key: cord-0010730-tciwtxud
authors: Singh, Nina; Gayowski, Timothy; Wagener, Marilyn M.; Marino, Ignazio R.
title: Outcome of patients with cirrhosis requiring intensive care unit support: Prospective assessment of predictors of mortality
date: 1998
journal: J Gastroenterol
DOI: 10.1007/s005350050047
sha: e64c32b048827fcb17f87861d5005c2416cc262c
doc_id: 10730
cord_uid: tciwtxud

Determinants of outcome and the utility of the Child-Pugh score and the acute physiology and chronic health evaluation (APACHE) II score as predictors of outcome were prospectively assessed in 54 consecutive patients with cirrhosis requiring intensive care unit (ICU) management. Overall mortality in the ICU was 43% (23/54). Child-Pugh scores did not differ between survivors or nonsurvivors (12.8 versus 12.3, P = 0.26), however APACHE II scores (P = 0.007), acute physiology scores (P = 0.006), and Karnofsky scores (P = 0.001) were significant predictors of outcome. By univariate analysis, requirement of mechanical ventilation analysis (P = 0.001), duration of mechanical ventilation (P = 0.001), pulmonary infiltrates (P = 0.0001), infections (P = 0.047), gastrointestinal bleeding (P = 0.005), and serum creatinine ≥1.5 mg/dl (P = 0.0005) were significantly associated with mortality. By logistic regression analysis only pulmonary infiltrates (P = 0.0001) and renal dysfunction (P = 0.041) were independent predictors of mortality. When controlled for the severity of illness (APACHE II scores), the mortality in patients with cirrhosis caused by alcohol was significantly lower than that in patients with liver disease not caused by alcohol (P = 0.01). Our study not only identified predictors of poor outcome in patients with cirrhosis requiring ICU care but also provided data that may have implications for optimal timing for transplantation.

Intensive care unit (ICU) care is an integral part of the management of patients with complications of cirrhosis. A number of medical complications, e.g., hepatic encephalopathy, coagulopathy predisposing to hemorrhage, variceal bleeding, pulmonary edema, infections, or cardiac complications, may prompt admission to the ICU in patients with endstage liver disease. These critically ill patients also comprise a significant proportion of those who eventually undergo liver transplantation.

Factors that adversely influence survival in critically ill patients have been assessed in a number of medical conditions, including hematologic malignancies, 1 bone marrow and organ transplantation, 2-4 and respiratory diseases. 5 Child-Pugh scores were initially developed for the assessment of hepatocellular functional reserve in patients undergoing surgical treatment for variceal bleeding, and have since been widely used for the assessment of the severity of illness in patients with endstage liver disease. 6, 7 Child-Pugh scores predicted outcome in a retrospective study of patients with cirrhosis and chronic liver disease admitted to the ICU. 8 However, these criteria have not been uniformly shown to be optimal predictors of survival in patients with cirrhosis and were of limited value for discriminating the highrisk patients for transplantation. 9,10 Acute physiology and chronic health evaluation (APACHE) scores were determined to be of prognostic significance in a study in ICU patients with cirrhosis; however, neither Child-Pugh scores nor etiology of liver disease were assessed in that study. 11 No study, to our knowledge, has prospectively evaluated Child-Pugh scores and APACHE scores concurrently as predictors of outcome in critically ill patients with cirrhosis.

Pervious studies on ICU outcome in patients with endstage liver disease were conducted prior to the routine testing of hepatitis C virus (HCV). 8, [11] [12] [13] HCV has emerged as a significant cause of cirrhosis and is one of the leading indications for liver transplantation; an estimated 3.5 million people in the United States have chronic HCV infection and each year 8000-10 000 chronically infected patients with HCV die of liverrelated complications.

We prospectively followed consecutive patients with cirrhosis requiring ICU care at our institution over a 2year period. The primary objectives of this study were: to determine the outcome and the prognostic factors associated with poor outcome in patients with cirrhosis admitted to the ICU and to determine whether standardized severity-of-illness scoring systems, e.g., Child-Pugh score and APACHE II scores, predicted the outcome in these patients.

The study samples comprised patients with cirrhosis referred to the liver transplant service of the Pittsburgh Veterans Affairs Medical Center between January, 1993 and December, 1994. Consecutive patients requiring ICU admission at any time between their referral and transplantation or death were considered as an ICU admission. For the purpose of this study, each ICU admission was considered a separate patient, as reported previously. 14 All ICU admissions were followed until discharge from the ICU or until death.

The following variables were recorded for each patient on admission to the ICU: age, sex, etiology of liver disease, data of hospital admission, data and reason for ICU admission, presence of co-morbid illness (diabetes, hypertension, heart disease, malignancy, renal failure), and Child-Pugh score. Admission vital signs, Karnofsky score, and laboratory data (complete blood count, BUN, creatinine, total bilirubin, alkaline phosphatase, alanine aminotransferase, gamma glutamyl transpeptidase, prothrombin partial thromboplastin times, serum cholesterol, albumin, and electrolytes) were also recorded. The APACHE scoring system is the most widely employed severity-of-illness scale in the ICU studies. 15 The APACHE II score was calculated at the time of the ICU admission by adding the acute physiology score (APS), chronic health points, and age points. 15 It has also been shown that preadmission functional status of the patient is an important determinant of the outcome in critically ill patients. 15, 16 We employed the Karnofsky performance score to assess the functional status of the study patients. 17 During the ICU stay, the following variables were recorded: invasive procedures (requirement for pulmonary artery catheter, arterial line, tracheostomy, endoscopy, bronchoscopy); total parenteral or enteral nutrition; inotropic or vasopressor support; requirement for and duration of mechanical ventilation; dialysis; infections; and presence of pulmonary infiltrates and their etiology. The criteria for diagnosing nosocomial pneumonia, 16 adult respiratory distress syndrome, 18 pulmonary edema, 19 and peritonitis 20 were as previously reported. Statistical analysis. Patient demographics and laboratory values were entered into a database (Prophet, BBN Systems and Technologies, Cambridge, MA, USA). Contingency tables were analyzed using the 2 or Fisher exact test. Continuous variables were compared using the t-test or Mann-Whitney test. Multivariate logistic regression analysis was done in stepwise fashion with variables being entered to the equation if the significance level was less than 0.1 and removed from the equation if greater than 0.15. 21 The pool of variables available for analysis was chosen from variables that were significant by univariate analysis and considered to be of clinical importance (factors that could possibly be altered by management in the ICU).

There were 54 ICU admissions in 40 patients with cirrhosis during the study period; 30 patients had 1 admission, 8 patients had 2 admissions, 1 patient had 3 admissions, and 1 patient had 5 admissions to the ICU. The mean age of the patients was 48 years (range, 28-69 years). The underlying liver diseases in these patients are outlined in Table 1 . Comorbid illnesses included diabetes mellitus in 20% (11/54), hypertension in 9% (5/ 54), coronary artery disease in 9% (5/54), and malignancy in 11% (6/54) of the patients. The reasons for ICU admission were hepatic encephalopathy, 52% (28/ 54); gastrointestinal bleeding, 20% (11/54); status postsurgical procedure; 9% (5/54); hypotension/sepsis, 6% (3/54); respiratory distress, 6% (3/54); and miscellaneous, 7% (4/54), i.e., chest pain 1, platelet transfusion reaction 1, seizure 1, and disseminated intravascular coagulation 1.

The duration of ICU stay ranged between 1 and 39 days (mean, 9 days); 39% of the patients had an ICU length of stay greater than 7 days and 22% stayed for 14 days or more. The study patients had been hospitalized a mean of 23 days (range, 0-129 days) before being admitted to the ICU.

Overall, 43% (23/54) of the ICU admissions were associated with death. Mortality was 100% (3/3) in patients Table 2) . Respiratory rate and pulse rate on admission did not differ significantly between survivors and nonsurvivors; however, nonsurvivors were significantly more likely to have an abnormal temperature (P ϭ 0.039) and abnormal blood pressure (P ϭ 0.06) than survivors. Mental status was not a predictor of outcome; coma scores (part of APACHE II scores) and percentage of patients with ; P ϭ 0.035), BUN (54.6 mg/dl versus 9 mg/dl; P ϭ 0.004), and white blood count (13.1 versus 9.06/mm 3 ; P ϭ 0.05) and significantly lower hematocrit (25% versus 28.3%; P ϭ 0.042) compared with survivors. Mortality was significantly higher in patients with renal impairment, i.e., creatinine more than 1.5 mg/ dl (100% versus 58%; P ϭ 0.0005).

Forty-three percent (23/54) of the patients required mechanical ventilation. Requirement for mechanical ventilation and duration of mechanical ventilation were significant predictors of mortality. Seventyfour percent of the nonsurvivors, compared with 29% of the survivors, required mechanical ventilation (P ϭ 0.001). Patients who died had significantly longer ventilatory support (mean, 7 days versus 1 day; P ϭ 0.001).

Twenty percent (11/54) of the patients required dialysis. Patients requiring dialysis had significantly greater severity-of-illness (APACHE II scores, 18.9 versus 15.5; P ϭ 0.011), were more likely to develop pulmonary infiltrates in the ICU (91% versus 30%; P ϭ 0.0004), and were more likely to have had hypotension at any time during the ICU stay (64%, 7/11 versus 12%, 5/43; P ϭ 0.001).

Pulmonary infiltrates developed in 43% (23/54) of the patients. The etiology of pulmonary infiltrates was pulmonary edema in 65% (15/23), acute respiratory distress syndrome in 13% (3/23), atelectasis in 9% (2/23), pneumonia in 9% (2/23), and pulmonary fibrosis in 4% (1/23). Patients developing pulmonary infiltrates were more severely ill, i.e., had higher APACHE II scores (P ϭ 0.03), had longer ICU length of stay (P ϭ 0.034), and were more likely to have renal impairment, i.e., creatinine more than 1.5 mg/dl (P ϭ 0.032). When controlled for the severity of illness, the patients with pulmonary infiltrates were significantly more likely to die than those without pulmonary infiltrates (P ϭ 0.0002).

Infections occurred in 30% (20/54) of the patients and bacteremia in 20% (11/54) of the patients; 26% (8/31) of the survivors versus 52% (12/23) of the nonsurvivors developed infections (P ϭ 0.047). Spontaneous bacterial peritonitis accounted for 35% (7/20) of all infections in the ICU and was caused by culture-negative neutrocytic ascites in 43% of the episodes, neutrocytic ascites in 29% of the episodes, and bacteriascites in 29% of the episodes. The etiologic agents of bloodstream infections were Staphylococcus aureus (27%, 3/ 11, including one patient with endocarditis), Klebsiella pneumoniae (27%, 3/11), Citrobacter freundii (18%, 2/11), enterococci (18%, 2/11), and Pseudomonas aeruginosa (9%, 1/11). One patient (receiving corticosteroids) had fungemia caused by Candida albicans. Patients experiencing infections were more likely to have had gastrointestinal bleeding (P ϭ 0.035), creatinine more than 1.5 mg/dl (P ϭ 0.07), and to have required dialysis (P ϭ 0.07). Mortality was higher in infected (60%, 12/20) versus noninfected (32%, 11/34) patients (P ϭ 0.047).

When pulmonary infiltrates, infections, renal impairment, and gastrointestinal bleeding were entered into a logistic regression model, only pulmonary infiltrates (P ϭ 0.0001) and renal dysfunction (P ϭ 0.041) were independent predictors of mortality. Mortality was 86% when all the above four variables were present and 0% when none of the above was present. Mortality rate was 91% when pulmonary infiltrates and gastrointestinal bleeding were present and 86% when pulmonary infiltrates and renal dysfunction existed.

When controlled for the APACHE II scores (APACHE II more than 15), 35% (7/20) patients with alcoholic cirrhosis versus 83% (10/12) of the patients with cirrhosis not caused by alcohol died (P ϭ 0.01). Eighty-three percent of the patients without alcoholic cirrhosis had viral hepatitis (HCV 75% and HBV 8%). Serum creatinine (3.3 versus 2.4 mg/dl), bilirubin (5.3 versus 8.4 mg/dl), and frequency of gastrointestinal bleeding (30% versus 50%) were not different between patients with cirrhosis caused by alcohol versus nonalcoholic cirrhosis; however the patients with nonalcoholic cirrhosis were more likely to have required mechanical ventilation (83% versus 35%, P ϭ 0.008) and to have pulmonary infiltrates (92% versus 25%, P ϭ 0.0005).

Assessment of prognosis has become increasingly important in the medical management of critically ill patients. Such data can serve to elucidate factors that portend an unfavorable outcome, but more importantly, identify those variables that may be potentially modifiable or amenable to intervention. In critically ill patients with cirrhosis, such data also have implications regarding optimal timing of transplantation. Mortality in these critically ill patients in our study was 43%. Child-Pugh scores failed to predict the outcome, whereas APACHE II scores, APS scores, and functional status of the patient (Karnofsky score) were more accurate predictors of outcome. The lack of predictive accuracy of Child-Pugh scores is probably because extrahepatic severity-of-illness variables, e.g., renal function, vital signs, and pulmonary status, are not measured by Child-Pugh scores, although these scores do assess a number of different criteria related to the liver disease.

Of routinely available liver functions tests assessed, only bilirubin and prothrombin time, and not aminotransferases, alkaline phosphatase, albumin, or cholesterol were significant predictors of outcome. The prognostic significance of hyperbilirubinemia and elevated prothrombin time has also been recognized in patients with other medical conditions requiring ICU care. In a study assessing outcome in ICU patients treated with hemodialysis for acute renal failure, serum bilirubin and prothrombin time were significantly associated with mortality; hyperbilirubinemia was an independent predictor of mortality. 22 Bilirubin was also a significant predictor of outcome in transplant recipients requiring ICU care 4 and of early graft failure in liver transplant candidates undergoing transplantation. 23 We observed a high incidence of renal dysfunction in our patients with cirrhosis admitted to the ICU. Renal impairment (creatinine more than 1.5mg/dl) was present in 76% of the patients, and 20% of the patients required dialysis. Patients requiring dialysis were significantly more ill (as assessed by APACHE II scores). However, when controlled for the severity of illness, renal impairment (P ϭ 0.014) and dialysis (P ϭ 0.017) remained significantly associated with mortality in the ICU. The grave prognostic association of renal failure in endstage liver disease was also emphasized in a recent Spanish study. Renal impairment after spontaneous bacterial peritonitis in cirrhosis was reported to be the most significant predictor of hospital mortality. 24 These data have important implications regarding timing of transplantation. Patients with endstage liver disease and renal impairment are at high risk for mortality. However, transplantation should ideally be undertaken before renal failure and the requirement for dialysis ensue. Pretransplant renal dysfunction was an independent predictor of early graft failure after liver transplantation 25, 26 and pretransplant dialysis has been identified as a significant risk factor for early posttransplant mortality and infectious morbidity. [27] [28] [29] [30] Nearly one-third of our patients with cirrhosis in the ICU experienced infections. Spontaneous bacterial peritonitis (SBP) and bacteremia have been reported in 7%-15% and 4%-10% of patients with cirrhosis, respectively. 31 Our incidence of SBP (35%) and bacteremia (20%) was therefore somewhat higher than that reported in cirrhotic patients in general. This was likely reflective of the fact that our patients represented a subgroup of most critically ill patients with liver disease. Although mortality was higher in the infected patients, infections were not an independent predictor of mortality. Escherichia coli, Klebsiella, and enterococcus were the most frequent causes of culture-positive SBP or spontaneous bacteremia; however, culturenegative neutrocytic ascites was the most frequently observed variant of SBP. Patients with gastrointestinal bleeding, renal impairment, and those requiring dialysis were at significantly higher risk for infections in the ICU. Gastrointestinal hemorrhage has been shown to impair the reticuloendothelial function and facilitate bacterial translocation across the gut. 32, 33 Likewise, the association of renal failure with infections is well recognized.

In 20% (11/54) of our patients, a unique presentation characterized by marked leukocytosis (in the absence of any documented infection) with eventual progression to multiple organ system failure, was observed. The white blood cell count of these patients ranged from 16.3 to 25.6/mm 3 (mean, 18.9/mm 3 ). The mortality in these patients (81%) was comparable to that of patients with documented infections in the ICU (60%). We hypothesize that this clinical presentation may likely be caused by enhanced production and decreased metabolism of cytokines, e.g., tumor necrosis factor-alpha and interleukin (1L)-6, which have been demonstrated in patients with endstage liver disease. 34 A systemic inflammatory response syndrome in patients with cirrhosis of the liver, in the absence of infections, has recently been described, with leukocyte activation in the peripheral blood of patients with cirrhosis and systemic inflammatory response syndrome correlated with the serum levels of 1L-6. 35 Mental status has been shown to be a significant predictor of outcome in ICU patients in a number of clinical settings, e.g., patients with acute renal failure requiring dialysis and transplant recipients. 4, 36 Although encephalopathy was the predominant reason for ICU admission in our patients, no association between the presence or absence of encephalopathy, the degree of encephalopathy (stage I, II, or III) or coma scores (APACHE) was demonstrated in our study. These findings may likely reflect the reversible nature of hepatic encephalopathy with appropriate management in patients with liver disease.

An intriguing finding in our study was that, when controlled for the severity of illness (APACHE II scores), the mortality in patients with cirrhosis caused by alcohol was significantly lower than that in patients with liver disease not caused by alcohol (35% versus 83%; P ϭ 0.01). Although Child-Pugh scores, serum bilirubin, creatinine, or the incidence of gastrointestinal bleeding was not different, the patients with liver disease not caused by alcohol may have been more debilitated, as indicated by a significantly greater requirement for mechanical ventilation and a higher incidence of pulmonary infiltrates. Nevertheless, the association of underlying liver disease with outcome has also been noted in transplant recipients. Higher graft and patient survival was observed in patients undergoing liver transplantation for cholestatic and alcoholic liver disease compared with other liver diseases, including hepatitis. 23 The vast majority of our patients with nonalcoholic liver disease had viral hepatitis caused by hepatitis C virus. It has been proposed that hepatitis C virus is an immunodulatory virus. 37 Greater infectious morbidity in patients with hepatitis C virus compared with patients without hepatitis C virus infection has been documented in transplant recipients. 38 In the nontransplant setting, T-cell-derived cytokine levels were significantly higher in patients with hepatitis C virus compared with normal controls. 39 Whether differences in cytokine profile, level, or expression were contributory variables to differential disease severity in patients with and without viral hepatitis in our study remains speculative.

Despite the high mortality in critically ill patients with cirrhosis, ICU support is not entirely futile in these patients. Of patients in this study who underwent liver transplantation, the 1-year survival was 91% (10/11), the mean Karnofsky score of these transplant recipients at 1 year being 96.

In conclusion, among critically ill patients with cirrhosis, the mortality in the ICU was 43%. Child-Pugh scores, compared to APACHE III scores, were less than optimal predictors of the outcome. Pulmonary infiltrates and renal impairment were identified as independently significant medical complications that predicted mortality in our patients. Future studies validating our findings are warranted, since these data have implications not only for prognosis but for the selection of candidates and timing of transplantation.

Intensive therapy for life threatening medical complications of hematological malignancy

Outcome of recipients of bone marrow transplant who require intensive-care unit support

Mortality during intensive care after orthotopic liver transplantation

Intensive care unit management in liver transplant recipients: Beneficial effect on survival and preservation of quality of life

Mortality in intensive care patients with respiratory disease; is age important?

In: Child CG (ed) The liver and portal hypertension

Transection of the oesophagus for bleeding oesophageal varices

Prog-nosis of patients with cirrhosis and chronic liver disease admitted to the medical intensive care unit

Child-Pugh score and liver transplantation

Prognostic value to Child-Turcotte criteria in medically treated cirrhosis

Intensive care unit admissions with cirrhosis: Risk-stratifying patient groups and predicting individual survival

Prediction of mortality in cirrhosis of the liver

Survival and prognostic indicators in compensated and decompensated cirrhosis

Prognostic assessment of acute complications of bone marrow transplantation requiring intensive therapy

APACHEacute physiology and chronic health evaluation: A physiologically based classification system

Ventilator-associated pneumonia, a multivariate analysis

The clinical evaluation of chemotherapeutic agents in cancer

The American-European consensus conference on ARDS: Definitions, mechanisms, outcomes, and clinical trial coordination

Causes of fever and pulmonary densities in patients with clinical manifestations of ventilator-associated pneumonia

Culture-negative neutrocytic ascites: A variant of spontaneous bacterial peritonitis

Estimation of the multivariate logistic risk function: A comparison of the discriminant function and maximum likelihood approaches

Risk factors influencing survival in acute renal failure treated by hemodialysisl

Assessing risk in liver transplantation. Special reference to the significance of a positive cytotoxic crossmatch

Renal impairment after spontaneous bacterial peritonitis in cirrhosis: Incidence, clinical course, predictive factors, and prognosis

The preoperative assessment of risk in liver transplantation. A multivariate analysis in 2376 cases of the UW era

Risk factors and predictive indexes of early graft failure in liver transplantation

Risk factors associated with mortality and infectious morbidity after liver transplantation

Tumor necrosis factor alpha and interleukin 6 plasma levels in infected cirrhotic patients

Leukocyte activation in the peripheral blood of patients with cirrhosis of the liver and SIRS, correlation with serum interleukin-6 levels and organ dysfunction

Acute renal failure treated by hemofiltration: Factors affecting outcome

Hepatitis C virus and organ transplantation

Increased infections in liver transplant recipients with recurrent hepatitis C virus hepatitis

Immunoregulatory cytokines in chronic hepatitis C virus infection: Pre and posttreatment with interferon alpha

Perioperative risk factor assessment in liver transplantation

Prognostic value to preoperatively obtained clinical and laboratory data in predicting survival following ortotopic liver transplantation

Infectious complications in liver recipients on tacrolimus: Prospective analysis of 88 consecutive liver transplants

Spontaneous bacterial peritonitis

Sequential changes in reticuloendothelial system after acute hemorrhage

Hepatic reticuloendothelial protection against bacteremia in experimental hemorrhagic shock