Objective To systematically assess the diagnostic efficacy of circRNA in detecting CRC.

Methods Eligible studies were obtained by searching the databases, such as PubMed, EMBASE, CNKI, etc. in line with the inclusion and exclusion criteria. Study quality and bias were judged by QUADAS-2 tool. The overall effect size, including pooled diagnostic parameters, was combined based on STATA 12.0 software. The causes of heterogeneity were investigated by conducting the subgroup, sensitivity and meta-regression analyses.

Results We included 17 eligible studies which contained 1873 CRC cases and 1573 paired controls. For CRC detection, circRNA expression harbored a pooled sensitivity, specificity and AUC of 0.79 (95%CI: 0.76-0.81), 0.69 (95%CI: 0.64-0.73) and 0.81, respectively. Subgroup analysis showed that serum-derived circRNA testing (AUC=0.89, DOR=16.79) harbored higher diagnostic efficacy than plasma (AUC=0.79, DOR=7.08) and tissue (AUC=0.76, DOR=6.87); oncogenic circRNAs achieved higher diagnostic capacity in detecting CRC than the anti-cancer circRNAs (AUC: 0.81 vs. 0.76; DOR: 8.54 vs. 6.21); however, circRNA testing quantified based on GAPDH showed comparable diagnostic capacity with the overall pooled effect.

Conclusion Abnormally-expressed circRNAs have relatively high diagnostic efficacy in diagnosing CRC and may be rated as auxiliary biomarkers for CRC detection.