Introduction Acute-on-chronic liver failure is associated with a high mortality and difficulty in outcome prediction. Neither liver specific models such as the Model for End Stage Liver Disease (MELD) or intensive care prognostic schema accurately predict hospital mortality. Multivariate methods applied to nuclear magnetic resonance (NMR) spectroscopy of serum have demonstrated correlation with MELD score but whether metabolic profiling predicts outcome in these patients is unknown.
Methods 80 patients referred to Kings College Hospital with decompensated cirrhosis were studied with 20 healthy controls. Plasma was drawn on admission for subsequent 1H NMR spectroscopy using a Carr-Purcell-Meiboom-Gill sequence in a 600 MHz Bruker Avance spectrometer. Full resolution NMR spectra were used in an orthogonal projection to latent structures discriminant analysis (OPLS-DA) to predict hospital mortality.
Results Patients had a median (range) age of 55 (23–75) years with 51 (64%) male. Aetiology of cirrhosis was alcohol related in 40 (50%), autoimmune in 15, viral hepatitis in 10, non-alcoholic steatohepatitis (NASH) in 5, with haemochromatosis and cryptogenic cirrhosis the causes in the remainder. Median (range) MELD score was 14 (6–40) with 18 patients not surviving to hospital discharge. The OPLS-DA model accurately discriminated between patients and healthy controls (R2(X) =0.65, R2(Y) =0.84, Q2(Y) =0.76, cross-validated (leave-one out) sensitivity and specificity 100%, CV-ANOVA p=10−27). Metabolites increased in patients were lactate, tyrosine, and glucose with LDL, VLDL and phosphocholines being increased in controls. OPLS-DA accurately discriminated between survivors and non-survivors (R2(X)=0.63, R2(Y)=0.64, Q2(Y)=0.37, sensitivity 100%, specificity 95% CV ANOVA p=10−6. Metabolites increased in non-survivors included lactate, tyrosine and phenylalanine with lipid and phosphocholine resonances reduced in non-survivors. Removing 1/3 of patients at random from the learning set, remodelling and predicting outcome for these patients resulted in an AUROC of 0.95 (95% CI 0.87 to 1.00, sensitivity 93%, specificity 86%) for predicting survival using NMR profiling in comparison with 0.78 for MELD (p<0.001).
Conclusion Plasma metabolic profiling with 1H NMR spectroscopy in patients with cirrhosis and acute-on-chronic liver failure can accurately describe a metabolic phenotype of non-surviving patients. Validation of these techniques in larger datasets is required.
Competing interests M McPhail: Grant/Research Support from: Wellcome Trust, UK, D Shawcross: None Declared, I Coltart: None declared, E Want: None declared, K Veselkov: None declared, M Crossey: None declared, C Willars: None declared, G Auzinger: None declared, J O'Grady: None declared, W Bernal: None declared, E Holmes: None declared, J Wendon: None declared, S Taylor-Robinson: None declared.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.