Predicting survival outcomes in pancreatic mucinous adenocarcinoma (PMA) using machine learning.

e16430 Background: PMA accounts for 3% of pancreatic ductal adenocarcinomas and remains understudied, often leading to challenging management. We aimed to apply machine learning (ML) to enhance personalized medicine and improve outcomes in this entity. Methods: Data were obtained from the SEER database (2004-2021). Patients who met any of the following criteria were excluded: diagnosis not confirmed by histology, previous history of cancer or other concurrent malignancies, or unknown data. To identify prognostic variables, we conducted Cox regression analysis and constructed prognostic models using ML algorithms to predict the 5-year survival. Patient records were randomly divided into training (70%) and validation (30%) sets. A validation method incorporating the area under the curve (AUC) of the receiver operating characteristic curve was used to validate the accuracy and reliability of the ML models. We also investigated the role of multiple therapeutic options using Kaplan-Meier survival analysis. Results: A total of 102,088 patients were included. Most patients (51.5%) were male and 79.6% were White. The median patient age was 67 years, and the median tumor size was 3.5 cm. Most patients had T3 (46.9%) or T2 (31.1%) disease. Nodal involvement was observed in 42.8% of the cases, while M0 involvement accounted for 59.9%. Of the metastatic cases, 17.9% were in the liver and 2.8% in the liver and lungs. Among these patients, 40.1% underwent surgery, 21.7% received radiation, and 34.7% received chemotherapy. Patients aged < 71 years had better 5-year survival rates (OS: 62.7%, CSS: 67.5%) than those aged ≥ 71 years (OS: 45.4%, CSS: 64%). Patients who did not receive radiation had higher survival rates (OS: 54.2%, CSS: 66.3%) than those who did (OS: 37.1%, CSS: 39.9%). Survival was worse for patients with metastasis (OS: 9.5%, CSS: 10.7%) compared to those without (OS: 61.3%, CSS: 75.1%). Among racial groups, Asian/Pacific Islanders had the highest survival rates (OS: 59%, CSS: 66.4%), followed by Whites (OS: 53.8%, CSS: 66.4%) and Blacks (OS: 51.2%, CSS: 60.3%). Multivariate Cox regression analysis identified older age, male sex, metastasis, and White and Black race as poor prognostic factors, whereas surgery and chemotherapy were good prognostic factors. Gradient boosting and Random Forest were the most accurate models. The ML models identified TNM stage as the most significant prognostic factor, followed by race and sex. The performance metrics for all ML algorithms are summarized in Table. Conclusions: ML enhances personalized medicine by identifying the key prognostic factors in PMA. These insights may help to refine prognostic models and guide treatment decisions for better patient management. ML Model Accuracy Precision Recall F1 score AUC LR 86% 50.8% 61.2% 55.6% 0.85 KNN 89.3% 51.2% 45.1% 48% 0.75 RFC 76.8% 58.9% 60.2% 59.5% 0.9 GBS 90.1% 55.6% 52.6% 54.1% 0.86 MLP 85.7% 57.4% 53.7% 55.5% 0.88