Targeted next-generation sequencing of circulating-tumor DNA for tracking minimal residual disease in localized colon cancer.

Abstract

A high percentage of patients diagnosed with localized colon cancer (CC) will relapse after curative treatment. Although pathological staging currently guides our treatment decisions, there are no biomarkers determining minimal residual disease (MRD) and patients are at risk of being undertreated or even overtreated with chemotherapy in this setting. Circulating-tumor DNA (ctDNA) can to be a useful tool to better detect risk of relapse. One hundred and fifty patients diagnosed with localized CC were prospectively enrolled in our study. Tumor tissue from those patients was sequenced by a custom-targeted next-generation sequencing (NGS) panel to characterize somatic mutations. A minimum variant allele frequency (VAF) of 5% was applied for variant filtering. Orthogonal droplet digital PCR (ddPCR) validation was carried out. We selected known variants with higher VAF to track ctDNA in the plasma samples by ddPCR. NGS found known pathological mutations in 132 (88%) primary tumors. ddPCR showed high concordance with NGS (r = 0.77) for VAF in primary tumors. Detection of ctDNA after surgery and in serial plasma samples during follow-up were associated with poorer disease-free survival (DFS) [hazard ratio (HR), 17.56; log-rank P = 0.0014 and HR, 11.33; log-rank P = 0.0001, respectively]. Tracking at least two variants in plasma increased the ability to identify MRD to 87.5%. ctDNA was the only significantly independent predictor of DFS in multivariable analysis. In patients treated with adjuvant chemotherapy, presence of ctDNA after therapy was associated with early relapse (HR 10.02; log-rank P < 0.0001). Detection of ctDNA at follow-up preceded radiological recurrence with a median lead time of 11.5 months. Plasma postoperative ctDNA detected MRD and identified patients at high risk of relapse in localized CC. Mutation tracking with more than one variant in serial plasma samples improved our accuracy in predicting MRD.