Detection of recurrent colorectal cancer with high specificity using a reporting threshold for circulating tumor DNA methylated in BCAT1 and IKZF1.

Abstract

A blood assay measuring methylated BCAT1 and IKZF1 can detect recurrent colorectal cancer (CRC) with high sensitivity but suboptimal specificity. This study aimed to establish an upper reference limit (URL) of these biomarkers in a reference population without CRC, apply that threshold to detecting clinical recurrence in patients who had undergone definitive therapy for CRC, and compare the performance of the biomarkers with carcinoembryonic antigen (CEA). The level of methylation was reported as the aggregate methylated BCAT1 and IKZF1 expressed as a percentage of total plasma DNA. A reference population of patients confirmed to have no colorectal neoplasia (n = 857) was used to determine the URL. Test accuracy for clinical recurrence was determined in a post-treatment surveillance population (n = 549; 77 recurrence cases). A methylation level of 0.07%, corresponding to the 98th percentile in the reference population, was set as the URL. In the surveillance population, 60 patients had methylation levels above 0.07%, and 81.7% of these had recurrence. In comparison with no minimum threshold being applied, assay sensitivity with a URL of 0.07% yielded similar sensitivity (63.6% [CI, 51.9%-74.3%] vs 64.9% [CI, 53.8%-74.7%]; P = .87) and higher specificity (97.7% [CI, 95.9%-98.8%] vs 91.3% [CI, 88.4%-93.5%]; P < .001). The BCAT1/IKZF1 test was 2.5-fold more sensitive than CEA for detecting recurrences considered amenable to surgery with curative intent (50.0% vs 20.8%; P = .016). Applying a threshold for positivity to the methylated BCAT1/IKZF1 blood assay improved the specificity for CRC recurrence without compromising sensitivity. Both the sensitivity and the specificity were superior to those of CEA.