Combining FIB-4 and Liver Stiffness Into the FIB-5, a Single Model that Accurately Predicts Complications of Portal Hypertension

We aimed to combine the fibrosis (FIB)-4 score and fibroscan-derived liver stiffness (LS) into a single score (FIB-5) that predicts incident complications of portal hypertension (PH) in persons with compensated liver disease. In this retrospective cohort study, we identified 5849 US veterans who und...

Full description

Saved in:
Bibliographic Details
Published in:The American journal of gastroenterology Vol. 117; no. 12; pp. 1999 - 2008
Main Authors: Vutien, Philip, Berry, Kristin, Feng, Ziding, VoPham, Trang, He, Qianchuan, Green, Pamela K., Ioannou, George N.
Format: Journal Article
Language:English
Published: United States Wolters Kluwer 01.12.2022
Wolters Kluwer Health Medical Research, Lippincott Williams & Wilkins
Subjects:
Accuracy
Age
Ascites
Aspartate Aminotransferases
Biomarkers
Biopsy
Calibration
Elasticity Imaging Techniques
Female
Gastroenterology
Hepatitis C
Humans
Hypertension
Hypertension, Portal - complications
Hypertension, Portal - diagnosis
Laboratories
Liver - diagnostic imaging
Liver Cirrhosis - complications
Liver Cirrhosis - diagnosis
Liver diseases
Machine learning
Male
Middle Aged
Patients
Retrospective Studies
Variables
Veterans
United States > US
ISSN:0002-9270, 1572-0241, 1572-0241
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We aimed to combine the fibrosis (FIB)-4 score and fibroscan-derived liver stiffness (LS) into a single score (FIB-5) that predicts incident complications of portal hypertension (PH) in persons with compensated liver disease. In this retrospective cohort study, we identified 5849 US veterans who underwent LS measurement from May 01, 2014 to June 30, 2019, and laboratory tests enabling FIB-4 calculation within 6 months of LS measurement. Patients were followed up from the LS measurement date until February 05, 2020, for incident complications of PH. We combined LS values and the individual components of the FIB-4 score (i.e. age, aspartate aminotransferase, alanine aminotransferase, and platelet count) using multivariable Cox proportional hazards modeling and the machine learning algorithm eXtreme gradient boosting to develop the C-FIB-5 and X-FIB-5 models, respectively. Models were internally validated using optimism-corrected measures. Among 5,849 patients, the mean age was 62.8 years, 95.9% were men, and the mean follow-up time was 2.14 ± 1.21 years. Within 3 years after LS measurement date, 116 (2.0%) patients developed complications of PH. The X-FIB-5 (area under the receiver operating characteristic [AUROC] 0.845) and C-FIB-5 scores (AUROC 0.868) demonstrated superior discrimination over LS (AUROC 0.688) and FIB-4 (AUROC 0.672) for predicting incident complications of PH. Both the X-FIB-5 and C-FIB-5 models demonstrated higher classification accuracy across all sensitivity cutoffs when compared with LS or FIB-4 alone. We combined LS and the individual components of the FIB-4 into a single scoring system (FIB-5, www.fib5.net ), which can help identify patients with compensated liver disease at risk of developing complications of PH.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
7) George N. Ioannou: Study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript
4) Ziding Feng: Study concept and design, analysis and interpretation of data, drafting of the manuscript
5) Qianchuan He: Study concept and design, analysis and interpretation of data, drafting of the manuscript
Authorship Contributions
1) Philip Vutien: Study concept and design, analysis and interpretation of data, drafting of the manuscript
3) Trang VoPham: Study concept and design, analysis and interpretation of data, drafting of the manuscript
2) Kristin Berry: Study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, statistical analysis
6) Pamela K. Green: Study concept and design, acquisition of data, analysis and interpretation of data
ISSN:0002-9270
1572-0241
1572-0241
DOI:10.14309/ajg.0000000000001906