Decision Tree Model for Predicting the Overall Survival of Patients with Diffused Large B-Cell Lymphoma in the Central Nervous System.

OBJECTIVE

To identify the significant predictors of overall survival for patients living with diffused large B-cell lymphoma (DLBCL) in the central nervous system and establish a novel decision tree model to help predict survival status at several time points.

METHODS

Patients diagnosed with DLBCL were identified from the SEER database and randomly divided into training and test samples (6:4). Dichotomous decision trees were developed for survival status at 3, 12, 24, and 60 months. Sensitivity, specificity, positive predictive value, negative predictive value, accuracy rate, and area under the receiver operating characteristic curve were calculated to evaluate the model performance.

RESULTS

A total of 2998 patients were included, with 1799 and 1199 patients divided into the training and testing groups. Decision trees for 3, 12, 24, and 60 months survival status were generated. Chemotherapy and patient's age were of the primary importance for prognosis in the novel models. Favorable consistency between the predicted and actual survival status was presented. The accuracy rates were 0.79, 0.71, 0.68, and 0.86 for training sample at 3, 12, 24, and 60 months, respectively, and 0.75, 0.69, 0.58, and 0.84 for test sample at 3, 12, 24, and 60 months, respectively. The area under the receiver operating characteristic curve values ranged between 0.645 and 0.721 for the training sample and between 0.607 and 0.712 for the test sample.

CONCLUSIONS

Novel decision tree models were established for predicting the 3, 12, 24, and 60 months survival status of patients with DLBCL. The newly developed models were verified using training and test samples, showing favorable accuracy and predictive value on overall survival.