Aim Urine cytokeratin 20 (CK20) has been reported as a novel diagnostic biomarker for bladder cancer (BC). This meta-analysis aims to evaluate the diagnostic value of urine CK20 for BC. Methods A systematic search for literatures was performed till March 31, 2018. Sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR) and the area under the summary receiver operator characteristic curve (AUC) were analyzed. Results Twenty-seven studies from 22 articles met the eligible criteria. The pooled estimates were as follows: sensitivity, 0.79; specificity, 0.90; PLR, 8.17; NLR, 0.23; DOR, 35.29 and; AUC, 0.90. CK20 was more sensitive for the diagnosis of bladder urothelial carcinoma (BUC) than all types of BC (0.83 vs 0.75). For Tis/Ta tumors, the values were inferior to T1 and >= T2 stage tumors, that is, the sensitivity (0.74 vs 0.81 vs 0.86), PLR (7.95 vs 9.97 vs 10.75) and DOR (27.81 vs 47.09 vs 69.63). For grade 1 tumors, the values were inferior to grade 2 and >= grade 3 tumors, that is, the sensitivity (0.70 vs 0.82 vs 0.87), PLR (11.86 vs 13.63 vs 14.26) and DOR (36.63 vs 72.00 vs 101.71). RT-PCR or PCR showed superior sensitivity and specificity (0.80 vs 0.76, 0.91 vs 0.89) than immunostaining in detection of urine CK20 for BC. Conclusion Urine CK20 might be a potential noninvasive biomarker for BC, especially for BUC. The diagnostic accuracy of urine CK20 is improved with the progression of tumor stage and grade. In addition, RT-PCR or PCR shows better diagnostic performance in detecting urine CK20 than immunostaining.