Circulating tumor DNA as markers of dynamic recurrence risk and adjuvant chemotherapy benefit in resected non-small cell lung cancer.

Abstract

3028 Background: A significant proportion of non-small cell lung cancer (NSCLC) patients relapse after surgical resection with or without adjuvant therapy. The detection of molecular residual disease (MRD) has great potentials to stratify postoperative risk and facilitate early recurrence diagnosis. Here, we aim to evaluate the clinical utility of serial plasma circulating tumor DNA (ctDNA) in MRD detection, adjuvant therapy guidance, and recurrence risk prediction in resected NSCLC patients. Methods: This prospective cohort study recruited 116 patients with NSCLC following surgery and/or adjuvant therapy. Thirteen patients discontinued, leaving 103 patients for analysis. Tumor samples were obtained at surgery. Plasma samples were collected at baseline, after surgery, after adjuvant therapy and every 3 months thereafter and were analyzed by ultradeep (30000×) next-generation sequencing with molecular barcode and in-silico error correction. Results: Pretreatment ctDNA was detected in 69.8% of patients. ctDNA positivity after surgery and after completion of adjuvant chemotherapy (ACT) were significantly associated with worse recurrence-free survival (RFS) (HR: 4.0; 95% CI: 2.0-8.0; P<.001 and HR:3.2; 95% CI: 1.3-8.2; P <.05, respectively). In stage II-III patients who were positive for ctDNA after surgery, ACT-treated patients had a better RFS than those without ACT ( P<.05), whereas ACT did not confer significant clinical benefits in patients with negative postsurgical ctDNA. During surveillance after definitive therapy, ctDNA positivity was associated with worse RFS (HR: 8.5, 95% CI: 3.7-20, P <.001) and preceded radiological recurrence by a median of 88 days. Using joint modeling of serial ctDNA and time-to-recurrence, we accurately predicted patients’ 12-month and 15-month recurrence status, with areas under receiver-operating characteristics curves (AUROC) of 0.88 and 0.84, respectively. Conclusions: These results indicate that ultradeep ctDNA sequencing could sensitively detect MRD, thus identifying patients with high recurrence risk and guiding the adjuvant therapy decision in resected NSCLC. We also demonstrate that joint modeling of serial ctDNA levels and time to recurrent can provide an accurate dynamic risk prediction for NSCLC patients during surveillance. Clinical trial information: ChiCTR1900024656.