Tumor-informed or tumor-agnostic circulating tumor DNA as a biomarker for risk of recurrence in resected colorectal cancer patients.

Introduction: Circulating tumor DNA (ctDNA) has been increasingly recognized as a promising minimally-invasive biomarker that could identify patients with minimal residual disease and a high risk of recurrence after definitive treatment. In this study, we've compared the clinical utility and sensitivity of 2 different approaches to ctDNA analyses: tumor-informed and tumor-agnostic in the management of colorectal (CRC) patients. The clinical benefits of a single timepoint ctDNA analysis compared to serial ctDNA monitoring after definitive treatment were also evaluated to uncover the ideal surveillance protocol. Methods: Patient-paired resected tumor tissues, peripheral blood cells, and a total of 127 pre-operative and serial plasma cell-free DNA (cfDNA) samples after definitive treatment from 38 CRC patients that had undergone curative intent surgery were analyzed using a commercial NGS cfDNA panel. Results: Up to 84% (32/38) of the recruited patients were detected with at least 1 genomic alteration from the tumor tissues that could be monitored using the tumor-informed ctDNA approach and none of the detected alterations were clonal hematopoiesis (CH) related. In contrast, 37% (14/38) of patients were detected with at least 1 monitoring alteration after exclusion of CH mutations using the tumor-agnostic approach. Serial plasma samples after definitive therapy were available for 31 patients. In the landmark ctDNA analysis, 24% (7/29) of patients had detectable ctDNA and were more likely to relapse than ctDNA-negative patients (p < 0.05). The landmark analysis sensitivity and specificity for recurrence were 67% and 87%, respectively. The incorporation of longitudinal ctDNA analysis at 6-months intervals improved the sensitivity to 100%. The median variant allele frequency (VAF) of the ctDNA mutations detected during surveillance was 0.028% (range: 0.018-0.783), where up to 80% (8/10) of the mutations were detected at VAF lower than the tumor-agnostic detection limit of 0.1%. Utilizing the tumor-agnostic approach reduced the recurrence detection sensitivity to 67% (4/6). Serial ctDNA analyses predicted disease recurrence at a median of 5 months ahead of radiological imaging. Conclusion: Longitudinal monitoring using tumor-informed ctDNA testing shows high analytical sensitivity, low probability of false-positive results due to CH mutations, and improved sensitivity in detecting recurrence which may modify the clinical management of CRC.