Magnetic Resonance Imaging for Diagnosis of Mandibular Involvement from Head and Neck Cancers: A Systematic Review and Meta-Analysis

Background: Diagnosis of mandibular involvement caused by head and neck cancers is critical for treatment. We performed a meta-analysis to determine the diagnostic efficacy of MR for distinguishing mandibular involvement caused by head and neck cancers. Methods: Thirteen databases were searched electronically and hand-searching was also done. Two reviewers conducted study inclusion, data extractions, and quality assessment of the studies independently. Meta-disc 1.4 and STATA 11.0 were used to conduct the meta-analysis. Results: 16 studies involving a total of 490 participants underwent MR examinations and were accounted for in this meta-analysis. Among the included studies, 2 had high risk of bias, while the rest had unclear risk of bias. Meta-regression showed that the slight clinical and methodological heterogeneities did not influence the outcome (P>0.05). Meta-analysis indicated that the MR for the diagnosis of mandibular involvement had a pooled sensitivity (SEN) of 78%, specificity (SPE) of 83%, positive likelihood ratio (+LR) of 3.80, negative likelihood ratio (-LR) of 0.28, diagnostic odds ratio (DOR) of 28.94, area under curve (AUC) of 0.9110, and Q* of 0.8432. Two studies detected the diagnostic efficacy of MR for the mandibular medullar invasion, and only one study reported the inferior alveolar canal invasion, which made it impossible to include it in our meta-analysis. In comparing to CT, MR had a higher SEN without statistical significance (P = 0.08), but a significantly lower SPE (P = 0.04). The synthesized diagnostic efficacy (AUC and Q*) on mandibular involvement was similar between the two modalities (P>0.05). Conclusions: Present clinical evidence showed that MR had an acceptable diagnostic value in detecting mandibular involvement caused by head and neck cancers. MR exceeded CT in diagnosing patients with mandibular invasion (higher sensitivity than CT) but was less efficacious to exclude patients without the mandibular invasion (lower specificity than CT).