Circulating tumor DNA methylation-based method for noninvasive detection and stage stratification of colorectal tumor.

[BACKGROUND] Early detection and optimal treatment could improve the outcomes of patients with colorectal cancer (CRC). No adequate method has been developed to meet these two requirements. Here, we aimed to identify differential circulating tumor DNA (ctDNA) methylation biomarkers associated with CRC, and then establish models for detection, stage stratification and clinical decision-making.

[RESULTS] A total of 636 participants were included in this prospective study. To identify differential ctDNA methylation biomarkers, we first performed a genome-wide analysis between tumor and adjacent normal tissues using the α-value, which is sensitivity to ctDNA methylation signals. After filtering with PBMC samples, 4965 biomarkers were identified. A panel of 21 biomarkers was selected after shrinkage. A ctDNA methylation-based CRC diagnostic model (cMCD) was constructed. The cMCD yielded a sensitivity of 87.82% (83.39%-91.41%) for the combined detection of advanced adenomas and CRC, and a specificity of 91.88% (88.54%-94.49%). The overall accuracy was 89.85% (87.50%-92.30%), which was much greater than that of tumor markers [CEA: 67.30% (63.50%-70.93%); CA19-9: 59.91% (55.98%-63.74%); CA72-4: 55.66% (51.70%-59.57%); all P < 0.001]. As the risk score and sensitivity of the cMCD tended to increase with tumor progression, we constructed another ctDNA methylation-based model to stratify stage (cMCSS) and further guide treatment selection, specifically for discriminating the Early-stage patients eligible for curative endoscopic resection and avoiding overtreatment. The cMCSS could discriminate 75.86% (68.81%-82.02%) of Early-stage patients (advanced adenomas and T1N0M0 CRCs), for whom endoscopic resection could achieve curative intent, and 89.45% (84.59%-93.19%) of Advanced-stage patients. The accuracy [83.42% (79.36%-86.96%)] was significantly greater than that of tumor markers [CEA: 60.20% (55.17%-65.08%); CA19-9: 51.77% (46.71%-56.83%); CA724: 46.17% (41.16%-51.25%); all P < 0.001].

[CONCLUSIONS] The approach based on ctDNA methylation is a noninvasive and robust method for both early detection and tumor stratification of CRC and could benefit patients in clinical decision-making.