Potential conflict of interest: Nothing to report.
Presented in part to meetings of the American Association for the Study of Liver Diseases (Boston, MA, November 2006) and the French Association for the Study of the Liver (Paris, France, October 2006).
Although it is often functional at presentation, acute renal failure has a poor prognosis in patients with cirrhosis. The role of inflammation, a key event in the outcome of cirrhosis, has never been studied in this setting. We aimed to investigate the predictive factors of mortality in patients with cirrhosis and acute functional renal failure, specifically in relation to inflammatory events. One hundred consecutive patients with cirrhosis from 5 French hospitals were prospectively included at the day of onset of acute renal failure. Medical history, treatments, and procedures during the month before inclusion were recorded. Physical examination, blood and urinary chemistries, and renal ultrasound examination were performed. The presence of systemic inflammatory response syndrome (SIRS), infection, and sepsis was assessed. The primary outcome was in-hospital mortality. The mechanism of renal failure was functional in 83 patients. Causes of renal failure were hypovolemia (34%), hepatorenal syndrome without ongoing infection (17%), hepatorenal syndrome with ongoing infection (16%), nephrotoxicity (2%), and multifactorial (31%). SIRS was observed in 41% of patients, 56% of them with infection. In-hospital mortality was 68% in patients with SIRS and 33% in patients without (P = 0.001). In multivariate analysis, only model for end-stage liver disease score and presence of SIRS, but not infection, remained associated with a poor outcome. Conclusion: The presence of SIRS, with or without infection, is a major independent prognostic factor in patients with cirrhosis and acute functional renal failure. This suggests that preventing and treating SIRS could decrease mortality in patients with cirrhosis and acute renal failure. (HEPATOLOGY 2007.)
Acute renal failure is a frequent complication of cirrhosis, and it may have various causes.1 Most renal dysfunction is due to functional failure in this population.1, 2 Gastrointestinal bleeding, hepatorenal syndrome (HRS), and sepsis are common causes of renal hypoperfusion.1, 2 In these 3 conditions, there is growing evidence that systemic inflammation is one of the key factors causing circulatory disorders, including impaired effective arterial blood volume leading to renal hypoperfusion and failure.3, 4
Renal failure in patients with cirrhosis is a serious event with a poor prognosis. The development of acute renal failure during hospitalization for acute upper gastrointestinal bleeding is an independent predictive factor of death.5 Similarly, the development of renal dysfunction, even reversible, during sepsis is strongly associated with a poor outcome.6, 9 Finally, despite advances in intensive care, in-hospital mortality in patients with type 1 HRS is 75%.2 However, the predictive factors of mortality in functional renal failure in patients with cirrhosis have not yet been clearly identified. In particular, the role of systemic inflammatory events has not been studied.
The aim of this multicenter prospective study was to determine the prognostic factors in a population of patients with cirrhosis and functional renal failure, paying special attention to inflammatory events assessed according to the recommendations of the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference.10
HRS, hepatorenal syndrome; MARS, molecular adsorbent recirculating system; MELD, model for end-stage liver disease; SIRS, systemic inflammatory response syndrome.
Patients and Methods
Patients were enrolled consecutively at 5 centers in France between January 2005 and January 2006. Inclusion criteria were: (1) occurrence of acute renal failure defined as an increase in serum creatinine levels of more than 50% of the baseline value, to above 133 μmol/L (1.5 mg/dL)1; (2) at least 1 normal serum creatinine level assessment during the month before inclusion; (3) normal serum creatinine levels during the final assessment before inclusion; and (4) histologically confirmed cirrhosis or obvious clinical, biochemical, and radiological signs of cirrhosis. Exclusion criteria were: (1) absence of information on baseline renal function; (2) chronic renal failure; and (3) refusal to participate in the study.
Patients could have acute renal failure on the day of admission or develop it during hospitalization. In the first case, special attention was paid to confirm that the patient had a recent (<1 month) assessment of renal function within normal values.
Definitions of Systemic Inflammation and Sepsis
A modified version of the criteria defined by the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference10 and by Bernard et al.11 was used. These criteria have been previously validated in patients with cirrhosis.8, 9, 12
Patients were considered to have systemic inflammatory response syndrome (SIRS) if they fulfilled at least 2 of the following criteria: (1) a core temperature of ≥38°C (100.4°F) or ≤36°C (96.8°F); (2) a heart rate of ≥90 beats/minute; (3) a respiratory rate of ≥20 breaths/minute or partial arterial carbon dioxide pressure (PaCO2) of ≤32 mm Hg or the use of mechanical ventilation for an acute respiratory process; or (4) a white blood cell count of ≥12,000/mm3 or ≤4000/mm3 or a differential count showing >10% immature neutrophils.
Patients were considered to have an infection if there was a known or suspected infection based on 1 or more of the following: white blood cells in a normally sterile body fluid (eg, polymorphonuclear count >250/mm3 in ascitic fluid)13; perforated viscus; radiographic evidence of pneumonia associated with purulent sputum; or a syndrome associated with a high risk of infection (eg, ascending cholangitis).
Patients with infection were considered to have sepsis if they had SIRS, as defined above, in response to a proven or suspected microbial event.
Patients had septic shock if they presented with sepsis-induced hypotension despite adequate fluid resuscitation as well as organ dysfunction or perfusion abnormalities.
Medical history was recorded upon inclusion. Medical events (infection, bleeding, ascites, and paracentesis) and treatments administered within the month and 48 hours before enrollment were carefully evaluated. Any use of contrast medium was investigated. Biochemical and bacteriological parameters within 48 hours before admission were recorded when available.
Patients underwent physical examination, laboratory tests (blood and urine), and renal ultrasound examination on the day of inclusion. Urine output was determined. The presence and number of SIRS criteria and the presence of sepsis, severe sepsis, or septic shock were assessed.10 Severity of liver disease was assessed according to model for end-stage liver disease (MELD)14, 15 and Child-Pugh16 scores. Renal biopsy was not indicated for the study and was performed only when clinical assessment and laboratory and radiological investigations suggested a diagnosis of an intrinsic renal disorder other than acute tubular necrosis.
Patients were followed up prospectively until discharge or death. Urine output was recorded daily. Complications of cirrhosis (eg, ascites, gastrointestinal bleeding, and sepsis) were recorded, as was the use of renal replacement therapy [conventional dialysis or molecular adsorbent recirculating system (MARS)] and mechanical ventilation. The course of renal failure was noted.
At discharge, patients underwent another physical examination and laboratory tests (blood and urine). The detail of the items recorded and analysis performed before inclusion (retrospective assessment), at inclusion and during follow-up are detailed in Supplemental Table 1.
Assessment of Causes of Renal Failure
The cause of acute renal failure at inclusion was determined according to the following guidelines, which included characteristics before and on the day of inclusion and clinical outcome after treatment. Renal failure was suspected to be functional when the following conditions were present: evidence of fluid loss, gastrointestinal bleeding, shock, former diuretic therapy; proteinuria <500 mg/dL, urine red blood cell count <50 per high-power field, urine sodium <10 mmol/L, urine Na/K ratio <1, urine osmolality greater than plasma osmolality, and absence of obstruction or a dilated collecting system on renal ultrasound examination.
Functional renal failure was classified into 4 groups:
1True hypovolemia. This classification was assigned when the patient had gastrointestinal bleeding or diarrhea or was prescribed diuretics and had an improvement of serum creatinine after diuretic withdrawal and volume expansion.
2HRS. In this classification, HRS refers to type 1 HRS. In accordance with new diagnostic criteria,22 renal failure in the setting of ongoing bacterial infection but in the absence of septic shock is considered HRS. Therefore, 2 subgroups of patients were identified in the HRS group: HRS without ongoing infection and HRS with ongoing infection.
3Nephrotoxicity. This classification was assigned to patients who had been exposed to nephrotoxic drugs such as nonsteroidal anti-inflammatory drugs or contrast medium.
4Mixed. This classification was assigned in cases that seemed to be multifactorial.
Intrinsic renal failure included acute tubular necrosis and glomerulonephritis and was suspected when proteinuria was >5 g/dL, hematuria was assessed, and the Na/K ratio was >1. Obstructive cause of renal failure was suspected in case of obstruction or a dilated collecting system on renal ultrasound examination.
Renal failure was considered reversible when serum creatinine decreased to <30% above baseline during hospitalization; it was considered nonreversible when serum creatinine remained elevated (at least 30% above baseline) above 133 μmol/L until the end of hospitalization or death. Patients with intrinsic renal failure and/or obstructive cause of renal failure were excluded from the analysis, because the aim of the study was to examine the prognostic factors of functional renal failure.
Fisher's exact test, chi-square test, Student t test, Mann-Whitney test, and analysis of variance were used for statistical analysis. The main outcome was vital status at hospital discharge. Univariate logistic regression assessed the relationship between patient characteristics and vital status at hospital discharge. Variables with an associated P value <0.2 were entered into multivariate logistic regression. Model discrimination was assessed using the area under the receiver operating curve. Two-sided statistical tests were used for all analyses; a P value of ≤0.05 was considered significant. Statistical analysis was performed using Number Cruncher Statistical Systems 2003 software (NCSS, Kaysville, UT).
Of 147 patients screened, 47 did not have a serum creatinine assessment within the month before screening and thus were not included in the study. The remaining 100 patients were included and followed prospectively. Eighty-nine of these patients developed acute renal failure during hospitalization; 11 were admitted for complications of cirrhosis and presented with renal failure at the first blood sample. All patients had normal renal function, had not been admitted to another hospital during the month before admission, and had no renal failure at the last visit. Most of the patients had severe liver disease (mean MELD score 30 ± 9; 74% of Child-Pugh class C patients) and were oliguric. The detailed patient characteristics are shown in Supplemental Table 2.
Causes of Renal Failure
Eighty-three patients had functional renal failure according to the definition given in Patients and Methods, and 17 patients had intrinsic renal failure. None of the patients had obstructive cause of renal failure. Characteristics of renal function in patients with functional and intrinsic renal failure are shown in Table 1. Because the aim of this study was to determine the prognostic factors in patients with functional renal failure, patients with intrinsic renal failure were excluded from further analysis.
Table 1. Characteristics of Renal Function in Patients with Functional Renal and Intrinsic Renal Failure
Patients were managed according to common guidelines agreed upon by all 5 centers.17 Diuretics and nephrotoxic drugs were stopped at inclusion. In case of upper gastrointestinal hemorrhage, transfusion of packed red blood cells was administered to maintain the total haematocrit between 25% and 30%, and plasma expanders were added to maintain hemodynamic stability. Vasoactive drugs were used according to standard recommendations.18 When there was an obvious or suspected infection, the correction of volume depletion included crystalloids and/or colloids and intravenously administered albumin. The use of hydroxyethyl starch was contraindicated. In case of septic shock, catecholamines were used if hypotension persisted despite adequate fluid replacement. Antibiotic therapy was administered after blood and urine samples were taken and paracentesis was performed.19 When the patient met the criteria for HRS without ongoing infection,20 treatment included vasoconstrictors (terlipressin or noradrenalin) and albumin.21 Intravascular blood volume was optimized by fluid replacement (at least 1.5 L of isotonic saline) and albumin. When the cause of renal failure was not clearly determined, correction of hypoperfusion was obtained with fluid replacement therapy including crystalloids or albumin. Continuous renal replacement therapy was indicated in patients with hyperkaliemia or pulmonary edema.
The reasons for admission in the 83 patients were: gastrointestinal bleeding in 26 patients, worsening of hepatic function and jaundice in 18 patients, ascites in 24 patients, and hepatic encephalopathy in 15 patients. Figure 1 describes the causes of renal failure in those patients. Twenty-seven patients were classified with HRS. Thirteen (16%) patients had obvious ongoing infection at the onset of renal failure, with no gastrointestinal bleeding, no shock, and no use of diuretics or nephrotoxic drugs. Renal failure was considered “infection-related” in these 13 patients, and they were classified in the “HRS with ongoing infection” group. Fourteen (17%) patients had not received nephrotoxic drugs, contrast medium injections, or diuretics 48 hours before inclusion. No infection or gastrointestinal bleeding had occurred recently in any of these patients. All these patients had large ascites and were classified in the “HRS without ongoing infection” group. In the 69 remaining patients, 13 experienced gastrointestinal bleeding on the day of or within 48 hours before inclusion, did not take any diuretics, and did not have any other acute complications of cirrhosis; 14 received diuretics for ascites without any other complications of cirrhosis, and renal failure was reversible after the withdrawal of diuretics; and 1 had gastrointestinal bleeding and diuretics on admission. Renal failure was considered “hypovolemia-related” in these 28 patients (34%).
One patient had received an intravenous injection of a contrast medium 3 days before the onset of renal failure; another was prescribed nonsteroidal anti-inflammatory drugs for back pain. In these 2 patients (2%), renal failure was considered to be related to nephrotoxicity. Twenty-six (31%) patients had a combination of infection and/or gastrointestinal bleeding and/or the use of nephrotoxic drugs or diuretics; in these patients, renal failure was considered to be multifactorial (ie, mixed cause).
Thirty-four of 83 (41%) patients presented at inclusion with SIRS as defined in Patients and Methods. Thirty-two of these patients fulfilled 2 SIRS criteria, 1 patient fulfilled 3 criteria, and 1 patient fulfilled 4 criteria. The distribution of the different SIRS criteria was as follows: heart rate >90 beats/minute, n = 18; core temperature ≥38°C (100.4°F) or ≤36°C (96.8°F), n = 9; respiratory rate ≥20 breaths/minute, partial arterial carbon dioxide (PaCO2) pressure ≤32 mm Hg, or use of mechanical ventilation for an acute respiratory process, n = 4; white blood cell count ≥12,000/mm3 or ≤4000/mm3 or a differential count showing >10% immature neutrophils, n = 25.
Patient characteristics with and without SIRS are given in Table 2. As expected, heart rate, respiratory rate, and leukocyte count were higher in patients with SIRS. Median serum creatinine values were significantly higher in patients with SIRS. MELD and Child-Pugh scores did not differ between patients with and without SIRS. The cause of renal failure differed between patients with and without SIRS (Table 3). The proportion of patients who received beta-blockers or norfloxacin was not significantly different between patients with and without SIRS.
Table 2. Characteristics at Inclusion of the 83 Patients with Functional Renal Failure According to the Presence or Absence of SIRS
Table 3. Distribution of the Causes of Functional Renal Failure According to the Presence or Absence of Infection or SIRS
Causes of Renal Failure
Infection with SIRS (n = 19)
Infection Without SIRS (n = 12)
SIRS Without Infection (n = 15)
No SIRS and No Infection (n = 37)
Hypovolemia, n (%)
HRS, n (%)
Nephrotoxic, n (%)
Multifactorial, n (%)
Among patients with SIRS, 19 [56% (23% of total study population)] had obvious infection and thus sepsis according to the definition given in Patients and Methods. Four patients had septic shock at inclusion. Twelve patients with infection did not have SIRS as defined in Patients and Methods. Multifactorial renal failure was mainly found in infected patients. In the subgroup of patients without SIRS and without infection, hypovolemia was the leading cause of renal failure (Table 3).
The median duration of hospitalization was 17 days (range, 4-46). Renal failure was reversible in 54 (65%) patients within a median of 5 days. Patients with nonreversible renal failure had higher MELD scores than patients with reversible renal failure (33 ± 8 versus 28 ± 8; P = 0.01) and less “hypovolemia-related” renal failure (14% versus 86%; P = 0.003). The proportion of reversible renal failure was significantly lower in patients with SIRS (44% versus 80%; P < 0.001). Four patients underwent renal replacement therapy (3 via conventional intermittent dialysis, 1 via MARS), and 18 patients required mechanical ventilation.
Ten of 15 (66%) patients with SIRS and without infection developed an obvious infection during hospital course. Conversely, only 5 of 37 (13%) patients without SIRS and without infection at inclusion developed infection secondarily (P < 0.01).
Thirty-nine patients died during hospitalization (47%). The patients who died developed circulatory failure related to gastrointestinal bleeding or septic shock (or pulmonary edema in 1 patient). Six of these patients also developed respiratory failure related to hypoxic lung infection. Causes of death were uncontrolled bleeding in 7 patients, persistent infection in 10 patients, new infection in 12 patients, liver failure without obvious infection in 5 patients, and metabolic complications of acute renal failure in 5 patients (hyperkalemia in 4 patients, pulmonary edema in 1 patient). Considering the causes of renal failure, mortality was 59% in patients with HRS (79% in patients with HRS without ongoing infection and 39% in patients with HRS with ongoing infection; P = 0.03), 14% in patients with “hypovolemia-related” renal failure, and 73% in patients with multifactorial renal dysfunction (P < 0.001).
Among the 44 patients who left the hospital, follow-up data were available for 30 patients at 2 months. Twenty-five patients were still alive, and 5 had died (2 from uncontrolled bleeding, 3 from hepatic failure).
Predictive Factors of In-Hospital Mortality
Predictive factors of in-hospital mortality in univariate analysis are given in Table 4. Elevated Child-Pugh and MELD scores at inclusion and the need for mechanical ventilation were associated with a poor outcome (P = 0.0007 for MELD score). The type of renal failure influenced mortality; all patients with nonreversible renal failure [29/29 (100%)] died, compared with 10 of 54 (18%) with reversible renal failure (P = 0.001). Renal failure related to true hypovolemia was predictive of a good outcome. The need for mechanical ventilation was associated with significantly higher mortality.
Table 4. Predictive Factors of In-Hospital Mortality in Univariate Analysis in the 83 Patients with Functional Renal Failure
The presence of SIRS showed a strong association with poor survival (P = 0.001) at hospital discharge and 2 months after hospital discharge (Fig. 2). Among criteria defining SIRS, only heart rate >90 beats/minute was independently associated with in-hospital mortality (P = 0.04). The presence of shock, infection, and septic shock at inclusion were significantly associated with a poor outcome. Mortality was 68% in patients with SIRS and infection. In patients without infection, the presence of SIRS remained highly predictive of in-hospital mortality (67% versus 27% in patients without SIRS; P = 0.008). In patients without SIRS, the presence of infection was not significantly associated with higher mortality (50% versus 27%; P = 0.10).
Of the 34 patients with SIRS at inclusion, SIRS resolved in the 11 patients who survived (100%) and did not resolve in any patient who died. Moreover, 6 of 12 patients with infection but without SIRS at inclusion developed SIRS before dying. Of the 6 patients who survived, 2 developed SIRS, but SIRS resolved during hospitalization.
In multivariate analysis, only MELD score and presence of SIRS at inclusion had an independent prognostic value (Table 5). In another model including Child-Pugh score, only Child-Pugh score and presence of SIRS at inclusion had an independent prognostic value (Table 5). A model that included presence of SIRS and MELD score had a higher prognostic value than MELD score alone (area under the curve = 0.81 ± 0.05 versus 0.73 ± 0.06; P = 0.04).
Table 5. Predictive Factors of In-Hospital Mortality in Multivariate Analysis in the 83 Patients with Functional Renal Failure
Odds Ratio (95% Confidence Interval)
Presence of SIRS
Presence of SIRS
This study shows that (1) functional renal failure is the major mechanism of acute renal failure in patients with cirrhosis; (2) functional renal failure has a poor outcome, with an in-hospital mortality reaching 50%; and (3) in addition to the prognostic factors related to the degree of liver failure (ie, MELD and Child-Pugh score), a systemic inflammatory response has independent prognostic value in this setting. This last finding raises the question of the role of inflammatory events in the mechanisms of functional renal failure in cirrhosis.
Although the diagnosis of acute renal failure itself is clearly defined and relatively easy in patients with cirrhosis,1 determining the causes of renal dysfunction is not. The mechanism of renal dysfunction must first be identified. We chose to distinguish 2 mechanisms of renal failure: functional renal failure and intrinsic or obstructive renal failure. Functional renal failure is caused by renal hypoperfusion with no cell injury. The main causes of functional renal failure are true hypovolemia related to gastrointestinal bleeding or the use of diuretics; infection-related renal failure (“HRS with ongoing infection” in the new classification of HRS of the recent consensus workshop), in which the mechanism is clearly an impairment of effective arterial blood volume9; and HRS without ongoing infection. Nephrotoxic drugs such as nonsteroidal anti-inflammatory drugs or iodinated agents may also cause functional renal failure by inducing renal vasoconstriction.1 Conversely, intrinsic renal failure may be due to glomerulonephritis or acute tubular necrosis. However, functional azotemia may in all cases lead to ischemic tubular injury and acute tubular necrosis. Thus, the causes of acute renal failure are time-dependent and can vary with the clinical course. To address this issue, we carefully selected patients to include those with acute renal failure at the onset based on our definition. This explains the higher incidence of functional failure in this study compared with others, ranging from 60% to 70%.2, 23 The second difficulty in describing the causes of renal failure is that they may be multifactorial. We therefore precisely described each patient's clinical condition when renal failure occurred and classified patients with several possible causes in the “multifactorial” group. Despite the strict selection of our population, the cause of renal failure was often multifactorial if the current definitions were strictly applied.20
As expected, most of the patients in this study had severe liver disease, 74% of whom were Child-Pugh class C. This confirms several published studies emphasizing that severe cirrhosis is an independent factor of the development of renal failure in cases of gastrointestinal bleeding,24 spontaneous bacterial peritonitis,25 and bacterial infections in general.26
The in-hospital mortality in this study was almost 50%, confirming the very poor outcome of acute renal failure in patients with cirrhosis. The development of acute renal failure during hospitalization for acute upper gastrointestinal bleeding is known to be an independent predictive factor of death.5 Similarly, in case of sepsis, renal dysfunction, even if it is reversible, is strongly associated with a poor outcome,6–9 with a mortality of 60%. The in-hospital mortality in patients with type 1 HRS reaches 75%.2
To our knowledge, only one study has attempted to determine the predictive factors of mortality in patients with cirrhosis and renal failure.23 The authors focused on standard prognostic factors and found that, in addition to MELD score, presence of HRS was associated with a poor outcome. However, the role of inflammatory events such as infection, sepsis, and SIRS was not investigated. As expected, the present study identified well-known prognostic factors related to the degree of liver failure (ie, Child-Pugh score and MELD score). As in the previous study, the cause of renal failure was found to affect the prognosis. Not surprisingly, reversible renal failure was predictive of a good outcome,9, 24 whereas the need for mechanical ventilation was associated with higher in-hospital mortality.27–30 A recent study investigated the course of sepsis-related renal failure in patients with cirrhosis.9 The mortality was high, reaching 66%—unlike the present study, in which mortality in patients with infection-related renal failure (the “HRS with ongoing infection” group) was only 39%. However, our definition of infection-related renal failure arbitrarily was very restrictive, because it did not include patients with infection and other potential causes of renal failure (multifactorial). If the patients with HRS with ongoing infection and multifactorial renal failure were taken together, the mortality was higher, about 50%, consistent with the previous study.
In the present study, in which patients with severe cirrhosis where enrolled, a high proportion of patients (40%) had SIRS at inclusion. Why SIRS occurs in end-stage cirrhosis is an interesting question. We know that noninfected patients with advanced cirrhosis have a spontaneous increased proinflammatory response compared with noncirrhotic patients because of an imbalance between proinflammatory (enhanced) and anti-inflammatory (inhibited) signaling pathways in immune cells.31, 32 This may explain why SIRS occurs in noninfected patients with cirrhosis. Interestingly, Cirera et al.33 showed that noninfected patients with advanced cirrhosis have significant intestinal translocation of gram-negative bacteria. In a study by Chan et al.,34 blood levels of endotoxin (ie, lipopolysaccharide, a component of the external wall of gram-negative bacteria) were found to be significantly associated with Child-Pugh scores >10 and with poor survival. Moreover, noninfected patients with Child-Pugh class C cirrhosis have activated monocytes that produce tumor necrosis factor α.31, 32 Finally, noninfected patients with cirrhosis treated with the antibiotic norfloxacin (which induces selective intestinal decontamination resulting in decreased gram-negative bacterial translocation) have decreased endotoxin blood levels and numbers of tumor necrosis factor α–producing monocytes.35 Together, these data suggest that noninfected patients with advanced cirrhosis have gram-negative bacterial translocation resulting in increased systemic levels of endotoxin and subsequent immune cell activation indicated by SIRS.
Interestingly, among the 31 patients with infection, 19 (60%) had SIRS and the remaining 12 (40%) did not. In a previous study,12 46% of infected patients with cirrhosis did not have SIRS. Together, these findings suggest that not all infected patients with cirrhosis develop a systemic inflammatory response to pathogens. Further studies are needed to investigate the immune response in these patients.
The causes of renal failure differed between patients with and without infection or SIRS. Hepatorenal syndrome was distributed between infected patients with or without SIRS and those with SIRS but without infection. Multifactorial renal failure was mainly found in infected patients. Interestingly, HRS and multifactorial renal failure were very rare in patients without infection or SIRS. In this subgroup, hypovolemia was the leading cause of renal failure.
The most important and original finding in this study is that a systemic proinflammatory response, assessed by the presence of SIRS, is an independent factor of a poor prognosis, even if there is no obvious or suspected infection. Interestingly, the resolution of SIRS seems to be very important, because SIRS never resolved in patients with SIRS at inclusion who died, and it developed in all the patients without SIRS at inclusion who died. Some authors have already suggested that the systemic proinflammatory response may be an indicator for the bad prognosis of patients with cirrhosis. Chan et al.34 showed that the level of endotoxinemia, which is a direct sign of systemic proinflammatory response, was significantly associated with poor survival, even in patients without any evidence of infection. However, in that study, the levels of endotoxinemia increased significantly with the Child-Pugh score and may only have indicated the severity of liver disease. Similarly, SIRS was already found to be an important and independent predictive factor of death in patients with cirrhosis with gastrointestinal bleeding.3 Nevertheless, presence of SIRS was neither adjusted to the severity of liver disease nor to the existence of an infection.
In patients without cirrhosis, about one-third of patients with SIRS have or evolve to sepsis,36 then to severe sepsis and septic shock in some.37 Insults from bacteria, trauma, pancreatitis, and burns can lead to SIRS.38, 39 In this study, SIRS seemed to be closely related to infection, because no patient presented with trauma, pancreatitis, or burn. Interestingly, 10 of the 15 patients with SIRS and without infection at inclusion developed an obvious infection during hospitalization. Conversely, only 5 of 37 patients without SIRS and without infection at inclusion developed secondary infection, suggesting that SIRS was a strong predictor of infection. This could explain why patients with SIRS have a poor prognosis, because we know well that sepsis has a dramatic outcome in patients with cirrhosis.40
The results of the present study suggest that in patients with cirrhosis and renal failure, the presence of SIRS should be considered a clinical warning of preinfection. SIRS could be used as a simple marker to select the most severe patients with a higher risk of mortality. This raises the question of whether modulating inflammation could be useful in that particular population. The concept that anti-inflammatory drugs could improve renal function in patients with cirrhosis has been described previously. In patients with acute alcoholic hepatitis, the administration of pentoxifylline has been associated with an increase in renal blood flow within 24 hours,41 and the use of MARS improves creatinine levels.42 In patients with spontaneous bacterial peritonitis, the use of albumin, by reducing oxidative stress, reduces the incidence of renal impairment and death compared with antibiotics alone.25 In patients with HRS, the administration of albumin in addition to terlipressin reverses renal impairement, with a low proportion of recurrence after stopping treatment, and may improve survival.43, 44 Furthermore, MARS has been shown to be useful in decreasing creatinine levels and perhaps mortality; however, the study that reported these findings is controversial and included a very small number of patients.45 In the present study, the use of anti-inflammatory drugs or MARS could also be beneficial in a particular subset of patients. Moreover, anti-bioprophylaxis could also improve outcome. It is noteworthy that approximately one-third of patients were receiving norfloxacin at the onset of renal failure. Although lower, this proportion did not significantly differ in patients with SIRS, probably because of the small sample size of our study.
One limitation of this study is the difficulty in interpreting each component of SIRS in patients with cirrhosis.46 Because of hyperdynamic circulatory syndrome, these patients have an elevated heart rate; encephalopathy can cause hyperventilation; the elevation of body temperature is often blunted; and the polymorphonuclear count is often reduced because of hypersplenism.46 Moreover, the administration of beta-blockers can reduce heart rate. Despite these drawbacks, we were able to identify patients with a poor prognosis, especially because of heart rate and polymorphonuclear count.
Unfortunately, we were unable to conduct long-term follow-up in our patients. The short-term follow-up data are rather good: 2-month survival was 81% among patients who left the hospital, suggesting that reversible renal failure has a good short-term prognosis in patients with cirrhosis. Further studies are warranted to evaluate the long-term follow-up in these patients.
Another limitation of the study is the fact that we preselected MELD as a useful prognostic marker, because MELD incorporates creatinine as a criterion for calculation. This is particularly true because the creatinine levels were significantly higher in patients who died. Thus, in order to see if liver failure was still a marker of bad prognosis, we performed a second multivariate analysis without the MELD score but with the Child-Pugh score and creatinine. Child-Pugh score and SIRS were still the 2 predictive factors associated with higher in-hospital mortality, emphasizing the important prognostic role of liver failure.
Finally, this study lacks a control group that has severe liver disease and SIRS but no renal dysfunction, which would help clarify the role of inflammation. Further controlled studies are warranted to clarify this point.
In conclusion, this study shows that functional renal failure is the main cause of acute renal failure in patients with cirrhosis and has a very poor outcome. In addition to the common prognostic factors (MELD and Child-Pugh score), the presence of SIRS was shown to have an important and independent prognostic value in these patients. The results of this study confirm the importance of preventing and treating SIRS to decrease mortality in patients with cirrhosis and acute renal failure.
We thank Professor Corine Bagnis for her advice and careful reading of the protocol.