肝移植术后早期感染危险因素分析及预测

目的探讨肝移植术后早期(≤1个月)感染发生的危险因素并建立列线图预测模型。 方法回顾性分析2016年1月至2020年12月宁波大学附属李惠利医院肝移植中心200例同种异体肝移植受者临床资料。根据纳入和排除标准共收集181例受者的人口学数据、临床资料和病原菌检测结果，根据术后早期是否发生感染分为感染组(n＝96)和非感染组(n＝85)。分析受者术后早期感染菌群分布特点和相关危险因素，构建列线图并评价其拟合度、区分度和临床实用性。正态分布计量资料采用独立样本t检验比较，不符合正态分布计量资料采用Mann-Whitney U检验比较。分类变量采用χ²检验或Fisher确切概率法。采用Logistic回归分析进行多因素分析。采用R语言(4.1.2)软件rms包构建列线图模型，并通过Bootstrap自抽样法对模型进行内部验证；采用Hosmer-Lemeshow检验、校准曲线、受试者工作特征(ROC)曲线下面积、一致性指数(C指数)及临床决策曲线分析来评价列线图的校准度、区分度及临床实用性。P<0.05为差异有统计学意义。 结果纳入研究的181例受者中，肝移植术后早期感染发生率为53.0%(96/181)，96例感染组受者共检出病原菌132株，以革兰阴性菌最为常见(42.4%)。受者术后2周内感染发生率最高(70.8%，68/96)，感染常见部位为肺部和血行感染。多因素Logistic回归分析结果显示，受者女性(OR＝4.235，95%CI：1.577～11.370)、MELD评分≥20(OR＝3.742，95%CI：1.296～10.805)、Chlid-Pugh分级C级(OR＝3.346，95%CI：1.263～8.862)、术后呼吸机使用时间(OR＝1.036，95%CI：1.009～1.063)是肝移植术后早期感染的独立危险因素。根据上述独立危险因素建立列线图预测模型，经Bootstrap法进行内部验证，Hosmer-Lemeshow检验无统计学意义(χ²＝7.236，P>0.05)，校正曲线贴近于理想曲线，预测模型与观测值具有较好的拟合度。C指数和ROC曲线下面积均为0.800(95%CI：0.735～0.865)，模型具有良好的区分度。模型临床决策曲线在较广的阈值概率范围内(0.2～1.0)，高于采用单一危险因素预测，显示该模型具有临床实用性。 结论肝移植受者性别、MELD评分、Child-Pugh分级和术后呼吸机使用时间是移植术后早期感染的独立危险因素，列线图对移植术后早期感染的预测效果良好。

Risk factors analysis and predicition on early infection after liver transplantation

ObjectiveTo investigate the risk factors of early infection (≤1 month) after liver transplantation and to construct a Nomogram model for early infection prediction after liver transplantation. MethodsThe clinical data of 200 allogeneic liver transplant recipients in the Liver Transplant Center of Li Huili Hospital Affiliated to Ningbo University from January 2016 to December 2020 were retrospectively analyzed. According to the inclusion and exclusion criteria, a total of 181 recipients′ demographic data, clinical data and pathogen test results were collected. They were divided into the infection group and the uninfected group early after liver transplantation according to diagnostic criteria for infection. The distribution characteristics and related risk factors of early infection after liver transplantation were analyzed. A Nomogram was constructed and its fit, discrimination and clinical utility were evaluated. Independent samples t-test was used to compare normally distributed continuous variables and Mann-Whitney U test was used to compare non-normally distributed continuous variables. Categorical variables were analyzed using Chi-square test or Fisher′s exact test. Logistic regression analysis was used for multivariate analysis. The Nomogram model was constructed using rms package of R software (version 4.1.2) and verified by the Bootstrap internal verification method. Hosmer-Lemeshow test, calibration curve, areas under the receiver operating characteristic (ROC) curves, concordance index (C-index) and decision-curve analysis (DCA) were used to assess the calibration, discrimination and clinical utility of the nomograms. P<0.05 was considered statistically significant. ResultsAmong the 181 recipients, the incidence of early infection after liver transplantation was 53.0% (96/181). A total of 132 strains of pathogens were detected in the infection group, among which Gram-negative bacteria were the most common (42.4%). Recipients had the highest incidence of infection within two weeks after surgery (70. 8%, 68/96). The most common infection sites were as follows: lung and bloodstream. Multivariate Logistic regression analysis showed that female recipients (OR=4.235, 95% CI: 1.577-11.370), model for end-stage liver disease(MELD) score ≥20 (OR=3.742, 95% CI: 1.296-10.805); Child-Pugh grade C (OR=3.346, 95% CI: 1.263-8.862), postoperative ventilator supporting time (OR=1.036, 95% CI: 1.009-1.063) were independent risk factors associated with early infection after liver transplantation. According to the above independent risk factors, a Nomogram model was constructed and verified by the Bootstrap internal verification method, and the Hosmer-Lemeshow test was was not statistically significant (χ²=7.236, P>0.05). The calibration curve was close to the ideal curve, and the predictive model and the observed value showed good fit. The C-index and areas under the ROC curves were both 0.800 (95% CI: 0.735-0.865), and the model showed good discrimination. DCA of the model was higher than the single risk factor prediction in a wide range of threshold probability (0.2-1.0). The model showed the good clinical utility. ConclusionsFemale recipients, MELD score, Child-Pugh grade C and postoperative ventilator supporting time were independent risk factors for early infection after liver transplantation, and a Nomogram has good predictive effect on early infection after liver transplantation.