慢性乙型肝炎肝损伤与肝再生失衡评估模型构建研究

发布时间：2018-06-18 19:45

本文选题：慢性乙型肝炎 + 肝损伤　；参考：《湖北中医药大学》2017年硕士论文

【摘要】：目的本研究旨在针对部分早期无症状的慢性乙型肝炎患者,通过建立慢性乙型肝炎肝损伤与肝再生失衡状态的生物学评价指标体系,构建肝损伤与肝再生失衡状态的量化评估模型,为“无症可辨”的慢性乙型肝炎患者提供临床监测和中医药干预依据,防止病情进一步发展。方法结合文献计量学方法和Delphi专家咨询法构建慢性乙型肝炎肝损伤与肝再生失衡状态评价生物学指标体系。在此基础上,根据纳入、排除标准,从湖北省某三甲医院中西医结合治疗慢性乙型肝炎病例中筛选数据纳入研究,采用单因素方差分析和Logistic回归方法筛选特征变量;结合研究数据实际问题(样本量小,变量共线性问题),从多种分级评估算法中筛选两种合适的算法构建模型,并根据灵敏度、准确度、特异度三个指标比较两种模型预估结果,获得最优模型。在临床上纳入多例慢性乙型肝炎“肝肾精虚”证患者,进行为期一年的治疗和动态监测,分别在首诊、半年和一年后对患者进行肝脏穿刺,同时利用模型对患者的肝损伤与肝再生失衡程度进行量化评估,与以“肝穿”为金标准的结果比较,检验模型预测的灵敏度/准确度。结果本研究在546篇临床试验研究文献计量结果基础上,制定了Delphi专家咨询表,其中涵盖了25个与肝损伤和肝再生失衡状态相关的生物学评价指标备选项,结合5位临床肝病专家的问卷调查结果和意见,形成了评价慢性乙型肝炎肝损伤与肝再生失衡状态的生物学指标体系,该体系包含肝损伤评价与肝再生评价两个指标集,共纳入ALT、AST、TNF-α、IL-6等16项指标,可有效反映肝损伤与肝再生失衡状态评价的6个要素(肝组织病变程度、保肝作用、抗氧化能力、降酶作用、抗炎作用、肝再生能力)。本研究共纳入慢性乙型肝炎患者114例,其中86例纳入模型构建训练集,其余28例为验证集。模型构建纳入ALT、ALB、TBIL、PDGF-B、G-CSF、TNF-α、IL-6和IL-18八个危险因素,分别采用Fisher判别算法和主成分回归算法构建模型,比较预估结果获得最优模型——Fisher判别模型(灵敏度:97.06%、特异度:100%、准确度:97.67%),该模型的非标准化典型判别函数为:(1)F1=-0.514*Ln(ALT_(实测值))+4.826*Ln(ALB_(实测值))-1.454*Ln(TBIL_(实测值))-0.403*Ln(PDGF-B_(实测值))+0.239*Ln(G-CSF_(实测值))-0.118*SQRT(TNF-α_(实测值))+0.549*Ln(IL-6_(实测值))-0.121*Ln(IL-18_(实测值))-15.556;(2)F2=-0.166*Ln(ALT_(实测值))+0.391*Ln(ALT_(实测值))+0.087*Ln(TBIL_(实测值))+0.328*(PDGF-B_(实测值))-0.878*Ln(G-CSF_(实测值))+0.021*SQRT(TNF-α_(实测值))+0.934*Ln(IL-6_(实测值))-0.077*Ln(IL-18_(实测值))-5.152;实证研究中,与“肝穿刺”金标准判断结果相比,模型对慢性乙型肝炎“肝肾精虚”证患者在首诊、半年和一年后的肝损伤与肝再生失衡程度量化评估准确度为82.14%(23/28)、85.71%(24/28)和89.29%(25/28)。结论本研究建立的慢性乙型肝炎肝损伤与肝再生失衡状态评价生物学指标体系具有结构简单、覆盖率高、重复率低等特点,涉及指标易于获得,便于临床采用。构建的慢性乙型肝炎肝损伤与肝再生失衡状态评估Fisher判别模型在训练和实证研究阶段对慢性乙型肝炎患者肝损伤和肝再生失衡状态分级诊断的准确性较高,可为后续深入研究提供基础性依据。但由于本次纳入研究样本量较少,模型稳定性不够高,后续研究除需扩大样本量外,还可通过改变Fisher算法判别准则或引入权重因子等方法进一步改进,以提升模型的稳定性和准确率。
[Abstract]:Objective to establish a quantitative evaluation model for the unbalanced state of liver injury and liver regeneration by establishing a biological evaluation index system of the unbalanced state of liver injury and liver regeneration in chronic hepatitis B patients with asymptomatic hepatitis B, and to provide clinical monitoring for the patients with "symptomless discrimination" of chronic hepatitis B. Based on the Bibliometrics and Delphi expert consultation, the biological index system of the evaluation of liver injury and liver regeneration in chronic hepatitis B was constructed by bibliometrics and Delphi expert consultation. Based on the inclusion and exclusion criteria, the treatment of chronic hepatitis B from a three a hospital in Hubei province was combined with Chinese and Western Medicine. Screening data in hepatitis cases were included in the study. Single factor analysis of variance and Logistic regression were used to screen feature variables. Combined with the actual data of research data (small sample size, variable colinearity), two kinds of appropriate algorithms were selected from a variety of classification evaluation algorithms, and three index ratios were based on sensitivity, accuracy and specificity. The optimal model was obtained from the two models. Many patients with chronic hepatitis B, "liver kidney essence deficiency", were included in the treatment and dynamic monitoring for one year. The patients were punctured in the first visit, half a year and one year later, and the model was used to quantify the degree of liver injury and the imbalance of liver regeneration. The evaluation, compared with the results of the "liver wear" as the gold standard, tested the sensitivity / accuracy of the model prediction. Results based on the bibliometric results of 546 clinical trials, a Delphi expert consultation table was developed, covering 25 biological assessment options related to liver injury and liver regeneration imbalance, and 5 The results and suggestions of the questionnaire survey of clinical liver disease experts have formed a biological index system to evaluate the unbalanced state of liver injury and liver regeneration in chronic hepatitis B. This system includes two indexes of evaluation of liver injury and evaluation of liver regeneration, including 16 indexes such as ALT, AST, TNF- A and IL-6, which can effectively reflect the imbalance of liver injury and liver regeneration. The 6 elements of the evaluation (liver lesion degree, liver preservation, antioxidant capacity, antioxidation, anti-inflammatory, liver regeneration). This study included 114 patients with chronic hepatitis B, of which 86 were included in the model construction training set and the other 28 were the validation set. The model construction was included in ALT, ALB, TBIL, PDGF-B, G-CSF, TNF- a, IL-6 and IL-18. Risk factors, using the Fisher discriminant algorithm and the principal component regression algorithm to construct the model respectively, compare the prediction results to obtain the optimal model Fisher discriminant model (sensitivity: 97.06%, specificity: 100%, accuracy: 97.67%), and the model's non standardized typical discriminant function is (1) F1=-0.514*Ln (ALT_ (measured value)) +4.826*Ln (ALB_ (measured value)) -1.454* Ln (TBIL_ (measured value)) -0.403*Ln (measured value)) +0.239*Ln (G-CSF_ (measured value)) -0.118*SQRT (TNF- alpha) +0.549*Ln (IL-6_ (measured value)) -0.121*Ln (IL-18_ (measured value)); (2) (measured value)) +0.021*SQRT (measured value)) +0.934*Ln (IL-6_ (measured value)) -0.077*Ln (IL-18_ (measured value)) -0.077*Ln (measured value)) -5.152. In the empirical study, compared with the results of "liver puncture" gold standard, the model was used to quantify the quantitative assessment of the imbalance of liver injury and liver regeneration after the first diagnosis of the patients with "liver kidney essence deficiency" in the first diagnosis. The accuracy was 82.14% (23/28), 85.71% (24/28) and 89.29% (25/28). Conclusion the biological index system for the evaluation of liver injury and liver regeneration in chronic hepatitis B has the characteristics of simple structure, high coverage and low repetition rate, which are easy to obtain and facilitate clinical use. The Fisher discriminant model of regenerative imbalance state is more accurate for the diagnosis of liver injury and liver regeneration in chronic hepatitis B patients during the training and empirical study stage, which can provide a basic basis for further research. However, because the sample size is less, the stability of the model is not high, and the follow-up study needs to be expanded. Besides, large sample size can be further improved by changing the criterion of Fisher algorithm or introducing weighting factors to improve the stability and accuracy of the model.
【学位授予单位】：湖北中医药大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R259

【参考文献】