Construction and validation of anomogram prediction model for chemotherapy combined with trastuzumab induced cardiotoxicity in breast cancer
Objective To explore the influencing factors of cardiotoxicity in breast cancer patients treated with chemotherapy combined with trastuzumab,and to establish and compare the performance of four predictive models using machine learning.Methods A total of 1 030 breast cancer patients treated in our hospital were selected for the study objects and randomly divided into the model-building group(n=721)and the validation group(n=309)in a 7:3 ratio.Independent risk factors for cardiotoxicity were identified using Least absolute shrinkage and selection operator(LASSO)regression and Logistic regression,and four predictive models were established(Logistic regression,extreme gradient boosting,Bayesian networks and random forest).The models'predictive effectiveness and clinical application value were assessed using the receiver operating characteristic curve(ROC)and the area under the curve(AUC),calibration curves,and decision curves.Results In the model-building group,age>60 years,hypertension,smoking,and anthracycline-containing chemotherapy regimens were identified as independent risk factors for cardiotoxicity.Among the four models constructed,Logistic regression performed the best and a nomogram was further drawn.The AUC for the model-building group and the validation group were 0.781(95%CI:0.713～0.848)and 0.805(90%CI:0.715～0.896),respectively.Calibration curves showed good consistency between predicted and actual results in both groups.Decision curves indicated that Logistic regression model had good clinical utility.Conclusion The nomogram constructed in this study can accurately predict the individualized risk of cardiotoxicity in breast cancer patients treated with chemotherapy combined with trastuzumab.
Breast cancerTrastuzumabChemotherapyCardiotoxicityNomogram
万方数据
维普
