Meta-analysis of magnetic resonance imaging accuracy for diagnosis of oral cancer

Abstract

Objective

To establish the diagnostic accuracy of magnetic resonance imaging (MRI) as an auxiliary means for the diagnosis of oral cancer through a systematic review and meta-analysis.

Methods

An exhaustive search of publications from 1986 to 2016 was performed of Medline, Embase and Cochrane (and related databases), including grey literature. Primary diagnostic accuracy studies that assessed oral cancer (target condition) using MRI (index test) were included. Diagnostic threshold, sensitivity and meta-regression analyses were performed. A meta-analysis was performed using Meta-DiSc® v. 1.4 software.

Results

A total of 24 primary studies were assessed, comprising 1,403 oral cancer lesions. Nine studies used diffusion-weighted MRI, with a diagnostic odds ratio (DOR) of 30.7 (95% confidence interval [CI]: 12.7–74.3) and area under the curve (AUC) of 0.917 (95% CI: 0.915–0.918); seven studies used dynamic contrast-enhanced MRI, with a DOR of 48.1 (95%CI: 22.4–103.2) and AUC of 0.936 (95% CI: 0.934–0.937); and 13 studies used traditional MRI, with a DOR of 23.9 (95%CI: 13.2–43.3) and AUC of 0.894 (95% CI: 0.894–0.895). Meta-regression analysis indicated that the magnetic field strength may have influenced the heterogeneity of the results obtained (p = 0.0233) using traditional MRI. Sensitivity analysis revealed a discrete reduction of inconsistency in some subgroups.

Conclusion

The three types of MRI assessed exhibited satisfactory accuracy compared to biopsy. Considering the relevance of early treatment and screening and that better health care results in improved survival rates and quality of life for oral cancer patients, we suggest the use of MRI as a part of the pre-treatment and monitoring protocol at public health services.

About the Authors:

Marcelo Aldrighi Moreira

Affiliations Graduate Program in Public Health (PPGSCol), Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil, Laboratory of Information and Communications Technology in Health (TISaude), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Luiza Silveira Lessa

Affiliations Graduate Program in Public Health (PPGSCol), Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil, Dentistry Course, Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Francieli Regina Bortoli

Affiliation: Universidade Federal de Santa Catarina (UFSC), Florianópolis, SC, Brazil

Abigail Lopes

Affiliations Graduate Program in Public Health (PPGSCol), Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil, Laboratory of Information and Communications Technology in Health (TISaude), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Eduardo Picolo Xavier

Affiliation: Laboratory of Information and Communications Technology in Health (TISaude), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Renan Antonio Ceretta

Affiliation: Dentistry Course, Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Fernanda Guglielmi Faustini Sônego

Affiliation: Dentistry Course, Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Cristiane Damiani Tomasi

Affiliation: Graduate Program in Public Health (PPGSCol), Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Patricia Duarte Simões Pires

Affiliation: Dentistry Course, Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Luciane Bisognin Ceretta

Affiliations Graduate Program in Public Health (PPGSCol), Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil, Laboratory of Information and Communications Technology in Health (TISaude), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil, Dentistry Course, Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Ingrid Dalira Schweigert Perry

Affiliation: Graduate Program in Public Health (PPGSCol), Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil

Priscyla Waleska Simões

* E-mail: [email protected]

Affiliations Graduate Program in Public Health (PPGSCol), Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil, Laboratory of Information and Communications Technology in Health (TISaude), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil, Dentistry Course, Unidade Acadêmica de Ciências da Saúde (UNASAU), Universidade do Extremo Sul Catarinense (UNESC), Criciúma, SC, Brazil, Engineering, Modeling and Applied Social Sciences Center (CECS), Universidade Federal do ABC (UFABC), São Bernardo do Campo, SP, Brazil

ORCID http://orcid.org/0000-0003-1782-0542

Introduction

Oral cancer is currently considered to be a public health problem worldwide. Although it represents little more than 2% of the global incidence of cancer, its lethality of 50% is a cause of much concern [1]; moreover, this type of cancer is the sixth most common globally [2]. Approximately 650,000 new cases occur every year, and approximately 40% of head and neck tumours are squamous cell carcinomas [3].

The high mortality rate of oral cancer has many causes and is likely associated with diagnostic delay. Oral lesions are easy to access and should be diagnosed early for treatment to be efficacious. Nevertheless, patients are often diagnosed in an advanced stage of disease. Diagnosis is late in most cases because patients do not seek treatment or do not have easy access to professionals who can establish the diagnosis [4].

Oral cancer has a subtle and asymptomatic onset, contributing to the diagnostic delay; for this reason, special attention is required from clinical practitioners, especially when risk factors, such as smoking, alcohol consumption and sun exposure, are present [5]. The prevention of oral cancer is intimately associated with early diagnosis and behavioural changes, such as quitting smoking and drinking [6].

Within this context, health services must simultaneously detect oral cancer in patients and maintain continued health education programmes that target risk factors. In addition, specific times and/or situations must be chosen to increase public opinion [7]. For example, dentists can play a significant role in the early detection of malignant and premalignant conditions and should assess all patients at high risk for disease [7, 8].

The current first line of investigation of oral abnormalities is visual inspection, which is a subjective method. Biopsy is the most widely accepted technique for accurate identification of lesions in the oral mucosa and is considered the gold standard for the detection of oral cancer [9].

Considering the aforementioned facts and the difficulties for early detection of oral cancer, the search for evidence and assessment of new rapid, precise and less invasive diagnostic methods is highly relevant [10]. Within this context, the use of imaging methods for the investigation of malignant head and neck tumours has increased substantially in parallel with the development of the modern, so-called last-generation methods [11].

Magnetic resonance imaging (MRI) is the technique that affords the highest-quality images of soft tissues without employing ionising radiation and without known biological risks [12]. This technique provides information on the extent of lesions, probable infiltration of major vessels and lymph node involvement, thus contributing to the determination of treatment and prognosis [13]. Some authors have noted the increasing use of diagnostic imaging techniques, such as MRI, among patients undergoing treatment for oral cancer [14–16].

One meta-analysis addressing this subject compared the accuracy of computed tomography and MRI for the diagnosis of cervical lymph node metastasis of head and neck cancer [17]. Another meta-analysis assessed the accuracy of MRI for the specific diagnosis of mandibular involvement of head and neck cancers [18]. However, none of these studies assessed the accuracy of MRI for the specific diagnosis of oral neoplasms.

Based on the aforementioned considerations and the lack of systematic reviews and meta-analyses assessing the accuracy of MRI for specific diagnosis of oral cancer, the present study sought to provide evidence on this subject. Oral cancer impacts the morbidity, mortality and quality of life of users of public health services, and early diagnosis can lead to a better prognosis. Moreover, methods such as MRI (i.e., non-invasive, easy to perform and reportedly efficacious for the diagnosis and precise delimitation of oral cancer lesions) are available at public health services. For these reasons, an assessment of the accuracy of these techniques for the diagnosis of oral cancer via a systematic review and meta-analysis will provide important evidence of their diagnostic validity. These data may serve as the foundation of proposals to include such methods in clinical practice, ultimately having a direct impact on the oral health of patients.

Materials and methods

A systematic review of diagnostic accuracy studies was performed. An exhaustive search strategy of studies published from 1986 to 2016 was performed in the databases Medical Literature Analysis and Retrieval System Online (Medline) via PubMed (S1 Fig), Cancer Literature (CANCERLIT), Latin American and Caribbean Health Sciences Literature (LILACS), Excerpta Medical Database (Embase) and related databases, including so-called “grey” literature”, using items for Systematic Reviews and Meta-analysis (PRISMA) guidelines (S1 File).

Search terms were selected from Medical Subject Headings (MeSH) and the Embase Emtree thesaurus and included “Oral Cancer” and synonyms associated to the index diagnostic test, which was designated “Magnetic Resonance” (and synonyms).

The symbol “*” was used to retrieve all variations of the search terms suffixes. The aforementioned terms were combined using the Boolean operators “AND”, “OR” and “NOT”.

The search was restricted to studies on humans; no limitation was established as to the language of publications. The references mentioned in all retrieved primary studies were analysed. In addition, authors of publications with incomplete data were directly contacted.

The titles and abstracts of retrieved studies were independently analysed by two investigators. Articles in English were assessed by two reviewers, and articles in other languages by a third independent reviewer; translations were performed when necessary. Instances of disagreement as to inclusion or exclusion of articles were solved by consensus; when consensus could not be attained, disagreement was solved by a fourth reviewer.

Primary diagnostic accuracy studies relating to diagnosis of oral cancer (target condition) by means of MRI (index test) were included. Studies reporting on oral cancer in adult patients were considered; oral cancer was defined as neoplasms affecting the lips and oral cavity (oral mucosa, gums, hard palate, tongue, salivary glands and oral floor). The MRI results (positive or negative) were considered as index tests.

Studies with oral cancer patients assessed based on the gold standard test (biopsy) and studies that included patients with histopathological diagnosis of oral cancer and without previous treatment were considered in the present review. Only studies in which the magnetic field strength (T) varied from 0.5 T to 3 T were considered because this is the most effective range for the identification of malignant tumours.

Studies with data that did not allow for the construction of a 2 x 2 contingency table were excluded, as were studies with patients who exhibited contraindications to MRI, including pacemakers, aneurysm clips, metallic fragments in the eyes, cochlear implants, ocular implants, internal drug infusion pumps, certain prosthetic heart valves, steel-containing bullets, bone growth stimulators and neurostimulators [6].

Studies conducted with patients exhibiting contraindications to the use of gadolinium contrast media were excluded. These conditions included the following: kidney failure, haemolytic anaemia, sickle-cell anaemia, pregnancy, breastfeeding, respiratory disorders, asthma and a history of allergy to the contrast media.

Next, the following data were extracted from the included studies: year of publication, country and continent where the study was conducted, type of oral cancer, study design and demographic data (e.g., age and gender) in addition to number of oral cancer patients and number of diagnostic hits on MRI.

These data were independently extracted by two reviewers. The data from articles published in languages other than English were independently extracted by a third reviewer; translations were performed when necessary. Instances of disagreement as to the extracted data were solved first by consensus or by another reviewer when consensus could not be reached.

The methodological quality of studies was assessed based on the criteria included in the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool [19, 20], which was designed to evaluate studies based on four key domains (patient selection; index test; reference standard; and flow and timing).

A 2 x 2 contingency table was constructed for each selected study; the results corresponding to the gold standard and MRI were entered as positive or negative. Sensitivity, specificity and likelihood ratio were calculated; the diagnostic odds ratio (DOR) was used as measure of diagnostic accuracy. A DOR value of 1 indicates a test without discriminatory power; the higher the DOR value, the greater the degree of relevance of the assessed diagnostic test [21].

For the studies in which one single cell in the 2 x 2 contingency table had a value of 0, 0.5 was added to all of the cells to enable calculation. However, studies in which 0 occurred in more than two cells were excluded from meta-analysis.

Meta-analysis was performed using software Meta-DiSc® version 1.4 (developed by Unit of Clinical Biostatistics, Ramón y Cajal Hospital, Madrid, Spain)[22]. Bivariate analysis was used to perform combined estimations of sensitivity, specificity and likelihood ratio with 95% confidence interval (CI) to estimate the summary results presented in the meta-analysis [23–25]. Summary measures were obtained using the DerSimonian and Laird’s random effects model (which considers the heterogeneity inherent to this type of systematic review) [26]. Heterogeneity was calculated using the chi-squared (X²) and Cochran’s Q test, inconsistency was assessed using I², and τ² was used to estimate variation between studies [23, 27].

Because of the heterogeneity that was found, sensitivity analysis was performed to identify associated cofactors (e.g., continent where the study was performed and the tumour primary site). Cofactors that were potentially associated with heterogeneity were also subjected to meta-regression [28].

To further investigate heterogeneity, diagnostic threshold analysis was performed based on the correlation (Spearman’s) between the logit of sensitivity and the logit of [1 –specificity] [22]. When a threshold effect occurs, the sensitivity and specificity of the investigated study exhibit negative correlation (or a positive correlation between sensitivity and [1 –specificity]. Therefore, a strong positive correlation between sensitivity and [1 –specificity] suggests the presence of a threshold effect [22].

When heterogeneity was observed, a summary receiver operating characteristic (SROC) curve was plotted. This method was appropriate given that the global sensitivity and specificity values were overestimated. In such cases, in addition to analysis of the ROC panel points, analysis of the SROC curve is also recommended [29, 30].

As is known, the various types of MRI techniques exhibit differences in threshold; therefore, to expand the information provided the study, the analysis was divided by MRI type (i.e., diffusion-weighted MRI [DW MRI], dynamic contrast-enhanced [DCE MRI] and traditional MRI).

The present study is registered at the PROSPERO database under number CRD42016043868.

Results and discussion

A total of 24 studies on the use of MRI as a diagnostic test for oral cancer were included in the systematic review, comprising 1,404 oral cancer lesions [31–54]. The study selection process is depicted in Fig 1. Search in databases retrieved 894 abstracts. Following assessment of titles and abstracts, the full text of 126 studies was analysed, and 24 were selected for meta-analysis.

[Figure omitted. See PDF.]

Fig 1. Study flow diagram.

https://doi.org/10.1371/journal.pone.0177462.g001

Nine studies performed DW MRI [32, 33, 36, 37, 41, 44, 49, 50, 52], seven used DCE MRI [31, 32, 39, 42, 43, 49, 53], and 13 applied traditional MR [33–35, 38, 40, 42, 45–48, 51, 52, 54]. It should be noted that some studies used more than one MRI type [32, 33, 42, 49, 52].

A full description of the studies, the standards and the tests used is presented in Table 1.

[Figure omitted. See PDF.]

Table 1. Characteristics of included studies–MRIs.

https://doi.org/10.1371/journal.pone.0177462.t001

Relative to DW MRI, the global sensitivity was 76.4% (95% CI: 67.3%–83.9%), and the global specificity was 91.3% (95% CI: 87.3%–94.4%). The global positive likelihood ratio was 8.0 (95% CI: 3.7–17.4), i.e., a positive DW MRI result increased by 8.0-fold the odds of an accurate diagnosis of patients who actually had oral cancer. Heterogeneity was found both with the Cochran’s Q test (p = 0.0067) and using τ² = 0.7296; inconsistency (I² = 62.2%) was moderate [27]. The global negative likelihood ratio was 0.3 (95% CI: 0.20–0.50). This result indicates the use of DW MRI given that it is close to zero (i.e., the odds of a false-positive result is only increased by a factor of 0.3). Cochran’s Q test did not detect heterogeneity, but a τ² of 0.1647 indicates heterogeneity; inconsistency (I² = 39.9%) was moderate [27]. Global DOR was 30.7 (95% CI: 12.7–74.3), i.e., the odds of a positive DW MRI result were 30.7-fold higher among individuals with oral cancer compared to those without disease. The area under the SROC was high [55] (AUC = 0.917; 95% CI: 0.915–0.918), and Q* was 0.85 for DW MRI for the diagnosis of oral cancer.

Relative to DCE MRI, the global sensitivity was 84.0% (95% CI: 76.2%–90.1%), and the global specificity was 89.5% (95% CI: 84.7%–93.2%). The global positive likelihood ratio was 7.2 (95% CI: 3.1–16.6); therefore, a positive DCE MRI result increased by 7.2-fold the odds of an accurate diagnosis of patients who actually had oral cancer. Heterogeneity was found both on Cochran’s Q test (p = 0.0002) and the τ² value (0.8916); inconsistency (I² = 77.1%) was high [27]. The global negative likelihood ratio was 0.2 (95% CI: 0.1–0.3). This result indicates the use of DCE MRI given that this value is close to zero. Specifically, the odds of a false-positive result are increased by only a factor of 0.22. Heterogeneity was not detected in Cochran’s Q test (0.2588) or by the τ² value (0.0951); inconsistency (I² = 22.4%) was low [27]. Global DOR was 48.1 (95% CI: 22.4–103.2), i.e., the odds of a positive DCE MRI result were 48.1-fold higher among individuals with oral cancer compared to those without the disease. The area under the SROC was high[55] (AUC = 0.936; 95% CI: 0.934–0.937), and Q* was 0.87 for DCE MRI for the diagnosis of oral cancer.

Regarding traditional MRI, the global sensitivity was 72.5% (95% CI: 66.4%–78.0%), and the global specificity 86.6% (95% CI: 84.0%–88.9%). The global positive likelihood ratio was 6.5 (95% CI: 3.8–11.0), i.e., a positive traditional MRI result increased by 6.5-fold the odds of an accurate diagnosis of patients who actually had oral cancer. Heterogeneity was found on Cochran’s Q test (p<0.001) and was confirmed by the τ² value (0.6480); and inconsistency (I² = 83.1%) was high[27]. The global positive likelihood ratio was 0.3 (95% CI: 0.2–0.4), a result favouring the use of traditional MRI given that the value was close to zero. Specifically, the odds of a false-positive result are increased by only 0.3-fold. Heterogeneity was detected using Cochran’s Q test (p = 0.0005) and was confirmed by the τ² value (0.2273); inconsistency (I² = 65.3%) was moderate[27]. The global DOR was 23.9 (95% CI: 13.2–43.3); therefore, the odds of a positive traditional MRI result were 23.9-fold higher among individuals with oral cancer compared to those without disease. The area under the SROC was high[55] (AUC = 0.894; 95% CI: 0.894–0.895), and Q* was 0.82 for traditional MRI for the diagnosis of oral cancer.

The data in Table 2 indicate that DCE MRI has the highest sensitivity, DOR and AUC and that DW MRI has the highest specificity. In turn, traditional MRI exhibited the lowest values for all of the parameters considered, in addition to a possible threshold effect (not confirmed by the ROC panel–Fig 2).

[Figure omitted. See PDF.]

Fig 2. Measures of diagnostic performance.

https://doi.org/10.1371/journal.pone.0177462.g002

[Figure omitted. See PDF.]

Table 2. Summary of diagnostic measures.

https://doi.org/10.1371/journal.pone.0177462.t002

The data in Table 2 suggest a threshold effect with traditional MRI, as shown by a positive correlation between the logit of sensitivity and [1 –specificity] (p = 0.027). However, visual inspection of the ROC panel does not indicate a threshold effect in any of the three types of MRI analysed (Fig 2).

According to meta-regression analysis (Table 3), the magnetic field strength might influence the accuracy (relative DOR [RDOR] = 0.32; 95% CI: 0.13–0.82; p = 0.0233) of traditional MRI. This finding may account for the heterogeneity exhibited by this MRI type. Regarding the other two MRI types, the results described in Table 3 do not indicate that the analysed cofactors are associated with heterogeneity.

[Figure omitted. See PDF.]

Table 3. Sensitivity analysis.

https://doi.org/10.1371/journal.pone.0177462.t003

Sensitivity analysis revealed a slight reduction in inconsistency in some subgroups (Table 3). For DW MRI, the heterogeneity of studies conducted in Europe and for tumours located in the parotid gland decreased (I² fell from 25.2% to 0%). As concerns DCE MRI, heterogeneity was maintained in the cases of salivary glands tumours and increased for tumours in other sites. Lastly, in regard to traditional MRI, heterogeneity decreased in the studies conducted in Asia (I² fell from 43.2% to 0%) and the Americas (I² decreased from 43.2% to 0%) and when the tumour was located in the salivary glands (I² decreased from 43.2% to 0%).

The methodological quality of the studies was very high, although some QUADAS-2 items scored negatively or the corresponding data were not presented by the authors.

The present is the first systematic review that assessed the accuracy of MRI for the diagnosis of oral cancer. The results show that the accuracy of the index test was good according to DOR (DW MRI– 30.7; DCE MRI– 48.1; traditional MRI– 23.9) and SROC-AUC (DW MRI– 0.91; DCE MRI– 0.93; traditional MRI– 0.89) analyses. In addition, MRI is non-invasive and can therefore be used as an auxiliary means for diagnosis and prevention of oral cancer, as well as for the planning of surgical interventions.

The 24 studies analysed in the present systematic review indicate the ability of MRI to diagnose malignant tumours with adequate sensitivity and specificity in patients with suspected oral cancer.

Our study extracted data from retrospective diagnostic accuracy studies. The methodological quality of the studies was very high, although some QUADAS-2 [20, 21] items scored negatively or the corresponding data were not presented by the authors (S1 Table). We performed bivariate analysis of sensitivity, specificity, likelihood ratio and DOR (with 95% CI); the latter considers intra- and inter-study variability. We applied the latest Cochrane guidelines for systematic reviews and diagnostic test accuracy studies [21–23, 25, 28].

MRI is the technique that provides the highest-quality images of soft tissues, without employing ionising radiation and without known biological risks [12, 56]. MRI also enables multiplanar studies (axial, coronal and sagittal) and provides images with high anatomical definition. The very low section thickness allows visualising structural lesions on the order of millimetres. Currently available clinical MRI devices use magnetic fields of 0.2 T to 3.0 T, and magnetic field strength is directly proportional to the definition of the obtained images and to the applicability of additional applications (e.g., software) [6].

Some researchers observed that the application of imaging tests to head and neck malignant neoplasms has substantially increased in recent years with the development of the modern, last-generation methods [11]. Diagnostic imaging techniques, such as computed tomography, ultrasound, MRI, bone scintigraphy and positron emission tomography (PET), must be used to complement the data obtained on clinical examination in terms of the tumour primary site, extension and invasion of adjacent structures [12].

Recently, DW MRI was suggested to be a highly sensitive method for assessment of the response to treatment, as well as for monitoring head and neck cancer [57]. Subsequently, several studies applied DW MRI as a diagnostic test for several brain conditions [58, 59]. In addition, this technique was applied in studies of the spinal cord and vertebral column [60], liver [61], kidneys and urinary tract [62].

Our study detected high heterogeneity among DW MRI results. A possible cause for such heterogeneity is variation in the magnetic field strength applied, which was 1.5 T in seven studies, 3 T in one study and unspecified in another. The reason for such heterogeneity that the magnetic field strength is directly related to the quality and wealth of detail of the obtained images. Padhani et al. [63] reported that accepted standards in DW MRI measurement or analysis methods have not been established. Moreover, current imaging analysis protocols suggested by different companies vary. This is the case even when considering the same manufacturer, which can alter its suggested methods following upgrades [61].

DCE MRI is a promising imaging technique for the assessment of microvascular parameters of tissue perfusion and consequently hypoxia [64]. This method is a non-invasive radiological technique that provides information on the microvascular environment of cancer lesions through the analysis of the contrast medium kinetics [65]. This test has potential for use in the detection and characterisation of tumours [66], as well as in the planning of treatment, the initial prediction of the response to treatment and assessment of treatment results [65]. We also found heterogeneity among results for this imaging modality, possibly due to variation in the cancer lesion sites assessed with DCE MRI. Specifically, the analysed region was the salivary glands in four studies, the palate in one and the tongue in another; the location was unspecified in one study.

In turn, traditional MRI has an outstanding role in the establishment of the origin, location and boundaries of lesions. Conventional radiographs have limitations, only being useful for screening lesions adjacent to mineralised tissue. This type of MRI does not employ ionising radiation and is one of the most indicated tests for the diagnosis of tumour lesions [67–73]. Considerable heterogeneity was also found among traditional MRI studies in the present meta-analysis, which might be accounted for by the various types of cancer assessed, different lesion sites and differences in calibration.

Conclusions

Providing integral care to users of public and private health services is a highly relevant goal, as are diagnostic confirmation in cases of oral cancer, pre-treatment assistance and control of the progression of disease once it is diagnosed. For these reasons, the present systematic review assessed the accuracy of MRI for the diagnosis of oral cancer, concluding this method is adequate for the intended purpose. The three types of MRI assessed (DW, DCE and traditional MRI) exhibited satisfactory accuracy compared to biopsy.

MRI is an adequate diagnostic option because it is available at public health services (in the case of Brazil); it can be used for monitoring patients as well as for obtaining a precise delimitation of lesions.

However, the variability among the currently available MRI methods has hindered their selection for clinical use; this bias was evident in the analysed studies given that the type of MRI used was not standardised.

Heterogeneity was found among the analysed studies, likely due to variation of the magnetic field strength used for traditional MRI as well as of the criteria for malignancy.

Early treatment and screening is important for oral cancer patients, and better health care results in improved survival rates, a better quality of life and the possibility of reducing the number of unnecessary biopsies. For these reasons and based on the results of the present study, we suggest the pilot use of MRI, as an auxiliary and complementary method to biopsy for pre-treatment and monitoring of oral cancer patients.

Supporting information

[Figure omitted. See PDF.]

S1 Fig. Full electronic search strategy–Pubmed.

https://doi.org/10.1371/journal.pone.0177462.s001

(DOCX)

S1 File. Prisma checklist.

https://doi.org/10.1371/journal.pone.0177462.s002

(DOC)

S1 Table. Results of the risk of bias assessment for each study, according to Quality Assessment of Diagnostic Accuracy Studies (QUADAS).

https://doi.org/10.1371/journal.pone.0177462.s003

(DOCX)

Author Contributions

1. Conceptualization: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

2. Data curation: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

3. Formal analysis: MAM IDSP PWS.

4. Funding acquisition: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

5. Investigation: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

6. Methodology: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

7. Project administration: IDSP PWS.

8. Resources: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

9. Software: PWS.

10. Supervision: PWS.

11. Validation: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

12. Visualization: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

13. Writing – original draft: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

14. Writing – review & editing: MAM LSL FRB AL EPX RAC FGFS CDT PDSP LBC IDSP PWS.

Citation: Moreira MA, Lessa LS, Bortoli FR, Lopes A, Xavier EP, Ceretta RA, et al. (2017) Meta-analysis of magnetic resonance imaging accuracy for diagnosis of oral cancer. PLoS ONE 12(5): e0177462. https://doi.org/10.1371/journal.pone.0177462

References

1. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int j cancer. 2010;127(12):2893–917. pmid:21351269

2. Warnakulasuriya S. Global epidemiology of oral and oropharyngeal cancer. Oral oncol. 2009;45(4–5):309–16. pmid:18804401

3. Mignogna MD, Fedele S, Lo Russo L. The World Cancer Report and the burden of oral cancer. Eur j cancer prev. 2004;13(2):139–42. pmid:15100581

4. Fontes KBFC, Cunha KSG, Rodrigues FR, Silva LE, Dias EP. Concordance between cytopathology and incisional biopsy in the diagnosis of oral squamous cell carcinoma. Braz oral res. 2013;27(2):122–7. pmid:23538424

5. Neville BW, Day TA. Oral cancer and precancerous lesions. Ca cancer j clin. 2002;52(4):195–215. pmid:12139232

6. Werning JW. Oral Cancer: Diagnosis, Management, and Rehabilitation. Stuttgart: Thieme; 2011.

7. Sciubba JJ. Oral cancer and its detection. History-taking and the diagnostic phase of management. J am dent assoc. 2001;133(4):12S–8S.

8. Eriksson AT, Corcuera MM, Trapero JC, Sanchez JC, Martinez AB. Analysis of new diagnostic methods in suspicious lesions of the oral mucosa. Med oral patol oral cir bucal. 2009;14(5):E210–6. pmid:19218907

9. Acha A, Ruesga MT, Rodríguez MJ, Martínez PMA, Aguirre JM. Applications of the oral scraped (exfoliative) cytology in oral cancer and precancer. Med oral patol oral cir bucal. 2005;10(2):95–102. pmid:15735540

10. Ram S, . Chemiluminescence as a diagnostic aid in the detection of oral cancer and potentially malignant epithelial lesions. Int j oral maxillofac surg. 2005;34(5):521–7. pmid:16053872

11. Prehn RB, Pasic TR, Harari PM, Brown WD, Ford CN. Influence of computed tomography on pretherapeutic tumor staging in head and neck cancer patients. Otolaryngol head neck surg. 1998;119(6):628–33. pmid:9852538

12. Lenz M, Greess H, Baum U, Dobritz M, Kersting-Sommerhoff B. Oropharynx, oral cavity, floor of the mouth: CT and MRI. Eur j radiol. 2000;33(3):203–15. pmid:10699737

13. Weber AL, Romo L, Hashmi S. Malignant tumors of the oral cavity and oropharynx: clinical, pathologic, and radiologic evaluation. Neuroimaging clin n am. 2003;13(3):443–64. pmid:14631684

14. Chikui T, Yuasa K, Inagaki M, Ohishi M, Shirasuna K, Kanda S. Tumor recurrence criteria for postoperative contrast-enhanced computed tomography after surgical treatment of oral cancer and flap repair. Oral surg oral med oral pathol oral radiol endod. 2000;90(3):369–76. pmid:10982961

15. Hayashi T, Tanaka R, Taira S, Koyama J-i, Katsura K, Kobayashi F. Non-contrast-enhanced CT findings of high attenuation within metastatic cervical lymph nodes in patients with stage I or II tongue carcinoma during a follow-up period. Ajnr am j neuroradiology. 2003;24(7):1330–3.

16. King AD, Tse GM, Yuen EH, To EW, Vlantis AC, Zee B, et al. Comparison of CT and MR imaging for the detection of extranodal neoplastic spread in metastatic neck nodes. Eur j radiol. 2004;52(3):264–70. pmid:15544904

17. Sun J, Li B, Li CJ, Li Y, Su F, Gao QH, et al. Computed tomography versus magnetic resonance imaging for diagnosing cervical lymph node metastasis of head and neck cancer: a systematic review and meta-analysis. Onco targets ther. 2015;8:1291–313. pmid:26089682

18. Li C, Yang W, Men Y, Wu F, Pan J, Li L. Magnetic Resonance Imaging for Diagnosis of Mandibular Involvement from Head and Neck Cancers: A Systematic Review and Meta-Analysis. PLoS ONE. 2014;9(11):e112267. pmid:25397614

19. Schueler S, Schuetz GM, Dewey M. The revised QUADAS-2 tool. Ann intern med. 2012;156(4):323. pmid:22351721

20. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann intern med. 2011;155(8):529–36. pmid:22007046

21. Reitsma H, Rutjes A, Whiting P, Vlassov V, Leeflang M, Deeks J. Assessing Methodological Quality. Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy The Cochrane Collaboration; 2009.

22. Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc: a software for meta-analysis of test accuracy data. Bmc med res methodol. 2006;6:31. pmid:16836745

23. Leeflang MM, Deeks JJ, Gatsonis C, Bossuyt PM. Systematic reviews of diagnostic test accuracy. Ann intern med. 2001;149(12):889–97.

24. Irwig L, Tosteson AN, Gatsonis C, Lau J, Colditz G, Chalmers TC, et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann intern med. 1994;120(8):667–76. pmid:8135452

25. Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J clin epidemiol. 2005;58(10):982–90. pmid:16168343

26. Der Simonian R, Laird N. Meta-analysis in clinical trials. Control clin trials. 1986;7(3):177–88. pmid:3802833

27. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557–60. pmid:12958120

28. Macaskill P, Gatsonis C, Deeks J, Harbord R, Takwoingi Y. Cochrane handbook for systematic reviews of diagnostic test accuracy. London: The Cochrane Collaboration; 2010.

29. Deeks JJ, Morris JM. Evaluating diagnostic tests. Baillieres clin obstet gynaecol. 1996;(10):613–30.

30. Knottnerus JA, Muris JW. Assessment of the accuracy of diagnostic tests: the cross-sectional study. J clin epidemiol. 2003;56(11):1118–28. pmid:14615003

31. Aghaghazvini L, Salahshour F, Yazdani N, Sharifian H, Kooraki S, Pakravan M, et al. Dynamic contrast-enhanced MRI for differentiation of major salivary glands neoplasms, a 3-T MRI study. Dentomaxillo fac radiol. 2015;44(2):20140166.

32. Ai S, Zhu W, Liu Y, Wang P, Yu Q, Dai K. Combined DCE- and DW-MRI in diagnosis of benign and malignant tumors of the tongue. Front biosci. 2013;18:1098–111.

33. Alibek S, Zenk J, Bozzato A, Lell M, Grunewald M, Anders K, et al. The value of dynamic MRI studies in parotid tumors. Acad radiol. 2007;14(6):701–10. pmid:17502260

34. Bartels S, Talbot JM, DiTomasso J, Everts EC, Andersen PE, Wax MK, et al. The relative value of fine-needle aspiration and imaging in the preoperative evaluation of parotid masses. Head neck. 2000;22(8):781–6. pmid:11084638

35. Christe A, Waldherr C, Hallett R, Zbaeren P, Thoeny H. MR imaging of parotid tumors: typical lesion characteristics in MR imaging improve discrimination between benign and malignant disease. Am j neuroradiol. 2011;32(7):1202–7. pmid:21724574

36. Eida S, Sumi M, Sakihama N, Takahashi H, Nakamura T. Apparent diffusion coefficient mapping of salivary gland tumors: prediction of the benignancy and malignancy. Am j neuroradiol. 2007;28(1):116–21. pmid:17213436

37. İnci E, Hocaoglu E, Kiliçkesmez Ö, Aydin S, Cimilli T. Quantitative diffusion-weighted MR imaging in the differential diagnosis of parotid gland tumors: is it a useful technique? Turkiye klinikleri j med sci. 2010;30(4):1339–45.

38. Fassnacht W, Schmitz S, Weynand B, Marbaix E, Duprez T, Hamoir M. Pitfalls in preoperative work-up of parotid gland tumours: 10-year series. B-ent. 2013;9(2):83–8. pmid:23909113

39. Hisatomi M, Asaumi J, Yanagi Y, Unetsubo T, Maki Y, Murakami J, et al. Diagnostic value of dynamic contrast-enhanced MRI in the salivary gland tumors. Oral oncol. 2007;43(9):940–7. pmid:17257881

40. Inohara H, Akahani S, Yamamoto Y, Hattori K, Tomiyama Y, Tomita Y, et al. The role of fine-needle aspiration cytology and magnetic resonance imaging in the management of parotid mass lesions. Acta otolaryngol. 2008;128(10):1152–8. pmid:18607904

41. Kato H, Kanematsu M, Watanabe H, Kajita K, Mizuta K, Aoki M, et al. Perfusion imaging of parotid gland tumours: usefulness of arterial spin labeling for differentiating Warthin's tumours. Eur radiol. 2015;25(11):3247–54. pmid:25921591

42. Lam PD, Kuribayashi A, Imaizumi A, Sakamoto J, Sumi Y, Yoshino N, et al. Differentiating benign and malignant salivary gland tumours: diagnostic criteria and the accuracy of dynamic contrast-enhanced MRI with high temporal resolution. Br j radiol. 2015;88(1049):20140685. pmid:25791568

43. Matsuzaki H, Yanagi Y, Hara M, Katase N, Hisatomi M, Unetsubo T, et al. Diagnostic value of dynamic contrast-enhanced MRI for submucosal palatal tumors. Eur j radiol. 2012;81(11):3306–12. pmid:22578835

44. Motoori K, Ueda T, Uchida Y, Chazono H, Suzuki H, Ito H. Identification of Warthin tumor: magnetic resonance imaging versus salivary scintigraphy with technetium-99m pertechnetate. J comput assist tomogr. 2005;29(4):506–12. pmid:16012309

45. Paris J, Facon F, Pascal T, Chrestian MA, Moulin G, Zanaret M. Preoperative diagnostic values of fine-needle cytology and MRI in parotid gland tumors. Eur arch otorhinolaryngol. 2005;262(1):27–31. pmid:14727124

46. Prades JM, Oletski A, Faye MB, Dumollard JM, Timoshenko AP, Veyret C, et al. [Parotid gland masses: diagnostic value of MR imaging with histopathologic correlations]. Morphologie. 2007;91(292):44–51. pmid:17652005

47. Rudack C, Jörg S, Kloska S, Stoll W, Thiede O. Neither MRI, CT nor US is superior to diagnose tumors in the salivary glands–an extended case study. Head face med. 2007;3(1):1.

48. Sakamoto M, Iikubo M, Kojima I, Sasano T, Mugikura S, Murata T, et al. Diagnostic value of capsule-like rim enhancement on magnetic resonance imaging for distinguishing malignant from benign parotid tumours. Int j oral maxillofac surg. 2014;43(9):1035–41. pmid:24703493

49. Sumi M, Nakamura T. Head and neck tumours: combined MRI assessment based on IVIM and TIC analyses for the differentiation of tumors of different histological types. Eur radiol. 2014;24(1):223–31. pmid:24013848

50. Sumi M, Van Cauteren M, Sumi T, Obara M, Ichikawa Y, Nakamura T. Salivary gland tumors: use of intravoxel incoherent motion MR imaging for assessment of diffusion and perfusion for the differentiation of benign from malignant tumors. Radiology. 2012;263(3):770–7. pmid:22447854

51. Takashima K, Tateda K, Matsumoto T, Iizawa Y, Nakao M, Yamaguchi K. Role of tumor necrosis factor alpha in pathogenesis of pneumococcal pneumonia in mice. Infect immun. 1997;65(1):257–60. pmid:8975920

52. Turner F, Cohen F, Varoquaux A, Giovanni A, Duflo S, Zanaret M. [Role of diffusion weighted MR in the diagnosis of parotid tumours]. Rev laryngol otol rhinol. 2008;129(3):175–80.

53. Yabuuchi H, Fukuya T, Tajima T, Hachitanda Y, Tomita K, Koga M. Salivary gland tumors: diagnostic value of gadolinium-enhanced dynamic MR imaging with histopathologic correlation. Radiology. 2003;226(2):345–54. pmid:12563124

54. Yerli H, Aydin E, Haberal N, Harman A, Kaskati T, Alibek S. Diagnosing common parotid tumours with magnetic resonance imaging including diffusion-weighted imaging vs fine-needle aspiration cytology: a comparative study. Dentomaxillofac radiol. 2010;39(6):349–55. pmid:20729184

55. Swets JA. Measuring the accuracy of diagnostic systems. Science. 1988;240(4857):1285–93. pmid:3287615

56. Langlais RP, van Rensburg LJ, Guidry J, Moore WS, Miles DA, Nortje CJ. Magnetic resonance imaging in dentistry. Dent clin north am. 2000;44(2):411–26. pmid:10740776

57. Kim S, Loevner L, Quon H, Sherman E, Weinstein G, Kilger A, et al. Diffusion-weighted magnetic resonance imaging for predicting and detecting early response to chemoradiation therapy of squamous cell carcinomas of the head and neck. Clin cancer res. 2009;15(3):986–94. pmid:19188170

58. Okamoto K, Ito J, Ishikawa K, Sakai K, Tokiguchi S. Diffusion-weighted echo-planar MR imaging in differential diagnosis of brain tumors and tumor-like conditions. Eur radiol. 2000;10(8):1342–50. pmid:10939505

59. Tsuruda JS, Chew WM, Moseley ME, Norman D. Diffusion-weighted MR imaging of the brain: value of differentiating between extraaxial cysts and epidermoid tumors. Am j roentgenol. 1990;155(5):1059–65; discussion 66–8.

60. Bammer R, Fazekas F. Diffusion imaging of the human spinal cord and the vertebral column. Top magn reson imaging. 2003;14(6):461–76. pmid:14872166

61. Vandecaveye V, De Keyzer F, Dymarkowski S. Perfusion-and diffusion-weighted imaging of hepatocellular carcinoma. Jbr-btr. 2007;90(6):492–6. pmid:18376763

62. Kim S, Naik M, Sigmund E, Taouli B. Diffusion-weighted MR imaging of the kidneys and the urinary tract. Magn reson imaging clin n am. 2008;16(4):585–96, vii-viii. pmid:18926424

63. Padhani AR, Liu G, Mu-Koh D, Chenevert TL, Thoeny HC, Takahara T, et al. Diffusion-Weighted Magnetic Resonance Imaging as a Cancer Biomarker: Consensus and Recommendations. Neoplasia. 2009;11(2):102–25. pmid:19186405

64. Bernstein JM, Homer JJ, West CM. Dynamic contrast-enhanced magnetic resonance imaging biomarkers in head and neck cancer: potential to guide treatment? A systematic review. Oral oncol. 2014;50(10):963–70. pmid:25116700

65. Lee FK, King AD, Ma BB, Yeung DK. Dynamic contrast enhancement magnetic resonance imaging (DCE-MRI) for differential diagnosis in head and neck cancers. Eur j radiol. 2012;81(4):784–8. pmid:21376492

66. Yuan J, Chow SK, Yeung DK, King AD. A five-colour colour-coded mapping method for DCE-MRI analysis of head and neck tumours. Clin radiol. 2012;67(3):216–23. pmid:21939962

67. Lloyd GA, Phelps PD. The demonstration of tumours of the parapharyngeal space by magnetic resonance imaging. Br j radiol. 1986;59(703):675–83. pmid:3015313

68. Som PM, Shugar JM, Sacher M, Stollman AL, Biller HF. Benign and malignant parotid pleomorphic adenomas: CT and MR studies. J comput assist tomogr. 1988;12(1):65–9. pmid:2826556

69. Som PM. Salivary Glands In: Som PM, Bergeron RT Head and Neck Imaging. 2 ed. ed. St. Louis: Mosby; 1991. p. 277–348.

70. Austin JR, Crockett DM. Pleomorphic adenoma of the palate in a child. Head neck. 1992;14(1):58–61. pmid:1320597

71. Silvers AR, Som PM. Salivary glands. Radiol clin north am. 1998;36(5):941–66, vi. pmid:9747195

72. Suenaga S, Indo H, Noikura T. Diagnostic value of dynamic magnetic resonance imaging for salivary gland diseases: a preliminary study. Dento-maxillo-facial radiol. 2001;30(6):314–8.

73. Ishii J, Nagasawa H, Wadamori T, Yamashiro M, Ishikawa H, Yamada T, et al. Ultrasonography in the diagnosis of palatal tumors. Oral surg oral med oral pathol oral radiol endod. 1999;87(1):39–43. pmid:9927078