Journal: NPJ precision oncology
Study type and population
- • Large case–control study developing a blood-based assay for lung cancer detection.
- • 1,600 lung cancer patients and 1,600 non-cancer controls, split into training and internal validation cohorts, plus an external validation cohort.
Assay and methods
- • Used a multi-omics approach based on whole‑genome features from cell‑free DNA (cfDNA).
- • Built a machine-learning model to distinguish lung cancer from non-cancer individuals.
Diagnostic performance
- • Internal training cohort: AUC 95.59%.
- • Internal validation cohort: AUC 95.74%.
- • Maintained strong performance across cancer stages and histologic subtypes.
- • External validation cohort: sensitivity 85.9% and specificity 94.78% for distinguishing cancer from non-cancer.
Comparisons and simulated screening
- • In simulated population screening scenarios, the ctDNA assay outperformed low-dose CT and a previously established blood-based method.
- • Results suggest the assay may be suitable for broader screening, likely as a complement rather than a replacement for existing LDCT programs.
Clinical implication
- • A highly sensitive and specific plasma cfDNA-based test shows promise for early lung cancer detection and risk stratification in the general population, potentially improving the effectiveness and efficiency of current screening strategies.