Development of a Machine Learning-Based Predictive Model for Lung Metastasis in Patients With Ewing Sarcoma

Area of Science:

• Oncology
• Machine Learning
• Biostatistics

Background:

• Ewing sarcoma (ES) is a rare bone cancer predominantly affecting children and young adults.
• Lung metastasis (LM) is a common and serious complication of ES, significantly impacting patient prognosis.
• Accurate prediction of LM is crucial for timely and effective treatment strategies.

Purpose of the Study:

• To develop and validate machine learning (ML)-based prediction models for lung metastasis (LM) in patients with Ewing sarcoma (ES).
• To identify the best-performing ML model for LM prediction in ES.
• To deploy the optimal model as an accessible open-access web tool for clinical use.

Main Methods:

• Retrospective analysis of data from the Surveillance Epidemiology and End Results (SEER) Database (2010-2016) and four medical institutions.
• Development of six ML-based models using demographic and clinicopathologic variables from 929 ES patients (training group).
• Internal validation using 10-fold cross-validation and external validation on 51 patients; model performance assessed by Area Under the Curve (AUC).

Main Results:

• The study included 980 ES patients (175 with LM).
• Multivariate logistic regression identified survival time, T-stage, N-stage, surgery, and bone metastasis as independent predictors of LM.
• The Random Forest (RF) model achieved the highest AUC of 0.705, demonstrating superior predictive power for LM in ES patients.

Conclusions:

• Machine learning models demonstrate significant utility in predicting lung metastasis in Ewing sarcoma.
• The Random Forest model emerged as the best-performing predictive tool.
• An accessible web-based tool utilizing the RF model can enhance personalized treatment strategies for ES patients at risk of LM.