Prediction of Early Mortality in Esophageal Cancer Patients with Liver Metastasis Using Machine Learning Approaches

Area of Science:

• Oncology
• Medical Informatics
• Biostatistics

Background:

• Esophageal cancer with liver metastasis carries a high risk of early mortality.
• Current predictive tools for early mortality in these patients are limited.
• Accurate prediction is essential for timely clinical decision-making and patient management.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting early mortality in patients with esophageal cancer and liver metastasis.
• To identify key prognostic factors associated with early mortality in this patient cohort.
• To compare the performance of various machine learning algorithms in predicting early mortality.

Main Methods:

• Utilized clinicopathological data from 1897 patients with esophageal cancer liver metastasis (2010-2020) from the SEER database.
• Employed univariate and multivariate logistic regression to identify prognostic factors.
• Developed and assessed seven machine learning models, including extreme gradient boosting (XGBoost) and support vector machine (SVM), using ROC curves, calibration curves, DCA, and F1 scores.

Main Results:

• 40% of patients experienced all-cause early mortality and 38% experienced cancer-specific early mortality.
• Significant predictors of early mortality included age, tumor location, chemotherapy, and lung metastasis.
• XGBoost demonstrated superior performance in predicting all-cause early mortality, while SVM was optimal for cancer-specific early mortality.

Conclusions:

• Machine learning models, specifically XGBoost and SVM, are effective tools for predicting early mortality in patients with esophageal cancer liver metastasis.
• These models can aid clinicians in identifying high-risk patients, thereby improving clinical decision-making and patient outcomes.
• Further validation and integration of these predictive tools into clinical practice are warranted.