INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Critically ill patients with advanced liver disease (ALD) constitute a significant percentage of patients in intensive care units (ICUs) at major referral centers. For many of these patients, orthotopic liver transplant (OLT) has provided the opportunity for cure from the underlying disease, and is currently the treatment of choice (1). However, the shortage of organ donors dictates careful consideration to ensure that organs are given to those most likely to benefit from transplantation. Screening and evaluating patients for possible OLT is a complicated, time-consuming process (4, 5) that may involve multiple consultations from various specialists as well as extensive imaging and physiological studies. Regrettably, many who present to the ICU with complications stemming from ALD are deemed ineligible for OLT. Patients may be denied consideration for OLT for reasons predating critical illness, such as ongoing alcohol abuse or new medical conditions that make the risk of the liver transplant procedure prohibitive. Such patients can be challenging to clinicians caring for them (6) and yet some do survive despite being denied the opportunity for OLT. Although much has been written regarding critically ill patients with liver disease (7), outcome predictors for those denied listing for OLT have not been previously described. Accordingly, we performed a detailed, retrospective analysis of patients with ALD that led to admission to our ICU over a 3¹/₂-yr period from 1994 to 1997. All patients were evaluated for OLT. This analysis focuses on those deemed ineligible for listing for OLT.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients and Data Abstraction

After receiving Institutional Review Board approval, we reviewed the medical records of all hospitalized adults with ALD admitted to the medical or surgical ICU at the University of Chicago Hospital. The hospital is a 480-bed tertiary care center with separate medical and surgical ICUs. The medical ICU is a closed unit managed by housestaff, a critical care fellow, and a critical care attending physician, whereas the surgical ICU is an open unit managed by the primary surgical team with critical care service available for consultation. Patients were identified by crossing the ICU database from May 1, 1994 to December 31, 1997 with ICD-9 codes describing all possible liver diseases. Only those patients who had active liver disease that necessitated their ICU admission were evaluated. Furthermore, only those patients deemed sick enough to merit consideration for liver transplantation were included in the database. For the purposes of this retrospective chart review, this included all those patients who received a consultation from the liver transplant surgery team. Furthermore, it excluded those patients deemed by the liver transplant team not sick enough to merit consideration for OLT. Table 1 lists indications for OLT at our institution. Patients were categorized according to whether or not they were placed on the liver transplant list. A patient's status with regard to listing for OLT was based on his or her ultimate disposition on the list. Patients initially placed on the list but subsequently removed after developing a complication or condition prohibiting transplantation were in the group "not listed for OLT." All charts were reviewed by three persons experienced in critical care chart abstraction (two critical care physicians and one critical care nurse-clinician). Information from charts was logged onto a standardized computer data collection form. All charts were reviewed independently by two separate data extractors. For those patients with more than one hospital admission over the study period, only the first admission was included in the analysis to ensure independence of observations.

Variable Definition

Demographic data (age, sex, etiology of liver disease, ICU admitting diagnosis, ICU length of stay, reason for ineligibility for OLT) were recorded for all patients. Clinical data while in the ICU included (1) presence of stage II or greater encephalopathy (stage IIdrowsiness, inappropriate behavior, inability to maintain sphincter control; stage IIIpatient sleeps most of the time but is arousable, speech is incoherent, confusion is marked; stage IVpatient may [IVA] or may not [IVB] respond to painful stimuli [12]); (2) presence of gastrointestinal (GI) bleeding (active upper and/or lower GI tract bleeding requiring at least 2 units packed red blood cells in the ICU); (3) sepsis (defined by clinical evidence of sepsis with a suspected source of infection as well as at least two of the following: temperature > 38.5° C or < 35.5° C; leukocyte count > 10,000/mm³ or < 3,500/mm³; heart rate > 100 beats/min; respiratory rate > 24 breaths/min; systolic blood pressure < 90 mm Hg, despite sufficient fluid replacement or the need for vasoactive agents to maintain systolic blood pressure  90 mm Hg; elevated serum lactate levels or metabolic acidosis not secondary to respiratory alkalosis; oliguria with urine output < 20 ml/h despite sufficient fluid replacement) (13); (4) spontaneous bacterial peritonitis (SBP), defined by at least one of the following: (a) a positive culture of ascitic fluid for bacteria, (b) a gram stain of ascitic fluid positive for the presence of bacteria, and (c) an ascitic fluid white blood cell count with greater than 300 polymorphonuclear cells per milliliter, or a total of 500 white blood cells per milliliter; (5) acute renal failure requiring hemodialysis (HD); and (6) respiratory failure requiring mechanical ventilation (MV) through an endotracheal tube. Other data collected included (7) Acute Physiology and Chronic Health Evaluation (APACHE) II score (14), (8) Glasgow Coma Scale (15), (9) type of respiratory failure (see below), and (10) multisystem organ failure scores (MSOF) on ICU admission. The cause of acute renal failure (acute tubular necrosis [azotemia, oliguria, renal ultrasound without evidence of postrenal obstruction, fractional excretion of sodium (FeNa) greater than 1%] versus hepatorenal syndrome [prerenal azotemia unresponsive to volume repletion, urine sodium less than 10 mEq/L]) was recorded when pertinent. Immediate cause of death and status with regard to "do not resuscitate" orders were also recorded for all patients. Organ failures were defined in the following manner: renal failurecreatinine > 3.4 mg/dl; hepatic failurebilirubin > 3.5 mg/dl, alkaline phosphatase > 350 U/L, and/or international normalized ratio (INR) > 2; cardiovascular failureneed for dobutamine, norepinephrine, or epinephrine at any dose, or dopamine at > 5 µg/ kg/min; hematologic failurewhite blood cells (wbc) < 2.0/µl and/or platelet count < 40,000/µl; respiratory failureneed for intubation and MV (16). The type of respiratory failure was divided into four categories. Type I: acute hypoxemic respiratory failure (AHRF)any alveolar filling process identified by chest X-ray associated with impaired oxygenation (e.g., pneumonia, pulmonary edema from any cause, alveolar hemorrhage). Included in this group were patients with acute respiratory distress syndrome (ARDS) as defined by the American-European Consensus Conference on ARDS (17). Type II: ventilatory failureany process associated with either a decreased drive to breathe, impaired neuromuscular competence of the respiratory system, or an excessive respiratory load (18). This group included reactive airways disease exacerbations, respiratory muscle weakness (any cause), as well as other causes of hypercapnic respiratory failure. Type III: postprocedure-related respiratory failureatelectasis due to sedative medications leading to respiratory failure. Type IV: shock-related respiratory failureany state of hypoperfusion leading to a need for MV. When more than one cause of respiratory failure was present, the patient was classified by the primary reason for MV as discerned from the ICU progress notes by the data abstractors. If a patient required endotracheal intubation solely for airway protection, but did not receive MV, he or she was not counted as having respiratory failure. Those with encephalopathy who needed airway protection as well as MV for respiratory failure were counted as having both encephalopathy and respiratory failure. ICU admission laboratory data were obtained and included INR, ammonia (µg/dl), bilirubin (mg/dl), albumin (g/dl), aspartate aminotransferase (AST) (IU/L), alanine aminotransferase (ALT) (IU/L), alkaline phosphatase (IU/L), and platelet count (k/µl). For the purpose of categorical comparisons, laboratory value abnormalities were categorized as "normal/mild," "moderate," and "severe" as summarized in Table 2.

Statistical Analysis

Parametric interval data were initially analyzed using a two-tailed Student's t-test. These data are reported as mean ± SD. Nonparametric data were initially examined using a Mann-Whitney U test or Kruskal-Wallis test as appropriate. These data are reported as median (interquartile range). Nominal data were analyzed by chi square analysis with Yates continuity correction or Fisher exact test where appropriate.

After the initial univariate analysis of the group of 183 patients, those variables found significant at p < 0.10 were subjected to stepwise backward logistic regression analysis, using hospital mortality as the dependent variable. Analyses to rule out interdependence between variables were included in the logistic regression analysis. Mortality odds ratios with 95% confidence intervals are presented. A p value of < 0.05 was considered significant. All statistical analyses were performed on a personal computer using SPSS version 8.0 software.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Demographics

A total of 667 ICU admissions for advanced liver disease were noted over the period of our data collection. One hundred and forty-three of these patients were not considered sick enough to merit evaluation for OLT, leaving 524 cases for transplant evaluation. Of these 524, 309 cases resulted in a transplant listing, whereas 215 were not listed. Sixty-nine and 32 patients respectively had more than one hospital admission, leaving 240 eligible and 183 ineligible patient admissions for the analysis (Figure 1). There were no age or sex differences between these two groups, though median ICU length of stay was shorter in the ineligible group compared with those listed for transplant (3 versus 5 d; p < 0.001). The incidence of hepatic failure (by laboratory criteria outlined in the METHODS section) was significantly higher in the group listed for OLT, compared with the group ineligible for OLT (88% versus 77%; p = 0.003). APACHE II scores and mortality were significantly higher in the ineligible group. ICU admission diagnoses varied greatly between those eligible and those ineligible for OLT. The vast majority of ICU admissions were for monitoring after OLT in the eligible group compared with GI bleeding and fulminant hepatic failure in the ineligible group (Table 3). The 56% mortality among those ineligible for OLT is substantially higher than the overall mortality of 20-30% in our medical ICU. However, it should be noted that the APACHE II scores for those ineligible for OLT were higher than the average APACHE II score for our ICU over the same period (20 versus 18; p = 0.003). The distribution of causes of liver failure varied when those ineligible for OLT were compared with those who were active on the transplant list. As seen in Table 4, there was a higher incidence of alcoholic cirrhosis and a lower incidence of hepatitis C, primary biliary cirrhosis, primary sclerosing cholangitis, and autoimmune hepatitis among those not listed. The remainder of the analysis focuses on the 183 patients deemed ineligible for OLT.

View larger version (16K): [in this window] [in a new window]   Figure 1.   Breakdown of patients with advanced liver disease evaluated for orthotopic liver transplantation.

For those denied listing for OLT, most admissions were to the medical ICU (90% versus 10% in the surgical ICU). Mortality rates were not different in the two ICUs (57% medical versus 50% surgical, p = 0.75). The most common reasons patients were denied listing for OLT included active ethanol consumption (35%), moribund state considered too unstable for surgery (25%), transplant evaluation in progress, but incomplete (14%), known malignancy (9%), sepsis (5%), and advanced age (4%). Incidences of various clinical end points for the 183 ineligible patients are summarized in Table 5.

For the group of patients ineligible for OLT, those with GI bleeding had a lower mortality than those without GI bleeding (44% versus 69%; p < 0.001 by chi square analysis). Among patients with GI bleeding, mortality was 39% for those with varices versus 58% for those without varices (p = 0.17). Causes of nonvariceal GI bleeding included upper GI bleeding, source unclear (16 patients), lower GI bleeding (4 patients), gastritis (1 patient), gastric cancer (1 patient), gastric polyp (1 patient), and esophageal desquamation (1 patient). Thirty-eight percent (25 of 65) of the ineligible patients with variceal bleeding underwent transvenous intrahepatic portosystemic shunt (TIPS). Those undergoing TIPS had a 41% mortality versus a 45% mortality for those not subjected to TIPS (p = 0.99). Survivors did not have a higher incidence of sclerotherapy compared with nonsurvivors (36% versus 23%; p = 0.24). Patients with varices receiving sclerotherapy had a 33% mortality versus 49% for those not receiving sclerotherapy (p = 0.25). Only three patients received balloon tamponade therapy for esophageal varices, with one survivor. Survivors received fewer transfusions of packed red blood cells (4 versus 7 units; p = 0.01) and fresh frozen plasma (4.5 versus 13.5 units; p = 0.02). All but three patients with GI bleeding presented to the ICU for this problem. Mortality for those with GI bleeding absent other comorbidities (e.g., MV, sepsis, HD) was only 6%.

The breakdown by type of respiratory failure was as follows: type I: acute hypoxemic respiratory failuren = 36 (all 36 met American-European Consensus Conference criteria for ARDS [15]); type II: ventilatory failuren = 46; type III: postprocedure-related respiratory failuren = 6; type IV: shock-related respiratory failuren = 17.

By univariate analysis, the following nine variables were significantly associated with mortality (p < 0.10) and were entered into the logistic regression analysis: APACHE II score, encephalopathy, GI bleed, sepsis, HD, MOSF score on ICU admission, need for MV, INR, and platelet count. For the purpose of the logistic regression analysis, INR and platelet count were evaluated by categories outlined in Table 1. Eighty percent of survivors versus 74% of nonsurvivors had hepatic failure as defined in the METHODS section (p = 0.42). When all of these variables were placed in the logistic regression model, APACHE II score, sepsis, need for MV, and GI bleeding remained significant. All were independent predictors of increased mortality, except GI bleeding, which was independently associated with decreased mortality. There was no evidence of interdependence between any of the variables reported. Table 6 summarizes results of the logistic regression analysis.

For those with acute renal failure requiring HD, 57% had acute tubular necrosis, 40% had hepatorenal syndrome, and 3% had other causes. Mortality was 90% in those with acute tubular necrosis versus 86% in those with hepatorenal syndrome (p = 0.87). Neither the presence of acute tubular necrosis nor hepatorenal syndrome was independently predictive of mortality by logistic regression.

In over 90% of cases, multiorgan system failure, fulminant hepatic failure, and/or sepsis were determined to be the cause of death. Among those patients who died, 76 of 103 (74%) received "do not resuscitate" orders. Only 6 of 80 (8%) survivors received "do not resuscitate" orders (p < 0.001).

As noted above, 80 patients who were denied listing for OLT survived their hospital stay. We were able to obtain 1-yr follow-up for 67 of these 80 patients and noted that 46 of these 67 were still alive at 1 yr. Of the 46 long-term survivors, 6 eventually received liver transplants. Reasons these six were initially denied listing for OLT were as follows: active ethanol consumption (n = 2), transplant evaluation in progress, but incomplete (n = 3), known malignancy (n = 1; hepatoma successfully resected). Even if all 13 of those for whom long-term follow-up was not obtained had died, we would have observed a 58% survival at 1 yr.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Critically ill patients with ALD frequently present to the ICU for evaluation and management of complications related to their disease. The advent of OLT has offered a lifesaving therapy for many of these patients. In 1997, there were 4,168 liver transplantations performed in the United States (19). For the period of our study (1994-1997), the yearly number of liver transplants at our institution ranged from 97 to 132. Transplant organ shortage is a problem well known to the medical community as well as the public at large. Such scarcity necessitates strict regulation of transplant organ dispersal, which is currently performed by the United Network for Organs Sharing (UNOS). Many patients with ALD, because of their underlying comorbidities and/or organ shortages, are deemed ineligible for liver transplantation. When such patients present to the ICU, treatment options are often limited. We sought to identify clinical outcome predictors for this challenging group of patients.

Most patients in the study (both listed and not listed for OLT) had hepatic failure as defined by laboratory criteria. However, laboratory definitions of organ failure may fail to capture all those with ALD, especially those whose chronic liver disease (e.g., alcoholic liver disease) may not manifest elevations in hepatic enzymes and whose INR levels, although abnormal, remain below two. This may explain the higher incidence of hepatic failure in those listed for OLT, as this group had fewer patients with alcoholic liver disease.

The results of our series of 183 patients show a significant increase in mortality for those patients not on the liver transplant list compared with those listed (56% versus 12%). This is not surprising, given that ineligibility for OLT often stems from comorbidities known to be associated with poor prognosis. Indeed, patients deemed ineligible had significantly higher APACHE II scores, predicting a higher mortality. Nevertheless, because nearly half of our patients survived, one of the challenges in their management is distinguishing those with a reasonable chance for survival from those likely to die.

The shorter ICU length of stay in the group ineligible for OLT is likely due to two factors. First, the higher mortality and acuity of illness may result in rapid demise and death in some ineligible patients. Second, those who were listed often remained in the ICU for extended time periods for close monitoring despite a relatively lower severity of illness.

The most common causes of liver failure in our patients were ethanol abuse and viral hepatitis, similar to that reported in previous series (8). Reasons for ICU admission likewise matched those of other studies (7, 9). Therefore, we believe this group is representative of the patient population at most referral centers.

Clinical conditions such as encephalopathy, GI bleeding, HD, MV, and multisystem organ failure have been found associated with increased mortality in patients with liver disease by others (8, 20). In our study, encephalopathy was not an independent predictor of mortality, though this may have been in part due to our definition. In some cases, we were unable to specify a patient's grade of encephalopathy because of imprecise descriptions in the daily progress notes. Thus, we chose to broadly define this category and likely included some milder cases of encephalopathy. Had we been able to more precisely characterize the degree of encephalopathy, only the more severe cases (grades III and IV) could have been included. Our more comprehensive approach, with its likely inclusion of some milder cases of encephalopathy, may in part explain this end point not reaching significance. Singh and coworkers (23) also found no mortality difference when comparing alert versus encephalopathic (any grade) patients.

Acute renal failure requiring HD is a strong predictor of mortality in patients with hepatic dysfunction (7, 8). However, when APACHE II scores were incorporated into the logistic regression model, acute renal failure requiring HD was not an independent predictor of mortality. Since creatinine levels are heavily weighted in the APACHE II scoring system, the presence of acute renal failure did not add significantly to the prognostication achieved by the APACHE II score. Overall mortality in those with acute renal failure requiring HD was 86% and all those requiring HD who also had GI bleeding died. Many patients develop renal dysfunction in conjunction with worsening hepatic function, which may progress to the hepatorenal syndrome. In this series of patients, there was no difference in mortality between renal failure secondary to hepatorenal syndrome and that caused by acute tubular necrosis. Clearly, the presence of acute renal failure requiring HD is an ominous prognosticator in this group of patients.

The APACHE II score was a strong predictor of mortality in this series of patients. This scoring system is among the most extensively studied and validated for predicting outcomes in critically ill patients (14). Our study validated the utility of APACHE II as a severity of illness scoring system for patients with ALD deemed ineligible for OLT. Additionally, we sought other outcome predictors for this cohort of patients.

Sepsis is an ominous but common finding in those with ALD. It may complicate and/or cause many of the conditions that contraindicate liver transplantation (e.g., active alcohol consumption, moribund state, malignancy). Even after controlling for severity of illness with APACHE II scores, the presence of sepsis was independently associated with increased mortality in patients denied consideration for OLT. Numerous human and animal studies have demonstrated a systemic inflammatory state with cytokine overactivity in the setting of acute liver dysfunction. Such a cytokine cascade appears to be instrumental to the development of multisystem organ failure in sepsis and liver failure (24).

The need for intubation and MV was also a strong predictor of mortality, as others have previously shown (7, 8, 20). Shellman and coworkers (7) reported a 91% mortality among mechanically ventilated patients with cirrhosis and chronic liver disease, with indications for intubation including hypoxia (type I), airway protection, respiratory arrest and coma (type II), and cardiac arrest (type IV). They did not report any patients with type III respiratory failure. Our mortality for patients with respiratory failure was likewise very high, except for those with type III respiratory failure. The presence of respiratory failure requiring endotracheal intubation is often a precursor to multisystem organ failure in these patients.

Organ dysfunction strongly impacted outcome, with mortality rates increasing progressively with each organ failure accrued. However, MOSF scores were not independent predictors of mortality and did not add to the prognostication achieved through APACHE II scoring. The overlap between APACHE II scores and MOSF scores likely explains the lack of additional prognostic value achieved from MOSF scores. Our findings lend further credence to previous observations that overall severity of illness impacts outcome in patients with liver disease more than the mere presence or severity of the liver dysfunction (27, 28).

Those patients presenting with gastrointestinal bleeding had surprising results, with GI bleeding being an independent predictor for decreased mortality relative to those ineligible for OLT who did not sustain GI bleeding. The 44% mortality rate for those with GI bleeding is high, yet similar to that noted by others (7, 19, 29); nevertheless, in our cohort of patients ineligible for OLT, the mortality odds ratio associated with GI bleeding was less than one, even when APACHE II scores were factored into the logistic regression model. This suggests that the decreased mortality in those with GI bleeding compared with those without GI bleeding was not simply explained by a lesser severity of illness. Gastrointestinal bleeding is unique compared with the other variables found associated with increased mortality, as this condition may be stabilized more readily using such measures as crystalloid and blood product administration, sclerotherapy, and TIPS. In these patients, the only variable associated with a decrease in mortality was the administration of fewer blood products. This is probably a marker for those whose bleeding was more readily controlled. Our data suggest that those deemed ineligible for OLT presenting with GI bleeding indeed benefit from aggressive resuscitative efforts. To our knowledge, this is the first time GI bleeding has been reported to be independently predictive of decreased mortality in a cohort of critically ill patients with advanced liver disease.

Long-term survival was acceptable for this cohort of patients. It is unlikely that all 13 of those for whom long-term follow-up was missing actually died; nevertheless, even this conservative assessment yields a 58% 1-yr survival (46 of 80) in those who lived to hospital discharge. Interestingly, 6 of these 46 long-term survivors eventually received a liver transplant after their condition precluding listing for OLT had been rectified. The notion that in this group of critically ill patients, most hospital survivors do not achieve long-term survival seems unwarranted, based on these results.

In conclusion, we have presented the first report of outcomes in critically ill patients with ALD denied listing for OLT. We have noted a substantially higher mortality in these patients compared with those who are listed for OLT and have determined several variables independently associated with mortality. Some of these predictors, such as APACHE II scores, sepsis, and respiratory failure requiring MV are known to predict mortality in other groups of patients with liver disease. Conversely, the presence of GI bleeding is an independent predictor of decreased mortality; therefore, in the absence of other comorbidities, these patients are a group for whom aggressive resuscitative measures are merited.