Personalized medicine in colorectal mucinous adenocarcinoma (MAC): Machine learning–based prognostic models.

Abstract

e15727 Background: MAC accounts for 10% of colorectal cancer cases and is a distinct subtype with poor prognosis and resistance to treatments such as adjuvant chemotherapy and chemoradiotherapy, complicating its management. We aimed to use machine learning (ML) to identify prognostic factors and improve survival outcomes. 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 62,725 patients were included. Most patients (54.2%) were female and 82.1% were White. The median patient age was 71 years. Most patients had T3 (60.1%), while T4 was rare (23.3%). Nodal involvement was observed in 43.5% of cases, while M0 accounted for 85.6%. Of the metastatic cases, 3.1% were in the liver and 0.5% in the liver and lungs. Among these patients, 98.2% underwent surgery, 1.9% received radiation, and only 34.9% received chemotherapy. Patients aged < 71 years had significantly better 5-year survival rates (OS: 62.7%, CSS: 67.5%) than those aged ≥ 71 years (OS: 45.4%, CSS: 64%). Survival was higher in patients who did not receive radiation (OS: 54.2%, CSS: 66.3%) compared to those who did (OS: 37.1%, CSS: 39.9%). Patients without metastasis had markedly better outcomes (OS, 61.3%; CSS, 75.1%) than those with metastasis (OS, 9.5%; CSS, 10.7%). Asian/Pacific Islanders and Whites had a higher CSS (66.4%) than Blacks (60.3%). Multivariate Cox regression analysis identified older age, radiation, metastasis, splenic flexure, sigmoid colon, and White and Black races as poor prognostic factors, whereas surgery and ascending colon surgery were good prognostic factors. The Random Forest and MLP classifiers were the most accurate models. The ML models identified TNM stage as the most significant prognostic factor, followed by age and race. The performance metrics for all the ML algorithms are summarized in Table. Conclusions: ML enhances personalized medicine by identifying key prognostic factors in MAC. Further investigation is needed to assess the benefits of radiation therapy in this distinct entity. ML Model Accuracy Precision Recall F1 score AUC LR 66.7% 64.2% 62.9% 63.5% 0.735 KNN 64% 60.8% 61.5% 61.% 0.701 RFC 67.9% 63.7% 70.2% 66.8% 0.763 GBC 66.8% 62.5% 69.5% 65.8% 0.741 MLP 67.4% 63.5% 68.8% 66% 0.747