Abstract
BACKGROUND: Patients with colorectal cancer often present with advanced disease and concomitant poor prognosis. The best known serum biomarker, carcinoembryonic antigen (CEA) is not recommended for screening because of its limited specificity and sensitivity. A number of other circulating proteins have been suggested to be diagnostically useful but individually none of these has proved to be of sufficient sensitivity or specificity to establish a role in routine clinical practice. Here, we test the hypothesis that combining several of these biomarkers will improve diagnostic efficacy. METHODS: To select the markers for our model we screened CEA and 26 other candidate biomarkers. Four candidates were selected and their concentrations determined in the serum of 239 patients (106 colorectal cancer patients and 133 non-cancer subjects). RESULTS: Class prediction models based on CEA, DR-70 and sCD26 produced a modest increase in detection accuracy over CEA alone, particularly for early stage cancers. The sensitivity and specificity required for a clinically useful test was not reached. CONCLUSION: It is unlikely that a biomarker panel comprised of the currently available serum markers will generate a clinically useful diagnostic test for colorectal cancer. Our findings reiterate the urgent need to discover novel biomarkers for the detection of colorectal cancer.