Wu Shanshan, Wang Shuren, Ding Yonghong, Zhang Zongwang
Department of Anesthesiology, Liaocheng People's Hospital, Shandong University, Liaocheng, People's Republic of China.
Department of Anesthesiology, Liaocheng People's Hospital, Liaocheng, People's Republic of China.
J Multidiscip Healthc. 2024 Aug 12;17:3889-3905. doi: 10.2147/JMDH.S470204. eCollection 2024.
Postoperative pain is a common complication in endoscopic submucosal dissection (ESD) patients. This study aimed to develop and validate predictive models for postoperative pain associated ESD.
We retrospectively constructed a development cohort comprising 2162 patients who underwent ESD at our hospital between January 2015 and April 2022. The dataset was randomly divided into a training set (n = 1541) and a validation set (n = 621) in a 7:3 ratio. The bidirectional stepwise regression with Akaike's information criterion (AIC) and multivariate logistic regression analysis were used to screen the predictors of post-ESD pain and construct three nomograms. We evaluated the model's discrimination, precision and clinical benefit through receiver operating characteristic (ROC) curves, calibration plots, Hosmer-Lemeshow (HL) goodness-of-fit test and decision curve analysis (DCA) in internal validation.
The proportion of patients developing postoperative pain in the training and testing data set was 25.6% and 28.5%, respectively. Three nomograms were constructed according to the final logistic regression models. The clinical prediction models for preoperative risks, preoperative and intraoperative risks, and perioperative risks consisted of seven, nine and six independent predictors, respectively, after bidirectional stepwise elimination. The models demonstrated the AUC of 0.794 (95% CI 0.768-0.820), 0.823 (95% CI 0.799-0.847) and 0.817 (95% CI 0.792-0.842) in the training cohort and 0.702 (95% CI 0.655-0.748), 0.705 (95% CI 0.659-0.752) and 0.747 (95% CI 0.703-0.790) in the validation cohort. The calibration plot, HL and DCA demonstrated the model's favorable clinical applicability.
We developed and validated three robust nomogram models, which might identify patients at risk of post-ESD pain and promising for clinical applications.
术后疼痛是内镜黏膜下剥离术（ESD）患者常见的并发症。本研究旨在建立并验证与ESD相关的术后疼痛预测模型。
我们回顾性构建了一个开发队列，包括2015年1月至2022年4月在我院接受ESD治疗的2162例患者。数据集以7:3的比例随机分为训练集（n = 1541）和验证集（n = 621）。采用基于赤池信息准则（AIC）的双向逐步回归和多因素逻辑回归分析筛选ESD术后疼痛的预测因素，并构建三个列线图。我们通过内部验证中的受试者工作特征（ROC）曲线、校准图、Hosmer-Lemeshow（HL）拟合优度检验和决策曲线分析（DCA）评估模型的辨别力、准确性和临床效益。
训练数据集和测试数据集中发生术后疼痛的患者比例分别为25.6%和28.5%。根据最终的逻辑回归模型构建了三个列线图。经过双向逐步剔除后，术前风险、术前和术中风险以及围手术期风险的临床预测模型分别由七个、九个和六个独立预测因素组成。模型在训练队列中的AUC分别为0.794（95%CI 0.768-0.820）、0.823（95%CI 0.799-0.847）和0.817（95%CI 0.792-0.842），在验证队列中的AUC分别为0.702（95%CI 0.655-0.748）、0.705（95%CI 0.659-0.752）和0.747（95%CI 0.703-0.790）。校准图、HL检验和DCA显示模型具有良好的临床适用性。
我们开发并验证了三个可靠的列线图模型，这些模型可能识别出有ESD术后疼痛风险的患者，具有临床应用前景。
