上海交通大学学报(医学版) ›› 2025, Vol. 45 ›› Issue (10): 1361-1371.doi: 10.3969/j.issn.1674-8115.2025.10.011

• 论著 · 临床研究 • 上一篇    下一篇

机器人辅助腹腔镜根治性前列腺切除术后患者尿失禁的在线风险计算器和列线图预测模型

敦译霆, 赵婧, 冯成领, 李行健, 崔迪, 韩邦旻()   

  1. 上海交通大学医学院附属第一人民医院,上海 200080
  • 收稿日期:2025-02-13 接受日期:2025-06-06 出版日期:2025-10-28 发布日期:2025-10-28
  • 通讯作者: 韩邦旻,教授,博士;电子信箱:Hanbm@163.com
  • 基金资助:
    上海申康医院发展中心临床科技创新项目(SHDC12021105)

Online risk calculator and nomogram prediction model for urinary incontinence after robot-assisted laparoscopic radical prostatectomy

DUN Yiting, ZHAO Jing, FENG Chengling, LI Xingjian, CUI Di, HAN Bangmin()   

  1. Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200080, China
  • Received:2025-02-13 Accepted:2025-06-06 Online:2025-10-28 Published:2025-10-28
  • Contact: HAN Bangmin, E-mail: Hanbm@163.com.
  • Supported by:
    Program of Shanghai Shenkang Hospital Development Center(SHDC12021105)

摘要:

目的·开发列线图预测模型和在线风险计算器,预测机器人辅助腹腔镜根治性前列腺切除术(robot-assisted radical prostatectomy,RARP)后患者尿控情况。方法·纳入2022年9月至2024年12月在上海交通大学医学院附属第一人民医院接受RARP手术且具备术前前列腺磁共振成像资料的604例前列腺癌患者。所有患者按照7∶3的比例随机重采样并分为训练集(n=420)和验证集(n=184)。自术后1个月起,每个月对患者的尿控情况进行随访。应用最小绝对收缩和选择算子回归(least absolute shrinkage and selection operator,LASSO)模型筛选预测特征;使用Logistic多因素回归分析建立从LASSO回归分析中选择的特征的预测模型;绘制受试者操作特征(receiver operator characteristic,ROC)曲线预测RARP术后患者尿控功能恢复情况;通过DeLong检验比较曲线下面积,评估模型的辨别力;通过校准曲线和决策曲线分析(decision curve analysis,DCA)评估模型的准确性和临床实用性。结果·根据患者术后的尿控随访数据,患者在术后3个月的尿控率为58.28%(352/604)。训练集的Logistic多因素回归分析结果显示,膜部尿道长度、右肛提肌厚度和术中失血量是术后早期(3个月)尿失禁的独立预测因素。基于该结果建立列线图预测模型。该模型显示出良好的区分度,训练集ROC曲线下面积为0.976(0.954,0.998),验证集ROC曲线下面积为0.977(0.945,1.000);DeLong检验证明训练集和验证集ROC曲线差异无统计学意义(P=0.949)。Hosmer-Lemeshow检验显示该模型具有良好的拟合优度(P=0.179)。DCA结果表明该列线图预测模型在临床上具有适用性。将列线图预测模型纳入在线计算器(https://yitingdun.shinyapps.io/DynNomapp/)。结论·该研究开发并验证了列线图预测模型,可以有效预测RARP术后早期患者的尿控情况;膜部尿道长度、右肛提肌厚度和术中失血量是术后早期尿失禁的独立预测因素。

关键词: 前列腺癌, 尿失禁, 机器人辅助腹腔镜根治性前列腺切除术, 预测因素, 列线图

Abstract:

Objective ·To develop a nomogram prediction model and an online risk calculator, and to predict the continence of patients after robot-assisted laparoscopic radical prostatectomy (RARP). Methods ·A total of 604 prostate cancer patients who underwent RARP and had preoperative prostate magnetic resonance imaging at the Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine from September 2022 to December 2024 were analyzed and included. All patients were randomly resampled and divided into a training set (n=420) and a validation set (n=184) at a ratio of 7∶3. The patients' continence was followed up every month from the first month after the operation. The least absolute shrinkage and selection operator (LASSO) model was applied to screen the features. A Logistic multivariate regression analysis was used to establish a prediction model integrating the features selected from the LASSO analysis. The receiver operator characteristic (ROC) curve was drawn to predict the recovery of continence in patients after RARP, and the areas under the curve were compared by the DeLong test to evaluate the discrimination of the model. Calibration curves and decision curve analysis (DCA) were used to evaluate the calibration and clinical utility the model. Results ·According to the postoperative continence follow-up data of the patients, the continence rate of the patients at 3 months after the operation was 58.28% (352/604). The length of the membranous urethra, the thickness of the right levator ani muscle, and blood loss were identified as independent predictors of early postoperative (3-month) incontinence by Logistic multivariate regression analysis of the training set. The area under the ROC curve was calculated as 0.976 (0.954, 0.998) for the training set and 0.977 (0.945, 1.000) for the validation set, demonstrating good discrimination of this model. No significant difference between the ROC curves of the training set and the validation set was confirmed by the DeLong test (P=0.949). A good goodness of fit of this model was demonstrated by the Hosmer-Lemeshow test (P=0.179). The clinical utility of the nomogram prediction model was indicated by the DCA plot. This nomogram prediction model was incorporated into an online calculator (https://yitingd.shinyapps.io/DynNomapp). Conclusion ·This study developed and validated a nomogram prediction model that can effectively predict the early continence of patients after RARP. The length of the membranous urethra, the thickness of the right levator ani muscle, and the intraoperative blood loss are significant independent predictors of early postoperative incontinence.

Key words: prostate cancer, urinary incontinence, robot-assisted laparoscopic radical prostatectomy (RARP), predictive factor, nomogram

中图分类号: