OBJECTIVETo systematically evaluate prostate health index (PHI) in predicting the results of prostate biopsy in the diagnosis of prostate cancer.

METHODSWe retrieved the literature relevant to the value of PHI in predicting prostate cancer at prostate biopsy published from the inception to February 2014 in the databanks of PubMed (1966 - 2014), CNKI (1982 - 2014), VIP (1989 - 2014), and Cochrane Library (1999 - 2014). According to the inclusion and exclusion criteria and retrieval strategies, we extracted the data, evaluated the quality of the included literature, and performed meta-analysis using the Meta-Disc 1.4 software.

RESULTSTotally, 64 articles were identified, of which 52 were excluded and 12 included with 1 430 cases of prostate cancer and 2 159 normal or BPH controls. There was heterogeneity among the included studies. Based on the random effect model, in predicting the results of prostate biopsy in the diagnosis of prostate cancer, serum PHI exhibited a pooled sensitivity of 55.1% (95% CI: 0.525 - 0.577), a pooled specificity of 71.5% (95% CI: 0.695 - 0.734), a positive likelihood ratio (P-LR) of 2.379 (95% CI: 1.922 - 2.943), a negative likelihood ratio (N-LR) of 0.515 (95% CI: 0.428 - 0.619), a diagnosis odds ratio (DOR) of 5.268 (95% CI: 3.870 - 7.170), a summary receiver operating characteristic curve--area under the curve (SROC AUC) of 0.7578, and a Q * index of 0.6999.

CONCLUSIONSerum PHI plays a role of auxiliary diagnosis in detecting prostate cancer and can be used in the prediction of the results of prostate biopsy.