摘要
目的 基于贝叶斯网络模型探讨乙型肝炎病毒滴度(HBV DNA)与慢性乙型肝炎(CHB)合并非酒精性脂肪肝(NAFLD)患者慢加急性肝衰竭(ACLF)发生的关系。方法 选择2018年9月—2022年1月广西河池市人民医院收治的175例合并NAFLD的CHB患者作为NAFLD组,另选择175例单纯CHB的患者作为对照组,比较两组患者的临床资料以及发生ACLF的概率。NAFLD组根据患者是否并发ACLF分为ACLF组(n=84)和非ACLF组(n=91)两个亚组。比较患者的临床资料;采用多因素logistic回归分析CHB合并NAFLD患者并发ACLF的影响因素;分析各影响因素与ACLF发生率以及各影响因素间的相关性;构建贝叶斯网络模型,使用Netica软件进行贝叶斯网络推理,并对模型效能进行评价。结果 NAFLD组ACLF的发生率为48.00%,显著高于对照组。ACLF组和非ACLF组在年龄、MELD评分、血脂异常、HBeAg阳性、lgHBV DNA、尿素氮(BUN)、总胆红素(TBIL)、血浆凝血酶原时间(PT)、部分凝血活酶时间(APTT)、国际标准化比率(INR)、血小板(PLT)、白蛋白(ALB)、胆碱酯酶(CHE)、凝血酶原活动度(PTA)方面的差异均具有统计学意义(P均<0.05)。多因素logistic回归分析显示:年龄、lgHBV DNA、TBIL、PT、INR是CHB合并NAFLD患者并发ACLF的独立危险因素(P<0.05),HBeAg阳性、CHE、PTA是保护因素(P<0.05)。相关性分析结果显示:年龄、lgHBV DNA、TBIL、PT、INR与ACLF均呈明显正相关(P<0.05),HBeAg阳性、CHE、PTA与ACLF均呈明显负相关(P<0.05);年龄、lgHBV DNA、TBIL、PT、INR之间均呈明显正相关(P<0.05),分别与HBeAg阳性、CHE、PTA呈明显负相关(P<0.05)。贝叶斯网络模型显示:年龄、HBeAg阳性、PT、TBIL、PTA、INR等变量通过复杂的网络关系与ACLF建立联系,lgHBV DNA、INR、CHE与ACLF的发生有直接联系;节点治疗方式通过影响TBIL、PTA等中间节点联系,间接影响ACLF的发生。结论 HBV DNA滴度是CHB合并NAFLD患者并发ACLF的独立危险因素,对预测ACLF的发生以及控制ACLF病情发展具有重要意义。贝叶斯网络模型分析表明,HBV DNA水平与ACLF的发生有直接联系,且与多个节点之间存在直接或间接的关联,可作为临床预判ACLF发生的重要指标,具有临床应用价值。
Objective To investigate the relationship between hepatitis B virus titer(HBV DNA) and acute on chronic liver failure (ACLF) in chronic hepatitis B(CHB) patients complicated with nonalcoholic fatty liver disease(NAFLD) based on Bayesian network model. Methods Overall, 175 CHB patients complicated with NAFLD were treated in Hechi People’s Hospital, Guangxi Zhuang Autonomous Region from September 2018 to January 2022. These patients were assigned to the NAFLD group. Additional 175 patients with CHB alone were assigned to control group. The clinical data and the probability of ACLF were compared between the two groups. The NAFLD group was stratified into ACLF stratum(n = 84) and non-ACLF stratum(n = 91) in terms of the incidence of ACLF. The clinical data of the two strata were compared. Multivariate logistic regression was used to analyze the risk factors of ACLF in the CHB patients complicated with NAFLD. The correlation between the incidence of ACLF and various factors and the correlation between risk factors and protective factors were analyzed. Bayesian network model was constructed. Netica software was used for Bayesian network inference. The performance of the model was evaluated. Results The incidence of ACLF was 48.00% in NAFLD group, which was significantly higher than that in the control group. Patient age, Model for End-Stage Liver Disease(MELD) score, dyslipidemia, HBeAg prevalence, lgHBV DNA, blood urea nitrogen, total bilirubin(TBIL), prothrombin time(PT), activated partial thromboplastin time(APTT), international normalized ratio(INR), platelet, albumin, cholinesterase(CHE) and prothrombin activity(PTA) showed statistically sinificant difference between ACLF stratum and non-ACLF strtum(all P < 0.05). Multivariate logistic regression analysis showed that age, lgHBV DNA, TBIL, PT, and INR were independent risk factors for ACLF in CHB patients complicated with NAFLD(P < 0.05), while HBeAg-positive, CHE, and PTA were protective factors(P < 0.05). The correlation analysis demonstrated that age, lgHBV DNA, TBIL, PT, INR were positively correlated with ACLF(P < 0.05), while HBeAgpositive, CHE, and PTA were negatively correlated with ACLF(P < 0.05). Age, lgHBV DNA, TBIL, PT, and INR were positively correlated(P < 0.05). These variables were negatively correlated with HBeAg-positive, CHE, and PTA, respectively(all P < 0.05). Bayesian network model indicated that age, HBeAg-positive, PT, TBIL, PTA, and INR were associated with ACLF through complex network relationships. lgHBV DNA, INR, and CHE were directly related to the occurrence of ACLF. Node-based treatment approach indirectly affected the incidence of ACLF by affecting the nexus of intermediate nodes such as TBIL and PTA. Conclusions HBV DNA titer is an independent risk factor for ACLF in CHB patients complicated with NAFLD. It is important for predicting the occurrence of ACLF and controlling the progression of ACLF. Bayesian network model analysis indicates that the level of HBV DNA is directly related to the occurrence of ACLF, and has an direct or indirect relationship with multiple nodes. It is a useful tool for predicting the occurrence of ACLF in clinical practice.
作者
黄洁婕
蓝婧
姚朝光
黄佳
HUANG Jiejie;LAN Jing;YAO Chaoguang;HUANG Jia(Department of Gastroenterology,Hechi People's Hospital,Hechi Guangxi 547000,China)
出处
《中国感染与化疗杂志》
CAS
CSCD
北大核心
2023年第2期173-180,共8页
Chinese Journal of Infection and Chemotherapy
基金
广西壮族自治区健康委员会自筹经费科研课题(Z20200690)。
