目的 探讨前列腺穿刺结果的相关预测因素，构建前列腺穿刺结果的预测列线图模型。方法 回顾性收集并分析2010—2017年于复旦大学附属华山医院行超声引导下经会阴前列腺穿刺患者的病例资料，根据穿刺结果将患者分为非肿瘤组与前列腺癌组。依据病理分化程度，将前列腺癌组进一步分为低级别前列腺癌亚组（low-grade prostate cancer，LGPCa，Gleason≤3+4）与高级别前列腺癌亚组（high-grade prostate cancer，HGPCa，Gleason≥4+3）。比较不同组别间临床特征的差异，分析不同参数对穿刺结果的预测价值。将所有患者的80%随机列入建模组，另20%列入验证组。建立多参数预测列线图模型，并进行内部人群验证。结果 选取穿刺患者共1 585人，排除研究数据缺失者、前列腺特异抗原（prostate specific antigen，PSA）>100 ng/mL者，最终纳入统计1 331人。病理确诊为前列腺癌共计519人，其中LGPCa患者249人，HGPCa患者270人。在非肿瘤组与前列腺癌组、LGPCa组与HGPCa组的比较中，患者PSA、前列腺体积、直肠指检和经直肠前列腺超声（transrectal ultrasound，TRUS）结果均存在显著差异。前列腺癌预测模型纳入年龄（OR=1.056）、PSA（OR=1.063）、前列腺体积（OR=0.960）、直肠指检（OR=5.991）和TRUS（OR=1.717），该模型的受试者工作曲线下面积（area under the receiver operating characteristic curve，AUC）达到0.895，显著高于PSA密度（PSA density，PSAD）和PSA。HGPCa预测模型纳入PSA（OR=1.032）、前列腺体积（OR=0.983）、直肠指检（OR=4.803）和TRUS（OR=1.987），其预测效力AUC达到0.872，亦显著高于PSAD和PSA。据此绘制穿刺预测列线图。验证组AUC同样提示对前列腺癌（0.846）和HGPCa（0.819）具有较高预测效力。结论 本研究发现年龄、PSA、前列腺体积、直肠指检和TRUS均为前列腺穿刺结果的独立预测因素，据此建立前列腺癌和HGPCa的多参数预测列线图模型，具有较高预测效力。列线图的应用有助于医患直观高效沟通，为科学制定个体化穿刺策略提供依据。

Objective To explore the predictors of prostate biopsy outcomes,and to develop the prediction nomogram. Methods The patients receiving ultrasound-guided transperineal prostate biopsy in Huashan Hospital,Fudan University from 2010-2017 were retrospectively enrolled. They were categorized into non-cancer group and prostate cancer group.The patients in prostate cancer group were further categorized into low-grade prostate cancer (LGPCa,Gleason ≤ 3+4) group and high-grade prostate cancer (HGPCa,Gleason ≥ 4+3) group according to the biopsy outcomes.The difference in clinical characteristics among different groups were compared, and the predictive value of different characteristics on prostate biopsy outcomes were analysed.A total of 80% of all patients were randomized to model-developing group,and the other 20% of the patients were randomized to validation group.The multivariate prediction models and related nomograms were developed,and internal validation was performed. Results A total of 1 585 patients receiving prostate biopsy were enrolled.The patients with missing data or with prostate specific antigen (PSA)>100 ng/mL were excluded.Finally,a total of 1 331 patients were enrolled for statistics.After biopsy,519 patients were diagnosed with prostate cancer. In specific,249 out of them were diagnosed with LGPCa,270 out of them were diagnosed with HGPCa.In comparison among different groups, the PSA,prostate volume,digital rectal examination and transrectal ultrasound (TRUS) were significantly different.The prediction model for prostate cancer included age (OR=1.056),PSA (OR=1.063),prostate volume (OR=0.960)、digital rectal examination (OR=5.991)and TRUS (OR=1.717) as variables,and the area under the receiver operating characteristic curve (AUC) of this model was 0.895,which was significantly higher than that of PSA density (PSAD) and PSA.The prediction model for HGPCa included PSA (OR=1.032),prostate volume (OR=0.983),digital rectal examination (OR=4.803) and TRUS (OR=1.987),and the AUC was 0.872,which was also significantly higher than that of PSAD and PSA.The nomograms were developed based on these two prediction models.The internal validation also confirmed the high value of these models in predicting prostate cancer (AUC 0.846) and HGPCa (AUC 0.819).Conclusions The present study discovered that age,PSA,prostate volume,digital rectal examination and TRUS were independent predictors for the outcomes of prostate biopsy.The prediction models and nomograms for prostate cancer and HGPCa based on these variables had high predicting abilities.The application of these nomograms would help to improve the communication between doctors and patients,as well as provide evidence for making individualized biopsy strategy.