Background: Absence of post-operative ctDNA identifies resected CRC pts with low recurrence risk for potential adjuvant chemotherapy (ACT) de-escalation. We present the largest resected CRC cohort with plasma-only MRD detection, facilitating fast turnaround times, rapid treatment decisions and circumventing the need for tissue analysis, a challenge in real world practice. Methods: TRACC included pts with stage I-III resectable CRC. Prospective longitudinal plasma collection for ctDNA occurred pre- and post-surgery, after completion of ACT, every 3m (yr 1), every 6m (yrs 2/3) with CT annually for 3yrs. Guardant’s Reveal assay evaluated genomic/methylation signals. The primary end point was 2 year recurrence free survival (RFS) by 3-12wk post-operative ctDNA detection. Secondary end points included landmark RFS (3-12wk post-definitive treatment; surgery/ACT), longitudinal (on/after landmark timepoint) sensitivity/specificity and lead time to recurrence. Survival outcomes were calculated using Cox regression. Results: Dec 16 - Aug 22, 1203 pts enrolled. Plasma samples (n = 997) from 214 pts were analysed. 143 pts were evaluable for the primary end point; 92 (64.3%) colon, 51 (35.7%) rectal cancer; 2 (1.4%) stage I, 64 (44.8%) stage II, 77 (53.8%) stage III. Median follow up was 30.4m (95% CI: 29.6-31.7). 2 year RFS is shown in the table. No association between CEA and ctDNA positivity was seen. 14 pts were landmark ctDNA positive, 8 of whom did not relapse. In-depth genomic review revealed true positives (positive in tissue), some with limited follow up and suspected false positives due to putative genomic (i.e., CHIP) or methylation calls. We observed dynamic genomic/methylation changes demonstrating cancer evolution over time. 9 landmark ctDNA negative pts relapsed/died; 5 had longitudinal samples available of whom 4 remained negative during follow up. Relapse sites included lung (n = 2), liver (n = 1) and nodal (n = 1). Longitudinal sensitivity and specificity were 50.0% (95% CI: 27.2-72.8) and 85.9% (95% CI: 78.9-91.3) respectively with a negative predictive value (NPV) of 92.1% (95% CI: 85.9-96.1) and median lead time from ctDNA detection to radiological recurrence of 5.2m (IQR: 3.2, 8.0) (n = 9). Conclusions: Plasma-only genomic/methylation MRD detection with longitudinal sampling can predict recurrence in pts with stage I-III CRC without need for tissue analysis. NPV is high supporting an ACT de-escalation strategy in post-operative ctDNA negative pts, now investigated in the ongoing UK TRACC C randomised study (NCT04050345). Clinical trial information: NCT04050345.