Machine Learning Models Predict Early Postoperative Relapse in Pancreatic Cancer

Abstract

115 Pancreatic ductal adenocarcinoma (PDAC) is a deadly disease even in the localized stage. The 5-year survival is approximately 10% in the USA, and 8085% of the patients present unresectable or metastatic cancer.1,2 Even for individuals eligible for surgery, the prognosis is dismal, with only 20% surviving 5 years.1-3 While recent advances in surgery and adjuvant chemotherapy have improved survival times, the outcomes are still far from desired. PDAC remains the fourth leading cause of cancer-related deaths in the world.4 A significant obstacle is the considerable heterogeneity of the disease and the lack of reliable clinical risk stratification. Early-stage colorectal cancer and breast cancer do not present these disadvantages, and hence, it is easy to tailor treatment selection. In the constantly changing paradigm of new treatments, it is essential to identify prediction strategies for selecting the best treatment for the right patient.5 However, current knowledge currently lacks such guiding ability.