Risk prediction models for selection of lung cancer screening candidates: A retrospective validation study
View abstract on PubMed
Summary
This summary is machine-generated.Individual risk models for lung cancer screening are more effective than age and smoking history alone. Further research is needed to assess the benefits, harms, and feasibility of implementing these risk-based screening strategies.
Area Of Science
- Pulmonary Medicine
- Oncology
- Biostatistics
Background
- Current lung cancer screening relies on age and smoking history (pack-years).
- Individual risk-based selection offers a potential alternative.
- Nine established risk models were evaluated for predicting lung cancer incidence and mortality.
Purpose Of The Study
- To assess the performance of nine individual risk models for lung cancer screening.
- To compare risk-based models against current age and pack-year criteria.
Main Methods
- Retrospective analysis of 53,452 participants from the National Lung Screening Trial (NLST) and 80,672 from the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO).
- Evaluated models for calibration, discrimination (Area Under the Curve - AUC), and clinical usefulness (decision curve analysis).
- Compared model sensitivities and specificities to NLST eligibility criteria.
Main Results
- Risk models showed satisfactory calibration but varied discrimination (AUCs 0.61-0.81).
- Models outperformed NLST criteria in decision curve analysis, demonstrating higher sensitivity and comparable specificity.
- PLCOm2012, Bach, and Two-Stage Clonal Expansion models exhibited the best performance, with AUCs >0.68 (NLST) and >0.77 (PLCO).
Conclusions
- Individual risk-based selection for lung cancer screening surpasses traditional age and pack-year criteria.
- Further assessment of benefits, harms, and implementation feasibility of risk prediction models is recommended.