Methylation-based circulating tumor DNA predicts recurrence in definitively treated non-small cell lung cancer (NSCLC)

Background: Circulating tumor DNA (ctDNA) is being utilized in detecting molecular residual disease (MRD) and predicting recurrence of non-small cell lung cancer (NSCLC). However, real-world implementation is yet to be reported. Methods: We retrospectively analyzed 240 longitudinal plasma samples from a real-world cohort of 78 NSCLC patients treated with curative intent, including select stage IV patients who underwent bilateral lung transplantation. MRD was assessed using a tissue-naïve, methylation-based assay (Guardant Reveal), and plasma-based genomic profiling was performed (Guardant360) at the time of MRD positivity. Recurrence-free survival (RFS) and overall survival (OS) were analyzed based on three MRD timepoints post-treatment: within 3 months (Subgroup A, n=42), within 6 months (Subgroup B, n=55), and at any time (Subgroup C, n=78). Recurrence date was determined by biopsy-proven recurrence date when available, otherwise, radiographic recurrence date was used. Results: 78 patients (median age at diagnosis 65; 33 stage I, 19 stage II, 19 stage III, 7 stage IV) received definitive treatment (surgery alone, n=37; surgery with perioperative therapy, n=33; definitive radiation, n=8). 56 were adenocarcinoma and 21 were squamous cell carcinoma, and 1 was adenosquamous. Targetable alterations (EGFR, ALK, etc.) were found in 31 patients. The median follow-up duration was 22.3 months (range: 2-80.6 months). Among 23 MRD(+) cases, 10 recurrences and 3 deaths occurred, whereas 55 MRD(-) cases showed 5 recurrences and no deaths. MRD(+) was associated with worse OS in all subgroups (p<0.01). It was also associated with inferior RFS in Subgroup B (HR = 4.35, 95% CI 1.30-14.50, p = 0.017) and Subgroup C (HR = 5.20, 95% CI 1.76-15.30, p<0.01), with a similar trend in Subgroup A (p=0.16). Among 11 patients who recurred and ever showed ctDNA positivity, ctDNA preceded radiographic relapse in 10 cases with a mean lead time of 5.4 months (range 2 days-20.3 months). In the subset of patients with ≥1 year follow-up from definitive treatment (n=66), MRD(+) showed a sensitivity of 63.6% (7/11) and specificity of 78.1% (43/55) in predicting recurrence. In 13 MRD(+) patients with paired plasma and tissue NGS, 3 (23%) exhibited shared mutations between baseline tissue NGS and post-treatment plasma NGS (matched group). The unmatched group showed a median number of 9 (range 1-14) genomic variations; however, none were biologically relevant. Conclusion: This is the first real-world analysis evaluating survival outcomes using a methylation-based ctDNA assay in definitively treated NSCLC. MRD(+) can be useful in predicting recurrence and was associated with worse outcomes. Integration of plasma-based NGS may reflect molecular heterogeneity and potential subclonal evolution, supporting its potential utility for personalized post-treatment surveillance.