Background: Recurrence rates after curative intent resection of colorectal cancer liver metastases (CRLM) are nearly 50%. Standard of care surveillance includes intensive computed tomographic (CT) imaging and CEA measurements. However, recurrence detection often happens too late to resume a second resection. Previous studies have shown that circulating tumor DNA (ctDNA) detection, as a biomarker of minimal residual disease (MRD), is a robust prognostic factor for patients with colorectal cancers. Here, we performed a prospective study to explore the potential use for improving postoperative CRLM surveillance by dynamic ctDNA detection and its impact on the second resection of recurrent patients. Methods: Enrolled patients received curative intent resection of CRLM and were confirmed at the status of no evidence of disease (NED) by CT about one month after surgery. All patients received stand-of-care postoperative surveillance and serial plasma samples were collected, including presurgical and postsurgical samples (4-6 weeks after surgery) and every 3rd month until relapse or up to 24 months after resection. A personalized tumor-informed assay (Signatera/HuaJianwei bespoke multiplex-PCR NGS) was used to quantify ctDNA in plasma samples. Results: A total of 318 plasma samples from 60 patients recruited from July 2020 to July 2021 were included in our analysis. Twenty-five (41.6%) patients received neoadjuvant chemotherapy, and 46 patients (76.7%) received adjuvant chemotherapy after surgery. The median follow-up time was 18.0 months (range, 7.5 to 25.3 months). ctDNA was detectable in 58/60 (96.7%) patients before surgery, in 35/60 (58.3%) patients after surgery, and in 23/58 (39.7%) patients at the end of treatment (EOT). Patients with ctDNA/MRD positive status post-surgery or EOT experienced a significant shorter recurrence-free survival (RFS) than patients with negative ctDNA/MRD [hazard ratio (HR) 4.2; P < 0.001 and HR 9.8; P < 0.001]. The recurrence rate in MRD-positive patients (aka. positive predictive value, PPV) was 91.4% (32/35) at post-surgery and 91.3% (21/23) at EOT, respectively. In 93.0% of postoperative MRD-positive patients, ctDNA detected early relapse with a lead time up to 9.0 months (median 3.4 months) compared with standard radiological imaging. Among 40 patients with recurrence, the total secondary resection rate was 52.5% (21/40) with 40% (21/40) cases reached secondary NED. While among 24 recurrent cases with only liver metastases, the secondary resection rate was 66.7% (16/24) with 58.3% (14/24) cases reached secondary NED. Conclusions: Postoperative ctDNA status is strongly associated with shorter RFS. Serial ctDNA monitoring assists early detection of relapse, and therefore, improves patient outcomes by secondary resection. Clinical trial information: ChiCTR2000035677.