Abstract： [Purpose]To construct risk prediction models for disease progress of prostate cancer.[Methods]The clinical and pathological data of 1 940 patients with prostate cancer treated in Army Medical Center of PLA from June 2000 to September 2022 were retrospectively analyzed.A total of 24 clinical data were collected to describe their characteristics and analyze the trend of disease characteristics over time.The patients were classified into localized or locally advanced prostate cancer group(M0 group)and metastatic prostate cancer group(M1 group).In M0 group,biochemical recurrence(BCR)was used as the endpoint event,while metastatic castration resistance prostate cancer(mCRPC)was used as the endpoint event in M1 group.Univariate and multivariate Cox proportional risk regression models were used to analyze the in-dependent risk factors for BCR in M0 group and mCRPC in M1 group.And then risk prediction models were established and their performance was evaluated.[Results]The median age of 1 940 patients was 72(66,77)years old,and 56.45％patients were aged＞70 years old.In M0 group,80.59％of the patients were classified as high-risk using D-Amic prognostic risk analysis,while in M1 group 55.30％of the pa-tients were found with high tumor load.In M0 group,high pathologic Gleason score(＞7,P=0.042),high pathologic stage(≥pT3,P=0.035),and positive margins(P＜0.001)were independent risk factors for BCR.A prediction model was constructed with above risk factors,the area under ROC curve(AUC)of the model in predicting BCR was 0.736(95％CI:0.675～0.797,P＜0.001)and C-index was 0.736(95％CI:0.678～0.793,P＜0.001).In M1 group,high tPSA(＞20 ng/mL,P=0.021),high Gleason score on needle biopsy(＞7,P＜0.001),and more bone metastases(＞3,P=0.026)were independent risk factors for mCRPC.A prediction model was constructed with above three indicators,AUC of the model in predicting mCRPC was 0.652(95％CI:0.590～0.714,P＜0.001),and the C-index was 0.680(95％CI:0.620～0.740,P＜0.001).The calibration curves of these 2 prediction models showed good predictive performance.[Conclusion]In localized or lo-cally advanced prostate cancer,patients with high Gleason score,high pathologic stage,and positive margins are more likely to have BCR.In metastatic prostate cancer,patients with high tPSA,high Glea-son score,and more bone metastases are more quickly to develop mCRPC.The prediction models based on the above risk factors show good predictive performance,which may provide a reference for clinical deci-sion-making in the treatment of prostate cancer patients and during the follow-up.