamous cell carcinoma overview, Oral Oncol 45 (2009), 301–308. doi: 10.1016/j.oraloncology.2009.01.004.

El-Naggar

A.K.

Chan

J.K.

Grandis

J.R.

Takata

and Slootweg

P.J.

, Oral potentially malignant disorders and oral epithelial dysplasia, in: WHO Classif. Head Neck Tumours, Fourth edition, International Agency for Research on Cancer, Lyon, France, 2017: pp. 112–115.

Ferlay

Colombet

Soerjomataram

Mathers

Parkin

D.M.

Piñeros

Znaor

and Bray

, Estimating the global cancer incidence and mortality in 2018: GLOBOCAN sources and methods, Int J Cancer 144 (2019), 1941–1953. doi: 10.1002/ijc.31937.

Sklenicka

Gardiner

Dierks

E.J.

Potter

B.E.

and Bell

R.B.

, Survival analysis and risk factors for recurrence in oral squamous cell carcinoma: does surgical salvage affect outcome? J. Oral Maxillofac. Surg. 68 (2010), 1270–1275. doi: 10.1016/j.joms.2009.11.016.

Van der Waal

, Potentially malignant disorders of the oral and oropharyngeal mucosa; terminology, classification and present concepts of management, Oral Oncol. 45 (2009), 317–323. doi: 10.1016/j.oraloncology.2008.05.016.

Van Der Waal

, Potentially malignant disorders of the oral and oropharyngeal mucosa; present concepts of management, Oral Oncol. 46 (2010), 423–425. doi: 10.1016/j.oraloncology.2010.02.016.

Petti

, Pooled estimate of world leukoplakia prevalence: a systematic review, Oral Oncol. 39 (2003), 770–780. doi: 10.1016/S1368-8375(03)00102-7.

Neville

B.W.

Damm

D.D.

Allen

C.M.

and Chi

A.C.

, in: Oral Maxillofac. Pathol., Saunders/Elsevier, St. Louis, MO, 2016: pp. 355–363.

Kujan

Oliver

R.J.

Khattab

Roberts

S.A.

Thakker

and Sloan

, Evaluation of a new binary system of grading oral epithelial dysplasia for prediction of malignant transformation, Oral Oncol. 42 (2006), 987–993. doi: 10.1016/j.oraloncology.2005.12.014.

10.

Fleskens

S.A.J.H.M.

Bergshoeff

V.E.

Voogd

A.C.

van Velthuysen

M.-L.F.

Bot

F.J.

Speel

E.-J.M.

Kremer

Takes

and Slootweg

, Interobserver variability of laryngeal mucosal premalignant lesions: a histopathological evaluation, Mod. Pathol. 24 (2011), 892–898. doi: 10.1038/modpathol.2011.50.

11.

Sperandio

Brown

A.L.

Lock

Morgan

P.R.

Coupland

V.H.

Madden

P.B.

Warnakulasuriya

Moller

and Odell

E.W.

, Predictive Value of Dysplasia Grading and DNA Ploidy in Malignant Transformation of Oral Potentially Malignant Disorders, Cancer Prev. Res. (Phila. Pa.) 6 (2013), 822–831. doi: 10.1158/1940-6207.CAPR-13-0001.

12.

Kaufman

and Lamster

I.B.

, The Diagnostic Applications of Saliva – A Review, Crit. Rev. Oral Biol. Med. 13 (2002), 197–212. doi: 10.1177/154411130201300209.

13.

Gröchl

, Saliva: a reliable sample matrix in bioanalytics, Bioanalysis 9 (2017), 655–668. doi: 10.4155/bio-2017-0010.

14.

Aro

Wei

Wong

D.T.

and Tu

, Saliva Liquid Biopsy for Point-of-Care Applications, Front. Public Health. 5 (2017), doi: 103389/fpubh.2017.00077.

15.

Lorenzo-Pouso

A.I.

Pérez-Sayáns

Bravo

S.B.

López-Jornet

García-Vence

Alonso-Sampedro

Carballo

and García-García

, Protein-Based Salivary Profiles as Novel Biomarkers for Oral Diseases, Dis. Markers 2018, (2018), 1-22. doi: 10.1155/2018/6141845.

16.

Amado

Lobo

M.J.C.

Domingues

Duarte

J.A.

and Vitorino

, Salivary peptidomics, Expert Rev Proteomics 7 (2010), 709–721. doi: 10.1586/epr.10.48.

17.

Siravegna

Marsoni

Siena

and Bardelli

, Integrating liquid biopsies into the management of cancer, Nat. Rev. Clin. Oncol. 14 (2017), 531–548. doi: 10.1038/nrclinonc.2017.14.

18.

Tyanova

Temu

Sinitcyn

Carlson

Hein

M.Y.

Geiger

Mann

and Cox

, The Perseus computational platform for comprehensive analysis of (prote)omics data, Nat. Methods. 13 (2016), 731–740. doi: 10.1038/nmeth.3901.

19.

Tyanova

and Cox

, Perseus: A Bioinformatics Platform for Integrative Analysis of Proteomics Data in Cancer Research, in: L.von Stechow (Ed.), Cancer Biol.

Syst.

, Springer New York, New York, NY, 2018: pp. 133–148. doi: 10.1007/978-1-4939-7493-1_7.

20.

Fabregat

Jupe

Matthews

Sidiropoulos

Gillespie

Garapati

Haw

Jassal

Korninger

May

Milacic

Roca

C.D.

Rothfels

Sevilla

Shamovsky

Shorser

Varusai

Viteri

Weiser

Stein

Hermjakob

and D’Eustachio

, The Reactome Pathway Knowledgebase, Nucleic Acids Res. 46 (2018), D649–D655. doi: 10.1093/nar/gkx1132.

21.

Kuhn

Szklarczyk

Pletscher-Frankild

Blicher

T.H.

von Mering

Jensen

L.J.

and Bork

, STITCH 4: integration of protein–chemical interactions with user data, Nucleic Acids Res. 42 (2014), D401–D407. doi: 10.1093/nar/gkt1207.

22.

Szklarczyk

Santos

von Mering

Jensen

L.J.

Bork

and Kuhn

, STITCH 5: augmenting protein–chemical interaction networks with tissue and affinity data, Nucleic Acids Res. 44 (2016), D380–D384. doi: 10.1093/nar/gkv1277.

23.

Muetze

Goenawan

I.H.

Wiencko

H.L.

Bernal-Llinares

Bryan

and Lynn

D.J.

, Contextual Hub Analysis Tool (CHAT): A Cytoscape app for identifying contextually relevant hubs in biological networks, (2016). doi: 10.12688/f1000research.9118.1.

24.

Gurevich

V.V.

, Protein multi-functionality: Introduction, Cell. Mol. Life Sci. 76 (2019), 4405–4406. doi: 10.1007/s00018-019-03271-6.

25.

Higgins

J.P.

and Green

, eds., Cochrane Handbook for Systematic Reviews of Interventions Version 51.0, (2011). www.cochrane-handbook.org.

26.

Liberati

Altman

Tetzlaff

Mulrow

Gøtzsche

Ioannidis

Clarke

Devereaux

P.J.

Kleijnen

and Moher

, The PRISMA Statement for Reporting Systematic Reviews and Meta-Analyses of Studies That Evaluate Health Care Interventions: Explanation and Elaboration, J. Clin. Epidemiol. 62 (2009), e1–34. doi: 10.1016/j.jclinepi.2009.06.006.

27.

Shamseer

Moher

Clarke

Ghersi

Liberati

Petticrew

Shekelle

Stewart

L.A.

, and the PRISMA-P Group, Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation, BMJ. 349 (2015), g7647–g7647. doi: 10.1136/bmj.g7647.

28.

Axell

Pindborg

J.J.

Smith

C.J.

Waal

, and an International Collaborative Group on Oral White Lesions*, Oral white lesions with special reference to precancerous and tobacco-related lesions: conclusions of an international symposium held in Uppsala, Sweden, May 18–21 1994, J. Oral Pathol. Med. 25 (1996), 49–54. doi: 10.1111/j.1600-0714.1996.tb00191.x.

29.

Van der Waal

, Historical perspective and nomenclature of potentially malignant or potentially premalignant oral epithelial lesions with emphasis on leukoplakia – some suggestions for modifications, Oral Surg. Oral Med. Oral Pathol. Oral Radiol. 125 (2018), 577–581. doi: 10.1016/j.oooo.2017.11.023.

30.

Leeflang

M.M.

Deeks

J.J.

Takwoingi

and Macaskill

, Cochrane diagnostic test accuracy reviews, Syst. Rev. 2 (2013), 82. doi: 10.1186/2046-4053-2-82.

31.

Ciapponi

, QUADAS-2: instrumento para la evaluación de la calidad de estudios de precisión diagnóstica, (2015), 5.

32.

Muruganujan

Huang

Ebert

Mills

Guo

and Thomas

P.D.

, Protocol Update for Large-scale genome and gene function analysis with PANTHER Classification System (v.14.0), Nat. Protoc. 14 (2019), 703–721. doi: 10.1038/s41596-019-0128-8.

33.

Szklarczyk

Morris

J.H.

Cook

Kuhn

Wyder

Simonovic

Santos

Doncheva

N.T.

Roth

Bork

Jensen

L.J.

and von Mering

, The STRING database in 2017: quality-controlled protein – protein association networks, made broadly accessible, Nucleic Acids Res. 45 (2017), D362–D368. doi: 10.1093/nar/gkw937.

34.

Dennis

Sherman

B.T.

Hosack

D.A.

Yang

Gao

Lane

H.C.

and Lempicki

R.A.

, DAVID: Database for Annotation, Visualization, and Integrated Discovery, Genome Biol. 4 (2003), P3.

35.

Higgins

J.P.T.

and Thompson

S.G.

, Quantifying heterogeneity in a meta-analysis, Stat. Med. 21 (2002), 1539–1558. doi: 10.1002/sim.1186.

36.

Zheng

Sun

Yuan

Wang

Sun

and Zeng

, Clinical value of Naa10p and CEA levels in saliva and serum for diagnosis of oral squamous cell carcinoma, J. Oral Pathol. Med. 47 (2018), 830–835. doi: 10.1111/jop.12767.

37.

Awasthi

, Role of salivary biomarkers in early detection of oral squamous cell carcinoma, Indian J. Pathol. Microbiol. 60 (2017), 464. doi: 10.4103/IJPM.IJPM_140_16.

38.

Gleber-Netto

F.O.

Yakob

Feng

Dai

Kao

H.-K.

Chang

Y.-L.

Chang

K.-P.

and Wong

D.T.W.

, Salivary Biomarkers for Detection of Oral Squamous Cell Carcinoma in a Taiwanese Population, Clin. Cancer Res. 22 (2016), 3340–3347. doi: 10.1158/1078-0432.CCR-15-1761.

39.

Rajkumar

Ramya

Nandhini

Rajashree

Ramesh Kumar

and Nirmala Anandan

, Salivary and serum level of CYFRA 21-1 in oral precancer and oral squamous cell carcinoma, Oral Dis. 21 (2015), 90–96. doi: 10.1111/odi.12216.

40.

Punyani

S.R.

and Sathawane

R.S.

, Salivary level of interleukin-8 in oral precancer and oral squamous cell carcinoma, Clin. Oral Investig. 17 (2013), 517–524. doi: 10.1007/s00784-012-0723-3.

41.

Sharma

Bairy

Pai

Satyamoorthy

Prasad

Berkovitz

and Radhakrishnan

, Salivary IL-6 levels in oral leukoplakia with dysplasia and its clinical relevance to tobacco habits and periodontitis, Clin. Oral Investig. 15 (2011), 705–714. doi: 10.1007/s00784-010-0435-5.

42.

Rhodus

N.L.

Miller

C.S.

Myers

and Ondrey

, NF-κB dependent cytokine levels in saliva of patients with oral preneoplastic lesions and oral squamous cell carcinoma, Cancer Detect. Prev. 29 (2005), 42–45. doi: 10.1016/j.cdp.2004.10.003.

43.

Airoldi

Piantino

Pacchioni

Mastromatteo

Pedani

and Gandolfo

, Gastrointestinal cancer-associated antigen (GICA) in oral carcinoma, Oral Surg. Oral Med. Oral Pathol. 61 (1986), 263–267. doi: 10.1016/0030-4220(86)90372-5.

44.

Wang

Chi

Xiang

Lin

Wang

Wei

and Feng

, Carcinoembryonic antigen cell adhesion molecule 1 inhibits the antitumor effect of neutrophils in tongue squamous cell carcinoma, Cancer Sci. 110 (2019), 519–529. doi: 10.1111/cas.13909.

45.

Molina

Agusti

Filella

Joseph

Giménez

and Ballesta

A.M.

, Study of a new tumor marker, CYFRA 21-1, in malignant and nonmalignant diseases, Tumour Biol. J. Int. Soc. Oncodevelopmental Biol. Med. 15 (1994), 318–325. doi: 10.1159/000217908.

46.

Camisasca

D.R.

da Rós Gonçalves

Soares

M.R.

Sandim

Nogueira

F.C.S.

Garcia

C.H.S.

Santana

de Oliveira

S.P.

Buexm

L.A.

de Faria

P.A.S.

Dias

F.L.

Pereira

D. de A.

Zingali

R.B.

Alves

and Lourenço

S.Q.C.

, A proteomic approach to compare saliva from individuals with and without oral leukoplakia, J. Proteomics. 151 (2017), 43–52. doi: 10.1016/j.jprot.2016.07.029.

47.

AlAli

A.M.

Walsh

and Maranzano

, CYFRA 21-1 and MMP-9 as salivary biomarkers for the detection of oral squamous cell carcinoma: a systematic review of diagnostic test accuracy, Int. J. Oral Maxillofac. Surg. (2020), S0901502720300369. doi: 10.1016/j.ijom.2020.01.020.

48.

Pitiyage

Tilakaratne

W.M.

Tavassoli

and Warnakulasuriya

, Molecular markers in oral epithelial dysplasia: review, J. Oral Pathol. Med. 38 (2009), 737–752. doi: 10.1111/j.1600-0714.2009.00804.x.

49.

Okamura

Takayama

Izumi

Harada

Furuyama

and Nakanishi

, Diagnostic value of CEA and CYFRA 21-1 tumor markers in primary lung cancer, Lung Cancer. 80 (2013), 45–49. doi: 10.1016/j.lungcan.2013.01.002.

50.

Barak

Meirovitz

Leibovici

Rachmut

Peretz

Eliashar

and Gross

, The Diagnostic and Prognostic Value of Tumor Markers (CEA, SCC, CYFRA 21-1, TPS) in Head and Neck Cancer Patients, Anticancer Res. 35 (2015), 5519–5524.

51.

Márton

I.J.

Horváth

Lábiscsák

Márkus

Dezső

, Szabó

Tar

Piffkó

Jakus

Barabás

Olasz

Kövér

Tőzsér

Sándor

Csősz

É.

Scholtz

and Kiss

, Salivary IL-6 mRNA is a Robust Biomarker in Oral Squamous Cell Carcinoma, J. Clin. Med. 8 (2019), 1958. doi: 10.3390/jcm8111958.

52.

Ralli

Grasso

Gilardi

Ceccanti

Messina

M.P.

Tirassa

Fiore

Altissimi

Salzano

F.A.

Vincentiis

M.D.

and Greco

, The role of cytokines in head and neck squamous cell carcinoma: A review, Clin. Ter. 171 (2020), e268–e274.

Usefulness of protein-based salivary markers in the diagnosis of oral potentially malignant disorders: A systematic review and meta-analysis

Abstract

Keywords

1. Introduction

2. Methods

2.1 PICO question

2.2 Criteria for considering studies for this review

2.2.1 Types of accuracy study

2.2.2 Participants

2.2.3 Control and target conditions

2.2.4 Reference standard

2.2.5 Index test(s)

2.3 Search methods for the identification and selection of studies

2.3.1 Information sources and search strategy

2.4 Data collection and analysis

2.5 Assessment of methodological quality

2.6 Qualitative analysis

2.7 Quantitative analysis

2.8 Functional networks, and pathway mapping

2.9 Assessment of heterogeneity

3. Results

3.1 Study selection

3.2 Characteristics of protein-based studies in saliva

Table 1 Descriptive data from the studies included in the meta-analysis

3.4 Synthesis of the qualitative and quantitative analyses of four salivary biomarkers

4. Discussion

Author contributions

Sources of support in the form of grants

Supplementary data

sj-pdf-1-cbm-10.3233_CBM-203043.pdf - Supplemental material

References

Table 1
Descriptive data from the studies included in the meta-analysis