Molecular residual disease (MRD) detection using an ultra-sensitive assay in a prospective colorectal cancer cohort: The VICTORI study.

3629 Background: Circulating tumor DNA (ctDNA) detection of molecular residual disease (MRD) is prognostic in colorectal cancer (CRC) but missed recurrences highlight constraints in assay performance. VICTORI is a prospective observational study investigating an ultrasensitive MRD assay (NeXT Personal) in CRC patients undergoing curative treatment. Methods: Whole-genome informed panels of ~1,800 variants interrogated plasma collected pre-curative loco-regional treatment, q2 weeks post-surgery (weeks 2–8, landmark window), and subsequently every 3 months (3-36 months, surveillance window). Results: 795 plasma samples from 109 patients were analyzed, with median follow-up of 21.1 months and 31 recurrences observed. Cohort characteristics: 61 (56.0%) rectal cancer, 48 (44.0%) colon cancer; 81 (74.3%) stage I-III; 56 (50.5%) had neoadjuvant treatment; 106 (97.2%) had surgery as curative procedure. Pre-surgical sensitivity (treatment-naïve) was 96.0% (n=48/50) and specificity was 100% (no detection in pathologic complete responders; 0/10). Landmark window overall sensitivity for recurrence was 83.9% (n=26/31) with 69.2% (18/26) of detections being <100 ppm. Sensitivity was similar weeks 4–8 (81.5% week 4 [n=22/27]; 82.1% week 6 [23/28]; 85.7% week 8 [24/28]), and lower at week 2 (52.4% [11/21]) when post-surgical cell-free DNA peaked (median 2.80 ng/mL pre-surgery to 9.40 at week 2; p=8.7×10⁻¹²). Sampling between weeks 4-8 yielded a sensitivity of 80.0% (n=24/30) at the first available timepoint and 89.7% (26/29) with incorporation of an additional subsequent sample. Detection of ctDNA at any landmark window timepoint was associated with inferior outcomes (adjusted HR [aHR]: 10.28 [95% CI: 4.03-26.23], p =1.06x10 -06 ). Week-4 detection <100 ppm was prognostic and conferred a 6.5 month longer lead time than detections ≥100ppm (HR 4.09, p=0.018). All patients who recurred with surveillance samples had ctDNA detected prior to recurrence (sensitivity: 100% (n=28/28). ctDNA identified 100% of lung- and liver-only metastases; most initial detections were <100 ppm (75% and 71.4%, respectively). In evaluable patients, lead time from ctDNA detection to recurrence was 176 days (including patients with expedited imaging upon receipt of a ctDNA detection result). If a higher threshold for ctDNA was used (≥100 ppm) in the same population, the lead time would be >two months shorter (105 days, p<0.001). All patients without ctDNA detected on surveillance remain disease free (NPV 100% [n=63/63]). Conclusions: NeXT Personal demonstrated high accuracy, prognostic value, and early detection at ultrasensitive levels with no loss in specificity. The ideal postoperative landmark time point for MRD testing is at 4-8 weeks. Our results show strong clinical potential for an ultrasensitive MRD assay for landmark MRD and longitudinal surveillance in CRC.