WCLC 2025 - Posters & ePosters-PT1.04.02 Performance of Lung Nodule and Cancer Risk Calculators in Identifying Primary Lung Cancer Across Screening and Incidental Nodule Populations

PT1.04.02 Performance of Lung Nodule and Cancer Risk Calculators in Identifying Primary Lung Cancer Across Screening and Incidental Nodule Populations

Back to course

Pdf Summary

This study evaluated the performance of lung nodule-based and clinical risk calculators in distinguishing primary lung cancer from non-investigated nodules across screening and incidental populations, including subgroups of never-smokers. A retrospective analysis of 288 individuals (2023–2024) with 74 confirmed primary lung cancers was conducted. The population was predominately older adults (~70 years), mostly female (55–66%), with nodules mostly sized between 9–29 mm.<br /><br />Eight models were assessed: nodule-based calculators (Brock, Mayo Clinic, Herder, Veterans Affairs [VA], Cleveland) and cancer risk-based calculators (Liverpool Lung Project [LLP v3], PLCOm2012, Simplified PLCOm2012). Risk thresholds from literature were applied for high-risk classification, and statistical analyses included mean risk score comparisons and categorical assessments with chi-square tests.<br /><br />Key findings revealed that nodule-based calculators outperformed the cancer risk models overall. The VA model demonstrated the highest specificity (87.4%) and strongest discriminatory power (p=0.001), particularly effective in general, incidental nodule, and never-smoker cohorts. The Brock model exhibited the highest sensitivity (82.6%, p=0.001), making it useful for ruling out malignancy. Herder and Mayo models showed balanced intermediate sensitivity and specificity with significant results, supporting their use as complementary or triage tools. Conversely, the Cleveland and all clinical risk calculators (LLP, PLCOm2012 variants) generally showed limited predictive value and nonsignificant discrimination at predefined cutoffs, except for LLP performance in incidental nodules (p=0.025). Notably, in never-smokers, only the VA model maintained statistically significant performance.<br /><br />The study concludes that imaging-based nodule models, particularly the VA calculator, should be prioritized for lung nodule evaluation, with selective incorporation of demographic-based risk models in incidental findings. These findings help refine lung cancer risk stratification and guide clinical decision-making in diverse patient populations.

Asset Subtitle

Francisco Sarmento Neto

Meta Tag

Speaker Francisco Sarmento Neto

Topic Screening and Early Detection

Keywords

lung nodule evaluation

primary lung cancer

risk calculators

nodule-based models

clinical risk models

VA model

Brock model

never-smokers

lung cancer screening

risk stratification