Chemotherapy-Induced Nausea and Vomiting in Patients With Breast Cancer: Risk Factor and Predictive Model Using Classification and Regression Tree (CART).

Abstract

Introduction Chemotherapy-induced nausea and vomiting (CINV) is a common and debilitating adverse effect of breast cancer chemotherapy. The incidence of CINV in the first cycle of chemotherapy is essential, as it sets the tone for anticipatory CINV and the overall patients' treatment experience. We aimed to investigate the risk factors of first cycle CINV in breast cancer patients and to develop a classification and regression tree (CART) model to predict its occurrence. Methods This is a cross-sectional study that nested in a prospective cohort. One hundred and thirty-seven female breast cancer patients receiving highly emetogenic chemotherapy were included. We used the Common Toxicity Criteria for Adverse Events (CTCAE) version 4.0 to assess patient-reported nausea and vomiting in the first chemotherapy cycle. The proportional difference of CINV between sociodemographic and clinicopathologic variables was analyzed using chi-square, and the strength and direction of the relationship with CINV were analyzed using bivariable logistic regression analysis. Multivariable logistic regression and CART analysis included variables with a p-value <0.250. Results The incidence of first-cycle CINV was 43.1%. The chi-square test revealed a significant association between insurance status and CINV (p<0.001) and between the stage at diagnosis and CINV (p<0.001). Underweight to normal body mass index (BMI) patients are significantly associated with an increased risk of first-cycle CINV (OR =2.17, 95% CI 1.03-4.56, p =0.041). In hierarchical order, three variables (stage at diagnosis, BMI, and age) were included in the CART model, which significantly influenced the probability of first cycle CINV. With an accuracy of 61.3%, the CART model had a sensitivity of 28.8%, a specificity of 85.9%, a positive predictive value of 60.7%, a negative predictive value of 61.5%, and an area under curve (AUC) of 0.602. Conclusion Breast cancer patients with an underweight to normal BMI have a higher risk of developing first-cycle CINV. Our CART model was better at identifying patients who would not develop CINV than those who would. The CART model may provide a simple and effective way to individualize patient care for first-cycle CINV.