FibroGENE: A gene-based model for staging liver fibrosis

详细信息    查看全文

• 作者：Mohammed Eslam¹ ; Ahmed M. Hashem² ; Manuel Romero-Gomez³ ; Thomas Berg⁴ ; ⁵ ; Gregory J. Dore⁶ ; ⁷ ; Alessandra Mangia⁸ ; Henry Lik Yuen Chan⁹ ; William L. Irving¹⁰ ; David Sheridan¹¹ ; ¹² ; Maria Lorena Abate¹³ ; Leon A. Adams¹⁴ ; Martin Weltman¹⁵ ; Elisabetta Bugianesi¹³ ; Ulrich Spengler¹⁶ ; Olfat Shaker¹⁷ ; Janett Fischer⁴ ; Lindsay Mollison¹⁸ ; Wendy Cheng¹⁹ ; Jacob Nattermann¹⁶ ; Stephen Riordan²⁰ ; Luca Miele²¹ ; Kebitsaone Simon Kelaeng¹ ; Javier Ampuero³ ; Golo Ahlenstiel¹ ; Duncan McLeod²² ; Elizabeth Powell²³ ; ²⁴ ; Christopher Liddle¹ ; Mark W. Douglas¹ ; David R. Booth²⁵ ; Jacob George¹ ; ^{jacob.george@sydney.edu.au" class="auth_mail" title="E-mail the corresponding author} ; for the International Liver Disease Genetics Consortium (ILDGC)^&dagger ;
• 关键词：Chronic hepatitis C ; Chronic hepatitis B ; Non-alcoholic steatohepatitis ; NASH ; IFNL ; Fibrosis ; Data mining analysis
• 刊名：Journal of Hepatology
• 出版年：2016
• 出版时间：February 2016
• 年：2016
• 卷：64
• 期：2
• 页码：390-398
• 全文大小：862 K

文摘

The extent of liver fibrosis predicts long-term outcomes, and hence impacts management and therapy. We developed a non-invasive algorithm to stage fibrosis using non-parametric, machine learning methods designed for predictive modeling, and incorporated an invariant genetic marker of liver fibrosis risk.

Methods

Of 4277 patients with chronic liver disease, 1992 with chronic hepatitis C (derivation cohort) were analyzed to develop the model, and subsequently validated in an independent cohort of 1242 patients. The model was assessed in cohorts with chronic hepatitis B (CHB) (n = 555) and non-alcoholic fatty liver disease (NAFLD) (n = 488). Model performance was compared to FIB-4 and APRI, and also to the NAFLD fibrosis score (NFS) and Forns’ index, in those with NAFLD.

Results

Significant fibrosis (⩾F2) was similar in the derivation (48.4%) and validation (47.4%) cohorts. The FibroGENE-DT yielded the area under the receiver operating characteristic curve (AUROCs) of 0.87, 0.85 and 0.804 for the prediction of fast fibrosis progression, cirrhosis and significant fibrosis risk, respectively, with comparable results in the validation cohort. The model performed well in NAFLD and CHB with AUROCs of 0.791, and 0.726, respectively. The negative predictive value to exclude cirrhosis was >0.96 in all three liver diseases. The AUROC of the FibroGENE-DT performed better than FIB-4, APRI, and NFS and Forns’ index in most comparisons.

Conclusion

A non-invasive decision tree model can predict liver fibrosis risk and aid decision making.