Detection of molecular residual disease and risk stratification as early as possible may improve the treatment of patients with cancer. Efficient pragmatic tests are therefore required. To measure circulating tumor DNA (ctDNA) with 6 DNA methylation markers in blood samples and to evaluate the association of the presence of ctDNA with colorectal cancer (CRC) recurrence throughout the disease course. In this multicenter prospective longitudinal cohort study performed from December 12, 2019, to February 28, 2022, 350 patients with stage I to III CRC were recruited from 2 hospitals for collection of blood samples before and after surgery, during and after adjuvant chemotherapy, and every 3 months for up to 2 years. A multiplex, ctDNA methylation, quantitative polymerase chain reaction assay was used to detect ctDNA in plasma samples. A total of 299 patients with stage I to III CRC were evaluated. Of 296 patients with preoperative samples, 232 (78.4%) tested positive for any of the 6 ctDNA methylation markers. A total of 186 patients (62.2%) were male, and the mean (SD) age was 60.1 (10.3) years. At postoperative month 1, ctDNA-positive patients were 17.5 times more likely to relapse than were ctDNA-negative patients (hazard ratio [HR], 17.5; 95% CI, 8.9-34.4; P < .001). The integration of ctDNA and carcinoembryonic antigen tests showed risk stratification for recurrence with an HR of 19.0 (95% CI, 8.9-40.7; P < .001). Furthermore, ctDNA status at postoperative month 1 was strongly associated with prognosis in patients treated with adjuvant chemotherapy of different durations and intensities. After adjuvant chemotherapy, ctDNA-positive patients had a significantly shorter recurrence-free survival than did the ctDNA-negative patients (HR, 13.8; 95% CI, 5.9-32.1; P < .001). Longitudinal ctDNA analysis after the postdefinitive treatment showed a discriminating effect in that ctDNA-positive patients had poorer recurrence-free survival than did the ctDNA-negative patients (HR, 20.6; 95% CI, 9.5-44.9; P < .001). The discriminating effect was enhanced (HR, 68.8; 95% CI, 18.4-257.7; P < .001) when ctDNA status was maintained longitudinally. Postdefinitive treatment analysis detected CRC recurrence earlier than radiologically confirmed recurrence, with a median lead time of 3.3 months (IQR, 0.5-6.5 months). The findings of this cohort study suggest that longitudinal assessment of ctDNA methylation may enable the early detection of recurrence, potentially optimizing risk stratification and postoperative treatment of patients with CRC.