Early detection of colorectal neoplasia: Combination of eight cancer-associated blood-based protein biomarkers.

Abstract

488 Background: Assessment of 8 blood based protein markers for identifying colorectal neoplasia. Methods: 4,698 subjects scheduled for first time ever colonoscopy with symptoms potentially attributable to colorectal neoplasia were accrued. Plasma levels of AFP, CA19-9, CEA, hs-CRP, CyFra21-1, Ferritin, Galectin-3, and TIMP-1 were determined by using the Abbott ARCHITECT i2000 automated immunoassay platform. Three clinical endpoints were chosen for this presentation: 1. CRC, 2. CRC and high risk adenoma, 3. CRC and other cancers. Logistic regression was used for statistical analysis of each endpoint. A final reduced model was selected choosing the best combination of 4 markers with the highest likelihood score statistic. Results: 512 CRC’s were identified, 323 CC and 189 RC. Malignancies other than CRC were detected in 177 subjects, 699 adenomas were identified including 298 with high risk, 1342 subjects had non-invasive bowell disease, and 1978 had a “clean” colon. Univariate analysis of each endpoint demonstrated that all markers were statistically significant. Multivariable logistic regression showed that the blood based markers in combination significantly predicted the 3 endpoints. The reduced model resulted in the selection of CEA, hs-CRP, CyFra21-1 and Ferritin for the 3 endpoints. The area under the receiver operating characteristic curves were 0.83, 0.74 and 0.82 respectively. The postive predictive value (PPV) at 90% sensitivity was 18% for endpoint 1 and the negative predictive value (NPV) was 97%. For endpoints 2 and 3, the PPV’s were 25% and 20% and the NPV’s were 93% and 96%. Conclusions: A subset of 4 biomarker in combination identify subjects with a high risk of CRC.