Selection criteria for lung-cancer screening

Area of Science:

• Oncology
• Radiology
• Epidemiology

Background:

• The National Lung Screening Trial (NLST) established lung cancer screening criteria based on smoking history and time since quitting.
• An improved risk-prediction model could enhance the identification of individuals who would benefit from lung cancer screening.

Purpose of the Study:

• To modify and validate the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial risk-prediction model for use with NLST data.
• To compare the accuracy of the modified PLCO(M2012) model with existing NLST criteria for lung cancer detection.

Main Methods:

• The PLCO Cancer Screening Trial risk-prediction model was adapted for NLST data, creating the PLCO(M2012) model.
• The PLCO(M2012) model was developed and validated using data from 80,375 ever-smokers in the PLCO trial.
• Model discrimination (AUC) and calibration were assessed, and its performance was compared against NLST criteria in detecting lung cancer.

Main Results:

• The PLCO(M2012) model demonstrated strong discrimination, with an AUC of 0.797 in the validation dataset.
• Compared to NLST criteria, PLCO(M2012) criteria showed significantly higher sensitivity (83.0% vs. 71.1%) and positive predictive value (4.0% vs. 3.4%).
• The PLCO(M2012) model missed 41.3% fewer lung cancers than NLST criteria, with similar specificity.

Conclusions:

• The PLCO(M2012) risk-prediction model is more sensitive than the NLST criteria for detecting lung cancer.
• This enhanced model may improve the selection of individuals for lung cancer screening, potentially leading to earlier diagnosis.