ABSDELL Model: Development and Internal Validation of a Risk Prediction Model of LVEF Decline in Breast Cancer Patients Treated With Trastuzumab

Abstract

Introduction/BackgroundThis study aims to establish an integrated model for predicting trastuzumab-associated decline of Left ventricular ejection fraction (LVEF) during drug administration. MethodsA retrospective study of 212 women who diagnosed with HER2-positive breast cancer and treated with chemotherapy and trastuzumab was conducted. Medical records were collected from 6 months before staring trastuzumab to 3 years afterwards. The least absolute shrinkage and selection operator (LASSO) regression analysis was used to select variables, time-dependent receiver operating characteristic (ROC) curve and calibration plots were used to evaluate the model. The adjusted C-index and Brier scores were calculated using a bootstrap internal validation procedure. ResultsThe median age of participants is 53.2 years old. The median length of follow-up was 336 days. There were 72 patients (33.96%) whose LVEF declined ≥ 10% (10 absolute percent points). Seven factors, namely age, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), mitral peak E-wave velocity (E-wave), left ventricular end-systolic diameter (LVESD) and LVEF, were selected. The name of the ABSDELL model was formed by the initials of each predictor. The area under the curve (AUC) of the model was 0.802 in 1 year and 0.881 in 3 years. Calibration plots indicate the predicted and actual probabilities were highly consistent. In the internal validation, 1-year and 3-year adjusted C-index was 0.801 and 0.881, and adjusted Brier score was 0.118 and 0.091, separately. ConclusionThe ABSDELL model can effectively predicts the probability of LVEF decline in breast cancer patients treated with trastuzumab.