超声引导下经会阴前列腺穿刺活检阳性风险预测模型的建立和验证

Development and validation of a predictive model for the risk of positive after transperineal ultrasound-guided prostate biopsy

目的建立超声引导下经会阴前列腺穿刺活检阳性风险预测模型,并验证模型的预测效能。方法回顾性分析2019年1月至2021年8月于皖南医学院弋矶山医院接受超声引导下经会阴前列腺穿刺活检的475例患者的临床病理资料。其中2019年1月至2020年12月的367例为开发队列,2021年1月至2021年8月的108例为验证队列。开发队列年龄(68.86±9.00)岁,总前列腺特异性抗原(tPSA)13.6(8.6,23.3)ng/ml,前列腺影像报告和数据系统(PI-RADS)评分4(3,4)分。验证队列年龄(68.89±8.67)岁,tPSA 13.1(8.7,25.6)ng/ml,PI-RADS评分4(3,5)分。采用单因素和多因素logistic回归分析开发队列中前列腺穿刺阳性的危险因素,根据危险因素建立列线图预测模型。在验证队列中验证模型的性能。在开发队列中采用受试者工作特征(ROC)曲线、校准图和决策曲线分析评估预测模型的性能。在验证队列中利用ROC曲线评估该模型对前列腺癌的诊断价值。结果本研究开发队列中前列腺癌180例(49.05%),非前列腺癌187例(50.95%);验证队列中前列腺癌53例(49.07%),非前列腺癌55例(50.93%)。基于开发队列的单因素和多因素logistic回归分析结果示,前列腺穿刺阳性的危险因素包括年龄(OR=1.059,P=0.003)、血小板与单核细胞比值(PMR)(OR=0.002,P=0.011)、f/t PSA(OR=0.009,P=0.020)及PI-RADS评分(OR=3.376,P<0.001)。基于上述危险因素构建的列线图模型具有良好的预测能力,其在开发队列的ROC曲线下面积(AUC)为0.845(95%CI 0.794~0.895),霍斯默-莱梅肖拟合优度检验结果提示该模型校准良好(χ^(2)=1.476,P=0.224)。决策曲线分析结果表明,在预测模型的阈概率为10%~90%时,有较高的临床应用价值。验证队列的ROC曲线示AUC为0.869(95%CI 0.797~0.941)。结论通过患者年龄、PMR、f/t PSA和PI-RADS评分指标构建的列线图模型,对前列腺癌有较高的预测效能,具有一定的临床应用价值。

Objective To develop a nomogram to predict the probability of prostate cancer after transeperineal prostate biopsy,and verify the diagnostic efficacy and clinical applicable value of the model.Methods The clinicopathologic data of 475 patients who underwent prostate biopsy at Yijishan Hospital of Wannan Medical College between January 2019 to August 2021 were retrospectively reviewed.Of all the patients,367 patients from January 2019 to December 2020 were in the development cohort and 108 patients from January 2021 to August 2021 were in the validation cohort.Patients in the development cohort were(68.86±9.00)years old.The tPSA level was 13.6(8.6,23.3)ng/ml,and Prostate Imaging Reporting and Data System(PI-RADS)score was 4(3,4)points.Patients in the validation cohort were(68.89±8.67)years old.The tPSA was 13.1(8.7,25.6)ng/ml,and PI-RADS score was 4(3,5)points.Univariate and multivariate logistic regression were used to analyze prostate cancer risk factors in the development cohort.Then the nomogram prediction model was established by the risk factors.The prediction model's performance was evaluated using receiver operating characteristic(ROC)curves,calibration maps,and decision curve(DCA)analysis in the development cohort.The performance of the model was verified in the validation cohort.Results The pathological results showed 180 patients with prostate cancer and 187 patients without prostate cancer in the development cohort.The validation cohort included 53 patients with and 55 without prostate cancer.Based on the results of the univariate and multivariate logistic regression analysis,this model incorporates factors including age(OR=1.059,P=0.003),platelet-to-monocyte ratio(PMR)(OR=0.002,P=0.011),f/tPSA(OR=0.009,P=0.020),and PI-RADS score(OR=3.076,P<0.001).The calibration curve revealed a great agreement.Internal validation of the nomogram showed that the area under the ROC curve was 0.845(95%CI 0.794-0.895).The Hosmer-Lemeshow test was also performed(χ^(2)=1.476,P=0.224).The validation group with an area under the ROC curve was 0.869(95%CI 0.797-0.941).The results of the decision curve analysis indicated that the decision curve was located above the positive and negative lines in the threshold range of 10%to 90%,within which the model has clinical application.Conclusions The nomogram,which combines patient age,PMR,f/t PSA,and PI-RADS scores,has high predictive efficacy for prostate cancer and has clinical application value.