**2.2 Prognostic models**

254 Liver Transplantation – Basic Issues

Mortality rates correlate with the severity of hepatic encephalopathy (HE), reported at 30% for grade 2, 45-50% for grade 3 and 80-90% for grade 4 HE (Daas et al., 1995; Hoofnagle et al., 1995). A multicenter US series, in which 39% participants had APAP hepatotoxicity, showed a 52% 3-week transplant-free survival in patients with grade 1-2 encephalopathy, but only 33% with grade 3-4 HE survived without transplant (Ostopowicz et al., 2002). Conversely, 85% of patients with non-APAP ALF without HE experienced spontaneous recovery (Elinav et al., 2005). Paradoxically, those with more rapid development of HE (i.e., APAP-induced) appear to have a better outcome than those with a longer interval between the development of symptoms and HE (i.e., DILI) (Bernuau et al., 1986a; O'Grady et al., 1989; O'Grady et al., 1993). A distinctive feature ALF-induced HE is the development of cerebral edema, the complete pathophysiology of which remains poorly understood. Cerebral edema develops in nearly 80% of patients who progress to grade 4 HE, leading to intracranial hypertension with subsequent ischemic brain damage or brainstem herniation, accounting for up to 50% of ALF mortality (Clemmensen et al., 1999; Jalan et al., 2003). Intracranial pressure (ICP) monitoring is more often utilized in patients who are deemed candidates for OLT, and ICP may be more aggressively managed in these cases. ICP monitors may also be of significant value during the transplant operation, when fluctuations in ICP are common (Philips et al., 1998). ICP monitoring is associated with up to a 10% risk of intracranial hemorrhage, and it has not been shown to change 30 day post-OLT survival (Gasco et al., 2010). Thus, the indication and timing of use of ICP monitoring devices remain controversial (Vaquero et al., 2005). Intracranial hypertension may persist during the first 10-12 hours following liver transplantation, thus ICP monitoring, if utilized, should

continue during and after surgery (Bismuth et al., 1995; Jalan et al., 2003).

ALF-induced hemodynamic changes can be difficult to distinguish from infection and sepsis and are complicated by the fact that ALF patients may not develop leukocytosis or fever. Bacterial infection is the cause of death in up to 37%, with the most common sites of infection being pulmonary (47%), blood (26%), and urine (23%) (Bernal et al., 2003). Fungal infections, especially Candida sp., are seen in up to 32%, occur later in the course of disease, particularly after use of antibiotics or in the setting of renal dysfunction, and are often associated with bacterial infection (Rolando et al., 1991; Vaquero et al., 2003). Active infection is a contraindication to OLT. The empiric use of antibiotics is controversial. Prophylactic antibiotics decrease the number of infections, but do not change overall outcome (Rolando et al., 1990; Rolando et al., 1996; Stravitz et al., 2007). Some centers administer anti-infectives (antibacterial and antifungal) to patients who have significant isolates on surveillance cultures, have progression to Stage 3-4 HE, have refractory hypotension, or have clinical evidence of systemic inflammatory response syndrome (Stravitz et al., 2007). Periodic surveillance cultures and frequent chest radiographs can help

The burden of medical follow-up after OLT can be substantial, and quality of life can be significantly affected. Therefore, the decision to offer OLT to an individual patient also

**2.1.4 Hepatic encephalopathy** 

**2.1.5 Infection** 

detect bacterial and fungal infections early.

**2.1.6 Psychosocial predictors** 

Multiple prognostic models have been proposed to help determine the likelihood of spontaneous survival (Table 2) (Antoniades et al., 2007; Bailey et al., 2003; Bernuau et al., 1986b; Bernuau, 1993; Craig et al., 2010; Harrison et al., 1990; Itai et al., 1997; O'Grady et al., 1989; Pereira et al., 1992; Rolando et al., 2000; Schiodt et al., 2005; Van Thiel, 1993). However, many of these models are methodologically flawed and subject to bias. In addition, many equate OLT with death, which falsely elevates the positive predictive value of these prognostic systems (Craig et al., 2010).


Table 2. Comparison Between the Various Prognostic Scoring Systems for Acute Liver Failure APAP-acetaminophen; MELD-model for end-stage liver disease; INR-international normalized ratio; T-temperature; BP-blood pressure; HR-heart rate; RR-respiratory rate; Na-sodium, K-potassium; WBC-white blood cell count. Clichy (Bernuau et al, 1986); King's Criteria (Bernal et al., 2002; O'Grady et al., 1989), Apache II (Mitchell et al., 1998); MELD (Schmidt & Larsen, 2007; Villamil et al., 2007; Wiesner, 2004; Yantorno et al., 2004; Zaman et al., 2006)

## **2.2.1 King's college hospital criteria**

The most widely applied prognostic system are the King's College Hospital criteria (King's criteria) developed from a retrospective cohort of nearly 600 patients (Bernal et al., 2002; O'Grady et al., 1989). The Kings criteria incorporate both the etiology of ALF (APAP- versus non-APAP induced ALF) and clinical parameters of disease (O'Grady et al., 1989). In a metaanalysis of studies using the Kings criteria, the pooled sensitivity and specificity was 69% and 92%, respectively (Bailey et al., 2003). The Kings criteria appear to have high positive predictive values (80% in APAP induced ALF and 70-90% in non-APAP induced ALF) but poorer negative predictive values (70-90% and 25-50%, respectively). A recent meta-analysis found that the Kings criteria for non-APAP induced ALF have good specificity, especially for patients with high grade encephalopathy (McPhail et al., 2005). The Kings criteria are helpful in identifying those who may need OLT, but up to 20% of those meeting criteria potentially could have survived without OLT, and those not meeting criteria may still require transplantation. The addition of arterial blood lactate levels to the model has improved its sensitivity (Bernal et al., 2002; MacQuillan et al. 2005).

### **2.2.2 Other models and predictors**

Other Models and Predictors. The Clichy criteria were developed in a cohort of French patients with acute hepatitis B virus infection (Bernuau et al., 1986b). These criteria suggest that a serum factor V level of <20% in patients younger than 30 years or <30% in any patient with grade 3-4 HE has validity as a marker of mortality. The criteria predicted a poor outcome with a sensitivity and specificity of 86% and 76%, respectively. Factor V level measurements are less readily available to the clinician than are the measures in the Kings criteria, therefore, this prognostic model is not commonly utilized (Izumi et al., 1996; Pauwels et al., 1993). In addition, this model has not been validated in the non-HBV population. A factor V <10% has been shown to predict a poor outcome with a sensitivity of 91% and a specificity of 100%; while a factor VIII : V ratio of >30 similarly predicts outcome (91% sensitivity, 91% specificity) (Pereira et al., 1992). The admission Acute Physiology and Chronic Health Evaluation (APACHE) II is ineffective in predicting who will survive without transplantation, since many patients who do not meet the severity criteria will ultimately die of subsequent complications (Mitchell et al., 1998).

Elevated arterial ammonia levels increase the risk of developing intracranial hypertension. A level of >150 µmol/L predicts development of intracranial hypertension with a sensitivity of 60% and a specificity of 84% (Kitzberger et al., 2009). Concentrations of more than 100- 150 µmol/L have been positively correlated with cerebral herniation (Bernal et al., 2007; Bhatia et al., 2006; Clemmensen et al., 1999; Toftent et al., 2006).

Serum alpha-fetoprotein (AFP) is generally considered a marker of hepatocellular regeneration. There has been no consistent correlation seen between the absolute AFP level and outcome in ALF (Tofteng et al., 2006). However, an increasing AFP level has been strongly associated with a more favorable outcome (Schiodt et al., 2006; Yang et al., 2002). A threshold AFP of ≤3.9 μg/L at 24 hours following the peak ALT identified nonsurvivors with a sensitivity and specificity of 100% and 74%, respectively, and a negative predictive value of 100% (Schmidt et al., 2005). In addition, it has been shown that a rising AFP level between day 1 and day 3 from presentation predicted survival without transplantation, whereas a decreasing level was seen in 80% of those who died (Schiodt et al., 2006).

Persistently elevated phosphate levels may be associated with a poorer prognosis in the setting of acetaminophen-induced ALF. A serum phosphate level >1.2 mmol/L on day 2 or

O'Grady et al., 1989). The Kings criteria incorporate both the etiology of ALF (APAP- versus non-APAP induced ALF) and clinical parameters of disease (O'Grady et al., 1989). In a metaanalysis of studies using the Kings criteria, the pooled sensitivity and specificity was 69% and 92%, respectively (Bailey et al., 2003). The Kings criteria appear to have high positive predictive values (80% in APAP induced ALF and 70-90% in non-APAP induced ALF) but poorer negative predictive values (70-90% and 25-50%, respectively). A recent meta-analysis found that the Kings criteria for non-APAP induced ALF have good specificity, especially for patients with high grade encephalopathy (McPhail et al., 2005). The Kings criteria are helpful in identifying those who may need OLT, but up to 20% of those meeting criteria potentially could have survived without OLT, and those not meeting criteria may still require transplantation. The addition of arterial blood lactate levels to the model has

Other Models and Predictors. The Clichy criteria were developed in a cohort of French patients with acute hepatitis B virus infection (Bernuau et al., 1986b). These criteria suggest that a serum factor V level of <20% in patients younger than 30 years or <30% in any patient with grade 3-4 HE has validity as a marker of mortality. The criteria predicted a poor outcome with a sensitivity and specificity of 86% and 76%, respectively. Factor V level measurements are less readily available to the clinician than are the measures in the Kings criteria, therefore, this prognostic model is not commonly utilized (Izumi et al., 1996; Pauwels et al., 1993). In addition, this model has not been validated in the non-HBV population. A factor V <10% has been shown to predict a poor outcome with a sensitivity of 91% and a specificity of 100%; while a factor VIII : V ratio of >30 similarly predicts outcome (91% sensitivity, 91% specificity) (Pereira et al., 1992). The admission Acute Physiology and Chronic Health Evaluation (APACHE) II is ineffective in predicting who will survive without transplantation, since many patients who do not meet the severity criteria will

Elevated arterial ammonia levels increase the risk of developing intracranial hypertension. A level of >150 µmol/L predicts development of intracranial hypertension with a sensitivity of 60% and a specificity of 84% (Kitzberger et al., 2009). Concentrations of more than 100- 150 µmol/L have been positively correlated with cerebral herniation (Bernal et al., 2007;

Serum alpha-fetoprotein (AFP) is generally considered a marker of hepatocellular regeneration. There has been no consistent correlation seen between the absolute AFP level and outcome in ALF (Tofteng et al., 2006). However, an increasing AFP level has been strongly associated with a more favorable outcome (Schiodt et al., 2006; Yang et al., 2002). A threshold AFP of ≤3.9 μg/L at 24 hours following the peak ALT identified nonsurvivors with a sensitivity and specificity of 100% and 74%, respectively, and a negative predictive value of 100% (Schmidt et al., 2005). In addition, it has been shown that a rising AFP level between day 1 and day 3 from presentation predicted survival without transplantation,

Persistently elevated phosphate levels may be associated with a poorer prognosis in the setting of acetaminophen-induced ALF. A serum phosphate level >1.2 mmol/L on day 2 or

whereas a decreasing level was seen in 80% of those who died (Schiodt et al., 2006).

improved its sensitivity (Bernal et al., 2002; MacQuillan et al. 2005).

ultimately die of subsequent complications (Mitchell et al., 1998).

Bhatia et al., 2006; Clemmensen et al., 1999; Toftent et al., 2006).

**2.2.2 Other models and predictors** 

3 following APAP overdose carries a sensitivity of 89% and a specificity of 100% for predicting poor outcome (Baquerizo et al., 2003; Chung et al., 2003; Schmidt et al., 2003). The level of Gc-globulin, a protein which is markedly reduced in the setting of tissue injury, has not been shown to reliably predict survival in those with APAP-induced ALF. In non-APAP ALF; however, using a cutoff value of ≤80 mg/L, the Gc-globulin level carries a positive predictive value of 74% and a negative predictive value of 81% (Schiodt et al., 2005; Schiodt et al., 2007)). An elevated arterial blood lactate level following volume resuscitation predicts worse survival in APAP-induced ALF (Bernal et al., 2002; Cholongitas et al., 2008; MacQuillan et al. 2005).

The model for end-stage liver disease (MELD) scoring system, is an excellent prognostic model for chronic liver disease (Schmidt & Larsen, 2007; Villamil et al., 2007; Wiesner, 2004; Yantorno et al., 2004; Zaman et al., 2006). However, it has been limited as a prognostic model in ALF, because it does not account for most of the extremely important outcome predictors in ALF, including age, etiology of ALF and duration of jaundice. The MELD score has a sensitivity and specificity of <75% for predicting outcome in all forms of ALF (Bernal et al., 2007; Dhiman et al., 2007; Riorden & Williams, 2003). Whether modification of the MELD with these important factors would improve the MELD as standard scoring system in prognosis of ALF remains to be seen. Dhiman and colleagues compared clinical predictors of MELD and Kings criteria in patients with ALF. Clinical predictors were superior to both MELD and Kings criteria in predicting prognosis of ALF. Significant factors included age >50 years, jaundice to encephalopathy time greater than 7 days, grade 3-4 encephalopathy, cerebral edema, prothrombin time ≥35 seconds, and serum creatinine ≥1.5mg/dL) are associated with a poor prognosis. The presence of 3 or more of these factors is associated with a poorer prognosis (Dhiman et al., 2007; O'Grady et al., 2007). Molecular markers of cell apoptosis have also been found to be helpful prognostic factors. Bechmann looked at replacing the bilirubin value in the MELD score with the ratio of CK18/M65, a marker of cell death. This model was found to be associated with higher sensitivity and specificity. Although this study was limited by a small number of patients, the idea of using molecular markers of cell death in predicting prognosis of ALF is promising and needs to be studied in larger cohorts (Bechmann et al., 2010). Cytokeratin 18-based modification of the MELD score improves prediction of spontaneous survival after acute liver injury.

In effort to develop a functional scoring model for non-APAP induced ALF, Miyaki and colleagues looked at 4 prognostic factors – etiology of ALF, hepatic coma grade (III or IV), systemic inflammatory response syndrome, and ratio of total to direct bilirubin (>2.0). The authors found these factors to be predictors of 2-week outcome with high positive and negative predictive values, 93.3%, and 81.8%, respectively (Miyake et al., 2005). This prognostic model would help the clinician predict prognosis and consideration for liver transplantation for patients with non-APAP ALF, however it requires validation before it can be widely clinically applied.

A liver volume of <1000 mL on computed tomography (CT) imaging is also associated with a high mortality rate, a finding which has been validated (Shakil et al., 2000; Yamagishi et al., 2009). Based upon these findings, a prognostic formula has been proposed, but is not widely utilized. Liver biopsy may also be helpful in determining the cause of the ALF and, theoretically, the severity and extent of liver damage. Hepatic necrosis of more than 70% was associated with a transplant free survival of <10% in one analysis (Scotto et al., 1973). However, there is a great degree of sampling error, and more recently, a multivariate analysis of 97 consecutive patients found that the amount of necrosis was not predictive of mortality (Miraglia et al., 2006; Voigt et al., 2007).

Based upon the available data, the current prognostic scoring systems have not consistently demonstrated reliable accuracy in predicting outcome from ALF and the subsequent need for OLT. Therefore, the American Association for the Study of Liver Diseases (AASLD) does not recommend reliance on any one of these systems (W Lee & Larson, 2005).
