Lung cancer in primary care: Development & validation of a prediction algorithm.

Area of Science:

• Oncology
• Public Health
• Biostatistics

Background:

• Lung cancer (LC) is a leading global cause of cancer mortality.
• Early detection in primary care is crucial for improving patient outcomes.
• Predictive models can enhance early referral and management of lung cancer.

Purpose of the Study:

• To develop and validate a multivariable prediction algorithm for 5-year lung cancer risk.
• To create a decision support tool for primary care physicians.
• To identify key risk factors for lung cancer prediction.

Main Methods:

• A large cohort study of 3,454,735 patients aged ≥30 years from Italian general practitioners' data (2002-2021).
• Utilized a Cox proportional hazard model, assessing performance with pseudo-R², AUC, and calibration metrics.
• Employed bootstrap methodology for model overfitting assessment and established risk thresholds.

Main Results:

• Identified smoking (HR=14.75), COPD (HR=2.3), and advanced age (HR=1.29) as significant lung cancer predictors.
• The model achieved a pseudo-R² of 0.609 and an AUC of 0.822, indicating strong predictive performance.
• Established risk groups: low (<0.11%), intermediate (0.11-0.89%), and high (≥0.9%) for 5-year lung cancer risk.

Conclusions:

• The developed lung cancer risk model is a viable predictive tool for primary care settings.
• The tool supports timely referral and resource prioritization for lung cancer management.
• Clinical decision support systems incorporating this algorithm can enhance early lung cancer detection.