Predictive effectiveness of circulating tumor DNA in recurrent early-stage non-small cell lung cancer: An updated meta-analysis.

e15025 Background: Circulating tumor DNA (ctDNA)-based minimal residual disease (MRD) detection has been integrated into the perioperative management of early-stage non-small cell lung cancer (NSCLC) recently. However, its efficacy varies across different studies and application scenarios. Methods: We conducted a meta-analysis of 30 studies on ctDNA for postoperative recurrence detection in NSCLC, identified through systematic searches of Medline, Embase and the Cochrane Library. Outcomes included disease-free survival, ctDNA surveillance time, and prediction sensitivity/ specificity/ positive predictive value (PPV)/negative predictive value (NPV). Results: A total of 3,287 patients from 30 studies were included for meta-analysis, with most having stages Ⅰa to Ⅲ disease. Analysis of studies with postoperative landmark detection revealed an overall sensitivity of 0.45 (95% CI: 0.39-0.52), specificity of 0.95 (95% CI: 0.92-0.97), and SROC curve area under the curve (AUC) of 0.74 (95% CI: 0.70-0.78)，while its diagnosis efficacy was significantly improved with multiple longitudinal monitoring. Among 21 studies with multiple surveillance instances, the overall sensitivity was 0.79 (95% CI: 0.72-0.86), specificity was 0.92 (95% CI: 0.86-0.96), and AUC was 0.91 (95% CI: 0.88-0.93). We also compared the performance based on different detection strategies. The tumor-informed strategy showed higher specificity (0.97 vs. 0.93) and AUC (0.81 vs. 0.65) compared to the tumor-agnostic strategy at landmark, while the latter had slightly higher sensitivity (0.47 vs. 0.40). However, in longitudinal observations, the tumor-agnostic strategy demonstrated higher sensitivity (0.84 vs. 0.74) and AUC (0.92 vs. 0.87) compared to the tumor-informed strategy, albeit with lower specificity (0.90 vs. 0.95). Conclusions: The timing and strategy of ctDNA-MRD detection significantly influence the predictive accuracy for NSCLC recurrence. Our findings suggest that longitudinal surveillance is more effective overall, with tumor-agnostic strategies offering higher sensitivity and tumor-informed strategies providing higher specificity.