Objective To systemically evaluate the accuracy of f/t-PSA for diagnosing prostate cancer with a t-PSA level of 4-10ng/mL through meta-analysis. Methods A literature search of CBM, VIP, CNKI and Wanfang Data from 1999 to 2009 was performed. Related journals were also searched manually. Two reviewers independently assessed trial quality according to QUADAS items. Heterogenous studies and meta-analysis were conducted by Meta-Disc1.4 software. The analysis was based on different critical values of f/t-PSA (0.1, 0.15, 0.2, 0.25, and 0.3). Results Total 18 studies involving 2217 subjects were included. No threshold effect was found. But there was heterogeneity due to other factors. The meta–analysis showed that, the sensitivity of f/t-PSA with the critical value of 0.15 for the diagnosis of prostate cancer with a t-PSA level of 4-10ng/mL was 75% (95%CI 70%-79%), and the specificity was 81% (95%CI 78%-84%). The area under SROC curve was 0.883 5, and the Q index was 0.814 0. Conclusion The f/t-PSA is a better index for diagnosing prostate cancer with t-PSA levels between 4 and 10ng/mL. And it is reasonable to consider 0.15 as a more suitable critical value.