Objective:To assess the value of MR spectroscopy (MRS) in the diagnosis of prostate cancer by meta analysis. Methods: Prospective studies were selected from the MEDLINE, Ovid, Embase databases, Springer, Elsevier, China National Knowledge Infrastructure. According to the suggestion, results were determined by the ratio of (Cho+Cr)/Cit . If this ratio was less than 0.75, negative result was respectively determined, and the definitive tumor was diagnosed as this threshold. According to the assessment standard of Evidence-based Medicine, English and Chinese literature in Grade A and B on MRS imaging was included. According to homogeneity test, different effect models were chosen to calculate different pooled weighted values of sensitivity, specificity and the corresponding 95％ confidence intervals (95％ CI). Summary receiver operating characteristic (SROC) curves were used to assess the results.Funnel plot was used to analyze publication bias. Results: According to the assessment standard of Evidence-based Medicine, only 5 papers in Grade B were included in this research. The pooled weighted sensitivity and its 95％ confidence interval is 82％ (73％, 89％) and the pooled weighted sensitivity and its 95％ confidence interval is 68％(58％, 76％). The AUC (area under curve) is 83. 40％. An asymmetric funnel plot suggested two missing studies leading to publication bias. Conclusion: If the ratio of (Cho+Cr)/Cit is regarded as the diagnostic criteria in detecting prostate cancer by MRS, meta-analysis suggests this method has a better diagnostic value to detect the malignant prostate mass but the sensitivity needs to be improved. We hope to support a method and requirement about diagnostic test. Performing perspective register and improving quality of study design is the only way to reduce the bias and get real information of disease.