Secondary Logo

Journal Logo


Ammonia Levels Do Not Guide Clinical Management of Patients With Hepatic Encephalopathy Caused by Cirrhosis

Haj, Mona MD1; Rockey, Don C. MD1

Author Information
The American Journal of Gastroenterology: May 2020 - Volume 115 - Issue 5 - p 723-728
doi: 10.14309/ajg.0000000000000343
  • Free



Hepatic encephalopathy (HE), also known as portosystemic encephalopathy, is a reversible neuropsychiatric complication of chronic liver disease or cirrhosis, found in up to 70% of patients (1,2). HE ranges from mild cognitive impairment to coma and is associated with increased morbidity and mortality, increasing health care resource utilization, as well as an increased burden of care on the patient, family, and society (3–5). Furthermore, the reversible nature of HE requires rapid recognition and initiation of treatment, typically with the nonabsorbable disaccharide, lactulose (1,6).

Although the pathophysiology of HE is complex, it is believed that ammonia plays a central role (2,7–9). Under normal conditions, ammonia is converted to urea and ultimately excreted by the kidneys, with small amounts cleared by the skeletal muscle (10). However, in patients with cirrhosis, the presence of portosystemic collaterals, sarcopenia, and possible comorbid renal dysfunction limits ammonia metabolism and excretion, leading to an elevated blood concentration with ammonia crossing the blood-brain barrier, presumably leading to cerebral edema with astrocyte malfunction and the development of HE (11,12).

As a result of the putative role ammonia may have in the pathogenesis of HE, and the development of multiple novel ammonia-lowering drugs, a number of studies have assessed the role of serum ammonia levels as diagnostic markers of HE in the clinical setting (13–15). Unfortunately, in the acute clinical setting, accurate measurement of serum ammonia is difficult. Samples may need to be collected on ice and rapidly processed to avoid falsely elevated values (16). Serum ammonia levels can also vary from person to person based on renal function, presence or absence of sarcopenia, and diet, factors that all lead to a high degree of variability (16,17). In addition, although blood ammonia levels have been found to be higher in patients with cirrhosis with a history of HE, they may not correlate with the severity of HE, as even patients with no neurocognitive changes were found to have elevated ammonia levels (15). Furthermore, as the role of ammonia in the pathogenesis of HE relies on its ability to cross the blood-brain barrier, venous ammonia levels—the most commonly measured in clinical settings—may not represent the concentration of ammonia exerting toxic effects at the level of the brain (14).

We have recognized that from a clinical practice standpoint, some clinicians routinely order ammonia levels in patients with HE or suspected HE, while others typically do not use ammonia levels as part of their practice. We have further observed that treatment with lactulose (the standard of care treatment in patients admitted to the hospital for HE) does not seem to vary based on ammonia levels. However, to assess the role that ammonia may have on HE management, we hypothesized that patients with elevated ammonia levels would receive more aggressive therapy with lactulose than those with normal ammonia levels or without an ammonia level drawn. Our aim was thus to investigate HE management as a function of ammonia level.


Patients and study design

This retrospective study of patients admitted to the Medical University of South Carolina (MUSC) examined consecutive patients admitted with cirrhosis and HE from 2005 to 2015. All admissions for cirrhosis patients to MUSC with a primary or secondary diagnosis of HE (ICD-9: 572.2) were included. The study was approved by the institutional review board (IRB) at MUSC and met all guidelines for good clinical practice (18).

HE was defined as altered mental status in a patient with cirrhosis after exclusion of other causes of mental status change, and the severity of HE was based on the West Haven Criteria (1). All records were reviewed to verify the diagnosis of HE. We included patients with known precipitants of HE including noncompliance with patient medical regimen, gastrointestinal bleeding, non–central nervous system infection, dehydration (defined by documented physical examination findings and/or an elevated BUN to creatinine ratio), or no identifiable precipitant (1,19). Patients with an obvious cause of altered mental status (i.e., primary central nervous system disease, toxin or drug ingestion, acute alcohol intoxication, alcoholic hepatitis, or electrolyte disturbance—such as hyponatremia or hypernatremia, hypokalemia, and hypercalcemia) were excluded. Patients younger than 18 years, or with HE secondary to acute liver failure, those with history of liver transplant, or patients with missing data regarding admission, lactulose therapy, or discharge were also excluded. All ammonia levels were venous samples drawn at admission or first onset of altered mental status and before treatment for HE with lactulose therapy had begun.


The primary outcome of the study was total oral lactulose given in the first 48 hours of HE management. The cutoff of 48 hours was chosen based on our clinical experience, in which appropriate treatment of HE (i.e., 3–4 bowel movements per day) typically leads to improvement in mental status within this time frame. Secondary endpoints included time to resolution of HE (defined as a physical examination of West Haven stage 0 or return to patient baseline mental status), overall length of stay, admission to the intensive care unit (ICU), and mortality. The Charlson Comorbidity Index (CCI) was used to evaluate the severity of underlying comorbid conditions, as previously reported (20). For analysis, patients were placed into 2 categories: those with ammonia drawn at the time of admission and those without ammonia drawn. Ammonia levels were based on local MUSC laboratory values, with hyperammonemia considered to be >72 μmol/L.

Statistical methods

Statistical analyses were performed using the Statistical Analysis System (SAS) version 9.4 (SAS Institute, Cary, NC) and GraphPad Prism Version 7 (La Jolla, CA).

Comparisons between groups were evaluated using the Fisher exact test for categorical variables and the Student t test for continuous variables, with results presented as percentages or mean values with SDs, where appropriate. All statistical tests were 2-sided, and a P value <0.05 was considered statistically significant. Multivariable logistic regression was used to determine effect of multiple variables on clinical outcomes. Propensity score matching was implemented to match for severity of HE in the group with ammonia levels drawn and those without (the control group). The full set of 1,202 subjects was matched according to age, admission MELD score, and CCI score with 1:1 sampling using the “Greedy Nearest Neighbor” propensity matching algorithm. Based on similar probabilities of having an ammonia level drawn at onset of HE, a total of 592 subjects were matched using the above criteria.


Patient demographics and baseline characteristics

Between 2005 and 2015, 1,202 cirrhosis patients with HE were identified (Figure 1). Forty percent were women, and the most common causes of cirrhosis were nonalcoholic steatohepatitis (NASH) (38%) and alcohol (30%) (Table 1). Within this group, 551 patients (46%) had an ammonia level drawn at admission and 651 patients (54%) did not. Of the patients with an ammonia result, 328 (60%) had elevated ammonia levels—defined as >72 μmol/L (based on the MUSC laboratories upper limit of normal), while 223 (40%) had a normal ammonia level.

Figure 1
Figure 1:
Consort diagram of the study cohort. Query of MUSC's Clinical Data Warehouse for cirrhosis patients with a primary or secondary diagnosis of hepatic encephalopathy (ICD-9: 572.2) from 2005 to 2015 resulted in 1,440 unique admissions. After excluding patients who had an erroneous diagnosis of HE, without recorded lactulose dose or missing clinical data, 1,202 unique admissions were identified.
Table 1
Table 1:
Study cohort

The severity of liver disease was similar in patients with and without ammonia levels drawn, with similar MELD and Child-Pugh scores (Table 1). Patients were also similar in regards to comorbidities, as evidenced by their CCI (4.6 and 4.4) (Table 1). The majority of patients (84%) had previous admissions for overt HE (Table 1). Overall, lactulose noncompliance was the most common precipitant of HE (38%).

Patients with elevated ammonia and normal ammonia levels were also similar in severity of liver disease with MELD scores of 22 and 23, respectively, and Child-Pugh scores of 11 and 12, respectively (Table 2). Patients with elevated ammonia had a mean CCI of 4.5, similar to the CCI of 4.4 in the normal ammonia group (Table 2). In patients with normal ammonia levels, infection represented 42% of precipitant causes compared with 26% in the elevated ammonia group (P < 0.001) (Table 2).

Table 2
Table 2:
Baseline clinical features and outcomes based on the initial ammonia level

Relationship between lactulose therapy and ammonia level

All 1,202 patients received oral lactulose, with the overall average oral dose in the first 48 hours being 172 mL. Patients without ammonia levels and those with ammonia levels drawn did not differ in with regard to lactulose treatment, with lactulose doses of 167 and 171 mL in the first 48 hours, respectively (P = 0.08). In patients with elevated ammonia and normal ammonia, the average of lactulose given in the first 48 hours was identical (171 mL) (Table 2). Analysis of lactulose dose based on the ammonia level at the time of admission revealed that there was no correlation between ammonia level and treatment with lactulose (R2 = 0.0026) (Figure 2).

Figure 2
Figure 2:
Ammonia level and lactulose dose. The Spearman correlation coefficient was determined for the 551 patients who had ammonia levels drawn to investigate the potential relationship between ammonia level and total lactulose dose given in the first 48 hours of HE management. HE, hepatic encephalopathy.

Relationship between ammonia levels and secondary clinical outcomes

There was no difference in the time to clearing of HE among patients in the different ammonia groups (Figure 3). The overall time to resolution of HE (defined as West Haven stage 0 or return to baseline) was 3.7 days, with a resolution time of 3.6 days in those without ammonia levels taken and 3.6 days in those with ammonia levels drawn (P = 0.15) (Figure 3a). Time to resolution was similar in patients with elevated ammonia (3.4 days) and normal ammonia (3.8 days) (P = 0.07), as well (Figure 3b).

Figure 3
Figure 3:
HE clearing as a function of ammonia level. Clearing was defined as achieving West Haven stage 0 or returning to baseline mental status. Shown are Kaplan-Meier plots of time to clearing (days) in patients with ammonia levels drawn compared to those without an ammonia level drawn (a), as well as clearing rates in those with normal and elevated ammonia levels (b). HE, hepatic encephalopathy.

Overall, 35% of all patients required ICU admission, and there was no difference in ICU admissions among those with and without ammonia levels drawn. Patients with elevated ammonia and normal ammonia levels also had similar rates of ICU admission (Table 2). There was no difference in duration of stay in the ICU, overall length of stay, or mortality among patients with or without ammonia levels drawn. In the subset of subjects with ammonia levels drawn (n = 551), lactulose noncompliance was the most common HE precipitant in those with elevated ammonia levels (43%) (P = 0.03), with infection being the most common precipitant in those with normal ammonia levels (42%) (P = <0.001).

To further evaluate the role of ammonia in impacting both management and outcomes of cirrhotic HE inpatients, we performed logistic regression analysis to identify clinical variables associated with in-hospital mortality. Higher MELD scores and more severe HE (i.e., higher West Haven stages of 3 or 4) were associated with increased mortality (odds ratio = 1.09, 1.06–1.12 95% confidence interval; P < 0.001 and odds ratio = 3.03, 1.77–5.20 95% confidence interval; P < 0.001, respectively).

Propensity-matched cohort

We additionally performed propensity matching based on age, admission MELD score, and CCI score to control for potential bias related to possible variation in severity of both liver disease and comorbid conditions among patients. This provided a cohort of 592 patients (Table 3). In the propensity-matched cohort, 292 (50%) and 292 (50%) had and did not have ammonia levels drawn, respectively, at admission or onset of encephalopathy. West Haven stage and Child-Pugh score were also similar between groups. The groups did not differ based on sex, race, or etiology of cirrhosis (Table 3).

Table 3
Table 3:
Propensity-matched study cohort (n) = 592

More patients who did not have ammonia levels drawn had HE secondary to dehydration (21.6%) compared to those with ammonia levels obtained (11.5%); P = 0.0013 (Table 3), but no other differences were noted in HE etiology between the groups. There was also no difference in the dose of lactulose given in the first 48 hours between the 2 groups; those without ammonia levels drawn received an average of 174 mL lactulose in the first 48 hours compared with 172 mL in those with ammonia levels drawn (P = 0.6887). Rifaximin administration was also similar in the 2 groups.

Outcomes were the same in the 2 groups, with no differences noted in time to HE resolution, overall length of stay, ICU needs, ICU length of stay, or inpatient mortality (Table 3).


Although currently available literature suggests an important role for ammonia in the pathogenesis of HE, we demonstrate here that there is no evidence that ammonia levels are important in management of HE. We speculate that based on the data presented here and our own clinical experience, the choice to obtain an ammonia level at admission in patients with HE is largely random and likely based on practice habits. We have also demonstrated that both management of HE with lactulose and overall outcomes did not vary according to whether ammonia levels were drawn or not.

A previous study reported a correlation between ammonia levels and the severity of HE, which could ultimately serve to guide clinical management (15). However, the study included a far smaller patient cohort than in this study (121 vs 1,202 in our prematched cohort and 592 in our matched cohort) and did not report clinical outcomes of HE resolution, inpatient length of stay, or mortality, nor did it refer to management and treatment options used and whether they differed based on HE severity. Furthermore, this previous study noted that despite there being a correlation between higher ammonia levels and more severe HE, there was substantial variation in ammonia levels across all stages, including some patients with West Haven stage 0 who had elevated ammonia levels; this was likely important as these patients represented 25% of the study's cohort (15). In addition, since we believe that patients without clinical evidence of HE would be managed differently than those with clinical HE, such patients were excluded (i.e., those with grade 0 encephalopathy).

It should be emphasized that regardless of attempts to use ammonia as a potential diagnostic marker of HE, the diagnosis continues to be a clinical one, using staging systems such as the West Haven Criteria in combination with assessment of neuromotor dysfunction (21,22). Our data raise important questions about the cost-effectiveness of using ammonia levels to guide clinical management. We did not study the use of serial ammonia levels as a guide to management and therefore cannot comment on this practice. In addition, we did not study the use of ammonia levels in patients with unclear diagnosis, and we cannot exclude the possibility that ammonia levels may be useful in diagnostic dilemmas.

Despite the fact that standard practice for many institutions may include use of ammonia levels at or near admission of a cirrhosis patient with altered mental status in an attempt to diagnose HE, relying on an ammonia level may also lead to misdiagnoses. One study reported that the use of ammonia in guiding diagnosis and management of altered mental status in patients with cirrhosis presenting to the emergency department led to a 40% rate of misdiagnosis of HE when compared with scoring systems such as West Haven Criteria (23). The reason for this is likely because patients with normal ammonia levels may be misdiagnosed as not having HE, therefore impeding proper clinical care. Therefore, just as the diagnosis of HE should be based on clinical grounds, we speculate that management should follow the same premise.

We recognize limitations of this study. First, this study was retrospective, and this could introduce selection bias, particularly with regard to which patients had an ammonia level drawn. For example, patients who had an ammonia level drawn may have been viewed by their physicians to have had altered mental status of unclear etiology, and thus, ammonia was obtained as a diagnostic tool. It is also possible that patients with more severe HE may have had ammonia levels drawn with the idea that ammonia levels would be elevated and thus lead to more aggressive therapy. However, we found that the severity of liver disease (based on MELD and Child-Pugh score) was similar regardless of whether ammonia was drawn or not. Another potential limitation of our study is that all ammonia levels were assessed from venous blood draws, despite studies reporting that arterial ammonia levels and measurement of partial pressure of ammonia may better represent the concentration of ammonia reaching the brain's circulation (13). In addition, fasting ammonia levels are considered superior to those in postprandial patients; yet, this is something nearly impossible to collect from all patients who present acutely to the hospital (16). However, these potential concerns do not detract from the core finding of the study in that management did not depend on ammonia level.

In summary, we have found that in patients admitted with HE, regardless of whether an ammonia level was obtained or not, or whether it was normal or abnormal, management was the same. Although this study did not assess the potential utility of serial ammonia levels to monitor the response to therapy, our study suggests that the routine use of ammonia as either an initial diagnostic test or as a test that might direct therapy should be abandoned.


Guarantor of the article: Don C. Rockey, MD.

Specific author contributions: M.H.: study concept and design; acquisition of data; analysis; and interpretation of data; drafting of the manuscript; and critical revision of the manuscript for important intellectual content. D.C.R.: study concept and design; analysis and interpretation of data; drafting of the manuscript; and critical revision of the manuscript for important intellectual content and supervisory activities.

Financial support: This project was internally funded by the Medical University of South Carolina Department of Medicine: Division of Gastroenterology & Hepatology.

Potential competing interests: None.

Previous presentation: This work was presented at the 2017 Digestive Disease Week (DDW) Meeting in Chicago, Illinois.

Study Highlights


  • ✓ Lactulose is the standard treatment of hepatic encephalopathy (HE).
  • ✓ The mechanism of lactulose is often attributed to its ammonia-lowering abilities.


  • ✓ The presence or absence of an abnormal ammonia level did not guide lactulose therapy.
  • ✓ Ammonia levels were not predictive of clinical outcomes, time to resolution of HE, or severity of liver disease.
  • ✓ Ammonia levels do not guide management of HE.


1. Rahimi RS, Rockey DC. Hepatic encephalopathy: How to test and treat. Curr Opin Gastroenterol 2014;30(3):265–71.
2. Patidar KR, Bajaj JS. Covert and overt hepatic encephalopathy: Diagnosis and management. Clin Gastroenterol Hepatol 2015;13(12):2048–61.
3. Stepanova M, Mishra A, Venkatesan C, et al. In-hospital mortality and economic burden associated with hepatic encephalopathy in the United States from 2005 to 2009. Clin Gastroenterol Hepatol 2012;10(9):1034–41.e1.
4. Bajaj JS, O'Leary JG, Tandon P, et al. Hepatic encephalopathy is associated with mortality in patients with cirrhosis independent of other extrahepatic organ failures. Clin Gastroenterol Hepatol 2017;15(4):565–74.e4.
5. Schmidt ML, Barritt AS, Orman ES, et al. Decreasing mortality among patients hospitalized with cirrhosis in the United States from 2002 through 2010. Gastroenterology 2015;148(5):967–77.e2.
6. Rahimi RS, Rockey DC. Hepatic encephalopathy: Pharmacological therapies targeting ammonia. Semin Liver Dis 2016;36(1):48–55.
7. Prakash R, Mullen KD. Mechanisms, diagnosis and management of hepatic encephalopathy. Nat Rev Gastroenterol Hepatol 2010;7(9):515–25.
8. Atluri DK, Prakash R, Mullen KD. Pathogenesis, diagnosis, and treatment of hepatic encephalopathy. J Clin Exp Hepatol 2011;1(2):77–86.
9. Parekh PJ, Balart LA. Ammonia and its role in the pathogenesis of hepatic encephalopathy. Clin Liver Dis 2015;19(3):529–37.
10. Pantham G, Mullen KD. Practical issues in the management of overt hepatic encephalopathy. Gastroenterol Hepatol (NY) 2017;13(11):659–65.
11. Ahluwalia V, Betrapally NS, Hylemon PB, et al. Impaired gut-liver-brain Axis in patients with cirrhosis. Sci Rep 2016;6:26800.
12. Coltart I, Tranah TH, Shawcross DL. Inflammation and hepatic encephalopathy. Arch Biochem Biophys 2013;536(2):189–96.
13. Kramer L, Tribl B, Gendo A, et al. Partial pressure of ammonia versus ammonia in hepatic encephalopathy. Hepatology 2000;31(1):30–4.
14. Nicolao F, Efrati C, Masini A, et al. Role of determination of partial pressure of ammonia in cirrhotic patients with and without hepatic encephalopathy. J Hepatol 2003;38(4):441–6.
15. Ong JP, Aggarwal A, Krieger D, et al. Correlation between ammonia levels and the severity of hepatic encephalopathy. Am J Med 2003;114(3):188–93.
16. Vierling JM, Mokhtarani M, Brown RS Jr, et al. Fasting blood ammonia predicts risk and frequency of hepatic encephalopathy episodes in patients with cirrhosis. Clin Gastroenterol Hepatol 2016;14(6):903–6.e1.
17. Ghabril M, Zupanets IA, Vierling J, et al. Glycerol phenylbutyrate in patients with cirrhosis and episodic hepatic encephalopathy: A pilot study of safety and effect on venous ammonia concentration. Clin Pharmacol Drug Dev 2013;2(3):278–84.
18. World Medical Association. World Medical Association Declaration of Helsinki: Ethical principles for medical research involving human subjects. JAMA 2013;310(20):2191–4.
19. Pantham G, Post A, Venkat D, et al. A new look at precipitants of overt hepatic encephalopathy in cirrhosis. Dig Dis Sci 2017;62(8):2166–73.
20. D'Hoore W, Sicotte C, Tilquin C. Risk adjustment in outcome assessment: The Charlson comorbidity index. Methods Inf Med 1993;32(5):382–7.
21. Nabi E, Thacker LR, Wade JB, et al. Diagnosis of covert hepatic encephalopathy without specialized tests. Clin Gastroenterol Hepatol 2014;12(8):1384–9.e2.
22. Nabi E, Bajaj JS. Useful tests for hepatic encephalopathy in clinical practice. Curr Gastroenterol Rep 2014;16(1):362.
23. Gundling F, Zelihic E, Seidl H, et al. How to diagnose hepatic encephalopathy in the emergency department. Ann Hepatol 2013;12(1):108–14.
© The American College of Gastroenterology 2020. All Rights Reserved.