预测贲门黏膜下肿瘤内镜切除技术困难评分系统的建立与验证

Establishment and validation of a difficulty scoring system for endoscopic resection of submucosal tumors in the cardia

摘要:

目的 建立预测贲门黏膜下肿瘤（submucosal tumors， SMT）内镜切除技术困难的临床评分系统并验证。方法 纳入复旦大学附属中山医院2017年7月至2021年3月246例患贲门SMT并接受内镜切除治疗的患者，分为技术困难组（n=21）和非困难组（n=225），通过单因素分析，筛选两组间存在显著差异的潜在风险因素。然后将246例患者随机分为训练队列（n=123）和内部验证队列（n=123），基于训练队列的多元逻辑回归分析结果确定影响技术困难的独立危险因素并赋予相应分值，构建评分系统并评估其效能。以张家港市第一人民医院31例患者的临床数据对该评分系统进行外部验证。结果 单因素分析结果显示，性别、生长模式、形态、长径及胃镜方向是贲门SMT内镜切除技术困难的潜在风险因素（P<0.05）。训练队列多元逻辑回归分析进一步证实，男性（P=0.026，OR=0.154，95%CI：0.023~0.683）、腔外生长（P=0.005，OR=0.040，95%CI：0.004~0.366）、长径≥3 cm（P=0.005，OR=0.072，95%CI：0.009~0.392）是技术困难的独立危险因素，分别赋2分、3分和3分。最终按总分将技术难度分为容易（0分）、中等（1~3分）、困难（4~6分）、非常困难（>6分）四级，训练队列各难度等级对应的技术困难发生率分别为0.0%（0/39）、6.8%（4/59）、33.3%（8/24）、100.0%（1/1），验证队列分别为0.0%（0/49）、5.6%（3/54）、22.2%（4/18）、50.0%（1/2）。评分系统效能验证显示，内部验证受试者工作特征曲线下面积为0.860（95%CI：0.763~0.958），Hosmer‑Lemeshow检验拟合度良好（P=0.979），外部验证受试者工作特征曲线下面积为0.798（95%CI：0.632~0.965）。结论 基于男性、腔外生长和长径≥3 cm风险因素构建的评分系统可用于预测贲门SMT内镜切除的技术困难程度，为临床医师治疗规划提供参考。

Abstract:

Objective To establish and validate a clinical scoring system for predicting the technical difficulty of endoscopic resection of submucosal tumors (SMT) in the cardia. Methods A total of 246 patients with cardiac SMT who underwent endoscopic resection at Zhongshan Hospital, Fudan University from July 2017 to March 2021 were enrolled. They were divided into technically difficult (n=21) and non‑difficult (n=225) groups, potential risk factors with significant differences between the two groups were screened by univariate analysis. All 246 patients were randomly divided into training (n=123) and internal validation (n=123) cohorts. Based on the training cohort, the independent risk factors influencing technical difficulty were identified by multivariate logistic regression analysis, and corresponding scores were assigned to construct the scoring system and its efficacy was evaluated. External validation of the scoring system was performed using clinical data from 31 similar patients at Zhangjiagang First People""s Hospital. Results Univariate analysis showed that gender, growth pattern, morphology, long diameter, and gastroscopic viewing direction were potential risk factors for technical difficulty in endoscopic resection of cardiac SMT (P<0.05). Multivariate logistic regression analysis in the training cohort further confirmed that male sex (P=0.026, OR=0.154, 95%CI: 0.023‑0.683), extraluminal growth (P=0.005, OR=0.040, 95%CI:0.004‑0.366), and long diameter ≥3 cm (P=0.005, OR=0.072, 95%CI: 0.009‑0.392) were independent risk factors for technical difficulty. Two points were assigned to male sex, 3 to extraluminal growth, and 3 to long diameter ≥3 cm. The technical difficulty was classified into four grades based on the total score: easy (0 point), moderate (1‑3 points), difficult (4‑6 points), and very difficult (>6 points). The incidence rates of technical difficulty in each grade of the training cohort were 0.0% (0/39), 6.8% (4/59), 33.3% (8/24), and 100.0% (1/1), respectively, while those in the internal validation cohort were 0.0% (0/49), 5.6% (3/54), 22.2% (4/18), and 50.0% (1/2), respectively. Efficacy verification of the scoring system showed that the area under the receiver operating characteristic curve (AUC) for internal validation was 0.860 (95%CI: 0.763‑0.958) with good calibration in the Hosmer‑Lemeshow test (P=0.979). The AUC for external validation was 0.798 (95%CI: 0.632‑0.965). Conclusion The scoring system constructed based on the male sex, extraluminal growth, and long diameter ≥3 cm can effectively predict the technical difficulty of endoscopic resection of cardiac SMT, and offers a practical tool for clinical decision‑making.