Diagnostic accuracy of blood-based cfDNA fragmentomic assays for lung cancer detection: A meta-analysis of clinically validated cohorts.

Abstract

10577 Background: Early detection of lung cancer remains limited by suboptimal sensitivity and specificity of current screening and diagnostic strategies. Blood-based cell-free DNA (cfDNA) fragmentomic assays that leverage genome-wide fragmentation patterns have emerged as a promising noninvasive approach for lung cancer detection. However, reported diagnostic performance varies across studies and clinical settings. A meta-analysis was conducted to evaluate the pooled diagnostic accuracy of cfDNA fragmentomic assays in clinically validated cohorts. Methods: We systematically searched PubMed, Cochrane Library, and Scopus for studies evaluating blood-based cfDNA fragmentomic assays for lung cancer detection. Eligible studies were required to be lung cancer-specific, use fragmentomics-based features derived from plasma cfDNA, include an independent validation or test cohort, and report extractable binary diagnostic accuracy data. Training-only studies, multi-omic assays, and studies lacking sensitivity and specificity at a prespecified threshold were excluded. Sensitivity and specificity were jointly pooled using a bivariate random-effects model based on 2×2 contingency tables, accounting for threshold effects and between-study heterogeneity. Results: Five studies with independent validation cohorts comprising 725 lung cancer cases and 435 non-cancer controls were included, spanning screening-eligible populations, LDCT-positive pulmonary nodules, early-stage disease, and mixed-risk cohorts. Using a bivariate random-effects model, pooled sensitivity was 87.5% (95% CI, 77.4%–93.4%) and pooled specificity was 85.7% (95% CI, 63.6%–95.4%), corresponding to a pooled positive likelihood ratio of 6.14 (95% CI, 2.34–19.26) and negative likelihood ratio of 0.146 (95% CI, 0.074–0.295), and diagnostic odds ratio of approximately 42. Moderate inter-study heterogeneity was observed (I²=62%), largely attributable to differences in clinical setting, disease spectrum, and control composition; however, diagnostic performance remained directionally consistent across cohorts. The summary receiver operating characteristic curve demonstrated excellent overall discrimination (AUC 0.92). Conclusions: Across clinically validated cohorts, blood-based cfDNA fragmentomic assays demonstrated strong discriminative performance for lung cancer detection. Despite heterogeneity across study populations and clinical use cases, pooled sensitivity and specificity were robust, supporting the potential role of cfDNA fragmentomics as a noninvasive adjunct to existing screening and diagnostic strategies. Diagnostic performance of cfDNA fragmentomic assays for lung cancer detection. Measure Estimate 95% CI Sensitivity 87.5% 77.4–93.4 Specificity 85.7% 63.6–95.4 LR+ 6.14 2.34–19.26 LR– 0.15 0.07–0.30 AUC 0.92 —