Validation and comparison of seventeen noninvasive models for evaluating liver fibrosis in Chinese hepatitis B patients.

ABSTRACT

BACKGROUND & AIMS: To avoid liver biopsy, many noninvasive models comprised of serum markers for liver fibrosis assessment have been developed. Given that most of them were developed in hepatitis C cohorts and few of them have been validated in Chinese hepatitis B patients, we aim to conduct this validation and compare their diagnostic accuracies in such a population. METHODS: A total of 937 HBV-infected patients who underwent liver biopsy were included in this single-centre retrospective study. The diagnostic accuracies of the 17 noninvasive models were assessed by areas under the receiver-operating characteristic curves (AUROCs), using histologically evaluated fibrotic stages of the biopsy specimens as standards. To compare efficiencies of the models, a grading system based on AUROC levels was developed. RESULTS: For discriminating significant fibrosis in all patients, the best three noninvasive models were King's score (AUROC = 0.756), Virahep-C model (AUROC = 0.756) and GPR (AUROC = 0.744); and for diagnosing cirrhosis, Lok index (AUROC = 0.832), FI (AUROC = 0.820) and FIB-4 (AUROC = 0.818) got the first three places. AUROCs in HBeAg-positive group were generally higher than those in HBeAg-negative group. In addition, based on the grading system, Virahep-C and GPR outstood others in evaluating liver fibrosis in all patients. CONCLUSIONS: In Chinese HBV-infected patients, Virahep-C models and GPR had high accuracies in diagnosing liver fibrosis and cirrhosis, while the most discussed models like APRI and FIB-4 did not outstand. Assessment should take into account the HBeAg sero-status, since these noninvasive models were more appropriate for HBeAg-positive patients than HBeAg-negative ones.