Integrative prognostic modeling for stage III lung adenosquamous carcinoma post-tumor resection: machine learning...

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Prognosis for stage III Lung Adenosquamous Carcinoma (ASC) post-resection is not well understood.
• A predictive model can guide clinical decisions and treatment strategies for ASC patients.

Purpose of the Study:

• To develop and validate a machine learning model for predicting Overall Survival (OS) in stage III ASC.
• Identify independent prognostic factors influencing OS in ASC patients.

Main Methods:

• Utilized Surveillance, Epidemiology, and End Results (SEER) database (2000-2018).
• Employed Boruta analysis for factor identification and Gradient Boosting, Random Forest, and Neural Network algorithms for model construction.
• Assessed model performance using AUC, calibration plots, Brier scores, and Decision Curve Analysis (DCA).

Main Results:

• Seven factors identified: age, sex, tumor size, N stage, primary site, chemotherapy, and systemic therapy.
• Poorer survival linked to older age, male sex, larger tumors, and lack of chemotherapy/systemic therapy.
• Random Forest model showed superior performance (3/5-year AUCs: 0.80/0.79 training, 0.74/0.65 validation) and was deployed as a web tool.

Conclusions:

• Developed a robust machine learning model for prognostic assessment in stage III ASC.
• Created a web-based tool to support personalized clinical decision-making and optimize treatment for ASC patients.