Risk prediction model for severe potential drug-drug interactions in colorectal cancer patients: a real-world data study

Objective: The potential drug-drug interactions (pDDIs) seriously affecting the prognosis of colorectal cancer (CRC) patients. This study aimed to identify the risk factors of pDDIs in hospitalized CRC patients and construct a risk prediction model to provide a reference for clinical rational drug use. Research design and methods: A retrospective cohort study was conducted, enrolling 2,868 patients from a tertiary hospital. Medscape was used to assess pDDIs, and a risk prediction model was constructed based on independent risk factors. Results: A total of 1,790 (62.41%) patients experienced at least one pDDIs, with 1,458 (50.84%) cases of severe pDDIs. The number of drug varieties, hypoalbuminemia, and treatment were independent risk factors. The area under the receiver operating characteristic curve (AUC) of the model in the training and validation sets were 0.826 and 0.824, respectively. The calibration curve showed a good agreement between the predicted probability and the actual occurrence probability. Decision curve analysis (DCA) demonstrated that the model had a positive net clinical benefit within a wide range of 10%-90%. Conclusion: The constructed model has good predictive performance and can be used to identify high-risk patients with pDDIs in clinical practice, thereby improving the safety of drug use.