Guo Jiamei, Zhang Kecheng, Ji Gang, Wang Wei, Li Gang, Liu Zhiqiang, Yang Zuli, Ye Zaisheng, Tian Yantao, Zhang Tao, Wang Xiangyu, Yang Kun, Zhou Tong, You Qi, Li Yong, Ren Peng, Zhang Rupeng, Deng Jingyu
Department of Gastric Surgery, Tianjin Medical University Cancer Institute & Hospital; National Clinical Research Center for Cancer; Key Laboratory of Cancer Prevention and Therapy, Tianjin; Tianjin Key Laboratory of Digestive Cancer; Tianjin's Clinical Research Center for Cancer, Tianjin, China.
Department of General Surgery, the First Medical Center, Chinese PLA General Hospital, Beijing, China.
Int J Surg. 2025 Jun 1;111(6):4068-4073. doi: 10.1097/JS9.0000000000002405. Epub 2025 Apr 18.
Recent years have witnessed a proliferation of studies aimed at developing clinical models capable of predicting lymph node metastasis (LNM) in early gastric cancer (EGC), yet tools for prediction grounded in the Lauren classification remain scarce.
Data of 6468 patients diagnosed with EGC from fifteen Chinese high-volume cancer centers between January 2005 and December 2015 were retrospectively analyzed. Utilizing multivariate logistic regression analysis and the multilayer perceptron (MLP) prediction algorithm, a nomogram and an artificial neural network (ANN) model were developed and validated, respectively, for predicting the likelihood of LNM in non-intestinal EGC cases. The models' performances were evaluated and a comparative analysis of their parameters was undertaken. Subsequently, in-depth risk stratification analyses were performed around the two models.
Non-intestinal type EGC demonstrated an elevated LNM rate and inferior prognosis compared to the intestinal type. Both nomogram and ANN model were developed and performed well in discrimination, calibration and clinical utility. Notably, the ANN model surpassed the nomogram in specificity (95.8% vs. 71.3%, P < 0.001), positive predictive value (PPV) (62.0% vs. 36.2%, P < 0.001) and overall accuracy (82.7% vs. 70.5%, P < 0.001). Patients with different risk strata derived from the nomogram, ANN model, and their combined application exhibited significantly different outcomes. The extent of lymph node dissection significantly influenced prognoses in high-risk patients identified by the combined model, whereas the anatomical location of metastatic lymph nodes did not.
The ANN model established in this study can screen the patients at high risk of LNM in non-intestinal type EGC more accurately. Considering the high extragastric LNM rate observed in the high-risk stratum, radical gastrectomy combined with D2 lymph node dissection is strongly recommended.
近年来，旨在开发能够预测早期胃癌（EGC）淋巴结转移（LNM）的临床模型的研究激增，但基于劳伦分类法的预测工具仍然稀缺。
回顾性分析了2005年1月至2015年12月期间来自中国15家大型癌症中心的6468例EGC患者的数据。利用多因素逻辑回归分析和多层感知器（MLP）预测算法，分别开发并验证了列线图和人工神经网络（ANN）模型，以预测非肠型EGC病例发生LNM的可能性。评估了模型的性能，并对其参数进行了比较分析。随后，围绕这两个模型进行了深入的风险分层分析。
与肠型EGC相比，非肠型EGC的LNM率更高，预后更差。列线图和ANN模型均已开发完成，且在区分度、校准度和临床实用性方面表现良好。值得注意的是，ANN模型在特异性（95.8%对71.3%，P<0.001）、阳性预测值（PPV）（62.0%对36.2%，P<0.001）和总体准确率（82.7%对70.5%，P<0.001）方面超过了列线图。来自列线图、ANN模型及其联合应用的不同风险分层的患者表现出显著不同的结果。淋巴结清扫范围对联合模型识别出的高危患者的预后有显著影响，而转移性淋巴结的解剖位置则没有。
本研究建立的ANN模型能够更准确地筛查非肠型EGC中发生LNM的高危患者。考虑到高危分层中观察到的较高的胃外LNM率，强烈建议行根治性胃切除术联合D2淋巴结清扫术。
