Li Wenjie, Chen Zhe, Chen Han, Han Xu, Zhang Guoxin, Zhou Xiaoying
Department of Gastroenterology, The First Affiliated Hospital of Nanjing Medical University, Nanjing, Jiangsu, China.
J Cancer. 2022 Aug 15;13(10):3103-3112. doi: 10.7150/jca.74772. eCollection 2022.
To establish and validate a model to determine the occurrence risk of colorectal ademomatous polyps. A large cohort of 3576 eligible participants who were treated in the Department of Gastroenterology, the First Affiliated Hospital of Nanjing Medical University from June 2019 to December 2021, were enrolled in our study and divided into discovery and validation cohorts at a ratio of 7:3. LASSO regression method was applied for data dimensionality reduction and feature selection. The nomogram for the occurrence risk of colorectal ademomatous polyps was constructed based on multivariate logistic regression. The predictive performance of the model was evaluated regarding its discrimination, calibration, and clinical applicability. A total of 10 high-risk factors were independent predictors of the colorectal ademomatous polyps occurrence and incorporated into the nomogram, including older age, male, hyperlipidemia, smoking, high consumption of red meat, high consumption of salt, high consumption of dietary fiber, Helicobacter pylori infection, non-alcoholic fatty liver disease and chronic diarrhea. The model showed favorable discrimination values, with the area under the curve of the discovery and validation cohorts 0.775 (95% confidence interval (CI), 0.755-0.794) and 0.776 (95% CI, 0.744-0.807) respectively. The model was also well-calibrated, with Hosmer-Lemeshow test = 0.370. In addition, the decision curve analysis revealed that the model had a higher net profit compared with either the screen-all scheme or the screen-none scheme. In this prospective study, we established and validated a prediction model that incorporated a list of high-risk features related to colorectal ademomatous polyps occurrence, showing favorable discrimination and calibration values.
建立并验证一个用于确定大肠腺瘤性息肉发生风险的模型。纳入了2019年6月至2021年12月在南京医科大学第一附属医院胃肠科接受治疗的3576名符合条件的参与者组成的大型队列，并按7:3的比例分为发现队列和验证队列。采用LASSO回归方法进行数据降维和特征选择。基于多因素逻辑回归构建大肠腺瘤性息肉发生风险的列线图。从区分度、校准度和临床适用性方面评估该模型的预测性能。共有10个高危因素是大肠腺瘤性息肉发生的独立预测因素，并纳入列线图，包括年龄较大、男性、高脂血症、吸烟、红肉摄入量高、盐摄入量高、膳食纤维摄入量高、幽门螺杆菌感染、非酒精性脂肪性肝病和慢性腹泻。该模型显示出良好的区分度值，发现队列和验证队列的曲线下面积分别为0.775（95%置信区间（CI），0.755 - 0.794）和0.776（95%CI，0.744 - 0.807）。该模型校准度也良好，Hosmer - Lemeshow检验值 = 0.370。此外，决策曲线分析显示，与全筛查方案或无筛查方案相比，该模型具有更高的净利润。在这项前瞻性研究中，我们建立并验证了一个包含与大肠腺瘤性息肉发生相关的一系列高危特征的预测模型，显示出良好的区分度和校准度值。
