Prediction of prognosis in breast cancer patients using lapatinib based on real-world evidence
AIM Based on real world data and machine learning technology,a predictive model of progression free survival(PFS)of patients with breast cancer treated with lapatinib was constructed.METHODS A retrospective collection of 150 patients admitted to the Fudan University Shanghai Cancer Center from July 2016 to June 2017 was conducted.The outcome indicator of the prediction model was whether the patient's PFS was≤1 year.Using sequential forward selection algorithms for feature selection,and comparing the predictive performance of 9 algorithms for building models,including extreme gradient boost(XGBoost),classification boost(CatBoost),random forest(RF),light gradient boost(LightGBM),gradient boost decision tree(GBDT),logistic regression(LR),support vector regression(SVR),artificial neural network(ANN),and TabNet.RESULTS Important variables included medication regimen,age,frequency of chemotherapy,anthracycline drugs,platinum drugs,estrogen receptor,disease stage,and number of metastatic sites.The XGBoost model had the best prediction performance,with a prediction accuracy of 93％and a recall rate of 87％for PFS ≤ 1 year.And a prediction accuracy was 71％,and a recall rate was 83％for PFS＞1 year.CONCLUSION The performance and robustness of the prognosis prediction model for patients with breast cancer treated with lapatinib established are good,which can provide a better auxiliary decision-making basis for clinical treatment of breast cancer.
lapatinibbreast neoplasmsmachine learningreal world study
万方数据
维普
