Development and Validation of Lung Cancer Risk Prediction Models in the Singapore Chinese Population

Area of Science:

• Oncology
• Epidemiology
• Biostatistics

Background:

• Lung cancer is often diagnosed late, particularly in never-smokers who are not covered by current screening guidelines.
• Effective screening technologies exist for smokers, but risk stratification for all populations remains a challenge.

Purpose of the Study:

• To develop 10-year lung cancer risk prediction models tailored to specific subgroups (sex and smoking status).
• To support more precise identification of high-risk individuals within clinical settings.

Main Methods:

• A Cox regression model was developed using the Singapore Chinese Health Study (n=63,187) to predict 10-year lung cancer risk.
• Predictors were identified using Elastic Net regularization with cross-validation, including sociodemographic, smoking, environmental, reproductive, and dietary factors.
• External validation was performed on the Singapore Multi-Ethnic Cohort (n=12,944), and sex- and smoking-specific models were developed.

Main Results:

• The overall risk model showed strong external validation (C-index = 0.821).
• Key predictors included age, smoking exposure, body mass index, education, sex, environmental tobacco smoke, and diet (tea/fruit consumption).
• Specific models for ever-smokers (C-index=0.734) and never-smokers (C-index=0.702) were developed, with never-smoker models incorporating factors like personal cancer history and allergic rhinitis. Sex-specific models also demonstrated good discrimination.

Conclusions:

• Subgroup-specific lung cancer risk models offer insights beyond traditional smoking history.
• These models may aid in tailored risk stratification in clinical settings after further real-world evaluation.