Objective: To evaluate the role of the serum testosterone level as an independent predictor of bone metastasis of prostate cancer., Methods: This study included 165 male patients with prostate cancer confirmed by biopsy. The patients were aged 58－78 (66.6±5.3) years and none had received androgen-deprivation therapy, chemotherapy or radiotherapy previously. We obtained the baseline clinical data from the patients, including prostate biopsy Gleason scores and the levels of serum prostate-specific antigen (PSA), total testosterone (TT), luteinizing hormone (LH), follicle-stimulating hormone (FSH), estradiol (E2), alkaline phosphatase (ALP) and prolactin. According to the results of bone scanning, we divided the patients into a bone metastasis and a non-bone metastasis group and screened out the differential factors by univariate analysis and the independent predictor of bone metastasis using the multivariate non-conditional logistic regression model., Results: Univariate analysis showed no statistically significant differences between the bone metastasis and non-bone metastasis groups in age (P = 0.126) or the levels of serum LH (P = 0.930), FSH (P = 0.763) and E2 (P = 0.256), but that the former had remarkably higher Gleason scores (P < 0.01), total PSA (P <0.01) and ALP (P <0.01) but a lower TT level than the latter (P = 0.013). According to the results of multivariate logistic regression analysis, serum ALP (P <0.01, OR = 1.018 ［1.011－1.026］) and total PSA (P <0.01, OR = 1.029 ［1.015－1.044］) could be regarded as independent predictors of bone metastasis of prostate cancer but not low serum TT (P = 0.531, OR = 0.999 ［0.996－1.002］) or biopsy Gleason score (P = 0.898, OR = 0.787 ［0.412－1.9559］)., Conclusions: The low level of serum testosterone is closely associated with but not an independent predictor of bone metastasis of prostate cancer.