Previous studies have evaluated the accuracy of serum and urinary measurements of cytokeratin-19 fragment (CYFRA 21-1) for the diagnosis of bladder cancer; however, the results have been inconsistent. The aim of this study was to evaluate the overall accuracy of CYFRA 21-1 for the diagnosis of bladder cancer. We performed a search for English-language publications reporting on the detection of CYFRA21-1 levels for the diagnosis of bladder cancer through November 2, 2014, using public medical databases, including EMBASE, Web of Science, and Medline. The quality of the studies was assessed by revised QUADAS tools. The performance characteristics were pooled and analyzed using a bivariate model. Publication bias was explored with the Deek's test. Sixteen studies, with a total 1,262 bladder-cancer patients and 1,233 non-bladder-cancer patients, were included in the study. The pooled sensitivities for serum and urine CYFRA 21-1 were 0.42 (95 % confidence interval (CI), 0.33-0.51) and 0.82 (95 % CI, 0.70-0.90), respectively. The corresponding specificities were 0.94 (95 % CI, 0.90-0.96) and 0.80 (95 % CI, 0.73-0.86), respectively. The areas under the summary receiver-operating-characteristic curves for serum and urine CYFRA 21-1 were 0.88 (95 % CI, 0.85-0.91) and 0.87 (95 % CI, 0.84-0.90), respectively. The major design deficiencies of the included studies were participant-selection bias, potential review, and verification bias. Therefore, we concluded that both serum and urine CYFRA 21-1 served as efficient indexes for bladder-cancer diagnosis. Additional, well-designed studies should be performed to rigorously evaluate the diagnostic value of CYFRA 21-1 for bladder cancer.