Tissue biopsy with histopathological examination is still considered the gold standard to diagnose oral squamous cell carcinoma (OSCC). This systematic review explored the diagnostic test accuracy of two salivary biomarkers in adults suspected of OSCC. The Cochrane Library, MEDLINE, and Embase databases were searched for clinical studies evaluating the diagnostic accuracy of salivary biomarkers in detecting OSCC. Studies were eligible for inclusion if only singular salivary biomarkers were evaluated in three or more studies. Studies investigating combined salivary biomarkers or assessing patients with oral potentially malignant disorders only were excluded. The reporting of the review follows the PRISMA checklist. Six studies, recruiting 775 participants, were included in this review for only two salivary biomarkers, cytokeratin 19 fragment (CYFRA 21-1) and matrix metalloproteinase 9 (MMP-9). The sensitivity and specificity (with 95% confidence intervals) for CYFRA 21-1 studies ranged from 0.84 (0.75–0.91) to 0.94 (0.83–0.99) and from 0.84 (0.71–0.93) to 0.96 (0.80–1.00), respectively. In MMP-9 studies, sensitivity (with 95% confidence intervals) ranged from 0.76 (0.67–0.83) to 1.00 (0.78–1.00) and specificity from 0.27 (0.12–0.46) to 1.00 (0.78–1.00). The overall quality of the included studies was poor. Due to a lack of strong and high-quality evidence, considerable uncertainty remains surrounding the use of singular salivary biomarkers for the detection of OSCC.
Many different techniques and tests delivered in a variety of settings are available for the detection of oral squamous cell carcinoma (OSCC). Conventional oral examination performed by a general dental practitioner has traditionally been recommended by various health and medical/dental organizations as a part of oral cancer screening programmes, e.g., the British Dental Association . A Cochrane systematic review has evaluated the diagnostic accuracy of a variety of tests as adjunct tests to the conventional oral examination in a dental setting, including vital staining, oral cytology, light-based detection, and oral spectroscopy in patients presenting with clinically evident lesions . The results showed that cytology tests appeared to have the greatest potential and warranted further investigation. However, the overall quality of the evidence was variable, with insufficient evidence to justify the replacement of the current ‘gold standard’ diagnostic test of scalpel biopsy with histopathological examination in addition to the conventional oral examination.
Comparatively, saliva, also known as the ‘mirror of the body’, is an easy fluid to collect for bioanalysis. Sample/specimen collection is non-invasive, inexpensive, and simple, which makes it an ideal method for diagnostic purposes . Saliva contains an extensive number of compounds, is safe to handle and store, does not clot, and as such, it is deemed a promising diagnostic biofluid . Two systematic reviews have evaluated the performance of a wide range of salivary biomarkers (biomarker expression) in OSCC . However, these systematic reviews mainly included ‘case–control’ type accuracy studies, which can be prone to bias. Studies that compare altered expression of a specific salivary biomarker between ‘cases’ with OSCC and healthy ‘control’ participants can deliver misleading results when the methods of participant recruitment are not taken into consideration . If participants representing only subgroups of the spectrum of disease and spectrum of non-diseased are included, the resulting estimates of diagnostic test accuracy may not be applicable to the clinical question .
OSCC is an increasing disease globally. Several factors have been implicated in the development of the disease, such as tobacco, betel quid, and alcohol use, as well as others like human papillomavirus (HPV) and Epstein–Barr virus (EBV) infections. The survival rate of individuals with OSCC is approximately 90% when detected early, compared to only 50% after delayed diagnosis . Hence, both prevention and early detection of disease are crucial in reducing OSCC mortality.
The aim of this systematic review was to identify and summarize the diagnostic accuracy of salivary biomarkers currently being used in the detection of OSCC in adults. The reporting of the review follows the PRISMA checklist .
Materials and methods
Studies were eligible for inclusion in the review if they evaluated salivary biomarkers and reported measures of diagnostic test accuracy such as sensitivity and specificity in detecting OSCC in adults (age ≥16 years) suspected of having OSCC, with no prior treatment such as chemotherapy or radiotherapy.
Inclusion and exclusion criteria
Only studies evaluating the performance of singular salivary biomarkers (index test) against any type of tissue biopsy with histopathological examination (reference standard) for the diagnosis of OSCC were eligible for inclusion. Studies were eligible for inclusion in the review if only singular salivary biomarkers were evaluated and where the performance of a salivary biomarker was assessed in three or more studies. Studies investigating combined salivary biomarkers or assessing patients with oral potentially malignant disorders (PMDs) only were excluded.
Search protocol and study selection
An electronic search was performed on June 14, 2018 for studies published within the last 5 years (from 2014 to 2018) using the following databases: Cochrane Library, MEDLINE via OVID, and Embase via OVID. The search strategy used was as follows: (oral cancer OR mouth cancer OR oral squamous cell carcinoma OR OSCC).mp AND (diagnos* OR detect*).mp AND (saliva* AND (marker* OR biomarker* OR test)).mp. Following the removal of duplicates, the titles and/or abstracts of the retrieved records were screened for relevancy, and full-text articles were obtained if necessary. Next, full-text articles were assessed for eligibility. Studies were included following consensus among all three authors.
The following data were extracted from each included study: (1) study information and setting (country, design, sample size, test used), (2) type of salivary biomarker, (3) study results: true-positive (sensitivity) and true-negative (specificity), and positive and negative predictive values (PPV and NPV) where reported. A 2 × 2 table was constructed for each study (reference standard against the index test) using Review Manager 5.3 . The corresponding author of a study was contacted via email where further information or explanation was needed.
Assessment of methodological quality
Using a tailored QUADAS-2 tool, the quality of the included studies (risk of bias and applicability concerns) was assessed through its four domains: patient selection, index test, reference standard, and flow and timing of participants through the study . The indicators of the assessment of quality (QUADAS-2) are summarized in Table 1 , as adapted from Macey et al. . The results were then graphically presented using RevMan .
|Domain||Patient selection||Index test||Reference standard||Flow and timing|
(Yes, No, or Unclear)
|Was a consecutive or random sample of patients enrolled?||Were the index test results interpreted without knowledge of the results of the reference standard?||Is the reference standard likely to correctly classify the target condition?||Was there an appropriate interval between the index test and reference standard?|
|Yes: if consecutive patients or a random sample of individuals were recruited
No: if non-consecutive patients or a non-random sample of individuals were recruited
Unclear: if patient selection was not clearly described
|Yes: if interpreters of the index test results clearly did not know the results of biopsy/histopathology
No: if interpreters of the index test results clearly knew the results of biopsy/histopathology
Unclear: if the study did not provide any information on whether interpreters of the index tests were blinded to biopsy/histopathology
|Yes: if the biopsy was independently confirmed by at least two qualified pathologists
No: if the biopsy was not independently confirmed by at least two qualified pathologists, or there was a lack of agreement between pathologists
Unclear: if the study did not state who confirmed the biopsy
|Yes: if the delay between the index test and reference standard was considered acceptable for the majority of participants
No: if the delay between the index test and reference standard was considered unacceptable for the majority of participants
Unclear: if the delay between the index test and reference standard was not explicitly stated
|Was a case–control design avoided?||If a threshold was used, was it pre-specified?||Were the reference standard results interpreted without knowledge of the results of the index tests?||Did all patients receive the same reference standard?|
|Yes: if a case–control design was avoided
No: if a case–control design was not avoided
Unclear: if no clear type of study design was reported
|Yes: if the threshold was pre-specified
No: if the threshold was not pre-specified
Unclear: if it was unclear whether the threshold was pre-specified
|Yes: if pathologists clearly did not know the index test results when interpreting biopsied tissues
No: if pathologists knew the results of the index test when interpreting biopsied tissues
Unclear: if the study did not provide any information on whether the pathologists were blinded to the index test results
|Yes: if the same reference standard was used for all participants
No: if the same reference standard was not used for all participants
Unclear: if it was unclear whether different reference standards were used
|Did the study avoid inappropriate exclusions?||Was conflict of interest avoided?||Were all patients included in the analysis?|
|Yes: if patients with any stage of OSCC regardless of their other systemic diseases (with no prior chemotherapy or radiotherapy) were recruited
No: if patients with common diseases/drugs seen in practice were excluded
Unclear: if no clear inclusion and exclusion criteria were reported
|Yes: if the study declared no conflict of interest
No: if the study declared a conflict of interest
Unclear: if there was no information on conflict of interest
|Yes: if all patients were included in the analysis
No: if only some patients were included in the analysis
Unclear: if it was unclear whether all patients were included in the analysis
|Overall risk of bias||Could the selection of patients have introduced bias?||Could the conduct or interpretation of the index test have introduced bias?||Could the reference standard, its conduct, or its interpretation have introduced bias?||Could the patient flow have introduced bias?|
|High: if answered ‘No’ to any question
Low: if answered ‘Yes’ to all questions
Unclear: if answered ‘Unclear’ to all questions or accompanied by any ‘Yes’
|Concerns regarding applicability: High, Low, or Unclear||Are there concerns that the included patients do not match the review question?||Are there concerns that the index test, its conduct, or interpretation differ from the review question?||Are there concerns that the target condition, as defined by the reference standard, does not match the review question?|
Statistical analysis and data synthesis
The target condition (disease positive) was OSCC. The true-positive, true-negative, false-positive, and false-negative values were used to construct 2 × 2 tables for each study . The diagnostic test accuracy results for each biomarker were expressed as sensitivity and specificity with 95% confidence intervals (CI). Results were presented graphically as a coupled forest plot for each salivary biomarker.
Following the removal of duplicates, the electronic search yielded a total of 250 records, of which 224 were excluded; 26 were further assessed for eligibility ( Supplementary Material Appendix 1). Only six studies met the inclusion criteria in this review, as illustrated in the PRISMA flow diagram in Fig. 1 . The main reason for exclusion was the low number of studies that investigated each salivary biomarker ( Supplementary Material Appendix 2). Only if a biomarker had three or more studies was it eligible to be included in the review. Thus, only six studies reporting the diagnostic accuracy of salivary biomarkers cytokeratin 19 fragment (CYFRA 21-1) and matrix metalloproteinase 9 (MMP-9) in detecting OSCC were included.
Description of studies
The included studies evaluated data from a total of 775 patients (355 patients for CYFRA 21-1 and 420 patients for MMP-9). Of the total 775 patients, 356 (46%) were diagnosed with OSCC according to the reference standard of biopsy and histopathological examination: 51% (180/356) patients in the CYFRA 21-1 studies and 176 (42%) patients in the MMP-9 studies. The remaining 419 patients were classed as disease-free or non-OSCC (175 patients for CYFRA 21-1 and 244 patients for MMP-9). The three CYFRA 21-1 studies were from India , while the three MMP-9 studies were from Egypt , Germany , and Taiwan . Salivary sample collection was conducted in a hospital setting (academic or dental/medical centre) between 2008 and 2014.
The classification and reporting of OSCC varied across studies, making meta-analysis inappropriate: two studies reported OSCC stages only (I–IV) , one study reported OSCC grades only (1 and 2) , two studies reported OSCC stage (0–IVA) with grade (1–3) , and one study reported OSCC grade (1–4), but reported stages in the TNM system separately .
Methodological quality of included studies
Figs. 2 and 3 summarize the results of the tailored quality assessment of the included studies for CYFRA 21-1 and MMP-9, respectively, utilizing the QUADAS-2 tool . The figures show a summary of the risk of bias and applicability concerns according to the authors’ judgement in each of the four domains for each included study ( Supplementary Material Appendix 3 includes the comprehensive quality assessment performed for each study).
None of the included studies could be classified as being at low risk of bias for all four domains. Patient selection was considered as high risk of bias in all of the studies, which was mainly due to the method of patient enrolment , the nature of the study design , and implementing inappropriate exclusions .
The index test was considered to be at low risk of bias in only one study , unclear in one study , but high in four studies . The variation across the level of bias differed due to insufficient details reported as to whether the results of the index test were interpreted without prior knowledge of the reference standard results, lack of pre-specification of a test-positive threshold, and statement of conflict of interest. Similarly, the reference standard domain was considered to be at low risk of bias in only one study , and unclear in the other five studies . All of the five studies were judged as unclear due to the inadequate reporting of whether the biopsy and histopathological diagnoses were made without any prior knowledge of the index test results. Additionally, three studies failed to comprehensively report the details of the reference standard .
The flow and timing domain was considered as having an unclear risk of bias across all of the studies due to insufficient details reported regarding the interval period between the index and reference standard (inadequately reported in all six studies), whether all patients received the same reference standard , and whether all patients had been included in the analysis . Only three studies were assessed as being of low concern for applicability according to patient selection, the index test, and the reference standard . The remaining three studies were assessed as being of unclear concern for applicability due to the lack of detail regarding patient selection and conduct of the reference standard .
The sensitivity and specificity of CYFRA 21-1 (with 95% CI) ranged from 0.84 (0.75–0.91) to 0.94 (0.83–0.99) and from 0.84 (0.71–0.93) to 0.96 (0.80–1.00), respectively. The sensitivity and specificity of MMP-9 (with 95% CI) ranged from 0.76 (0.67–0.83) to 1.00 (0.78–1.00) and from 0.27 (0.12–0.46) to 1.00 (0.78–1.00), respectively. Table 2 shows a summary of the data extracted from the included studies. The overall quality of the included studies for both CYFRA 21-1 and MMP-9 was poor and mainly limited by selective patient selection, a lack of assessors to the results of previous assessments, and small sample sizes. Table 3 shows a summary of the findings and quality assessment for both CYFRA 21-1 and MMP-9 studies.
|Biomarker||Study design||City, country||Sample size (OSCC a /control)||Technique/test used||Sensitivity||Specificity||PPV||NPV||Cut-off value||AUC|
|CYFRA 21-1||Awasthi N. Role of salivary biomarkers in early detection of oral squamous cell carcinoma. Indian J Pathol Microbiol 2017: 60: 464–468.
Years conducted: 2010 to 2011
|Lucknow, India||55 (30/25)||ELISA||90%||97%
|Malhotra R, Urs A, Chakravarti A, Kumar S, Gupta V, Mahajan B. Correlation of Cyfra 21-1 levels in saliva and serum with CK19 mRNA expression in oral squamous cell carcinoma. Tumour Biol 2016: 37: 9263–9271.
Year conducted: Unknown
|New Delhi, India||100 (50/50)||ECLIA||93.8%
|85.5% c||93.3% c||8.5 ng/ml||NR|
|Rajkumar K, Ramya R, Nandhini G, Rajashree P, Ramesh Kumar A, Nirmala Anandan S. Salivary and serum level of CYFRA 21-1 in oral precancer and oral squamous cell carcinoma. Oral Dis 2015: 21: 90–96.
Years conducted: 2009 to 2012
|Chennai, India||200 (100/100)||ELISA||83.6%
|95%||94.4% c||85.6% c||>7.91 ng/ml||0.895|
|MMP-9||Ghallab N, Shaker O. Serum and salivary levels of chemerin and MMP-9 in oral squamous cell carcinoma and oral premalignant lesions. Clin Oral Investig 2017: 21: 937–947.
Year conducted: 2014
|Cairo, Egypt||30 (15/15)||ELISA||100%||100%||100%||100%||260.32 ng/ml||1.000|
|Peisker A, Raschke G, Fahmy M, Guentsch A, Roshanghias K, Hennings J, Schultze-Mosgau S. Salivary MMP-9 in the detection of oral squamous cell carcinoma. Med Oral Patol Oral Cir Bucal 2017: 22: e270–e275.
Year conducted: Unknown
|Jena, Germany||60 (30/30)||ELISA||100%||26.7%
|57.7% c||100% c||>0.104
|Yu J, Chen Y, Chiang W, Hsiao Y, Chu L, See L, et al. Saliva protein biomarkers to detect oral squamous cell carcinoma in a high-risk population in Taiwan. Proc Natl Acad Sci 2016: 113: 11549–11554.
Years conducted: 2008 to 2013
|Taiwan||330 (131/199)||Multiplex LC-MRM-MS||75.6%
|60.3%||55.6% c||79% c||NR||0.726|