Predicting the Outcome of Fulminant Hepatic Failure

Fulminant hepatic failure is a rare disorder, with an estimated 2300 to 2800 cases in the United States per year. It is responsible for an estimated 0.1% of all deaths and 7% of liver transplants among adults.1 With its high associated morbidity and mortality, the diagnosis of “fulminant hepatic failure” has grim implications. OPTN, Organ Procurement and Transplantation Network; SRTR, Scientific Registry of Transplant Recipients; ALFSG, Acute Liver Failure Study Group; INR, International Normalized Ratio; APACHE, Acute Physiology and Chronic Health Status Evaluation; MELD, Model for End-Stage Liver Disease. Prior to the advent of liver transplantation, reported mortality was in excess of 80%.2 With modern intensive care and, particularly, the application of liver transplantation, the outcome of such patients has steadily improved. Among patients diagnosed with acute liver failure at the University of Pittsburgh Medical Center during the period 1983 to 1995, 14% survived without liver transplantation.3 The US Acute Liver Failure Study Group (ALFSG) evaluated the outcome of 295 patients evaluated at 13 hospitals during the period 1994 to 1996 and found the spontaneous survival rate to be 25%.4 In the most recent analysis of an ongoing prospective study, ALFSG enrolled 684 patients with acute liver failure at 25 sites. Among them 44% survived with medical therapy and 26% underwent liver transplantation; the overall survival rate was 67% (William Lee, personal communication, March 2005). For the multidisciplinary team managing a patient with fulminant hepatic failure, it is crucial that an early decision is taken regarding transplantation before the patient develops life-threatening complications of worsening cerebral edema and multiorgan failure.5 The window for decision-making is thus short compared to the majority of patients in need of a liver allograft. A prognostic system of high diagnostic accuracy that could predict outcome early in the clinical course would be an invaluable tool for the management team. Such a system would not only help identify patients in need of emergent transplantation but would also ascertain spontaneous survivors who would thus be spared a major procedure and lifelong immunosuppression. Over the years, several prognostic instruments have been developed. The Clichy criteria were developed in patients with fulminant hepatitis B.6 The prognostic markers included factor V level, patient age, absence of HBsAg, and serum alfa-fetoprotein level. The model, however, included a single etiological group and required factor V measurement, which is not commonly available. The most widely applied are the King's College Hospital criteria, which have stood the test of time since their publication 15 years ago.2 The study was based on a retrospective analysis of 585 patients managed during the period 1973 to 1985, before the general application of liver transplantation. The findings were validated in a separate cohort of 175 patients. Two sets of prognostic criteria were developed, one for acetaminophen-induced hepatotoxicity and the other for nonacetaminophen etiologies. Among patients with acetaminophen-induced fulminant hepatic failure, an arterial pH less than 7.3 or a combination of International Normalized Ratio (INR) greater than 6.5 and serum creatinine higher than 3.4 mg/dL in a patient with grade III or IV hepatic encephalopathy were indicative of a poor outcome. In nonacetaminophen cases, an INR higher than 6.5 or any two of the following variables indicated worse outcome: age younger than 10 or older than 40 years, etiology (non-A non-B hepatitis, halothane, drug hepatotoxicity, Wilson's disease), interval between jaundice and encephalopathy more than 7 days, INR greater than 3.5, and serum bilirubin higher than 17.5 mg/dL. The criteria have since been validated by other investigators.3, 7, 8 The King's criteria are also notable for their ease of use, as readily available variables were included. Several investigators have, however, noted that although the criteria are highly predictive of a poor outcome when fulfilled, a lack of fulfillment does not assure survival.3, 7, 8 It has been suggested that serial measurements of APACHE III scores might be helpful in identifying nonsurvivors among patients who do not fulfill King's criteria.8 Several other indices and measurements have been proposed as prognostic markers, such as coagulation factor V and factor VIII–factor V ratio,9 arterial ketone body ratio,10 galactose elimination capacity,11 plasma Gc protein concentration,12 and most recently Model for End-Stage Liver Disease (MELD),13 but none have been validated in large prospective studies. The search goes on! The study in this issue by MacQuillan et al.14 needs to be viewed in this context. The authors performed a retrospective analysis of 83 consecutive patients with fulminant hepatic failure admitted to their institution during a period of 3 years between 2000 and 2003. Fifty-eight percent survived with medical management, 23% underwent liver transplantation and 19% died without receiving an allograft. Survivors with medical management (spontaneous survivors) were younger, were more likely to have acetaminophen-induced hepatotoxicity, and had significantly lower serum bilirubin levels on admission. Although serum phosphate levels were significantly higher on admission among nonsurvivors (those who died or had a liver transplant), this difference was mainly attributed to those who underwent transplantation. Arterial blood lactate levels were significantly lower among survivors compared with nonsurvivors on admission and over the subsequent 12 hours. When acetaminophen and nonacetaminophen cases were analyzed separately, blood lactate levels remained significantly different between survivors and nonsurvivors. On multivariate analysis, the 12-hour arterial blood lactate level was the only significant predictor of survival among acetaminophen cases, and also in nonacetaminophen cases after exclusion of patients who underwent transplantation. The authors concluded that serum phosphate levels had limited prognostic value, but 12-hour blood lactate level in conjunction with admission serum bilirubin level might have prognostic utility in patients with fulminant hepatic failure. Hypophosphatemia was first described among patients with fulminant hepatic failure almost 35 years ago.15 There are many reasons for hypophosphatemia to develop in fulminant hepatic failure. The most likely reason is the increased intracellular shift of serum phosphate resulting from an increased synthesis of phosphorylated adenine nucleotides due to enhanced hepatocyte regeneration. Hypophosphatemia may thus be reflective of increased liver regenerative activity following massive or submassive hepatic necrosis, and therefore may be an indicator of better outcome. Conversely, a high serum phosphate level may be due to suboptimal hepatic regeneration or a direct result of phosphate release from liver necrosis. Among patients who develop renal insufficiency, a further increase in serum phosphate could develop from reduced renal excretion. Several investigators have examined the prognostic utility of serum phosphate, and some have proposed that hypophosphatemia was predictive of spontaneous survival.16-18 In contrast, MacQuillan et al. did not find the serum phosphate level to be of prognostic value.14 The authors acknowledge that their results could have been affected by a type II error in view of the small number of patients studied. Blood lactate is frequently elevated in patients with fulminant hepatic failure.19 Hyperlactatemia and high brain lactate levels have been noted in animal models of fulminant hepatic failure as well.20-23 In addition, high blood lactate levels develop commonly in critically ill patients and are associated with poor outcome.24, 25 The underlying reasons for hyperlactatemia in fulminant hepatic failure are poorly understood. Lactate production is closely linked to carbohydrate metabolism. Central to this process is glycolysis, which results in the breakdown of each molecule of glucose to two pyruvate molecules. Under aerobic conditions, pyruvate is oxidized to form the acetyl component of acetyl-coenzyme A, which is further oxidized to CO2 via the citric acid cycle.26 However, under anaerobic conditions pyruvate is reduced to lactate via lactate dehydrogenase, and this process allows the body to regenerate NAD+, which is necessary for glycolysis to continue. With adequate oxygenation, lactate is converted back to pyruvate, which is then used for gluconeogenesis. Although there is increased release of lactate from liver, lungs, and intestine in patients with fulminant hepatic failure, it is not always due to tissue hypoxia.27, 28 It has been suggested that increased energy requirements in such patients stimulates glycolysis, which results in increased pyruvate production. Excess pyruvate is then converted to lactate, although there is no tissue hypoxia.28 Blood lactate has previously been reported to be an early indicator of outcome in acetaminophen-induced fulminant hepatic failure.29 A value of 3.5 mmol/L early after admission was noted to have a sensitivity of 67% and specificity of 95% for nonsurvival. Another study reported a predictive model for acute liver failure that incorporated serum albumin, lactate, alanine, and pyruvate and that had a sensitivity of 94% and specificity of 86% for death or liver transplantation.30 However, validation studies involving large number of patients are lacking. The current study provides further evidence of the prognostic value of arterial blood lactate level during the early phase of fulminant hepatic failure. If validated by well-powered prospective studies, blood lactate level may serve as an additional discriminator to the well-tested King's College Hospital criteria. Even the best-developed prognostic model has its limitations.31 Despite belonging to a well-defined disease state such as “fulminant hepatic failure,” there is considerable variation among individual patients, particularly when a large number of variables are assessed. Because of the heterogeneity of data, precise prediction of outcome in an individual patient may not be possible by application of a predictive model. Clinicians, beware that such prognostic tools are meant to be used not as sole determinants of clinical decisions but as complementary devices along with all other available information.