Predicting prognosis in acute liver failure: Ammonia and the risk of cerebral edema


  • Timothy J. Davern

    Corresponding author
    1. Gastroenterology Division, University of California at San Francisco, San Francisco, CA
    • Gastroenterology Division, University of California at San Francisco, 513 Parnassus Avenue, Room S-357, San Francisco, CA 94143-0538
    Search for more papers by this author

  • See Article on Page 1844.

  • Potential conflict of interest: Nothing to report.

Caring for patients with acute liver failure (ALF) is arguably the most challenging task facing the practicing hepatologist. In particular, accurate prediction of prognosis in patients with ALF is difficult but critically important because liver transplantation is potentially life saving. Complications of ALF, including refractory intracranial hypertension (ICH) and multiorgan failure, often develop rapidly and may prevent the possibility of successful liver transplantation. On the other hand, some patients with ALF may recover completely without progression to chronic liver disease. Subjecting these patients to liver transplantation not only uses a valuable donor organ unnecessarily but also exposes these patients to the morbidity of a major surgical procedure as well as lifelong immunosuppression.


ALF, acute liver failure; CE, cerebral edema; HE, hepatic encephalopathy; ICH, intracranial hypertension; MELD, Model for End-Stage Liver Disease.

Cerebral edema (CE) resulting in ICH and its complications has long been recognized as a common and potentially devastating consequence of ALF.1 The pathogenesis of CE in the setting of ALF remains murky, but the prevailing model focuses on astrocyte swelling, at least in part induced by ammonia toxicity, as the major component of brain edema2–4 (Fig. 1). Supporting the critical importance of ammonia in this process, Clemmensen and colleagues5 from Copenhagen reported a strong correlation of arterial ammonia and CE in patients with ALF. They studied 44 ALF patients, 14 of whom developed cerebral herniation, diagnosed clinically by development of dilated pupils unresponsive to light, and loss of brain stem reflexes. The median arterial ammonia was significantly higher in the group that developed CE (230 μmol/L) than in the group that did not (118 μmol/L), but there was overlap between the groups. Importantly, however, all of the ALF patients in their study that suffered CE had an arterial ammonia > 146 μmol/L. More recently, Bhatia and colleagues6 from New Delhi generally confirmed these findings in 80 consecutive ALF patients who were followed to either spontaneous recovery or death as liver transplantation was not available at their center. In their patients, an arterial ammonia concentration of >124 μmol/L was associated with deeper hepatic encephalopathy (HE) and development of CE and predicted mortality with an accuracy of almost 80%. Thus, although common wisdom dictates that blood concentrations of ammonia correlate relatively poorly with the severity of HE in patients with cirrhosis, the story seems to be different with ALF.7, 8

Figure 1.

Acute liver failure leads to increases in blood ammonia levels. The ammonia is taken up by brain astrocytes that metabolize ammonia via amidation of glutamate to glutamine. Intracellular excess of glutamine results in osmotic stress and cell swelling via several mechanisms. Hyperammonemia also alters astrocyte gene expression, disrupts various signaling pathways, and contributes to the increase in cerebral blood flow commonly observed in the setting of acute liver failure. Inflammatory mediators appear to also contribute to astrocyte swelling and cerebral edema. Brain swelling within the rigid confines of the skull may result in increased intracranial pressure, brain ischemia, cerebral herniation, and brain death.

The report from Clemmensen and colleagues5 raised several important questions related to the diagnosis, prognosis, and treatment of patients with ALF. For diagnosis, should invasive monitoring (bolting) be reserved for patients with marked ammonia elevations? Given that monitoring of intracranial pressure is invasive and carries a substantial risk for morbidity, primarily from bleeding, the ability to accurately select patients at highest risk for CE would obviously be very useful.9 For prognosis, should arterial ammonia levels be incorporated into a Model for End-Stage Liver Disease (MELD)–like prognostic scoring system? This is broadly analogous to recent proposals to incorporate serum sodium into MELD in order to better predict prognosis in certain patients with end-stage cirrhosis, such as those with diuretic refractory ascites.10, 11 Finally, with respect to therapy, will rapidly lowering arterial ammonia—by cathartics (lactulose), antibiotics (rifaximin), conjugation (sodium phenylacetate/sodium benzoate), hypothermia, or bioartificial liver support devices—improve neurological outcome in patients with ALF? Clearly, prospective studies are needed to answer these important questions.

In this issue of HEPATOLOGY, Bernal and colleagues12 from King's College in London have made a major contribution to the understanding of CE in patients with ALF. They compared arterial ammonia levels on admission in 165 ALF patients with severe (grade IIII-IV) HE, 50 patients with ALF with milder (<grade III) HE, 33 patients with decompensated chronic liver disease, and 9 patients without cirrhosis hospitalized after undergoing elective hepatobiliary surgery. The demographics of their ALF population are generally similar to those in the United States, with most (56%) of the cases attributed to acetaminophen poisoning. Of the 58 ALF patients who died without liver transplant, cerebral herniation was the primary cause of death in 17 (29%), whereas the remainder succumbed to multiorgan failure, although the admission ammonia levels were nearly identical in these two groups. Likewise, there were no significant differences in arterial ammonia levels between patients who survived without liver transplantation, those who required liver transplantation, and those who died. However, ammonia levels > 200 μmol/L were clearly associated with ICH, and this confirmed the earlier studies suggesting that marked ammonia elevations increased the risk of developing complications of CE, including cerebral herniation and death.

Were there factors that predicted progression of HE and development of ICH in the group of patients with severe liver injury but modest or no HE on presentation? On multivariate analysis, ammonia level and MELD score were independently associated with progression of HE, and a combination of ammonia > 100 μmol/L and MELD > 32 appeared to predict this most accurately. More importantly, the risk of developing ICH was independently associated with several factors: ammonia concentration, requirement for vasopressors or renal dialysis, and young age (<45 years). Interestingly, this article confirms the prior observation that youth is a poor prognostic feature; parenthetically, protection from CE appears to be one of the few advantages that we derive from the cerebral atrophy that inevitably accompanies advancing age. Unfortunately, Bernal and colleagues could not determine an optimal threshold level of ammonia for predicting development of ICH; for example, a value of >200 μmol/L had excellent specificity (92%) but very low sensitivity (25%), thus failing to correctly identify most (75%) of the patients who would ultimately develop ICH. However, in the group of patients with serial ammonia levels measured, a falling level appeared to correlate with lack of ICH development, whereas a lack of fall in ammonia was more worrisome for this complication.

Thus, although the authors confirmed that arterial ammonia concentration is a strong independent risk factor for development of both severe HE and ICH, they were not able to confirm the prior observation by Clemmensen et al.5 that an arterial ammonia value of >150 μmol/L, or any specific level for that matter, accurately predicts cerebral herniation. Although the ammonia assays used in the two studies differed, this alone appears to be an unlikely explanation for the discrepant results. Bernal and colleagues found that although patients with an ammonia concentration of >200 μmol/L had a very high risk of developing ICH, this higher threshold failed to identify the majority of ALF patients who would ultimately develop ICH in their study. They appropriately conclude that factors other than ammonia concentration contribute to the development of ICH.

Does the article have any weaknesses? One potential criticism of the article is that intracranial pressure (ICP) monitoring was performed in only a minority of patients (55 of 165); in the remainder, the development of ICH was defined by changes on physical examination such as the development of pupillary abnormalities or by autopsy evidence of CE. Because physical examination has limited sensitivity and specificity for detecting ICH, this represents a weakness of the analysis. In addition, the details of why some patients were monitored and others were not are not explicated stated. Furthermore, many patients had only one or two ammonia measurements; this probably reflects the fact that this was, in essence, a retrospective analysis of prospectively collected data, rather than a prospective protocol per se. These weaknesses notwithstanding, this article represents the latest in a long list of important studies from the Liver Intensive Care Unit at King's College Hospital, arguably the most important single liver center anywhere when it comes to studying ALF and a model for the rest of us to emulate.

This study suggests that ammonia-lowering interventions should be pursued in patients with ALF even in the presence of mildly elevated arterial ammonia because such patients may still be at risk for developing ICH. Clearly, more work is needed in the context of large, multicenter, prospective studies to better define prognostic parameters of ALF, including those associated with development of ICH. In such studies, all patients should have serial, at least daily arterial ammonia levels measured, and ICH should be confirmed by direct ICP measurement rather than on clinical grounds alone. Hopefully, through such studies, we will obtain important insights into the pathogenesis of ALF in humans and be better prepared to determine prognosis and provide optimal therapy to patients with this life-threatening syndrome.


The author thanks Sadie McFarlane for her assistance with the graphics used in Fig. 1.