Premium
Evaluation of a logistic regression model for predicting liver necroinflammation in hepatitis B e antigen‐negative chronic hepatitis B patients with normal and minimally increased alanine aminotransferase levels
Author(s) -
Xie Yao,
Yi Wei,
Zhang Lu,
Lu Yao,
Hao HongXiao,
Gao YuanJiao,
Ran ChongPing,
Lu HuiHui,
Chen QiQi,
Shen Ge,
Wu ShuLing,
Chang Ming,
PingHu Lei,
Liu RuiYu,
Sun Lei,
Wan Gang,
Li MingHui
Publication year - 2019
Publication title -
journal of viral hepatitis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.329
H-Index - 100
eISSN - 1365-2893
pISSN - 1352-0504
DOI - 10.1111/jvh.13163
Subject(s) - medicine , alanine aminotransferase , youden's j statistic , gastroenterology , receiver operating characteristic , cutoff , liver biopsy , chronic hepatitis , logistic regression , hepatitis b , confidence interval , biopsy , immunology , virus , physics , quantum mechanics
Liver necroinflammation is the indicator for treating patients with chronic hepatitis B (CHB) infection. However, there is no suitable non‐invasive index for diagnosing liver necroinflammation. This study aimed to create a non‐invasive index to predict liver necroinflammation in patients who lack clear‐cut clinical inflammation parameters. Patients who were hepatitis B e antigen (HBeAg)‐negative and underwent liver histological diagnosis, had a normal or minimally increased alanine aminotransferase (ALT) level were enrolled. Liver necroinflammation was defined as histological active index ≥4. A logistic regression model (LRM) was established based on the parameters independently associated with liver necroinflammation. Of all 550 patients, 36.73% had necroinflammation. In patients with an abnormal ALT level, the rate of necroinflammation was 52.49%. The area under the curve (AUC) of the ALT level for predicting necroinflammation was 0.655 (95% confidence interval [CI], 0.609‐0.702), and that of the HBV DNA level ≥2000 IU/mL combined with an abnormal ALT level was 0.618. By using the LRM, the AUC improved to 0.769 (95% CI, 0.723‐0.815) with a Youden index of 0.519 and diagnostic accuracy of 75.3%. The cutoff value ≥0.7 in the LRM had a specificity of 97.4% and positive predictive value of 85.0% for predicting necroinflammation. By using the cutoff value <0.15 in the LRM, the presence of necroinflammation could be excluded with a negative predictive value of 90.8%. This study indicated that the LRM can be used to effectively diagnose liver necroinflammation in HBeAg‐negative patients with CHB who have normal or minimally elevated ALT levels.