A noninvasive model to predict histological liver cirrhosis in patients with chronic hepatitis B.
- Author:
Xiang-Lin TU
1
;
Ying-Qun XIAO
;
Fang CHEN
Author Information
- Publication Type:Journal Article
- MeSH: Adolescent; Adult; Biopsy; Female; Hepatitis B e Antigens; blood; Hepatitis B, Chronic; blood; complications; diagnosis; Humans; Liver; pathology; Liver Cirrhosis; blood; diagnosis; etiology; Male; Middle Aged; Models, Biological; Platelet Count; Predictive Value of Tests; Prognosis; ROC Curve; Sensitivity and Specificity; Severity of Illness Index; Young Adult; gamma-Glutamyltransferase; blood
- From: Chinese Journal of Hepatology 2009;17(1):28-32
- CountryChina
- Language:Chinese
-
Abstract:
OBJECTIVETo construct a noninvasive model to predict histological liver cirrhosis in patients with chronic hepatitis B.
METHODS275 patients with chronic hepatitis B were divided into a training group (206 cases) and a validation group (69 cases). The constituent ratios of patients in the fibrosis stages S0-S3, fibrosis stage S4 (early cirrhosis) and active cirrhosis stage were calculated according to the liver biopsy results. 30 noninvasive variables, including age-platelet index (API), aspartate aminotransferase to platelet ratio index (APRI), spleen-platelet ratio index (SRPI) and age-spleen-platelet ratio index (ASPRI), were analyzed by univariate analysis and multivariate logistic regression. Variables that were significantly different between patients with and without cirrhosis were used to construct a noninvasive prediction model, and the model was then tested in the validation group.
RESULTS(1) Of the 275 patients with chronic hepatitis B, 193 (70.2%) were in the fibrosis stages S0-S3, 42 (15.3%) in fibrosis stage S4, 40 (14.5%) in active cirrhosis stage. (2) There were 23 variables that are significantly different between patients with and without cirrhosis by univariate analysis. The 23 variables were further analyzed by multivariate logistic regression, and 4 independent factors, including international normalized ratio (INR), gamma glutamyltranspeptidase (GGT), ASPRI, hepatitis B e antigen (HBeAg) were used to construct a noninvasive prediction model. (3) By receiver operating characteristic curves (ROC) analysis, to discriminate patients in stages S0-S3 from patients in stage S4 and patients in active cirrhosis stage, the area under ROC (AUROC), cut-off value, sensitivity, specificity and accuracy of the model were 0.871, 0.458, 84.4%, 75.7%, and 79.7% respectively. To discriminate patients in active cirrhosis stage from patients in other stages, the AUROC, cut-off value, sensitivity, specificity and accuracy were 0.753, 0.526, 81.8%, 62.9%, and 67.4% respectively. There was no significant difference in AUROC between the training group and the validation group (P less than 0.05).
CONCLUSIONINR, GGT, ASPRI and HBeAg are associated with early cirrhosis and active cirrhosis. Our model can be used to predict early cirrhosis and active cirrhosis.
