目的

探讨基于炎症反应评分系统构建预测根治性术后食管胃结合部腺癌患者预后的列线图模型的预测效果。

方法

回顾性病例系列分析+队列研究2013年1月至2020年11月行根治性手术的364例AEG患者的临床资料。利用计算机随机数字生成程序，将患者按1:1比例进行简单随机分组，分为训练队列（n=182例）和验证队列（n=182例）。采用多因素COX回归筛选潜在的AEG独立预后因素，构建预测AEG患者3年总生存期（OS）率的列线图。采用Bootstrapping法进行内部交叉验证，用以评估列线图的准确性；结合训练队列与验证队列，使用Harrell法C指数评估列线图模型的区分性能，绘制校准曲线、受试者工作特征（ROC）曲线及决策（DCA）曲线，进一步评估模型的预测能力。根据构建的列线图模型计算所有患者的评分，通过X-tile软件获取评分的最佳截断值，将总体队列、训练队列、验证队列按照最佳截断值划分为高危组（≥最佳截断值）和低危组（＜最佳截断值），对各个队列的高、低危组3年OS进行Kaplan-Meier分析。

结果

训练队列中，死亡组73例，生存组109例。两组AEG患者的BMI、TNM分期、T分期、N分期、肿瘤最大直径、脉管浸润、神经浸润、D-二聚体、NLR、PLR、AFR相比差异均有统计学意义（P＜0.05）。依据NPA-IRS评分系统，不同分值的AEG患者的OS率表现出一定的等级趋势差异（χ²=24.583，P＜0.001）；多因素COX回归分析显示：D-二聚体、神经浸润、TNM分期、N分期、NPA-IRS评分系统为影响AEG患者预后的独立危险因素（P＜0.05）；内部验证和外部验证的校准曲线显示列线图预测与实际吻合较好，时间ROC曲线显示，模型具有较好的预测性能（内部验证AUC为0.858、外部验证AUC为0.789），Kaplan-Meier分析显示列线图预测模型具有良好的OS风险分层性能。

结论

基于D-二聚体、神经浸润、TNM分期、N分期、NPA-IRS评分系统构建的列线图预测模型可为AEG患者的预后评估提供重要依据。

Objective

To investigate the predictive effect of constructing a nomogram model based on inflammatory response scoring system to predict the prognosis of all esophageal and gastric nodular adenocarcinoma patients after radical surgery.

Methods

A retrospective case-series analysis + cohort study of 364 patients with AEG who underwent radical surgery from January 2013 to November 2020.The patients were randomly divided into a training cohort (n=182 cases) and a validation cohort (n=182 cases) in a 1:1 ratio using a computer random number generator.Multivariate COX regression was used to screen potential independent prognostic factors of AEG, and a column graph was constructed to predict 3-year overall survival (OS) rate of AEG patients.Bootstrapping method was used for internal cross-validation to evaluate the accuracy of the nomogram.Combining the training cohort with the validation cohort, the Harrell method was used to evaluate the differentiation performance of the nomogram model, and the calibration curve, receiver operating characteristic(ROC) curve and decision making (DCA) curve were drawn to further evaluate the prediction ability of the model.The scores of all patients were calculated according to the constructed column graph model, and the optimal cut-off value of the scores was obtained through X-tile software.The overall cohort, training cohort and verification cohort were divided into high-risk group (≥optimal cut-off value) and low-risk group (＜optimal cut-off value) according to the optimal cut-off value.Kaplan-Meier analysis of 3-year OS was performed in high and low risk groups of each cohort.

Results

In the training cohort, 73 patients died and 109 survived.There were significant differences in BMI, TNM stage, T stage, N stage, maximum tumor diameter, vascular invasion,nerve invasion, D-dimer, NLR, PLR and AFR between the two groups (P＜0.05).According to the NPA-IRS scoring system, the OS rate of AEG patients with different scores showed a certain grade trend difference (χ²=24.583,P＜0.001).Multivariate COX regression analysis showed that D-dimer, nerve infiltration, TNM stage, N stage and NPA-IRS scoring system were independent risk factors for the prognosis of AEG patients (P＜0.05).Calibration curves of internal verification and external verification showed that the nomogram prediction was in good agreement with the reality.The time ROC curve showed that the model had good prediction performance (AUC of internal verification was 0.858 and AUC of external verification was 0.789).Kaplan-Meier analysis showed that the nomogram prediction model had good OS risk stratification performance.

Conclusions

The prediction model based on D-dimer, nerve infiltration, TNM stage, N stage and NPA-IRS scoring system can provide an important basis for the prognosis assessment of AEG patients.