Regency EF
A Machine-Based Approach to Preoperatively Identify Patients with the Most and Least Benefit Associated with Resection for Intrahepatic Cholangiocarcinoma (304091)
Madison Hyer, The Ohio State University James Cancer Center*Rittal Mehta, The Ohio State University James Cancer Center
Timothy Pawlik, The Ohio State University James Cancer Center
Kota Sahara, The Ohio State University
Diamantis Tsilimigras, The Ohio State University James Cancer Center
Keywords: machine learning; survival; prognostic factors; CART
While a relatively rare cancer, the incidence of intrahepatic cholangiocarcinoma (ICC) has increased 3-fold over the last three decades. The majority of patients recur within 2 year of surgery. As such, patient selection and risk stratification to identify patients who may be the optimal candidates for surgery has particular importance for patients with ICC. Our study utilized an international Multi-institutional database to identify preoperative factors that matter the most in terms of survival following surgery for ICC. A machine-based survival decision tree (CART model) was used to generate homogeneous groups of patients relative to overall survival based on preoperative factors. Our poster highlights the use of machine learning to better understand the prognostic weight and the hierarchical association of factors relative to survival. The CART model was able to risk stratify patients into 4 groups with a 5-year OS ranging from 60.5% to 3.8%. Thus, machine-based, CART model could provide an easy to interpret representation of heterogeneous outcomes relative to preoperative factors that could be used as a guide for preoperative patient selection and risk stratification.