Lee Su-Min, Liyanage Sidath H, Wulaningsih Wahyu, Wolfe Konrad, Carr Thomas, Younis Choudhry, Van Hemelrijck Mieke, Popert Rick, Acher Peter
Department of Urology, Southend University Hospital, Westcliff-on-Sea, Essex, UK.
Department of Radiology, Southend University Hospital, Westcliff-on-Sea, Essex, UK.
Urol Oncol. 2017 Nov;35(11):664.e11-664.e18. doi: 10.1016/j.urolonc.2017.07.018. Epub 2017 Aug 8.
To develop and internally validate a nomogram using biparametric magnetic resonance imaging (B-MRI)-derived variables for the prediction of prostate cancer at transperineal sector-guided prostate biopsy (TPSB).
SUBJECTS/PATIENTS AND METHODS: Consecutive patients referred to our institution with raised prostate-specific antigen (PSA), abnormal prostate examination, or persistent suspicion of prostate cancer after previous transrectal biopsy between July 2012 and November 2015 were reviewed from a prospective database. All patients underwent prebiopsy B-MRI with T2-weighted and diffusion-weighted imaging sequences, followed by 24 to 40 core TPSB with additional targeted cores using cognitive registration. Univariable and multivariable logistic regression analysis was used to determine predictors of prostate cancer outcomes. Multivariable coefficients were used to construct 2 MRI-based nomograms to predict any and significant (Gleason 4 or maximum cancer core length ≥6mm) prostate cancer at TPSB. Bootstrap resamples were used for internal validation. Accuracy was assessed by calculating the concordance index.
In total, 615 men were included in the study. Prostate cancer was diagnosed in 317 (51.5%) men with significant cancer diagnosed in 237 (38.5%) men. Age, Prostate Imaging Reporting and Data System (PI-RADS) score, PSA, PSA density, and primary biopsy were predictors of prostate cancer at TPSB on univariable analysis (P<0.0001). PSA showed strong correlation with PSA density and was excluded. The remaining variables were all independent predictors of prostate cancer on multivariable analysis (P<0.0001) and used to generate the nomograms. Both nomograms showed good discrimination for prostate cancer, with a concordance index of 87% for any cancer and 92% for significant disease. Using a nomogram-derived probability threshold of<15%, 111 (18.0%) biopsies can be saved, at the expense of 3 missed significant prostate cancers.
These internally validated MR-based nomograms were able to accurately predict TPSB outcomes for prostate cancer, especially significant disease. Our findings support the combination of prebiopsy MRI results and clinical factors as part of the biopsy decision-making process.
开发并在内部验证一种列线图，该列线图使用双参数磁共振成像（B-MRI）得出的变量来预测经会阴扇形引导前列腺穿刺活检（TPSB）时的前列腺癌。
受试者/患者与方法：回顾性分析2012年7月至2015年11月期间因前列腺特异性抗原（PSA）升高、前列腺检查异常或先前经直肠活检后仍怀疑患有前列腺癌而转诊至本机构的连续患者，数据来自前瞻性数据库。所有患者在活检前均接受了带有T2加权和扩散加权成像序列的B-MRI检查，随后进行了24至40针的TPSB，并使用认知配准法进行了额外的靶向穿刺。采用单变量和多变量逻辑回归分析来确定前列腺癌结局的预测因素。多变量系数用于构建两个基于MRI的列线图，以预测TPSB时的任何前列腺癌和显著前列腺癌（Gleason 4级或最大癌芯长度≥6mm）。采用自抽样重采样进行内部验证。通过计算一致性指数评估准确性。
本研究共纳入615名男性。317名（51.5%）男性被诊断为前列腺癌，其中237名（38.5%）男性被诊断为显著前列腺癌。单变量分析显示，年龄、前列腺影像报告和数据系统（PI-RADS）评分、PSA、PSA密度和初次活检是TPSB时前列腺癌的预测因素（P<0.0001）。PSA与PSA密度显示出强相关性，故将其排除。其余变量在多变量分析中均为前列腺癌的独立预测因素（P<0.0001），并用于生成列线图。两个列线图对前列腺癌均显示出良好的区分度，任何前列腺癌的一致性指数为87%，显著疾病的一致性指数为92%。使用列线图得出的概率阈值<15%，可减少111次（18.0%）活检，但代价是漏诊3例显著前列腺癌。
这些经过内部验证的基于MRI的列线图能够准确预测前列腺癌的TPSB结局，尤其是显著疾病。我们的研究结果支持将活检前MRI结果与临床因素相结合，作为活检决策过程的一部分。
