Prediction Model Of HBeAg Seroconversion In Patients With HBeAg-positive Chronic Hepatitis B Treated With Peg-interferonα: Establish And Evaluate

Objective: To establish a prediction model of HBeAg seroconversion in Patients withHBeAg-positive chronic hepatitis B treated with Peg-interferon α and evaluate the predict value ofthe model. Methods:Patients with HBeAg-positive chronic hepatitis B treated with Peg-interferon αduring March2010to March2013were collected in our unit consecutively.The patients’characteristics, such as age and sex, and both baseline and on-treatment quantitative HBsAg、HBeAg、HBV-DNA levels of12thweek、16thweek、24thweek were collected. Patients were dividedinto training group and the validation group randomly. Choosing the evidently relevant indicators byunivariate analyses, and using multiple logistic regression to analyse independent predictors, thenbuilt a model predicting for HBeAg seroconversion in Patients with HBeAg-positive chronichepatitis B treated with Peg-interferon α in training group. Analying the predict value of the model inthe validation group by receiver operating characteristic curve(ROC). Results: A total of129patients were enrolled in this retrospective study.There were90patients in training group, and39invalidation group respectively. All the clinical date including the HBeAg seroconverison rate werecamparable between the two groups. In the training group, we found that the baseline HBV-DNA（P=0.017）、HBeAg12（P<0.001）、HBV-DNA12（P=0.007）、HBeAg16（P<0.001）、HBV-DNA16（P=0.034）、HBeAg24（P<0.001）、HBV-DNA24（P=0.010）、S/D（P=0.037）、S/D12（P=0.007）、S/D24（P=0.044）、ΔE24/HBeAg（P=0.001）were evidently relevant predictive factors for HBeAgseroconversion in patients with HBeAg-positive chronic hepatitis B treated with Peg-interferon α.Multiple logistic regression showed that HBeAg16（P=0.009）and HBeAg24(P<0.001) wereindependent predictors. A model was established by the two factors: S=0.806+0.659ln(HBeAg16)-1.297ln (HBeAg24). ROC curve analysis revealed that the area under the curve(AUC) was0.873in the validation group using the prediction model with a cut-off score of <-0.917.Thepatients wound not achieve HBeAg seroconversion at the end of48week treatment could beidentified with high accuracy(100%negative predictive value (NPV),100%sensitivity，67.9%specificity).Conclusions: A model established by the on-treatment serum HBsAg quantification of16thweek and24thweek can identified the patients won’t achieve HBeAg seroconversion with highaccuracy.