OBJECTIVE: Infections are common and associated with complications and mortality in acute liver failure (ALF). The temporal relationship between ammonia and infection in ALF patients is unclear. We aimed to evaluate the predictors of infection and its relationship with arterial ammonia levels. MATERIALS AND METHODS: Consecutive ALF patients hospitalized between January 2004 and December 2015, without signs of infection at/within 48 h of admission, were included. Occurrence of infection after 48 h was documented and ammonia levels were estimated for five consecutive days. Multivariate logistic regression analysis was used to assess factors associated with development of infection. Generalized estimating equations (GEE) were used to evaluate five-day time trend of ammonia in patients with and without infection. RESULTS: Of 540 consecutive patients, 120 were infected at admission/within 48 h and were excluded. Of the rest 420 patients, 144 (34.3%) developed infection after 48 h and 276 (65.7%) remained non-infected. Infected patients had higher mortality than non-infected patients (61.8% vs 40.0%, p < .001). On multivariate analysis, presence of cerebral edema(HR 2.049; 95%CI, 1.30-3.23), ammonia level on day 3 of admission (HR 1.006; 95%CI, 1.003-1.008), and model for end stage liver disease (MELD) score (HR 1.051; 95%CI, 1.026-1.078) were associated with development of infection. GEE showed group difference in serial ammonia values between infected and non-infected patients indicating lack of ammonia decline in infected patients. CONCLUSIONS: Cerebral edema, elevated ammonia on day 3, and higher MELD score predict the development of infection in ALF. Ammonia persists at high levels in infected patients, and elevated ammonia on day 3 is associated with complications and death.
OBJECTIVE: Infections are common and associated with complications and mortality in acute liver failure (ALF). The temporal relationship between ammonia and infection in ALFpatients is unclear. We aimed to evaluate the predictors of infection and its relationship with arterial ammonia levels. MATERIALS AND METHODS: Consecutive ALFpatients hospitalized between January 2004 and December 2015, without signs of infection at/within 48 h of admission, were included. Occurrence of infection after 48 h was documented and ammonia levels were estimated for five consecutive days. Multivariate logistic regression analysis was used to assess factors associated with development of infection. Generalized estimating equations (GEE) were used to evaluate five-day time trend of ammonia in patients with and without infection. RESULTS: Of 540 consecutive patients, 120 were infected at admission/within 48 h and were excluded. Of the rest 420 patients, 144 (34.3%) developed infection after 48 h and 276 (65.7%) remained non-infected. Infected patients had higher mortality than non-infectedpatients (61.8% vs 40.0%, p < .001). On multivariate analysis, presence of cerebral edema(HR 2.049; 95%CI, 1.30-3.23), ammonia level on day 3 of admission (HR 1.006; 95%CI, 1.003-1.008), and model for end stage liver disease (MELD) score (HR 1.051; 95%CI, 1.026-1.078) were associated with development of infection. GEE showed group difference in serial ammonia values between infected and non-infectedpatients indicating lack of ammonia decline in infected patients. CONCLUSIONS:Cerebral edema, elevated ammonia on day 3, and higher MELD score predict the development of infection in ALF. Ammonia persists at high levels in infected patients, and elevated ammonia on day 3 is associated with complications and death.