Delayed Hyperbilirubinemia Is Associated with Mortality in Hypoxic Hepatitis across Shock Etiologies: derivation and independent validation in two critical care databases
Hideya Itagaki, Tomoyuki Endo
Abstract
Purpose. Serum bilirubin is a recognized prognostic marker in hypoxic hepatitis (HH), but prior studies treated it as a static value. We characterized the temporal dynamics of bilirubin in HH across shock etiologies and tested their reproducibility in an independent multicenter cohort.
Materials and methods. Retrospective cohort study using MIMIC-IV (derivation) and the eICU Collaborative Research Database (eICU-CRD; independent validation, no overlapping hospitals). HH was defined as peak alanine aminotransferase >500 U/L in septic, cardiogenic, or hypovolemic shock. Time zero was the first mean arterial pressure <65 mmHg. The exposure was peak total bilirubin ≥10 mg/dL and the outcome in-hospital mortality. Logistic regression adjusted for age, illness severity, and etiology.
Results. Among 1,352 derivation patients, peak bilirubin ≥10 mg/dL (n=209) was independently associated with in-hospital mortality (76.1% vs 52.8%; adjusted odds ratio [aOR] 2.48, 95% CI 1.74–3.55), consistently in septic and cardiogenic shock. Marked hyperbilirubinemia peaked late, at a median of 167–174 h after shock onset versus 55–57 h below 10 mg/dL, with prolonged INR and elevated creatinine. In hemorrhagic shock, this phenotype required massive transfusion and showed bimodal mortality. Both findings reproduced in validation (n=890): aOR 2.54 (95% CI 1.63–3.98) and a delayed peak (138–141 h). Direct bilirubin predominated (ratio 0.60), indicating a cholestatic pattern.
Conclusions. A marked peak bilirubin ≥10 mg/dL is independently associated with in-hospital mortality in HH and arises late as a cholestatic rise reflecting evolving parenchymal liver failure. Its timing, not merely its magnitude, may improve recognition of patients progressing to liver failure after shock.