Lung cancer screening: 4 risk models are better than the rest

Access to the full content of this site is available only to registered healthcare professionals. Access to the full content of this site is available only to registered healthcare professionals.

Takeaway

  • The Bach model, Ovarian Cancer Screening Trial Model 2012 (PLCO-M2012), Lung Cancer Risk Assessment Tool (LCRAT), and Lung Cancer Death Risk Assessment Tool (LCDRAT) best identify candidates for CT screening among smokers/former smokers.

Why this matters

  • In the absence of thresholds for screening, 2018 National Comprehensive Cancer Network guidelines advise using risk-prediction models.

Study design

  • 2010-2012 National Health Interview survey data simulated a US screening population.
  • Validity was assessed using data for 337,388 ever-smokers in the NIH-American Association of Retired Persons Diet and Health Study and 72,338 ever-smokers in the Cancer Prevention Study II Nutrition Survey cohort.       
  • Funding: NIH/National Cancer Institute.

Key results

  • At a 5-year threshold of 2.0%, screening populations ranged from 7.6 to 26 million.
  • Bach, PLCO-M2012, LCRAT, and LCDRAT showed greater agreement on screening population size (7.6-10.9 million) and achieved consensus on proportion of selected ever-smokers (73%).
  • These 4 models had the highest overall discrimination, highest sensitivity at a fixed specificity (and vice versa), and similar discrimination at a fixed risk threshold.
  • All 4 had limitations:
    • Bach did not account for race/ethnicity, COPD, or family history of lung cancer.
    • PLCO-M2012 underestimated risk in Hispanics by 2- to 3-fold.
    • LCRAT/LCDRAT underestimated risk in Asian/other subgroup.

Limitations

  • May not represent contemporary US population.