Application of fuzzy inference to European patients to predict cervical lymph node metastasis in carcinoma of the tongue.

In head and neck cancers, the presence of cervical lymph node metastasis is an important determinant of outcome. Many attempts have been made to predict cervical lymph node metastasis, but the accuracy of currently available techniques remains inadequate. We used fuzzy inference to predict cervical lymph node metastasis retrospectively in 75 patients with squamous cell carcinoma of the tongue and prospectively in 23 patients. Our model was based on three variables: tumor size, keratinization, and mode of invasion. The accuracy of fuzzy inference for the prediction of cervical lymph node metastasis in the 75 patients studied retrospectively was 86.7%, the sensitivity was 70.8%, and the specificity was 94.1%. In the 23 patients studied prospectively, the accuracy was 91.3%, the sensitivity was 50.0%, and the specificity was 95.2%. The accuracy obtained in this European series of patients was similar to that previously obtained in Japanese patients. We conclude that fuzzy inference may be a useful method for predicting cervical lymph node metastasis. Its high specificity is likely to reduce the number of unnecessary neck dissections. However, the current level sensitivity is inadequate for routine clinical use. Therefore, other predictors of lymph node metastasis should be identified to refine the current model.