A novel algorithm for lymph node status prediction of oral cancer before surgery.

An algorithm was developed to predict lymph node metastasis in patients with oral squamous cell carcinoma. To improve the diagnostic accuracy of lymph node status, we suggest a criterion for the prediction of lymph node status by combining clinical and molecular information. To develop a model to predict lymph node metastasis before surgery, principal component analysis (PCA) was applied. Patients were separated into a training set (n=60) and a test set (n=30), and four variables were evaluated: clinical T stage, clinical N stage, insulin-like growth factor II mRNA-binding protein 3 (IMP3), and cyclooxygenase-2 (COX2). PCA was applied to the training set to determine the weight of each variable. P score (patient score) was generated for each patient in the training set, and a cutoff defining lymph node metastasis was established by the classification accuracy. The model was then used to classify the test set and the classification accuracy, sensitivity, and specificity of the test set were calculated. For more reliable results, we conducted the whole process using re-sampled 100 training and test sets. The performance of the prediction model including IMP3 was better than the model using only clinical variables. However, COX2 was not influential on the test results. The prediction accuracy of the implemented model was 0.84 in the training set and 0.80 in the test set. The sensitivity and specificity in the test set were 0.70 and 0.88, respectively. Although IMP3 was not significantly related with lymph node status independently, it did improve the accuracy of the test when combined with clinical factors. The model was 84% accurate in predicting lymph node status. The accurate prediction of the presence of lymph node metastasis, before surgery, may help in deciding the need for surgical lymph node dissection or additional preoperative treatment modalities, which might improve survival.