基于PI-RADS v2评分在PSA 4~10ng/mL患者前列腺癌预测模型的构建及验证  

Establishment of the prediction model based on PI-RADS v2 score for prostate cancer in patients with PSA range from 4 to 10ng/mL

作　　者：罗志强 黄建文[1,2] 曹乃龙 胡晓勇[1,2] 张炯 郭辉[1,2] 宋鲁杰 傅强[1,2] LUO Zhiqiang;HUANG Jianwen;CAO Nailong;HU Xiaoyong;ZHANG Jiong;GUO Hui;SONG Lujie;FU Qiang(Department of Urology,Shanghai Jiao Tong University Affiliated Sixth Peopled Hospital,Shanghai,200233,China;Shanghai Eastern Urological Reconstruction and Repair Institute)

机构地区：[1]上海交通大学附属第六人民医院泌尿外科,上海200233 [2]上海东方泌尿修复重建研究所

出　　处：《临床泌尿外科杂志》2022年第2期109-113,118,共6页Journal of Clinical Urology

基　　金：上海市第六人民医院院级科学研究基金(No:YNLC201816)。

摘　　要：目的:联合核磁共振第2版前列腺影像报告和数据系统(prostate imaging reporting and data system version 2,PI-RADS v2)与其他临床指标,建立前列腺特异性抗原(PSA)"灰区"(4~10 ng/mL)患者的前列腺癌(prostate cancer,PCa)预测模型,并验证模型的准确性。方法:回顾性分析2016年1月—2020年12月在我院行前列腺穿刺活检的PSA"灰区"患者的临床资料,包括年龄、血清PSA、游离/总PSA比值(f/tPSA)、前列腺体积(prostate volume,PV)、PSA密度(PAS density,PASD)、经直肠前列腺超声(transrectal ultrasonography,TRUS)结果、PI-RADS v2评分等临床指标。其中2016年1月—2019年12月纳入病例作为模型构建组,2020年1月—2020年12月为模型验证组。采用二元logistic单因素及多因素分析计算PCa的独立预测因素,依据回归系数建立预测模型,通过ROC曲线下面积(area under curve,AUC)对比预测模型与临床指标对PSA"灰区"患者的PCa预测价值,并在模型验证组检验模型最佳预测值的准确性。结果:模型构建组共纳入患者349例,年龄(69.4±8.0)岁,PSA(7.9±1.9)ng/mL,f/tPSA(0.15±0.70),PSAD(0.18±0.10),PV(53.5±28.5)mL,TRUS异常90例(25.8%),PI-RADS v2(3.79±0.90)分。病理诊断PCa 122例(35.0%)。在恶性肿瘤组中,Spearman相关性分析结果显示Gleason评分与PI-RADS v2评分呈正相关,多因素分析显示年龄(OR=1.083,P=0.001)、TRUS(OR=1.832,P=0.040)、PV(OR=0.975,P=0.003)、PI-RADS v2评分(OR=3.160,P<0.001)是PCa的独立预测因素,建立预测模型为-8.665+0.080×年龄+0.157×PASD-0.025×PV+0.605×TRUS+1.151×PI-RADS v2,ROC-AUC比较示预测模型(0.855)>PI-RADS v2(0.794)>年龄(0.680)>TURS(0.650)>PSAD(0.630),差异有统计学意义(P<0.05),ROC曲线提示模型最佳预测界值为-0.6595时敏感度为83.6%,特异度为71.4%。在模型验证组,患者共118例,病理诊断PCa 34例(28.8%),将临床指标带入预测模型计算预测值,ROC曲线示模型最佳预测界值>-0.6595的诊断效能高于PI-RADS v2评分(AUC 0.803 vs.0.71Objective:To establish a more efficient prediction model for prostate specific antigen(PSA)"grey area"(4-10 ng/mL)patients with prostate cancer(PCa)through combining prostate imaging reporting and data system version 2(PI-RADS v2)and other clinical data,and test the accuracy of the model.Methods:A retrospective study was performed from Jan.2016 to Dec.2020,and patients’clinical data including age,PSA,f/tPSA,prostate volume(PV),PSA density(PSAD),transrectal ultrasonography(TRUS)result and PI-RADS v2 score were collected through reviewing medical record system.Cases from Jan.2016 to Dec.2019 were used for model establishment and cases from Jan.2020 to Dec.2020 for model verification.The correlation between each factor and PCa was analyzed by binary univariate and multivariate logistic regression.The prediction models were established according to the regression coefficient.The area under the curve(AUC)was used to compare the prediction value of each factor and the prediction model.Meanwhile,in the verification group,the accuracy of the model’s best predictions was verified.Results:In total 349 were enrolled in this study,age(69.4±8.0)years old,PSA(7.9±1.9)ng/mL,f/tPSA(0.15±0.70),PSAD(0.18±0.10),PV(53.5±28.5)mL,PI-RADS v2 score(3.79±0.90),90 patients(25.8%)with TRUS positive results.Biopsy revealed 122 cases(35.0%)with carcinoma.In the malignant tumor group,the results of correlation analysis between Gleason score and PI-RADS v2 score showed that Gleason score was positively correlated with PI-RADS score.Multivariate analysis showed that age(OR=1.083,P=0.001),TRUS positive(OR=1.832,P=0.040),PV(OR=0.975,P=0.003)and PI-RADS v2 score(OR=3.160,P<0.001)were independent risk factors for PCa.The prediction model was-8.665+0.080×Age+0.157×PASD-0.025×PV+0.605×TRUS+1.151×PI-RADS v2.The comparation of ROC-AUC indicated prediction model(0.855)>PI-RADS v2(0.794)>age(0.680)>TURS(0.650)>PSAD(0.630),which were significant difference(P<0.05).When best prediction critical value was-0.6595,it has a sensitivity of 83.6%a

关 键 词：前列腺癌 核磁共振 前列腺特异性抗原 预测 

分 类 号：R737.25[医药卫生—肿瘤]