Oliver Tavabie, Constantine J. Karvellas, SIamak Salehi, Jaime Speiser, Christopher F. Rose, Krishna Menon, Andreas Prachalias, Michael A. Heneghan, Kosh Agarwal, William M. Lee, Mark J. McPhail, Varuna Aluvihare, U. S. Acute Liver Failure Study Group.
Background: Acute liver failure (ALF) remains a rare but life‐threatening condition which requires early prognostication for transplantation (LTx). Existing models such as the King’s College Criteria (KCC) lack sensitivity. We have previously demonstrated the potential for regeneration linked miRNA to perform as biomarkers in acute and chronic liver disease. The aim of this study was to develop miRNA‐based prognostic models for acetaminophen (APAP) ALF 21‐day transplant‐free mortality. Methods: We assessed serum miRNA expression from 193 patients (94 survivors, 89 non‐survivors defined as mortality within 21 days) with APAP‐ALF at two time points (early; day 1, late; day 3‐5) from the US Acute Liver Failure Study Group registry (1998‐2014). Patients who underwent liver transplantation for APAP‐ALF were excluded from this study. A panel of 24 miRNA identified from our previous studies were analysed. Multiple logistic regression was used to create outcome prediction models at both early and late time points. Clinical data were incorporated to improve prognostication. The DeLong method was used to compare models. Correction for multiple comparisons was performed using the Benjamini‐Hochberg procedure with a false discovery rate set at 0.05. Primary outcome was prediction of 21‐day transplant‐free mortality. Results: At the early time point; up‐regulation of miR‐150, down‐regulation of ‐16‐2, the detection of miR‐20a and absence of miR‐149 were associated with mortality. At the late time point; up‐regulation of miR‐30a, down‐regulation of ‐122, 16‐2 and ‐21 and the detection of miR‐149, ‐17 and ‐191 were associated with mortality. MiRNA based prognostic models were made for both early and late time points (Table 1). The early time point model contained miRNA associated with regeneration (area under the receiver operator curve (AUC) 0.78, 95% CI 0.71‐0.84, p<0.01). The late time point model contained miRNA associated with cell death (AUC 0.83, 95% CI 0.76‐0.89, p<0.01). Both models were enhanced when combined with the Model for End‐Stage Liver Disease score (MELD) and vasopressor requirements (early time point; AUC 0.83, 95% CI 0.78‐0.90, p<0.01, late time point; AUC 0.91, 95% CI 0.86‐0.96, p<0.01). These enhanced models significantly outperformed the KCC (early time point; AUC 0.54, 95% CI 0.49‐0.59, late AUC 0.59, 95% CI 0.54‐0.53) and early time point MELD (early; AUC 0.72 95% CI 0.64‐0.79), and performed comparably to MELD at the late time point (late; AUC 0.87, 95% CI 0.81‐0.92) and the ALF Study Group Prognostic Index (early AUC 0.76, 95% CI 0.69‐0.83, late AUC 0.88, 95% CI 0.82‐0.95). Conclusion: We demonstrate that specific serum miRNA have prognostic value as biomarkers to discriminate outcome in ALF. Our early model utilised regeneration linked miRNA whereas our late model utilised cell‐death linked miRNA; this may signify mechanistic differences at early and late time points which determine patient mortality.