Background: Circulating tumor DNA (ctDNA) is a biomarker for detecting molecular residual disease (MRD) and predicting recurrence in cancer patients. Our group previously reported the first real-world data of ctDNA-based MRD detection. We now investigate its long-term prognostic value with extended follow-up in the real-world cohort of non-small cell lung cancer (NSCLC) patients treated with curative-intent. Methods: Longitudinal ctDNA monitoring was performed using a commercially available personalized, tumor-informed assay (Signatera) in 116 patients with NSCLC. Plasma samples were collected post-treatment, and ctDNA status was evaluated within the 6-month MRD window, and longitudinally thereafter. Recurrence-free survival (RFS) and overall survival (OS) were assessed according to ctDNA status. For ctDNA(+) patients, plasma-based next-generation sequencing (NGS; Guardant360) at MRD timepoint was performed when available and compared with baseline tissue NGS results (PGDx or Tempus xT). Results: A total of 116 patients across stages I-IV (stage I: 41.4%, stage II: 30.2%, stage III: 21.6%, and stage IV: 6.9%) were identified (90 adenocarcinoma, 23 squamous cell carcinoma, 2 adenosquamous and 1 other). Median age at diagnosis was 66 years (range: 23-86 years) and median follow-up was 33 months (range: 4-120 months). MRD results were available for 99 patients within 6 months of curative-intent treatment. Within the 6-month MRD window, ctDNA(+) was associated with inferior RFS (aHR: 24.3; 95% CI: 3.2-186.8; p < 0.001), with a median RFS of 8.5 months. ctDNA(+) during longitudinal surveillance beyond MRD window was associated with inferior RFS (HR: 8.0, 95% CI: 3.4-18.9; p < 0.001) and OS (HR: 11.6, 95% CI: 3.4-39.8; p < 0.001), with median RFS of 10.8 months. Among patients with both tissue NGS (tNGS) and blood NGS (bNGS) profiling results, 8 patients were ctDNA(+). Among them, 5 patients had at least one overlapping gene mutation at the variant level between tNGS and bNGS. All 5 of these patients recurred, while the other 3 patients (who did not have an overlapping gene mutation at the variant level) did not recur. The matched group consisted mostly of stage III cases (4 of 5), with one stage II, whereas the unmatched group included one each of stage I, II and IV cases. Conclusion: This first real-world report with extended follow-up provides real-world evidence that tumor-informed, personalized ctDNA monitoring is a clinically valuable tool for predicting long-term outcomes in patients with NSCLC after curative-intent treatment. Although interpretation is limited by the small sample size, blood NGS at ctDNA(+) may reflect potential subclonal evolution and molecular heterogeneity. These findings support the prognostic value of ctDNA and its potential integration into personalized cancer management.