|
摘要：
|
|
摘要：目的 构建基于中医因素的肝硬化合并轻微型肝性脑病（minimal hepatic
encephalopathy，MHE）风险预测模型。方法 选取秦皇岛市第三医院2019年1月至
2022年1月收治的408例肝硬化患者为训练集，并按照相同标准选取2022年2月至
2023年1月收治的125例肝硬化患者为验证集。分析患者的临床资料 [包括病程、病
因、终末期肝病模型（model for end-stage liver disease，MELD）评分、国际标准化
比值（international normalized ratio，INR）、总胆红素（total bilirubin，TBil）等]
和中医证候特点，采用Lasso回归和多因素Logistic回归分析肝硬化患者发生MHE的
危险因素，采用R（R3.6.3）软件包和rms程序包构建列线图模型，采用图形校准法
和Hosmer-Lemeshow（H-L）进行拟合优度检验。采用校正曲线、受试者工作特征
（receiver operating characteristic curve，ROC）曲线及决策曲线评价列线图模型的预
测效能。结果 训练集的408例肝硬化患者中44例发生MHE，其发生率为10.78%。训练
集MHE组和非MHE组患者中医证型分布、病程、肝性脑病病史、感染、碱中毒、肾
功能不全、脾脏肿大、TBil水平异常和ALB水平异常的比例差异均有统计学意义（P
均＜ 0.05）。Logistic回归分析表明肝肾阴虚（OR = 1.354，95%CI：1.053～1.740，
P = 0.018）、失代偿期肝硬化（OR = 2.776，95%CI：1.323～5.824，P = 0.007）、肝
性脑病病史（OR = 2.767，95%CI：1.339～5.720，P = 0.006）、感染（OR = 2.596，
95%CI：1.237～5.450，P = 0.012）、碱中毒（OR = 3.023，95%CI：1.447～6.317，
P = 0.003）、脾脏肿大（OR = 2.786，95%CI：1.345～5.770，P = 0.006）、TBil ＞
24 μmol/L（OR = 2.593，95%CI：1.261～5.333，P = 0.010）、ALB ≤ 35 g/L（OR =
0.426，95%CI：0.204～0.890，P = 0.023）是肝硬化患者并发MHE的独立危险因素。
训练集和验证集的一致性指数分别为0.822和0.807，校正曲线与理想曲线走势大致相
符，H-L拟合优度检验表明训练集（χ2 = 13.542，P = 0.082）和验证集（χ2 = 10.747，
P = 0.073）模型拟合效果较好，训练集和验证集的模型ROC曲线下面积分别为0.803
（95% CI：0.775～0.831）和0.796（95% CI：0.768～0.824），训练集阈值概率在
3%～65%时净获益值较高，验证集阈值概率在5%～70%时净获益值较高。结论 肝肾阴
虚、失代偿期、肝性脑病病史、感染、碱中毒、脾脏肿大、TBil ＞ 24 μmol/L、ALB ≤
35 g/L是肝硬化患者并发MHE的独立危险因素。构建的列线图模型可准确评估和量化
肝硬化患者发生MHE的风险。
|
|
Abstract: Objective To construct a risk prediction model for liver cirrhosis complicated with
minimal hepatic encephalopathy (MHE) based on the involvement of traditional Chinese
medicine factors. Methods A total of 408 patients with liver cirrhosis admitted to the Third
Hospital of Qinhuangdao from January 2019 to January 2022 were selected as the training
set, and 125 patients with liver cirrhosis from February 2022 to January 2023 were selected
as the validation set according to the same standard. The clinical data [including course of
disease, cause of disease, model for end-stage liver disease (MELD) score, international
normalized ratio (INR), total bilirubin (TBil), etc.] and traditional Chinese medicine syndrome
characteristics of the patients were analyzed. Lasso regression and multivariate Logistic
regression were used to analyze the risk factors for MHE in patients with liver cirrhosis. The
nomogram model was constructed by R (R3.6.3) software package and rms program package,
and the goodness of fit was tested by graphical calibration method and Hosmer Lemeshow (H-L)
method. The predictive performance of the nomogram model was evaluated by calibration
curve, receiver operating characteristic (ROC) curve and decision curve. Results Among the
408 patients with liver cirrhosis in the training set, 44 cases were confirmed with MHE, and
the incidence rate was 10.78%. There were statistically significant differences in the distribution
of traditional Chinese medicine syndrome types, disease duration, history of hepatic
encephalopathy, infection, alkalosis, renal dysfunction, splenomegaly, abnormal TBil and ALB
levels between patients in MHE group and non MHE group in the training set (all P ＜ 0.05).
Logistic regression analysis showed that liver and kidney yin deficiency (OR = 1.354, 95%CI:
1.053～1.740, P = 0.018), decompensated cirrhosis (OR = 2.776, 95%CI: 1.323～5.824,
P = 0.007), history of hepatic encephalopathy (OR = 2.767, 95%CI: 1.339～5.720, P = 0.006),
infection (OR = 2.596, 95%CI: 1.237～5.450, P = 0.012), alkalosis (OR = 3.023, 95%CI:
1.447～6.317, P = 0.003), splenomegaly (OR = 2.786, 95%CI: 1.345～5.770, P = 0.006),
TBil ＞ 24 μmol/L (OR = 2.593, 95%CI: 1.261～5.333, P = 0.010) and ALB ≤ 35 g/L (OR =
0.426, 95%CI: 0.204～0.890, P = 0.023) were independent risk factors for MHE in patients
with liver cirrhosis. The C-index of the training set and the validation set were 0.822 and 0.807,
respectively, and the calibration curve was roughly consistent with the ideal curve trend. The H-L
goodness of fit test showed that the models of the training set (χ2 = 13.542, P = 0.082) and validation
set (χ2 = 10.747, P = 0.073) had good fitting effects. The areas under the ROC curve of the
models of the training set and validation set were 0.803 (95%CI: 0.775～0.831) and 0.796
(95%CI: 0.768～0.824), respectively. The net benefit value was higher when the threshold
probability was 3%～65% in training set and 5%～70% in validation set. Conclusions Liver
and kidney yin deficiency, decompensated cirrhosis, history of hepatic encephalopathy,
infection, alkalosis, splenomegaly, TBil ＞ 24 μmol/L and ALB ≤ 35 g/L were independent
risk factors for complications of MHE in patients with liver cirrhosis. The constructed
nomogram model could accurately assess and quantify the risk of developing MHE in patients
with liver cirrhosis.
|
|
基金项目：
|
|
|
|
作者简介：
|
|
|
|
参考文献：
|
|
|
|
服务与反馈：
|
|
【文章下载】【加入收藏】
|
|
|
|