Undetectability of oral tongue cancer on magnetic resonance imaging; clinical significance as a predictor to avoid unnecessary elective neck dissection in node negative patients.

Methods:: We retrospectively reviewed early stage oral tongue cancer patients treated with radical surgery with clinically N0, between May 2009 and February 2016. Collected data include age, sex, pathological DOI, DOI on MRI, locoregional control rate, disease-free survival rate, and overall survival rate. These data were statistically compared between the detectable lesion (DL) group and undetectable lesion (UL) group on MRI. Interobserver agreement in evaluation of detectability of the oral tongue cancer was assessed by k statistics.

Results:: Total of 53 patients were studied, and 28 were DLs and 25 ULs. Pathological DOI in UL was significantly smaller than that of DL (average 1.7 vs 4.6 mm, p < 0.001). Cut-off value between UL group and DL group was 3.5 mm (sensitivity 96 %, specificity 75 %). 96 % of ULs had pathological DOI smaller than 4 mm, the recommended cut-off value for neck dissection. There was no significant difference in locoregional control rate (p = 0.24), disease-free survival rate (p = 0.24) or overall survival rate (p = 0.92). Interobserver agreement in evaluation of detectability on MRI was very good ( k-value = 0.89, p < 0.001).

Conclusions:: When oral tongue cancer is not detected on MRI, it indicates pathological DOI being smaller than 4 mm, which may imply that elective neck dissection is unnecessary.