Abstract
BACKGROUND: The current NCCN guidelines recommend considering elective neck dissection (END) for early-stage oral cavity squamous cell carcinoma (OCSCC) with a depth of invasion (DOI) exceeding 3 mm. However, this DOI threshold, determined by evaluating the occult lymph node metastatic rate, lacks robust supporting evidence regarding its impact on patient outcomes. In this nationwide study, we sought to explore the specific indications for END in patients diagnosed with OCSCC at stage cT2N0M0, as defined by the AJCC Eighth Edition staging criteria. METHODS: We examined 4723 patients with cT2N0M0 OCSCC, of which 3744 underwent END and 979 were monitored through neck observation (NO). RESULTS: Patients who underwent END had better 5-year outcomes compared to those in the NO group. The END group had higher rates of neck control (95% vs. 84%, p < 0.0001), disease-specific survival (DSS; 87% vs. 84%, p = 0.0259), and overall survival (OS; 79% vs. 73%, p = 0.0002). Multivariable analysis identified NO, DOI ≥5.0 mm, and moderate-to-poor tumor differentiation as independent risk factors for 5-year neck control, DSS, and OS. Based on these prognostic variables, three distinct outcome subgroups were identified within the NO group. These included a low-risk subgroup (DOI <5 mm plus well-differentiated tumor), an intermediate-risk subgroup (DOI ≥5.0 mm or moderately differentiated tumor), and a high-risk subgroup (poorly differentiated tumor or DOI ≥5.0 mm plus moderately differentiated tumor). Notably, the 5-year survival outcomes (neck control/DSS/OS) for the low-risk subgroup within the NO group (97%/95%/85%, n = 251) were not inferior to those of the END group (95%/87%/79%). CONCLUSIONS: By implementing risk stratification within the NO group, we found that 26% (251/979) of low-risk patients achieved outcomes similar to those in the END group. Therefore, when making decisions regarding the implementation of END in patients with cT2N0M0 OCSCC, factors such as DOI and tumor differentiation should be taken into account.