年龄≤30岁慢性HBV感染者启动抗病毒治疗无创预测模型的构建及分析

Establishment of a noninvasive predictive model for antiviral therapy in patients with chronic hepatitis B virus infection and an age of≤30 years

目的通过构建无创模型预测年龄≤30岁的慢性HBV感染者是否需要抗病毒治疗,并验证其诊断价值。方法回顾性分析2017年1月—2023年1月深圳市第三人民医院收治的175例符合纳入标准的慢性HBV感染者的临床资料,根据肝穿刺病理结果,将患者分为治疗组(有抗病毒治疗指征,n=41)和观察组(无抗病毒治疗指征,n=134)。分析两组患者的临床资料、影像学检查、血清生化学等指标,通过单因素及多因素Logistic回归分析影响患者具有抗病毒治疗指征的参数,通过相关参数构建不同的预测需要抗病毒治疗的模型,采用受试者工作特征曲线(ROC曲线)比较不同模型的诊断价值。符合正态分布的计量资料组间比较采用成组t检验;不符合正态分布的计量资料组间比较采用Mann-Whitney U秩和检验。计数资料组间比较采用χ^(2)检验或Fisher检验。结果治疗组和观察组两组间ALT、铁蛋白、总胆固醇(CHOL)、TG、PLT、肝硬度(STE)、Ⅲ型前胶原N末端肽(PⅢP)比较,差异均有统计学意义(P值均<0.05)。多因素Logistic回归分析显示,CHOL(OR=0.4,95%CI:0.2~1.0)、STE(OR=1.5,95%CI:1.0~2.1)、PⅢP(OR=1.1,95%CI:1.0~1.1)是具有抗病毒治疗指征的独立预测因子。模型1(STE+PⅢP+CHOL)、模型2(STE+PⅢP)、模型3(STE+CHOL)、模型4(PⅢP+CHOL)及STE、PⅢP、CHOL的ROC曲线下面积分别为0.908、0.848、0.725、0.725、0.836、0.725、0.634,模型1 ROC曲线下面积最大,特异度为77.34%,敏感度为96.36%,与STE、PⅢP、CHOL及模型2、模型3、模型4比较,差异均有统计学意义(Z值分别为0.21、3.08、3.06、3.23、0.89、0.88,P值均<0.05)。结论通过由CHOL、STE、PⅢP指标构建的无创模型对年龄≤30岁的慢性HBV感染者是否需要抗病毒治疗有较好的预测价值。

Objective To predict whether antiviral therapy is required in patients with chronic hepatitis B virus(HBV)infection and an age of≤30 years by establishing a noninvasive model,and to investigate the diagnostic value of this model.Methods A retrospective analysis was performed for the clinical data of 175 patients with chronic HBV infection who were admitted to Shenzhen Third People’s Hospital from January 2017 to January 2023 and met the inclusion criteria,and according to the results of liver biopsy,they were divided into treatment group with 41 patients(with indications for antiviral therapy)and observation group with 134 patients(without indications for antiviral therapy).The two groups were analyzed in terms of the indicators including clinical data,imaging examinations,and serum biochemical parameters.The univariate and multivariate Logistic regression analyses were used to investigate the parameters affecting the indication for antiviral therapy,and different models for predicting the need for antiviral therapy were constructed based on related parameters.The receiver operating characteristic(ROC)curve was used to compare the diagnostic value of different models.The independent-samples t test was used for comparison of normally distributed continuous variables between groups,and the Mann-Whitney U rank sum test was used for comparison of non-normally distributed continuous variables between groups;the chi-square test or the Fisher’s exact test was used for comparison of categorical data between groups.Results There were significant differences between the treatment group and the observation group in alanine aminotransferase,ferritin,total cholesterol(CHOL),triglyceride,platelet count,liver stiffness measured by sound touch elastography(STE),and procollagen III N-terminal propeptide(PIIIP)(all P<0.05).The multivariate Logistic regression analysis showed that CHOL(odds ratio[OR]=0.4,95%confidence interval[CI]:0.2—1.0),STE(OR=1.5,95%CI:1.0—2.1),and PIIIP(OR=1.1,95%CI:1.0—1.1)were independent predictive factors for the indications for antiviral therapy.Model 1(STE+PIIIP+CHOL),model 2(STE+PIIIP),model 3(STE+CHOL),model 4(PIIIP+CHOL)had an area under the ROC curve of 0.908,0.848,0.725,and 0.725,respectively,while STE,PIIIP,and CHOL used alone had an AUC of 0.836,0.725,and 0.634,respectively,suggesting that model 1 had the largest AUC,with a specificity of 77.34%and a sensitivity of 96.36%,and had a significant difference compared with STE,PIIIP,CHOL,and the models 2,3,and 4(Z=0.21,3.08,3.06,3.23,0.89,and 0.88,all P<0.05).Conclusion The noninvasive model established based on CHOL,STE,and PIIIP has a good value in predicting the need for antiviral therapy in patients with chronic HBV infection and an age of≤30 years.