SEARCH

SEARCH BY CITATION

Keywords:

  • APACHE;
  • cirrhosis;
  • intensive care unit;
  • MELD;
  • renal failure;
  • RIFLE;
  • SOFA

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Background and Aim:  To evaluate the association of the Risk, Injury, Failure, Loss and End-stage renal failure (RIFLE) score on mortality in patients with decompensated cirrhosis admitted to intensive care unit (ICU).

Methods:  A cohort of 412 patients with cirrhosis consecutively admitted to ICU was classified according to the RIFLE score. Multivariable logistic regression analysis was used to evaluate the factors associated with mortality. Liver-specific, Acute Physiology and Chronic Health Evaluation (APACHE) II, Sequential Organ Failure Assessment (SOFA) and RIFLE scores on admission, were compared by receiver–operator characteristic curves.

Results:  The overall mortality during ICU stay or within 6 weeks after discharge from ICU was 61.2%, but decreased over time (76% during first interval, 1989–1992 vs 50% during the last, 2005–2006, P < 0.001). Multivariate analysis showed that RIFLE score (odds ratio: 2.1, P < 0.001) was an independent factor significantly associated with mortality. Although SOFA had the best discrimination (area under receiver–operator characteristic curve = 0.84), and the APACHE II had the best calibration, the RIFLE score had the best sensitivity (90%) to predict death in patients during follow up.

Conclusions:  RIFLE score was significantly associated with mortality, confirming the importance of renal failure in this large cohort of patients with cirrhosis admitted to ICU, but it is less useful than other scores.


Abbreviations
APACHE

Acute Physiology and Chronic Health Evaluation

ARF

acute renal failure

AUC

area under the ROC curve

BIL

bilirubin

Cr

creatinine

CTP

Child–Turcotte–Pugh

FiO2

inspiratory concentration of oxygen

GFR

glomerular filtration rate

GI

gastrointestinal

HRS

hepatorenal syndrome

ICU

intensive care unit

MELD

Model for End-stage Liver Disease

RIFLE

Risk, Injury, Failure, Loss and End-stage renal failure

ROC

receiver–operator characteristic

SBP

spontaneous bacterial peritonitis

SOFA

Sequential Organ Failure Assessment.

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Development of renal dysfunction is associated with very poor prognosis in patients with decompensated cirrhosis,1,2 and renal failure is frequently associated with the development of severe complications of cirrhosis, such as spontaneous bacterial peritonitis (SBP) and hepatorenal syndrome (HRS),3–5 many of which are common reasons for admission to intensive care unit (ICU).

Although acute renal failure (ARF) is a well-known risk factor of mortality in patients with cirrhosis,6 there is a lack of agreement in the definitions used to classify ARF. This leads to different management protocols and makes comparative studies difficult. Some definitions are based on serum creatinine (Cr), as several studies have shown the impact of Cr on outcome. For example, patients with cirrhosis admitted to the intensive care units with Cr > 1.5 mg/dL (> 132.6 µmol/L) had a poorer prognosis compared to those with Cr within the normal range.7 Similarly, we showed that ARF (defined as Cr ≥ 3.5 mg/dL [≥ 309.4 µmol/L], oliguria, and/or presence of HRS8 or need for hemofiltration) was also significantly associated with poor outcome in critically ill patients with cirrhosis.9 However, it is known that Cr is not an accurate index of glomerular filtration rate (GFR) (i.e. renal function), particularly in patients with decompensated cirrhosis and/or acute illness.10

Recently, the Acute Dialysis Quality Initiative (ADQI) Working Group, comprised of experts in the fields of nephrology and critical care, published the RIFLE classification, a new consensus and evidence-based definition for ARF.11 The ability of the RIFLE score to predict mortality has been shown in a wide range of critically ill patients with ARF.12,13 However, only two studies14,15 have evaluated the prognostic impact of the RIFLE score in critically ill cirrhotic patients. In both reports14,15 the prognostic ability of the RIFLE score was also compared with ICU prognostic models (i.e. the Acute Physiology and Chronic Health Evaluation [APACHE] and Sequential Organ Failure Assessment [SOFA]) and liver specific scores (i.e. Child–Turcotte–Pugh [CTP] and Model for End-stage Liver Disease [MELD]). Du Cheyron et al.15 found that in 186 patients with cirrhosis, the RIFLE score together with sepsis and alcohol as etiology of cirrhosis were significantly associated with the ICU mortality. In the more recent study by Jenq et al.,14 RIFLE and SOFA scores were the independent factors of hospital mortality in 134 critically ill cirrhotic patients.

The aim of the present study was to validate the prognostic impact of ARF based on the RIFLE score on outcome in a large consecutive cohort of patients with cirrhosis admitted to ICU, as well as to assess its performance in comparison with well-established prognostic scores.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Consecutive cirrhotic patients with their first admission to the Royal Free Hospital (RFH) ICU, between January 1989 and December 2006 were evaluated retrospectively using a prospectively collected data-base and ICU chart, with a follow up that included up to 6 weeks after discharge from ICU. Pediatric patients, patients with end-stage renal disease under renal replacement therapy or those who had undergone liver transplantation (LT) or transplanted during the evaluable follow-up period were excluded. Non-survivors were defined as those who died during the follow-up period. The period of 6 weeks was chosen so as to include those who died on the general wards following their discharge from ICU and to encompass those patients in whom a decision had been made not to resuscitate further, and thus returned to the wards for terminal care. At admission, 40 variables, including demographic and clinical data including the year of admission (also grouped by intervals of 3 years), as well as laboratory parameters (including full blood count, lactate, biochemical and clotting profile, arterial blood gas and inspiratory concentration of oxygen [FiO2]) had been prospectively collected for each patient. During the ICU stay, gastrointestinal (GI) bleeding episodes, development of aspiration pneumonia and additional use of inotropes, mechanical ventilation or hemofiltration were also recorded. All these variables have been described in detail elsewhere.16

In the present study we used the RIFLE criteria at admission to ICU in order to classify our patients as those without ARF and as those with ARF (see Table 1). The latter patients were classified in three stages of severity: Risk, Injury, and Failure. The RIFLE score also classifies the patients into two clinical outcome categories: Loss and End-stage renal disease. However, no patient met the criteria for loss of renal function while patients with end-stage renal disease had been excluded from the study. Similarly to the previous study,14 the remaining four stages of RIFLE classification (Non-ARF, Risk, Injury and Failure) were scored from 0 to 3 points, respectively (Table 1). The prognostic performance of the RIFLE score was compared to the liver-specific prognostic models (CTP and MELD scores) as well as the general-ICU models (APACHE II and SOFA scores). All scores were evaluated on ICU admission, but the RIFLE score was also evaluated at 48 h after admission to ICU in the patients with available data on renal function at 48 h (e.g. in patients who survived or were not discharged within 48 h). The SOFA score was used for grading of organ dysfunction or failing organ systems (liver, kidney, hematological, respiratory, cardiovascular and neurological). All the above scoring systems were calculated as previously published.17–19

Table 1.  RIFLE classification for acute renal failure
 GFR criteriaUO criteriaRIFLE score
  1. ARF, acute renal failure; Cr, creatinine; GFR, glomerular filtration rate; RIFLE, Risk, Injury, Failure, Loss and End-stage renal failure;UO, urine output.

Non-ARFGFR decrease ≤ 25%UO ≥ 0.5 mL/kg/h0
RiskIncrease Crx1.5 or GFR decrease > 25%UO < 0.5 mL/kg/h × 6 h1
InjuryIncrease Crx2 or GFR decrease > 50%UO < 0.5 mL/kg/h × 12 h2
FailureIncrease Crx3 or GFR decrease > 75% or Cr > 4 mg/dLUO < 0.3 mL/kg/h × 24 h3
LossComplete loss of kidney function > 4 week
End-stage renal diseaseEnd-stage renal disease (> 3 months)

Statistical methods

All data were analyzed using the statistical package SPSS (version 13.0, SPSS Inc., Chicago, IL). χ2-test was used for categorical variables and the Mann–Whitney U-test was used for quantitative variables in order to compare the patients who died to those who survived during the follow-up period. Variables with normal distribution were expressed as mean ± standard deviation and in variables with abnormal distribution as median (range). Significance testing was two-sided and set to a P-value of less than 0.05. Univariate comparisons and then multivariate analysis (backward stepwise logistic regression analysis) of variables (with P < 0.05) at baseline were performed to identify independent factors associated with mortality. All scores were compared by receiver–operator characteristic (ROC) curves and the Lemeshow–Hosmer test. Thus, the discriminative ability of each prognostic score at baseline was evaluated by using the area under a ROC curve (AUC), for which values above 0.8 or more are considered clinically useful.20 To test the calibration (i.e. the degree of correspondence between predicted and observed mortality) we used goodness-of-fit testing to evaluate the calibration using the Lemeshow–Hosmer test (a P-value in this test close to 1.0 is considered good calibration).21

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Characteristics of patients

There were 412 patients with cirrhosis who were consecutively admitted during the study period: 244 (59.2%) were male and the mean age was 49.3 ± 12 years (Table 2). Alcohol was the main cause of cirrhosis in our cohort (69.4%), while viruses (HCV or HBV) were the second most frequent cause (17%). In the cohort of 412 patients, 10 had already been placed on a liver transplantation waiting list. Their characteristics were not significantly different from those who were not on a waiting list at the time of admission to ICU. By definition, no patient underwent liver transplantation within 6 weeks of ICU discharge, but 15 did so between 20 and 96 weeks. The remaining characteristics of the patients are shown in Table 1. Notably, 276 (67%) cirrhotics were transferred from RFH general wards and the clinical indications for admission to the ICU were: 165 (40%) patients for respiratory failure, 69 (17%) for gastrointestinal bleeding, 58 (14%) patients for renal dysfunction, and 26 (6.3%) for sepsis. The definitions regarding the indications for ICU admission have been described in detailelsewhere.16 The median CTP, APACHE II, MELD and SOFA scores on admission to ICU were 11 (5–15), 18 (0–45), 23 (6–40) and 11 (0–21), respectively. Finally, according to RIFLE criteria, 207 (51%) cirrhotics were classified as non-ARF (RIFLE score = 0), 91 (22%) were at risk (RIFLE score = 1), 37 (9%) had renal injury (RIFLE score = 2) and 77 (19%) had renal failure (RIFLE score = 3).

Table 2.  Clinical and laboratory characteristics of patients with cirrhosis on admission to ICU or during ICU stay
Variable (unit)All (n = 412)
  1. ICU, intensive care unit; RIFLE, Risk, Injury, Failure, Loss and End-stage renal failure; RFH, Royal Free Hospital.

Age (years)49.3 ± 12
Sex, n (%) 
 Male244 (59.2)
Transfer from RFH, n (%)276 (67)
Days in the ICU2 (1–120)
Indication for ICU admission, n (%) 
 Renal failure58 (14)
 Respiratory failure165 (40)
 Sepsis26 (6)
 Gastrointestinal bleeding69 (17)
 Multiple organ failure16 (4)
 Liver failure10 (2)
 Others68 (17)
During ICU stay, n (%) 
 Gastrointestinal bleeding86 (21)
 Mechanical ventilation363 (88)
 Inotropic support331 (81)
 Hemofiltration59 (14.4)
RIFLE score, n (%) 
 0207 (50.2)
 191 (22.1)
 237 (9)
 377 (18.7)
Mortality, n (%)252 (61.2)
Comparison of cirrhotics with mild ARF versus those with severe ARF

The 128 cirrhotics with mild ARF (i.e. those with RIFLE score 1 or 2), compared to 77 with severe ARF (i.e. RIFLE score = 3), were older (51.8 ± 11 vs 46.7 ± 11, P = 0.002), were more frequently men (68% vs 49%, P = 0.008), more frequently had variceal bleeding on admission to the ICU (64% vs 35%, P < 0.001), but they required cardiovascular support less often (45% vs 69%, P = 0.003). Finally, they had significantly higher pH (7.34 vs 7.23) and lower bilirubin (BIL) (65 µmol/L vs 104 µmol/L), MELD (28 vs 33) and SOFA (11.8 vs 13.6) scores on admission (P < 0.001) (Table 3).

Table 3.  Characteristics of patients with mild or severe ARF on admission ICU
Variable (unit)With mild ARF (RIFLE = 1 or 2) (n = 128, 31.1%)With severe ARF (RIFLE = 3) (n = 77, 18.7%)P
  1. APACHE II: Acute Physiology and Chronic Health Evaluation; ARF, acute renal failure; CTP, Child–Turcotte–Pugh; ICU, intensive care unit; MELD, Model for End-stage Liver Disease; RIFLE, Risk of renal failure, Injury to the kidney, Failure of kidney function, Loss of kidney function, and End-stage renal disease; SOFA, Sequential Organ Failure Assessment.

Age (years)51.8 ± 1146.7 ± 110.002
Sex, n (%)   
 Male87 (68)38 (49)0.008
Days in the ICU6 (1–37)8 (1–48)0.45
Cause of cirrhosis  0.36
 Alcohol8947
 Non-alcoholic3930
On admission to ICU, n (%)   
 Gastrointestinal bleeding58 (64)23 (35)< 0.001
 Inotropic support47 (45)53 (69)0.003
 Mechanical ventilation80 (88)59 (77)0.04
During ICU stay, n (%)   
 Gastrointestinal bleeding24 (18)13 (16)0.75
 Mechanical ventilation111 (86)63 (81.5)0.38
 Inotropic support106 (83)70 (91)0.18
Bilirubin (µmol/L)65 (5–922)104 (2–1058)< 0.001
Prothrombin time (sec)29 (14–120)31 (14–124)0.31
Lactate (mmol/L)2.7 (0.3–19.4)1.9 (0.1–21)0.44
pH7.34 ± 0.17.23 ± 0.1< 0.001
MELD28.6 ± 633 ± 7< 0.001
SOFA11.8 ± 313.6 ± 3< 0.001
APACHE II22 ± 823 ± 70.63
CTP11.7 ± 211.8 ± 20.78
Mortality97 (75)67 (87)0.05

Mortality

The overall mortality in ICU or 6 weeks after discharge from ICU was 61.2% (Table 2). Causes of death were: multiple organ failure in 45.4%, respiratory failure in 18.6%, renal failure in 11.9%, uncontrolled variceal bleeding in 8.8%, liver failure in 6.2% and other in 9.1%. The median length of stay in ICU was not different between survivors (2 [1–34] days) and non-survivors (two [1–120] days, P = 0.89) (Table 4). Furthermore, during the five time periods (1989–1992 [n = 46], 1993–96 [n = 105], 1997–2000 [n = 91], 2001–2004 [n = 70] and 2005–2006 [n = 100]) there was a progressive and significant decrease in mortality: 76% for the first period (1989–1992), compared to 53% and 50% for the 4th (2001–2004) and last (2005–2006) periods, respectively (P < 0.001) (Table 4). The clinical variables at admission or during the ICU stay that were significantly associated with a worse outcome are shown in Table 4. Interestingly, cirrhotics who died were less likely to have GI bleeding at admission to ICU compared to survivors (P < 0.001), and more frequently had severe encephalopathy (P = 0.002), ascites (P < 0.001) and need for cardiovascular support with inotropes at admission and during the ICU stay (P < 0.001). In addition, the cirrhotics who died, compared to those who survived had significantly higher median values of CTP (12 vs 10), APACHE II (21 vs 16), MELD (28 vs 18) and SOFA (12 vs 9) scores (P < 0.001) at admission.

Table 4.  Clinical and laboratory characteristics of survivor versus non-survivor patients with cirrhosis on admission to intensive care unit
Variable (unit)Survivors (n = 160, 38.8%)Non-survivors (n = 252, 61.2%)P
  1. aPTT, activated partial thromboplastin time; ICU, intensive care unit; INR, international normalized ratio; PLT, platelet count; RIFLE: Risk of renal failure, Injury to the kidney, Failure of kidney function, Loss of kidney function, and End-stage renal disease; WBC, white blood cell.

Age (years)48 ± 1251 ± 110.16
Sex, male n (%)96 (60)148 (58)0.79
Days in the ICU2 (1–34)2 (1–120)0.89
Admission year, n  0.003
 1989–19921135
 1993–19963075
 1997–20003655
 2001–20043337
 2005–20065050
Cause of cirrhosis, n  0.52
 Alcohol114172
 Non-alcoholic4680
On admission to ICU, n (%)   
 Gastrointestinal bleeding118 (73)137 (54)0.001
 Inotropic support35 (22)142 (56)< 0.001
 Mechanical ventilation138 (86)224 (89)0.47
 Encephalopathy (grade 3,4)31 (20)95 (38)0.002
 Severe ascites21 (13)83 (33)< 0.001
During ICU stay, n (%)   
 Gastrointestinal bleeding19 (12)67 (27)0.003
 Mechanical ventilation133 (83)230 (91)0.034
 Inotropic support104 (65)227 (91)< 0.001
On admission to ICU   
 Albumin (g/L)25 ± 7.127 ± 80.03
 WBC (109/L)8.1 (0.8–35)11.5 (2.1–52)0.01
 PLT (109/L)110 (11–586)82 (8–1166)0.07
 Bilirubin (µmol/L)51 (9–523)111 (2–758)< 0.001
 Prothrombin time (sec)22 (13–48)23 (14–124)< 0.001
 INR1.8 (0/9–4.3)2.5 (1.2–10.2)< 0.001
 aPTT (s)45 (30–180)57 (10–250)< 0.001
 Creatinine (µmol/L)79 (48–728)126 (21–1252)< 0.001
(mg/dL)0.89 (0.5–8.2)1.4 (0.2–14)
 Urea (mmol/L)7.5 (2–45)11.8 (0.8–57)< 0.001
 Na (mmol/L)138 ± 7.7136 ± 110.10
 K (mmol/L)4 ± 0.84.2 ± 10.06
 Lactate (mmol/L)2.4 (0.1–11.8)4.2 (0.1–21)< 0.001
 FiO2 (%)0.52 ± 0.10.62 ± 0.2< 0.001
 pH7.394 ± 0.17.32 ± 0.1< 0.001
 RIFLE score (0/1/2/3), n119/27/4/1088/64/33/67< 0.001

The laboratory variables, which had prognostic value in the univariate analysis were: white blood cell (WBC), albumin, BIL, Cr, Urea, pH, FiO2, lactate, PT, international normalized ratio (INR) and activated partial thromboplastin time (aPTT) (Table 4). The RIFLE score was also significantly associated with mortality (P < 0.001) (Table 4), and there was a progressive increase in mortality associated with increasing RIFLE score: non-ARF (42.5%), Risk (71%), Injury/Failure (88%) on admission to the ICU (P < 0.05) (Fig. 1). In addition, the patients with mild ARF (RIFLE score 1 or 2) had significantly lower mortality rates, compared to the patients with severe ARF (75% vs 87%, P = 0.05). Finally, a higher RIFLE score was significantly associated with higher MELD and SOFA scores (for example the patients with RIFLE score 0, 1, 2 and 3 had median SOFA scores 9 [0–15], 11.5 [3–19], 13 [6–18] and 14 [3–21], respectively [P < 0.001]). The variables that were significantly associated with mortality in the univariate analysis (including the year of admission, urea and creatinine) were included in the multivariate regression analysis. The latter revealed that higher FiO2, serum lactate, BIL and RIFLE score (odds ratio [OR]:2.1) on admission were the variables that were independently associated with mortality during the ICU stay or 6 weeks after discharge from ICU (Table 5). Interestingly, when the RIFLE score was introduced as a categorical variable in the regression analysis, it was found that a RIFLE score of 1 and 2/3 had an OR 2.5 and 7.4, respectively (P = 0.007 and P < 0.001, respectively) as compared to normal renal function. In addition, when we analyzed only the patients with abnormal renal function (RIFLE score = 1, 2 or 3), only bilirubin was an independent risk factor of mortality (P = 0.003), indicating that the type of ARF was less important than renal failure per se.

image

Figure 1. Survival curves in 412 consecutive critically ill patients with cirrhosis admitted to intensive care unit according to the Risk, Injury, Failure, Loss and End-stage renal failure (RIFLE) score (P < 0.001 by log–rank test for cirrhotics with RIFLE score = 0 (n = 207) versus those with RIFLE score = 1 (n = 91) versus RIFLE score = 2 or 3 (n = 114)). inline image APACHEII = 0.77; inline image CTP = 0.73; inline image RIFLE = 0.71; inline image SOFA = 0.84; inline image MELD = 0.81.

Download figure to PowerPoint

Table 5.  Multivariable analysis to identify the independent factors associated with mortality in patients with cirrhosis admitted to the intensive care unit
VariablesStandard errorBeta coefficientsP-valueOdds ratio95% Confidence interval95% Confidence interval
LowerUpper
  1. RIFLE, Risk, Injury, Failure, Loss and End-stage renal failure.

FiO20.87433.8347< 0.00146.28.34256.8
Lactate0.05360.20650.0011.231.11.3
Bilirubin0.00140.0062< 0.0011.0061.0031.09
RIFLE0.16340.7473< 0.0012.111.532.9
Prognostic factors associated with mortality—ROC curves and calibration

Based on the area under the ROC curves, the SOFA had the best discriminative accuracy for mortality (AUC = 0.84), followed by the MELD score (AUC = 0.81). The APACHE II and CTP score had an acceptable discrimination (AUC = 0.77 and 0.73, respectively), while RIFLE score on its own had the worst discrimination value (AUC = 0.71) (Fig. 2). Interestingly, 175 (42.5%) of 412 patients had available data regarding their renal function (serum creatinine and urine output) at 48 h after admission to ICU. In this cohort of patients, the discriminative accuracy of RIFLE score at baseline and 48 h was similar (AUC: 0.74 and 0.75, respectively) and not significantly different compared to the total cohort (AUC: 0.71, P > 0.05). Regarding the goodness-of-fit measured by the Lemeshow-Hosmer, the calibration of APACHE II (χ2 = 7.2, P = 0.51) was superior to SOFA (χ2 = 8.9, P = 0.34), MELD (χ2 = 10.7, P = 0.21) and RIFLE (χ2 = 4.5, P = 0.10). Table 6 shows the ROC curves, goodness-of-fit, sensitivity and specificity of these prognostic scores.

image

Figure 2. Receiver–operator characteristic (ROC) curves for baseline Child–Turcotte–Pugh (CTP), Model for End-stage Liver Disease (MELD), Sequential Organ Failure Assessment (SOFA), Acute Physiology and Chronic Health Evaluation (APACHE) II and Risk, Injury, Failure, Loss and End-stage renal failure (RIFLE) scores in 412 consecutive cirrhotics admitted to intensive care unit. RIFLE score: inline image 0; inline image 1; inline image 2; inline image 3.

Download figure to PowerPoint

Table 6.  Prediction of mortality in 412 cirrhotics admitted to intensive care unit
Prognostic scoreAUC95% CICut-off pointYouden indexSensitivitySpecificityPPVNPVGoodness-of-fit (P- value)
  1. APACHE II, Acute Physiology and Chronic Health Evaluation; CTP, Child–Turcotte–Pugh; MELD, Model for End-stage Liver Disease; NPV, negative predictive value; PPV, positive predictive value; RIFLE, Risk of renal failure, Injury to the kidney, Failure of kidney function, Loss of kidney function, and End-stage renal disease; SOFA, Sequential Organ Failure Assessment.

APACHE II0.7690.709 to 0.822180.43776678650.51
CTP0.7330.671 to 0.789110.35706576580.32
MELD0.8060.749 to 0.855210.51718085640.21
SOFA0.8350.780 to 0.881100.47767181650.34
RIFLE0.7040.640 to 0.76200.34904472740.10

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Patients with decompensated cirrhosis have a poor prognosis, particularly when they develop complications related with end-stage liver disease, such as SBP, gastrointestinal bleeding with or without ARF.1,2,6 The latter is associated with marked splanchnic arterial underfilling and it is a well-established adverse predictor associated with increased mortality.10 For example, cirrhotics who develop hepatorenal syndrome (HRS) have very high mortality (particularly for HRS type I).22–25 Similarly, patients with cirrhosis who suffer a decompensating event have a poor outcome, as most die from multi-organ dysfunction.26,27 The prognosis is worse when these patients have ARF at admission or develop ARF during their ICU stay.9

Being able to identify those patients with cirrhosis who may benefit most from admission to ICU or continued ICU care would be very useful.28 Although the evaluation of prognosis could be based on the accurate assessment and prompt detection of renal dysfunction, there is no accepted definition of ARF due to the absence of a consensus, particularly in critically ill patients. Serum Cr is the most widely used index for renal function evaluation and its importance is reflected by the inclusion of Cr in the MELD score.19 However, very few studies have assessed the outcome of cirrhotics admitted to the ICU with elevated levels of Cr. In addition, these few studies defined ARF using different thresholds of Cr. For example, Chen et al.7 evaluated cirrhotic patients considering ARF as the presence of Cr ≥ 1.5 mg/dL (≥ 132.6 µmol/L) (Table 7); ARF on admission was associated with a high ICU mortality rate (86.6%). Similarly, Fang et al.29 from the same centre, evaluated 111 different cirrhotics with renal dysfunction. ARF was defined as Cr ≥ 1.5 mg/dL (≥ 132.6 µmol/L) or a 50% rise in Cr from baseline or the need for hemofiltration (the first two criteria were based on the 1st stage of severity [Risk] of ARF according to the RIFLE classification) (Table 7). In this cohort the mortality of ARF so defined was 81.1%. However, Cr is an inaccurate marker to assess renal function, and thus, a more accurate classification would be preferable for reliable estimation of renal function. In our own centre9 ARF on admission or during the ICU stay was an independent factor for mortality. In our study, ARF was defined as a SOFA score of 3 or more points for kidney function (i.e. Cr ≥ 3.5 mg/dL [≥ 309.4 µmol/L] or oliguria [< 500 mL/day]), or the presence of HRS or need for hemofiltration. Although in contrast to the two previous studies7,29 (from the same centre), we used a combination of Cr and diuresis to define ARF, neither our study,9 nor these previous ones7,29 evaluated the impact of different degrees of renal dysfunction (from normal kidney function to complete renal failure) on mortality.

Table 7.  Review of published studies evaluating outcome of cirrhotic patients admitted to intensive care unit with acute renal failure
StudyNumber of patients with ARFAge (years)Liver disease (%)Definition of renal failureScores (mean)MortalityMultivariate analysis (odds ratio)Prognostic models (ROC curve)
  • *

    Based on Acute Dialysis Quality Initiative group, ∧ Based of SOFA score.

  • APACHE, Acute Physiology and Chronic Health Evaluation; ALD, alcoholic liver disease; ARF, acute renal failure; CHB, chronic hepatitis B; Cr, serum creatinine; CTP, Child–Turcotte–Pugh; ICU, intensive care unit; MAP, mean arterial pressure; MELD, Model for End-stage Liver Disease; RIFLE, Risk of renal failure, Injury to the kidney, Failure of kidney function, Loss of kidney function, and End-stage renal disease; SOFA, Sequential Organ Failure Assessment.

Chen et al.76758.9CHB (51)Cr ≥ 1.5 mg/dLCTP:11.4 APACHE III:118.5 APACHE II:28.1 SOFA:12.5ICU mortality: 86.6%APACHE III:0.87 APACHE II:0.79 SOFA:0.86 CTP:0.77
Cholongitas et al.9128 (total 312)48 ± 12ALD: 63Cr ≥ 3.5 mg/dL or HRS or hemofiltrationCTP: 12 APACHE II:22 MELD:31 SOFA: 13ICU mortality or 6 weeks after ICU: 91%FiO2, bilirubin, CrSOFA:0.85, MELD:0.81, APACHE II:0.78, CTP: 0.78
Jenq et al.1481 (total: 134)55CHB:41RIFLE*Total cohort CTP:11.5 MELD:28.4 APACHE II:24.2 APACHE III: 102 SOFA: 10.9Hospital mortality: RIFLE-R: 68.8%, RIFLE-I:71.4%, RIFLE-F: 94.8%Total cohort: RIFLE class: 2.13 SOFA: 1.95Total cohort: CTP: 0.69, MELD: 0.88, APACHE II:0.81, APACHE III: 0.88, SOFA: 0.91
Du Cheyron et al.1573 (total: 186)56.4ALD (72)RIFLE*APACHE II: 20.1 SOFA: 9.8 MELD: 14.3ICU mortality: 60%ALD (HR:6.3), severe sepsis (HR:3.5), severe ARF (HR: 5.3)
Fang JT et al.2911157 ± 12CHB: (41)Cr ≥ 1.5 mg/dL, a 50% rise in Cr or need for hemofiltrationCTP: 11.9 APACHE II:27 MELD:36 SOFA: 12Hospital mortality: 81.1%MAP, bilirubin, sepsis, respiratory failureSOFA:0.84, MELD:0.75, APACHE II:0.71, APACHE III: 0.81, CTP: 0.61
Present study205 (total: 412)49.3 ± 12ALD:(33)RIFLECTP: 11 APACHE II:19 MELD:25 SOFA: 11ICU mortality or 6 weeks after ICU: RIFLE-R: 71%% RIFLE-Injury/Failure:88%FiO2, bilirubin, RIFLE, lactateSOFA:0.84, MELD:0.81, APACHE II:0.77, CTP: 0.72 RIFLE:0.71

In the present study we evaluated renal dysfunction as classified by the RIFLE score, proposed by the Acute Dialysis Quality Initiative (ADQI) Working Group as an evidence-based definition for ARF. Although RIFLE score has been evaluated in several ICU populations, only two studies focused on patients with cirrhosis (Table 7). Our cohort of 412 patients with cirrhosis admitted to ICU is the largest in which the RIFLE score has been assessed. Our study showed a similar total mortality rate compared to the study by Jenq C et al.14 and higher compared to the study by Du Cheyron et al.15 (e.g. patients with a RIFLE score of 1 or 2 had 75%, 69.5% and 51% mortality rates, respectively). However, the recruitment rate in our study was lower, compared to the two previous studies (23 patients/year vs 90 patients/year14 and 31 patients/year,15 respectively). Similar to the two previous studies evaluating the RIFLE score,14,15 we found that the RIFLE score was an independent factor associated with mortality during the ICU stay or within 6 weeks after discharge from ICU (OR:2.1) (Table 5). The other independent factors, that is, FiO2, serum lactate and BIL (Table 4), are surrogate factors of respiratory insufficiency, multi-organ dysfunction/failure and liver dysfunction, respectively. Interestingly, in the patients with RIFLE score > 0, only bilirubin was an independent factor of mortality reflecting the fact that the outcome of cirrhotics with established ARF, is associated mainly with the severity of the underlying liver disease.

Although most of our patients had alcoholic liver disease, in contrast to the study by Du Cheyron et al.,15 we did not find that etiology of cirrhosis was significantly associated with mortality. In addition, in contrast to Jenq C et al.,14 the RIFLE score at baseline had the worst discriminative ability (AUC = 0.71), significantly lower than SOFA score (AUC = 0.84, P < 0.001), and without significant improvement at 48 h after admission to ICU (AUC:0.75). Thus, the RIFLE score cannot substitute intensive care scores in cirrhosis. However, the RIFLE score had the highest sensitivity (90%) to detect the patients who died during the follow-up period (Table 6), suggesting renal function to be a very important factor in the prognosis of critically ill patients with cirrhosis. This was also shown in our previous study,16 in which renal function was the most important factor to be associated with mortality during ICU stay or within 6 weeks after discharge from ICU. Although Cr is not an accurate marker of renal function, that is, of glomerular filtration rate (GFR),10 in our previous study evaluating Cr at admission in the same cohort, the discriminative ability was similar to that of the RIFLE score.9 Nevertheless, similar to Jenq C et al.,14 the severity of renal dysfunction was significantly associated with the outcome, as the patients without ARF had a 57.5% survival, compared to 21% in those with Risk, and only 12% (P < 0.05) in those with the presence of Injury/Failure according to the RIFLE classification.

Although the time interval of admission was not an independent factor of mortality, we re-confirmed the findings of our previous study,16 as there was a significant improvement in survival over the years (24% during the 1st interval [1989–1992] vs 50% during the last [2005–2006], P < 0.001). This may reflect improvement in therapeutic interventions as well as earlier admission or stricter selection of the patients who are admitted to the ICU, as the median RIFLE score was progressively lower (better renal function) over time (e.g. 2 vs 0.5 for the 1st and the last periods, respectively, P < 0.001). Nevertheless, despite the importance of renal function as a prognostic indicator in patients with cirrhosis, RIFLE was not as good as the SOFA score in our population.

A limitation of our study is that our cohort derived from a specialized single referral centre, with patients referred for specific therapy, particularly for bleeding (therapeutic endoscopy, transjugular intrahepatic portosystemic shunt). The second potential limitation is the relatively low number of cirrhotics with stage 2 (Injury) ARF (n = 37), while no patient had Loss or End-stage renal failure according to the RIFLE score.

In conclusion, in our study, half of the patients with cirrhosis requiring ICU admission had ARF according to RIFLE classification. The latter was an independent factor associated with mortality. In particular, the cirrhotics with Injury or Failure (RIFLE score 2 or 3) had a very poor outcome (mortality 88%) despite therapeutic intervention. This emphasizes the need for preventive measures in patients with cirrhosis (such as SBP prophylaxis, avoidance of nephrotoxic agents)5 in order to reduce the occurrence of ARF in patients with decompensated cirrhosis.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References