Published in

Public Library of Science, PLoS ONE, 10(8), p. e76813, 2013

DOI: 10.1371/journal.pone.0076813

Links

Tools

Export citation

Search in Google Scholar

Diagnosing Fatty Liver Disease: A Comparative Evaluation of Metabolic Markers, Phenotypes, Genotypes and Established Biomarkers

This paper is made freely available by the publisher.
This paper is made freely available by the publisher.

Full text: Download

Green circle
Preprint: archiving allowed
Green circle
Postprint: archiving allowed
Green circle
Published version: archiving allowed
Data provided by SHERPA/RoMEO

Abstract

BACKGROUND: To date, liver biopsy is the only means of reliable diagnosis for fatty liver disease (FLD). Owing to the inevitable biopsy-associated health risks, however, the development of valid noninvasive diagnostic tools for FLD is well warranted. AIM: We evaluated a particular metabolic profile with regard to its ability to diagnose FLD and compared its performance to that of established phenotypes, conventional biomarkers and disease-associated genotypes. METHODS: The study population comprised 115 patients with ultrasound-diagnosed FLD and 115 sex- and age-matched controls for whom the serum concentration was measured of 138 different metabolites, including acylcarnitines, amino acids, biogenic amines, hexose, phosphatidylcholines (PCs), lyso-PCs and sphingomyelins. Established phenotypes, biomarkers, disease-associated genotypes and metabolite data were included in diagnostic models for FLD using logistic regression and partial least-squares discriminant analysis. The discriminative power of the ensuing models was compared with respect to area under curve (AUC), integrated discrimination improvement (IDI) and by way of cross-validation (CV). RESULTS: Use of metabolic markers for predicting FLD showed the best performance among all considered types of markers, yielding an AUC of 0.8993. Additional information on phenotypes, conventional biomarkers or genotypes did not significantly improve this performance. Phospholipids and branched-chain amino acids were most informative for predicting FLD. CONCLUSION: We show that the inclusion of metabolite data may substantially increase the power to diagnose FLD over that of models based solely upon phenotypes and conventional biomarkers.