Biomarker Panel May Aid Early Detection of Colorectal Cancer
A panel of 4 blood-derived biomarkers showed promise as an aid to early detection of colorectal cancer (CRC). The panel yielded a negative predictive value exceeding 90% for CRC, the combination of CRC and high-risk adenomas, and colorectal plus other cancers. The biomarker assay demonstrated fair to good performance characteristics, associated with the receiver operating characteristic (ROC) curve ranging between 0.70 and 0.80.
“We have demonstrated that a panel of blood-based markers can significantly predict the probabilities of the chosen end points,” said Ib J. Christensen, MSc, University of Copenhagen, Denmark. “In the future, we hope to improve the results by including new markers and performing alternative analytic approaches, such as the adaptive index model.”
Mr Christensen presented findings at the 2016 Gastrointestinal Cancers Symposium from a study designed to evaluate combinations of existing biomarkers to aid in the detection of colorectal neoplasia. After optimizing the biomarker combination, investigators evaluated biomarker-based approaches to early detection of CRC.
The investigators identified 8 biomarkers with evidence suggesting associations with CRC. The 8 markers were carcinoembryonic antigen (CEA), high-sensitivity C-reactive protein (hs-CRP), alpha-fetoprotein (AFP), ferritin, galectin-3, cytokeratin fragment 21-1 (CyFra21-1), cancer antigen 19-9 (CA19-9), and tissue inhibitor of metalloproteinase-1 (TIMP-1).
The investigators recruited patients referred for a first-ever colonoscopy in follow-up to symptoms suggesting the presence of colorectal neoplasia. Patients with a history of CRC were excluded.
Blood samples were obtained from almost 4700 patients. Colonoscopy identified colon cancer in 319 patients, rectal cancer in 193, colonic adenomas in 515, rectal adenomas in 174, other types of cancer in 177, and nonmalignant findings in 1342. The remaining 1978 patients had what Mr Christensen described as “clean colons.”
Assessment of the blood samples produced a separation among the biomarkers in terms of their ability to detect the chosen end points. A panel of 4 biomarkers—CEA, hs-CRP, ferritin, and CyFra21-1—performed as well as a panel comprising all 8 markers.
The final model included the 4 protein markers plus patient age and sex. For the end point of CRC plus high-risk adenomas, the model had an ROC of 0.71. For CRC only, the performance improved to 0.81.
The investigators calculated specificity and positive and negative predictive values (PPV, NPV) for different levels of sensitivity. With 90% sensitivity for CRC plus high-risk adenomas, the model had a specificity of 33%, PPV of 25%, and NPV of 93%. For 90% sensitivity to detect only CRC, the model had a specificity of 48%, PPV of 18%, and NPV of 97%. For CRC and other cancers, the panel had a PPV of 20% and 96%.