Predictors of Hepatorenal Syndrome in Alcoholic Liver Cirrhosis

Background. Alcoholic liver disease is a major cause of liver cirrhosis and the hepatorenal syndrome is a serious complication. Risk factors for hepatorenal syndrome (HRS) in alcoholic liver cirrhosis are not entirely explored. Aim. To assess the risk factors for hepatorenal syndrome in alcoholic liver cirrhosis. Patients and Methods. Consecutive patients with alcoholic liver disease were followed for two months, development of renal failure, classified either as HRS or renal failure not fulfilling criteria of HRS, was the main outcome. Multivariate analysis adjusted for sex and age showed that sodium together with creatinine are the strongest HRS predictors, followed by bilirubin with respective odds´ratios (95% CI) of 1.041 (1.012-1.072) for creatinine, 0.870 (0.766-0.988) for serum sodium and 1.005 (1.001-1.010) for serum bilirubin. Conclusion. Serum levels of sodium, creatinine and bilirubin are important predictors of the hepatorenal syndrome.


INTRODUCTION
Hepatorenal syndrome occurs in roughly 25% of patients with acute alcoholic hepatitis 1 , 20-30% of patients with spontaneous bacterial peritonitis 2 and in 15% of patients undergoing large volume paracentesis (over 5 liters of evacuated fluid) without adequate plasma expansion 3 .
The actual risk factors for HRS are much less studied.There are several papers documenting that patients with HRS have lower mean values of serum sodium [4][5][6] but this association does not necessarily mean that serum sodium also has any predictive value.
The diagnosis of HRS is based on exclusion of other types of renal failure.Currently most accepted, are the diagnostic criteria proposed by the International Ascites club 7 .
There are at least five papers which report data on the predictive value of hyponatremia in HRS, but the results are controversial.Two studies found serum sodium was not a significant, independent HRS predictor 5,8 and two did 3,9 .The fifth paper was only a case-control design 10 .The foregoing formed the rationale for confirming the predictive value of hyponatremia in HRS development and quantifying its risk level relative to parameters of liver and kidney function in a methodically robust cohort design study which included a sufficient number of participants.

PATIENTS AND METHODS
Consecutive patients admitted to the 1st department of internal medicine, university hospital, Košice, Slovakia in 2002 to 2007 with decompensated alcoholic liver cirrhosis (ICD-10 code K70.3) were retrospectively analyzed.Diagnostic criteria for liver cirrhosis were: 1) morphologic -typical appearance of the liver on ultrasound or computed tomography, 2) laboratory parameters of impaired liver function, 3) clinical signs of liver cirrhosis and portal hypertension.Decompensation was defined as the presence of ascites, variceal bleeding, encephalopathy or icterus 11 .Alcoholic etiology was based on reported alcohol use on average more than 60 g daily for men and 40 g daily for women for at least five years, and on negative serology screening for viral hepatitis B or C. Patients with serum creatinine over 133 micromoles/L at the time of presentation were excluded.Other exclusion criteria were coronary heart disease, dilatation cardiomyopathy, acute coronary syndrome, acute surgical abdomen, chronic obstructive pulmonary disease with respiratory failure, prior hepatorenal syndrome.
Relevant data, which included age, sex, etiology of liver cirrhosis, comorbidities, reason for admission, duration of hospitalization and follow-up, serum creatinine, INR, bilirubin, serum and urine sodium and measured glomerular filtration (mGF) were extracted from patients' charts.All results were determined in Košice University hospital laboratories; glomerular filtration was measured by creatinine clearance.
All patients were followed up for two months, which is the maximum time between scheduled check-ups for patients with advanced liver cirrhosis in our institution.Development of renal failure was the main outcome.Hepatorenal syndrome was diagnosed according to International Ascites Club guidelines from 2007 (ref. 12), but we also accepted plasma expansion with 1.5 litres of normal saline with diuretic treatment stopped for 24 h (ref. 13).Elevation of creatinine over 133 μmol/L without fullfilling other criteria for HRS was regarded as "non-HRS renal failure".
The study was approved by institutional ethics committee and performed according to the Declaration of Helsinki.Informed consent was obtained from all patients at the time of admission.
Data are presented as means ± standard deviation, if not stated otherwise.Statistically significant difference among continuous variables was analyzed by ANOVA or Kruskall-Wallis test, respecting these tests' data assumptions.Differences in categorical variables were assessed by the Chi-square test.Univariate and multivariate regression analysis was performed by logistic regression analysis.Second multivariate regression was performed with the standardized variables (Z-scores), so that the relative effect of the variables on the outcome would be easily comparable.Probability of development of HRS was assessed by the Kaplan-Meier method with log-rank test comparison of the curves.Statistical analyses were performed in SPSS statistical software package, version 17, IBM Corporation, New York.

RESULTS
The cohort consisted of 171 patients, 65% males.Eighty three patients were taking furosemide (mean dosage 21.3 ± 26.8 mg), 125 patients were taking spironolactone (mean dosage 56.5 ± 44.9).Baseline parameters of the study cohort are summarized in Table 1.Acute alcoholic hepatitis superimposed on liver cirrhosis was present in 24 cases (14%), hepatic encephalopathy in 35 cases (20.5%) and variceal bleeding in 19 cases (11.1%).Of 24 patients with acute alcoholic hepatitis at the time of admission, 4 (16.7%)developed renal failure.Two of these patients had HRS and two had renal failure which did not fulfill the criteria of HRS.In total, 120 (70.2%) patients had ascites and 66 (38.6%) patients had at least one infectious complication.
Patients were hospitalized for a median of 12 days (range 1 to 54); all patients were censored at day 60.Overall, the 60-day mortality in the cohort was 19.3% (33 patients out of 171), group-specific mortality is presented in Table 2.
All patients received treatment with albumin and terlipressin.Albumin dosage was 20 g per day, terlipressin 2 mg per day divided to 4 doses.One patient required catecholamine treatment for hypotension and two patients underwent renal replacement therapy.We observed complete response to the therapy in 5 (35.7%) and partial response in 4 (28.6%)patients.
Mean values of selected parameters in the patients without renal failure, in patients who developed other renal failure and in patients who developed hepatorenal syndrome are summarized in Table 2.We found significant differences in serum creatinine, serum sodium, urine sodium, serum bilirubin, INR, MELD score and MELD-Na score.Patients who developed hepatorenal syndrome were younger than patients without failure or non-HRS renal falure.There were no significant differences in body weight or diuretic dosing.
Logistic regression adjusted for age and sex, confirmed serum sodium, urine sodium, MELD, MELD-Na, total bilirubin, albumin, INR, and serum creatinine as significant predictors of HRS development (Table 3.) Multivariate logistic regression analysis was performed to assess the predictive value of an individual variable in combination with other variables in a mathematical model.Because of presumed high inter-correlation among similar parameters (bilirubin, albumin, MELD score, urine sodium and serum sodium) which could have inflated the variance and reduced the validity of the model, we only selected age, creatinine, serum sodium and bilirubin for the multivariate regression.All of the selected parameters analyzed together were found to be significant and independent predictors of hepatorenal syndrome.Odds ratios (95% CI) were 0.929 (0.862-1.000) for age, 1.041 (1.012-1.072)for creatinine, 0.870 (0.766-0.988) for serum sodium and 1.005 (1.001-1.010)for serum bilirubin.Comparison of the relative effect of variables shows that in this predictive model serum creatinine explains 28.1%, serum sodium 24.7%, age 24.6% and bilirubin 23.1% of the model performance.Optimal cut-off value had 100% sensitivity and 70.1% specificity with positive predictive value of 23.0% and negative predictive value of 100%.Model-generated probabilities were used to sort patients to high and low HRS probability group.The incidence of HRS between these groups was compared by Kaplan-Meier analysis (Fig. 1).The difference of survival between low and high risk groups was 23%, P < 0.001.

DISCUSSION
Hepatorenal syndrome is a potentially preventable, but serious complication of advanced liver cirrhosis 14 .
Although several complex predictive tools exist for mortality of liver cirrhosis and its complications, no such tool is available for hepatorenal syndrome.For mortality of liver cirrhosis the most widely used scoring systems are MELD and Child-Pugh, out of which MELD score is slowly becoming the gold standard in liver transplantation 15 .Occasionally, data for unconventional mortality predictors in liver cirrhosis are published 16 .The predictors of some liver cirrhosis complications, such as variceal bleeding have also been described 17 .
The predictors of hepatorenal syndrome are much less studied.That is very unfortunate, because there are several strategies for decreasing the probability of HRS in high risk patients.It is well-known that intravenous albumin infusion in high risk patients with spontaneous bacterial peritonitis, decreases the risk of renal impairment 18 .Another classic paper by Akriviadis reported that in high risk patients with acute alcoholic hepatitis pentoxifylline decreases the probability of hepatorenal syndrome 1 .This aside, new data are emerging that show that prophylaxis with rifaximin also decreases the risk of hepatorenal syndrome in decompensated cirrhotics.The authors suggest that the probable mechanism is reduction of endotoxin production with subsequent decrease in tumor necrosis alpha levels, which have positive effect on portal hypertension 19 .
Unfortunately, the tools for identifying this group of high risk patients are inadequate.The most widely cited paper was published almost 20 years ago by Gines et al.In the cohort of 234 patients, with a one year HRS incidence of 18%, the authors determined the individual prognostic value of 16 variables.In the multivariate analysis, only hyponatremia, high plasma renin activity and lack of hepatomegaly were found to be independent significant predictors of HRS.Surprisingly, no variable reflecting liver function was found to be a significant predictor in this paper 3 .Liver function parameters (MELD, Child-Pugh, INR, bilirubin, albumin) were not found to be different in patients with and without subsequent development of HRS in a study published by Seo et al.There was also no difference in serum sodium between these two groups of patients 7 .These results are opposed to a large body of evidence showing that liver function parameters are the main mortality predictors in hepatorenal syndrome.A study by Yang et al. showed that serum bilirubin, albumin, creatinin and MELD score were significantly associated with prognosis of patients with hepatorenal syndrome 20 .
In our study, in contrast to the above papers, the results of univariate regression analysis suggest that parameters of liver function (total bilirubin, albumin, INR, MELD and MELD-Na score) as well as age, serum and urine sodium and creatinine are significant predictors of hepatorenal syndrome.To assess the independent prognostic value of these variables we selected four for the multivariate analysis.The predictive model constructed from these four variables was very good at predicting HRS and furthermore each variable was an independent predictor.Serum creatinine contributed the most to the overall performance of the model followed by serum sodium, age and bilirubin, but the differences were small.Deranged intravascular volume distribution with effective arterial hypovolemia on the one hand and volume overload in splanchnic vasculature on the other, with activation of compensatory mechanisms take place at the outsetof HRS development.For this reason, dilution hyponatremia and low sodium output in the urine are expected to be very common in patients with HRS.This association has been confirmed by published data [4][5][6] .It is however necessary to rule out, pseudohyponatremia in acute on chronic liver failure 21 .
Data on the actual prognostic value of serum sodium are relatively controversial.Porcel et al. reported a casecontrol study which included 44 patients with cirrhosis and hyponatremia in the study group, and 20 controls with cirrhosis without hyponatremia.Reported odds ratio for developing HRS in hyponatremic group was 45.125 (95% CI 5.56-366.307)vs. eunatremic group.Unfortunately, as this was a case-control study, it carried inherent selection bias, manifested as a very high HRS incidence rate (52% in roughly 5 months), which is approximately 5 times higher than the generally reported HRS incidence 10 .
It is necessary to rule out pseudohypoanatremia in this setting, which could by caused hypercholesterolemia.
Other published papers, what include data for hyponatremia, concentrated mostly on novel parameters of renal failure, such as cystatin C. Ahn et al. studied 112 patients with a 21% 1-year HRS incidence and reported that serum sodium, along with MELD score and cystatin C, were significant predictors of HRS in univariate (OR 0.894, 95% CI 0.841-0.851)and multivariate (HR 0.93, 95% CI 0.87-0.99)analysis 9 .
Younger age was also significantly associated with risk of HRS in our data.This association has never been published to the best of our knowledge.In fact, in some papers higher age was associated with higher risk of HRS (ref. 5).We believe that this discrepancy was caused by the fact that we had a priori ruled out patients with elevated serum creatinine and who were generally older.
Our results reported here indicate that the baseline risk of HRS development is dependent on liver and kidney function (in our data represented by serum creatinine and bilirubin) and on systemic/splanchnic circulation derangement represented by hyponatremia.
A predictive model based on serum sodium, bilirubin and creatinine, could be helpful in HRS risk assessment in cirrhotic patients.It allows clinicians to identify patients with higher HRS risk, who require closer followup, intensive management of precipitating factors (such as variceal bleeding or spontaneous bacterial peritonitis) and caution when performing invasive procedures, such as paracentesis.

Table 1 .
Basic characterization of the study cohort.
INR -international normalization ratio

Table 2 .
Mean values of parameters of interest in three outcome groups in the beginning of follow-up.

Table 3 .
Logistic regression analysis of predictors of hepatorenal syndrome.Parameters are adjusted for age and sex.
1.099-1.365Fig. 1.Comparison of HRS development rate between patients with and without hyponatremia.