Background: Circulating tumor DNA (ctDNA) has been reported to be valuable for minimal residual disease (MRD) detection, but there is a lack of studies comprehensively assessing whether dynamic ctDNA monitoring can inform whole-course precise postoperative cancer management. In the present study (LUNGCA-2), we analyzed data obtained from longitudinal ctDNA monitoring of stage I-III NSCLC patients in the LUNGCA cohort, and sought to assess the use of dynamic ctDNA monitoring for precise whole-course postoperative management of NSCLC patients. Methods: All eligible NSCLC patients underwent curative-intent surgery and subsequently received adjuvant therapies when indicated by clinical practice guidelines for NSCLC. The clinical follow-up was scheduled to be every 3-6 months mainly involving CT and/or MRI. Fresh frozen or paraffin-embedded tissues of primary tumors were collected intraoperatively for baseline somatic mutation profiling. Blood samples were collected before surgery, at 3 days and 1 month after surgery, and then every 3-6 months at clinical visits up to 3 years postoperatively. 233 patients with a total of 2123 plasma samples were ultimately included in the longitudinal ctDNA analysis in the current study. NGS assays were performed with a custom-designed panel that covered 769 cancer-related genes via the MinerVA platform (Genecast Biotechnology Co., Ltd). Results: Among the cohort, 166 patients harboring targetable mutations in the primary tumor tissue, 15 were preadjuvant ctDNA positive, and among them, only 25% patients treated with adjuvant TKI therapy experienced relapse during follow-up, and patients received ACT or no adjuvant therapy had a relapse rate of 100%. TKI treatment tended to improve RFS compared with no treatment (HR 0.1, P = 0.025), but ACT failed to improve RFS (HR 1.1, P = 0.834). Among all 82 patients receiving adjuvant therapy, the pre- and post-adjuvant therapy postoperative ctDNA status with a negative-negative or positive-negative pattern had similar and favorable RFS (P = 0.545) but that patients with a positive-positive or negative-positive pattern had significantly inferior RFS than those with a negative-negative pattern (both P< 0.001). For recurrence monitoring, posttreatment ctDNA MRD status had a positive predictive value of 100% and a negative predictive value of 91.8%, and the sensitivity of MRD detection was affected by the organ of relapse. Among positive ctDNA MRD patients, faster increase of ctDNA concentration correlated with more imminent clinical relapse. The median lead time from ctDNA detection to radiological relapse was 273 days. Moreover, ctDNA negativity after relapse was associated with favorable survival. Conclusions: These findings highlight the clinical significance of ctDNA dynamics for the precise whole-course postsurgical management of NSCLC patients.