METHODS: We retrospectively collected the data from 81 patients with WD and hypersplenism who underwent splenectomy. Based on whether PVST occurred within a month after the operation, they were divided into the PVST group and the non-PVST group. The clinical data of the 2 groups were compared, and univariate analysis was used to select the statistically significant features and incorporated into the least absolute shrinkage and selection operator (LASSO) regression model for optimization. Multivariate logistic regression analysis was used to determine the independent risk factors for PVST after splenectomy, which were then applied to establish a personalized nomogram. We calculated the concordance (C)-index and drew the receiver operating characteristic (ROC) curve, the model calibration curve, and the clinical decision analysis (DCA) curve to evaluate the accuracy, calibration, and clinical applicability of the model, respectively. We used bootstrapping for internal validation of the model.
RESULTS: Univariate analysis showed that the differences in preoperative portal vein diameter and velocity of portal blood flow, postoperative mean platelet volume (MPV), mean platelet distribution width (PDW), D-dimer, prothrombin time (PT), and the increase of platelet count (PLT) were of statistical significance (P<0.05). According to the results of the LASSO and multivariate logistic regression analyses, a model including preoperative portal vein diameter, preoperative portal blood flow velocity, postoperative D-dimer, and the increase of PLT was established to predict the risk of PVST after splenectomy. The model showed good accuracy with a C-index of 0.838 (95% CI: 0.750-0.926) and had a well-fitted calibration curve. Furthermore, internal validation showed it achieved a moderate C-index of 0.805. The DCA curve indicated that the model has clinical applicability when patients are treated at thresholds of 2-100%.
CONCLUSIONS: Establishing a predictive model for the risk of PVST in patients with WD and hypersplenism after splenectomy can help clinicians identify patients at high risk of PVST who require intervention measures.
方法:我们回顾性收集了81例接受脾切除术的WD和脾功能亢进患者的数据。根据手术后一个月内是否发生PVST,分为PVST组和非PVST组.比较2组患者的临床资料,单变量分析用于选择具有统计学意义的特征,并将其纳入最小绝对收缩和选择算子(LASSO)回归模型进行优化。采用多因素logistic回归分析确定脾切除术后PVST的独立危险因素。然后应用它们来建立个性化的列线图。我们计算了一致性(C)指数,并绘制了接收器工作特性(ROC)曲线,模型校准曲线,和临床决策分析(DCA)曲线来评估准确性,校准,以及该模型的临床适用性,分别。我们使用引导进行模型的内部验证。
结果:单因素分析显示,术前门静脉直径和门静脉血流速度的差异,术后平均血小板体积(MPV),平均血小板分布宽度(PDW),D-二聚体,凝血酶原时间(PT),血小板计数(PLT)升高有统计学意义(P<0.05)。根据LASSO和多变量逻辑回归分析的结果,包括术前门静脉直径的模型,术前门静脉血流速度,术后D-二聚体,建立PLT升高预测脾切除术后PVST的风险。该模型显示出良好的准确性,C指数为0.838(95%CI:0.750-0.926),并且具有拟合良好的校准曲线。此外,内部验证显示其达到了0.805的中等C指数.DCA曲线表明,当以2-100%的阈值治疗患者时,该模型具有临床适用性。
结论:建立脾切除术后WD合并脾功能亢进患者PVST发生风险的预测模型可帮助临床医师识别需要采取干预措施的PVST高危患者。