The role of MRI-derived depth of invasion in staging oral tongue squamous cell carcinoma: inter-reader and radiological-pathological agreement.

Abstract

Background The 8th edition of tumor, node, metastasis (TNM) classification incorporates depth of invasion evaluation in the staging of oral cavity squamous cell carcinoma, since it is a predictor of nodal metastasis and an independent prognostic factor. Although the histopathological definition of depth of invasion is clear, an accurate method for its radiological assessment has not yet been validated. Purpose To investigate the role of MRI-derived depth of invasion evaluation in staging oral tongue squamous cell carcinoma and to assess the inter-reader agreement and the radiological–pathological correlation. Material and Methods We retrospectively reviewed 43 patients with oral tongue squamous cell carcinoma who underwent preoperative MRI. The MRI-derived depth of invasion was measured by two radiologists, each with a different degree of experience in head and neck imaging. The pathological depth of invasion was recorded from histopathological reports. The inter-reader and the radiological–histopathological correlations for the depth of invasion were evaluated with Bland–Altman plots, the intraclass correlation coefficients (ICC), and the paired samples test; agreements for T staging were assessed using the Kappa coefficient. Results Inter-reader reliability was excellent for the MRI-derived depth of invasion (ICC = 0.91), very good between MRI-derived depth of invasion and pathological depth of invasion (ICC = 0.89 for the experienced reader, 0.86 for the inexperienced reader). Both readers reached a good agreement regarding T staging (kappa value = 0.70). Furthermore, the agreement between radiological and pathological T staging was good (kappa value = 0.74 for the experienced reader, 0.60 for the inexperienced reader). Conclusion MRI-derived depth of invasion should be measured in the pretreatment assessment of oral tongue squamous cell carcinoma as it has an excellent inter-reader reliability and nearly excellent radiological–pathological correlation.