Testing details
Almost 21,000 asymptomatic participants age 40 years and older undergoing screening colonoscopy were evaluated at 186 US sites during the period 2019 to 2023. Of the cohort, 98 had CRC, 2144 had advanced precancerous lesions, 6973 had nonadvanced adenomas, and 10,961 had nonneoplastic findings or negative colonoscopy.
Advanced precancerous lesions included one or more adenomas or sessile serrated lesions measuring at least 1 cm in the longest dimension, lesions with villous histologic features, and high-grade dysplasia. The new DNA test identified 92 of 98 participants with CRC and 76 of 82 participants with screening-relevant cancers. Among the findings for the new assay:
- Sensitivity for any-stage CRC was 93.9% (95% confidence interval [CI], 87.1- 97.7)
- Sensitivity for advanced precancerous lesions was 43.4% (95% CI, 41.3-45.6)
- Sensitivity for high-grade dysplasia was 74.6% (95% CI, 65.6-82.3)
- Specificity for advanced neoplasia was 90.6% (95% CI, 90.1- 91.0).
- Specificity for nonneoplastic findings or negative colonoscopy was 92.7% (95% CI, 92.2-93.1)
- Specificity for negative colonoscopy was 93.3 (95% CI, 92.8-93.9)
- No adverse events occurred.
In the comparator assay, OC-AUTO FIT by Polymedco, sensitivity was 67.3% (95% CI, 57.1-76.5) for CRC, 23.3% (95% CI, 21.5-25.2) for advanced precancerous lesions, and 47.4% (95% CI, 37.9-56.9) for high-grade dysplasia. In the comparator FIT, however, specificity was better across all age groups — at 94.8% (95% CI, 94.4-95.1) for advanced neoplasia, 95.7% (95% CI, 95.3- 96.1) for nonneoplastic findings, and 96.0% (95% CI, 95.5-96.4) for negative colonoscopy.
In another article in the same issue of NEJM, Guardant Health’s cell-free DNA blood-based test had 83% sensitivity for CRC, 90% specificity for advanced neoplasia, and 13% sensitivity for advanced precancerous lesions in an average-risk population.
An age-related decrease in specificity was observed with the new Cologuard test, but that did not concern Dr. Imperiale because the same observation was made with the current version. “In fact, the next-gen version appears to have less of an age-related decrease in specificity than the current version, although, again, the two versions were not tested head-to-head,” he noted.
The effect of age-related background methylation of DNA is well known, he explained. “Clinicians and older patients in the screening age range do need to be aware of this effect on specificity before ordering or agreeing to do the test. I do not see this as a stumbling block to implementation, but it does require discussion between patient and ordering provider.”
The new version of the DNA test is expected to be available in about a year.
According to Dr. Imperiale, further research is needed to ascertain the test’s acceptability and adherence rates and to quantify its yield in population-based screening. Determining its cost-effectiveness and making it easier to use are other goals. “And most importantly, the degree of reduction in the incidence and mortality from colorectal cancer,” he said.
Cost-effectiveness and the selection of the testing interval may play roles in adherence, particularly in populations with lower rates of screening adherence than the general population, John M. Carethers, MD, AGAF, of the University of California, San Diego, noted in a related editorial.
“Adherence to screening varies according to age group, including persons in the 45- to 49-year age group who are now eligible for average-risk screening,” he wrote. “It is hoped that these newer tests will increase use and adherence and elevate the percentage of the population undergoing screening in order to reduce deaths from colorectal cancer.”
This study was sponsored by Exact Sciences Corporation, which conducted the stool testing at its laboratories.
Dr. Imperiale had no competing interests to disclose. Several study co-authors reported employment with Exact Sciences, or stock and intellectual property ownership. Dr. Shaukat disclosed consulting for Freenome. Dr. Carethers reported ties to Avantor Inc. and Geneoscopy.