目的：分析肝硬化门静脉高压症脾切除术后门静脉血栓（PVT）形成的危险因素并建立PVT发生的Logistic回归预测模型。方法：对符合入选标准的236例脾切除术患者围术期相关临床因素进行单因素及多因素Logistic回归分析，根据多因素分析结果建立Logistic回归预测模型，计算Logit P并绘制各独立因素预测术后PVT的ROC曲线。结果：多因素结果分析显示术前门静脉血流速度（VPBF）、术后平均血小板体积（MPV）、术后D-二聚体（D-D）、术中门静脉自由压力差（FPPD）为术后发生PVT的独立危险因素，术后使用抗凝药物（UAT）为术后发生PVT的独立保护因素（均P<0.05）。根据上述指标建立Logistic回归预测模型为Logit P=5.715-0.558×VPBF（cm/s+0.592×MPV（fL）+0.707×D-D（mg/L）+0.573×FPPD（cmH2O）-0.872×UAT（是=1，否=0），Logit P的临界值为-0.96，ROC曲线下的面积为0.898，准确度为86.9%，VPBF、MPV、D-D、FPPD所对应的界值点分别为13.85 cm/s、10.92 fL、3.54 mg/L、6.99 cmH2O。结论：VPBF≤13.85 cm/s、MPV≥10.92 fL、D-D≥3.54 mg/L、FPPD≥6.99 cmH2O可增加脾切除术后PVT的发生风险，术后UAT可有效减少PVT的发生；所构建的预测模型对判断此类患者术后PVT形成有较高的准确度，有一定的临床参考价值。
Risk factors for portal vein thrombosis after splenectomy for portal hypertension due to liver cirrhosis and establishment of its prediction model
Objective: To analyzed the risk factors for the formation of portal vein thrombosis (PVT) after splenectomy for portal hypertension due to liver cirrhosis, and to establish a Logistic regression model for predicting the occurrence of PVT. Methods: The relevant perioperative factors in 236 eligible patients undergoing splenectomy were determined by univariate analysis and multiple Logistic regression analysis, respectively, and Logistic regression prediction model was established based on the results of the multivariate analysis. Subsequently, Logit P was calculated and the ROC curve of each independent factor for estimating PVT was drawn. Results: Logistic regression analysis showed that preoperative velocity of portal venous blood flow (VPBF), postoperative mean platelet volume (MPV), postoperative D-dimer (D-D), and intraoperative free portal venous pressure difference (FPPD) were the independent risk factors while postoperative usage of anticoagulation therapy (UAT) was an independent protective factor for postoperative PVT (all P<0.05). According to the above factors, Logistic regression prediction model was established and expressed as Logit P=–5.715–0.558×VPBF (cm/s)+0.592×MPV (fL)+0.707×D-D (mg/L)+0.573×FPPD (cmH2O)–0.872×UAT (yes=1, no=0), and the cut off value of Logit P was -0.96, the area under ROC (AUROC) and the accuracy were 0.898 and 86.9%, and the cut off value for VPBF, MPV, D-D and FPPD was 13.85 cm/s, 10.92 fL, 3.54 mg/L and 6.99 cmH2O, respectively. Conclusion: Factors that include VPBF≤13.85 cm/s, MPV≥10.92fL, D-D≥3.54 mg/L and FPPD≥6.99 cmH2O may increase the risk of postoperative PVT, while postoperative UAT may decreased the risk of postoperative PVT; the established prediction model has relatively high accuracy for predicting PVT in those patients, and has certain reference value in clinical practice.