Early serum ammonia variation in critically ill patients with cirrhosis: A multicentre cohort study

Filipe S. Cardoso, Minjee Kim, Rui Pereira, Luís Bagulho, Pedro Fidalgo, Anna Pawlowski, Richard Wunderink, Nuno Germano, Sean M. Bagshaw, Juan G. Abraldes, Constantine J. Karvellas

Serum ammonia variation in critically ill patients with cirrhosis has been poorly studied. To describe and assess the impact of serum ammonia variation in these patients' outcomes. We studied patients ≥18 years old admitted to the intensive care units (ICUs) at University of Alberta Hospital (Edmonton, Canada) and Curry Cabral Hospital (Lisbon, Portugal; derivation cohort, n = 492) and Northwestern University Hospital (Chicago, USA; validation cohort, n = 600) between January 2010 and December 2021. Primary exposure was ICU days 1-3 serum ammonia. Primary endpoint was all-cause hospital mortality. In the derivation cohort, 330 (67.1%) patients were male and median (IQR) age was 57 (50-63) years. On ICU day 1, median ammonia was higher in patients with grade 3/4 hepatic encephalopathy (HE) than those with grade 2 HE or grade 0/1 HE (112 vs. 88 vs. 77 μmoL/L, respectively; p < 0.001). Furthermore, medium ammonia was higher in hospital non-survivors than survivors (99 vs. 86 μmol/L; p < 0.030). Following adjustment for significant confounders (age, HE, vasopressor use and renal replacement therapy delivery), higher ICU day 2 ammonia was independently associated with higher hospital mortality (adjusted OR per each 10 μmoL/L increment [95% CI] = 1.11 [1.01-1.21]; p = 0.024). In the validation cohort, this model with serial ammonia (ICU days 1 and 3) predicted hospital mortality with reasonably good discrimination (c-statistic = 0.73) and calibration (R2  = 0.19 and Brier score = 0.17). Among patients with cirrhosis in the ICU, early serum ammonia variation was independently associated with hospital mortality. In this context, serial serum ammonia may have prognostic value.