A blood test combining cell-free and exosome-bound microRNAs (miRNAs) detected 81% of advanced adenomas and 92% of stage I to III colorectal cancers in the independent testing cohort of the DENEB study, with 85% specificity for advanced neoplasia. The case-control design enriched the study population for cancer, and the investigators cautioned that the observed performance may not hold in a routine screening population.
Blood-based tests could expand screening participation, but their preventive effect is limited if premalignant lesions are missed, Ajay Goel, PhD, of Beckman Research Institute of City of Hope, Duarte, California, and colleagues wrote in The Lancet Gastroenterology & Hepatology. Most existing blood tests were built primarily to find cancer, and their sensitivity for precursor lesions has stayed low. The two most recent cell-free DNA tests, Shield and Freenome, reported sensitivity of 13.2% and 12.5% for advanced adenomas, the authors wrote, while meta-analyses of the fecal immunochemical test (FIT) have reported values as low as 36%.
In a related commentary, Jean M. Winter, PhD, and Erin L. Symonds, PhD, both of Flinders University, Adelaide, Australia, described this as “an important step towards a longstanding goal for colorectal cancer prevention …”
Study details
DENEB was an outcome-enriched, retrospective, case-control, multicenter biomarker study conducted under the Early Detection Research Network phase 1 through 3 framework. Biospecimens were collected prospectively and participants in the screening streams were enrolled consecutively, but the final analytical cohorts were assembled as a case-control sample to ensure sufficient representation of colorectal neoplasia for machine-learning training and evaluation. Separate, nonoverlapping cohorts were used for biomarker discovery, assay training, and testing. A total of 1,613 participants contributed data.
Participants included average-risk adults with findings from screening or diagnostic colonoscopy, as well as treatment-naive patients with colorectal cancer recruited before surgery. Colonoscopy and histopathology findings classified participants as having colorectal cancer, advanced adenomas, low-risk adenomas, or no disease. Advanced adenomas were defined as five or more adenomas, at least one adenoma with high-grade dysplasia, or at least one adenoma 10 mm or larger.
For biomarker discovery, the investigators analyzed two clinical cohorts totaling 581 participants, along with three in silico cohorts totaling 267 participants used to filter candidates for biological validity and reproducibility. Candidate miRNAs overexpressed in the blood of patients with colorectal cancer and adenomas were identified by small RNA sequencing and retained if they showed significant overexpression in at least two of the discovery and in silico cohorts. The final panel comprised 19 cell-free and 20 exosomal miRNAs.
The selected biomarkers were then measured using reverse transcription quantitative polymerase chain reaction in the training and testing cohorts. Using a machine learning model, the investigators built and locked separate models for colorectal cancer and advanced adenomas and established diagnostic thresholds in a 535-participant training cohort. The thresholds were then tested in a separate 230-participant cohort.
The primary outcomes were sensitivity for colorectal cancer and advanced adenomas and specificity for advanced neoplasia among eligible participants aged 45 years or older.
Test performance
The test distinguished colorectal cancer and advanced adenomas from nondisease controls with an area under the receiver operating characteristic curve of 0.95 in the testing cohort, which comprised 55 participants with colorectal cancer, 58 with advanced adenomas, 58 with low-risk adenomas, and 59 with negative colonoscopy findings.
The investigators reported the following:
Colorectal cancer: The test identified 50 of 55 cancers overall, a sensitivity of 91%, and 36 of 39 screening-relevant stage I to III cancers, a sensitivity of 92%. By stage, sensitivity was 96% for stage 0 to I, 91% for stage II, and 67% for stage III. All nine stage IV cancers were detected.
Advanced adenomas: The test identified 47 of 58 advanced adenomas, a sensitivity of 81%. Among the 23 participants whose adenomas qualified as advanced for reasons other than size — high-grade dysplasia or five or more adenomas — the test detected 17, a sensitivity of 74%. Sensitivity was 67% for high-grade dysplasia, 67% for sessile serrated lesions, and 82% for adenomas with villous histology.
Specificity: Of 117 participants without advanced colorectal neoplasia, 99 tested negative, a specificity of 85%. Among the 59 participants with negative colonoscopy findings, 49 tested negative, a specificity of 83%.
The assay met its prespecified performance targets for any-stage and stage I to III colorectal cancer. For advanced adenomas, the 81% point estimate cleared the study’s 80% threshold but the lower bound of the confidence interval around that estimate did not. That lower-bound requirement mirrors criteria traditionally applied to invasive colorectal cancer, the authors wrote, and whether a lower statistical bar is clinically acceptable for advanced adenomas — given the extended window for malignant transformation — remains a topic of ongoing discussion, provided the assay maintains high sensitivity and specificity for early-stage cancer.
Limitations
The investigators noted several limitations. The outcome-enriched design included a higher prevalence of cancer and advanced adenomas than would occur in an unselected average-risk population, and the authors wrote that this may have inflated the observed performance metrics. The study population comprised primarily individuals of Asian and White backgrounds, limiting generalizability. Because all participants underwent colonoscopy, the design could not determine whether blood-based testing would increase screening participation. In a routine screening population, where colorectal cancer and advanced adenoma prevalence are roughly 0.7% and 7% to 10%, the 85% specificity would produce higher colonoscopy referral rates than current fecal tests — rates that would “overwhelm colonoscopy services in many countries,” Drs. Winter and Symonds wrote. The thresholds were set using a method that assigns equal weight to sensitivity and specificity, which the authors wrote may require reweighting for clinical implementation. Data on lifestyle factors, comorbidities, and medications were unavailable, and the cohorts were classified by colonoscopy and histopathology findings rather than recruitment stream, so performance could not be stratified by how participants entered the study. The authors also acknowledged that confidence intervals for subgroups, including early-stage colorectal cancer, were relatively wide.
Looking ahead
The findings provide a rationale for advancing the assay into large-scale prospective studies to determine its performance, clinical utility, and potential role within population-based screening strategies, the investigators wrote. Drs. Winter and Symonds added that the test must demonstrate its accuracy, establish its effectiveness relative to current blood and fecal tests, and ultimately show it can reduce colorectal cancer incidence and mortality.
If prospectively validated, Dr. Goel told GI & Hepatology News that he envisions the test not as a replacement for colonoscopy or every other established screening option, but as an additional option for average-risk, screening-eligible adults who might otherwise remain unscreened. Positive results would still require follow-up colonoscopy.
Against the backdrop of the study’s enriched design, Dr. Goel said the biggest unanswered question is whether the observed performance, particularly the high sensitivity for advanced adenomas, will hold up in a large, real-world screening population.
Addressing how the specificity trade-off should be evaluated, he said, “I think the more meaningful question for the prospective study is not, ‘Can we reach a particular specificity number?’ It is: Can we find a clinically acceptable operating point that preserves high detection of advanced adenomas and early cancers while producing a colonoscopy referral rate that patients and health systems can realistically manage?”
He concluded, “The real test … will be whether using the test helps us find and remove more precancerous lesions in people who otherwise might never have been screened.”
The National Institutes of Health and AIRC Foundation funded the study. The study authors declared no competing interests. Among the commentary authors, Dr. Symonds reported institutional funding from Eiken Chemical Company and Health First Systems (UK) for investigator-led research on colorectal cancer screening tests; Dr. Winter reported no conflicts of interest.