Our systematic review summarizes the evidence concerning the accuracy of serum diagnostic and prognostic tests for colorectal cancer (CRC). The databases MEDLINE and EMBASE were searched iteratively to identify the relevant literature for serum markers of CRC published from 1950 to August 2012. The articles that provided adequate information to meet the requirements of the meta-analysis of diagnostic and prognostic markers were included. A 2-by-2 table of each diagnostic marker and its hazard ratio (HR) and the confidence interval (CI) of each prognostic marker was directly or indirectly extracted from the included papers, and the pooled sensitivity and specificity of the diagnostic marker and the pooled HR and the CI of the prognostic marker were subsequently calculated using the extracted data. In total, 104 papers related to the diagnostic markers and 49 papers related to the prognostic serum markers of CRC were collected, and only 19 of 92 diagnostic markers were investigated in more than two studies, whereas 21 out of 44 prognostic markers were included in two or more studies. All of the pooled sensitivities of the diagnostic markers with > = 3 repetitions were less than 50%, and the meta-analyses of the prognostic markers with more than 3 studies were performed, VEGF with highest (2.245, CI: 1.347-3.744) and MMP-7 with lowest (1.099, CI: 1.018-1.187)) pooled HRs are presented. The quality of studies addressing the diagnostic and prognostic accuracy of the tests was poor, and the results were highly heterogeneous. The poor characteristics indicate that these tests are of little value for clinical practice.