Prediction of lymph node metastasis in T1/T2 tongue squamous cell carcinoma.

OBJECTIVES: In T1, T2, and clinically NO squamous cell carcinoma of the tongue, there is no reliable predictive variable to determine whether or not neck dissection is needed. Thus, we established a predictive score model based on tumour depth and other pathological variables. METHODS: We retrospectively reviewed 115 patients with T1 and T2 stage squamous cell carcinoma of the tongue. Their pathological variables were used to construct a score model for predicting the risk of cervical lymph node metastasis. RESULTS: A predictive score model was proposed using multivariate logistic regression analysis: Score = (2.694 x tumour depth (cm)) + (1.814 x lymphovascular invasion (yes = 1, no = 0)) + (1.175 x perineural invasion (yes = 1, no = 0)). The cutoff point was set at 2.7427. This predictive score model has a sensitivity of 91.2% and specificity of 65.4%. CONCLUSION: A predictive score model was built and a two-stage surgical approach was suggested for T1 and T2 squamous cell carcinoma of the tongue.