Abstract
Background Renal dysfunction is a common and potentially life-threatening complication in hospitalized patients with cirrhosis.
Aims To determine the prevalence, cause, and outcome of patients with cirrhosis and acute kidney injury (AKI) and/or chronic kidney disease (CKD).
Methods This retrospective analysis examined hospital records of 152 consecutive patients with cirrhosis and creatinine levels of 1.5 mg/dL or greater. Multiple clinical and laboratory variables were abstracted for each subject. Precise definitions were used to define cirrhosis and etiologies of renal dysfunction. Univariate and multivariable logistic regression analyses were performed to identify features with prognostic value for hospital mortality.
Results The most common type of renal dysfunction was AKI, present in 107 patients (70%). Acute kidney injury plus CKD was found in 26 patients (17%), and CKD alone was present in 19 patients (13%). Prerenal azotemia was the most common cause of AKI (69%), often occurring secondary to gastrointestinal hemorrhage. The overall mortality for the cohort was 31%, with the highest mortality occurring in patients with type 1 hepatorenal syndrome (HRS) (11/14, 79%). We were unable to identify any patient meeting diagnostic criteria for type 2 HRS. The development of AKI on preexisting CKD did not infer worse prognosis than AKI alone. The presence of upper gastrointestinal bleeding, bacteremia, and HRS-1 predicted mortality.
Conclusions Both AKI and CKD are common in hospitalized patients with cirrhosis, often occurring simultaneously. Type 2 HRS was not identified, suggesting that its diagnostic criteria may need reevaluation or that this syndrome may not represent a unique functional kidney disorder.
Kidney dysfunction is a common and potentially life-threatening event in patients with cirrhosis, and underlying mechanisms for renal dysfunction are highly variable.1–3For example, it is well established that portal hypertension and its accompanying hemodynamic abnormalities lead to increased synthesis of endogenous vasodilatory compounds, such as nitric oxide,4,5in turn leading to exaggerated arterial vasodilation in the splanchnic and systemic circulatory systems.6–8These circulatory abnormalities lead to activation of the renin-angiotensin-aldosterone system, fluid retention, and renal hypoperfusion, which may precipitate the development of what has been termed in the literature functional renal failure.7Although appealing for its pathophysiological logic, this model does not take into account all potential causes of renal dysfunction in patients with cirrhosis, as kidney function also may be impaired because of common clinical events such as prerenal azotemia from gastrointestinal bleeding, diarrhea, and excessive use of diuretics or by underlying chronic kidney disease (CKD).6
Although previous studies have provided valuable information regarding kidney injury in patients with cirrhosis, most studies have examined acute kidney injury (AKI) in the setting of acute decompensation (ie, sepsis, intensive care)9–11or development of AKI throughout the course of hospitalization.12Few studies have evaluated renal dysfunction in noncritical care settings, and far fewer have evaluated the prevalence and/or clinical implications of CKD in this population.13–15Because CKD represents an important comorbid condition in essentially all hospitalized patients, it is likely that the combination of chronic liver and kidney dysfunction would result in poor patient outcomes. Moreover, the development of AKI in patients with cirrhosis and underlying CKD may be particularly hazardous.
In this study, we aimed to determine the prevalence of renal dysfunction in patients admitted with cirrhosis to a large urban hospital. In our clinical experience, we have noticed that many hospitalized patients with cirrhosis have underlying CKD and, thus, postulated that, although AKI would likely be the most common type of renal dysfunction encountered, many also would have underlying CKD. Moreover, we hypothesized that patients who develop AKI in the setting of preexisting CKD would have poorer outcomes than those with AKI alone. Therefore, the goals of the present investigation were as follows: (1) to summarize the clinical and laboratory characteristics of inpatients with renal insufficiency and cirrhosis, (2) to determine the precise etiologies of kidney injury in this population, and (3) to explore the association of kidney injury with clinical outcome.
MATERIALS AND METHODS
This retrospective cohort analysis included all consecutive patients with cirrhosis admitted to Parkland Memorial Hospital from 2003 to 2005, with a creatinine level of 1.5 mg/dL or greater (Fig. 1). We collected approximately 300 unique variables for each patient at the time of presentation including demographic, clinical, and historical data such as the presence of previous renal disease, medical comorbidities, ascites, edema, gastrointestinal bleeding, hepatic encephalopathy, and bacteremia defined by a positive blood culture upon hospital admission, as well as laboratory data including bilirubin, albumin, prothrombin time, aspartate aminotransferase, alanine aminotransferase, blood urea nitrogen, serum electrolytes, and urine chemistry. Clinical and laboratory variables reported were recorded at the time of hospital admission. Because the use of diuretics was pervasive in the cohort, an attempt was not made to catalogue their use. Outcomes were reported through the course of index hospitalization. The study was approved by the institutional review board at the University of Texas Southwestern Medical Center and met all criteria for good clinical research.16
Definitions
Cirrhosis was defined based on clinical features, including a history consistent with chronic liver disease as well as a documented complication of chronic liver disease (ie, ascites, varices, and hepatic encephalopathy) and/or imaging consistent with cirrhosis and/or liver histology consistent with cirrhosis.9,11The cause of cirrhosis was determined according to the following criteria: (1) hepatitis C cirrhosis was defined by the presence of cirrhosis in a person with hepatitis C virus (HCV) RNA, (2) hepatitis B cirrhosis was defined by the presence of cirrhosis in patients with hepatitis B surface Ag, (3) alcoholic cirrhosis was determined from the provider’s note in the presence of a history of alcohol abuse or dependence and the absence of other potential causes of liver disease, (4) other causes of cirrhosis were determined using standard diagnostic criteria (ie, serologies, histology, iron studies), and (5) patients without any known cause of primary liver disease were considered to have cryptogenic cirrhosis.
The classification of AKI was defined according to the Acute Kidney Injury Network (AKIN) diagnostic criteria by an abrupt (within 48 hours) increase in serum creatinine (either ≥0.3 mg/dL absolute increase or ≥50% increase over baseline) to a serum creatinine value of 1.5 mg/dL or greater.17A cutoff value of 1.5 mg/dL was used because previous studies have shown that patients with creatinine values of 1.5 mg/dL or greater have significantly decreased renal function, and it is also a cutoff used to define the hepatorenal syndrome.18,19Acute kidney injury was further divided into stages based upon AKIN diagnostic criteria: stage 1 (increase in serum creatinine ≥0.3 mg/dL or ≥150% to 200% from baseline), stage 2 (increase in serum creatinine ≥200% to 300% from baseline), or stage 3 (increase in serum creatinine >300% from baseline or serum creatinine ≥4.0 mg/dL with an acute increase of at least 0.5 mg/dL). Baseline creatinine was defined as the patient’s creatinine value for a consistent 3 months before the current hospital admission. Patients without a baseline creatinine who did not meet diagnostic criteria for AKI by absolute increase in serum creatinine were excluded. Chronic kidney disease was defined as an elevated serum creatinine level (≥1.5 mg/dL) present for more than 3 months before hospitalization.20The etiology of CKD was extracted from previous nephrology service notes and records. Acute and chronic kidney dysfunction was defined as elevated but stable creatinine levels (≥1.5 mg/dL) for at least 3 months, followed by a sudden increase in serum creatinine at time of hospital admission as in AKIN diagnostic criteria.
Patients were further divided into the following etiologies of renal dysfunction: prerenal azotemia, intrinsic renal disease, type 1 hepatorenal syndrome (HRS-1), type 2 hepatorenal syndrome (HRS-2), and postrenal dysfunction.
Prerenal azotemia was defined using clinical and laboratory parameters. Urinary markers necessary for diagnosis included a urine specific gravity greater than 1.020 as well as a fractional excretion of sodium less than 0.01 (1%) in oliguric patients not treated with diuretics. Supporting clinical factors included tachycardia (heart rate, >90 beats per minute), hypotension (systolic blood pressure, <90 mmHg), orthostatic blood pressure changes, dry mucous membranes, other signs of decreased effective arterial blood volume in the setting of documented vomiting, prolonged diarrhea, excessive diuresis, acute hemorrhage, decreased oral intake, infection, or decreased cardiac output. In all cases, the diagnosis was dependent upon a decrease in serum creatinine with volume replacement, treatment of infection, or correction of volume status in those with clinical evidence of congestive heart failure.
Intrinsic renal dysfunction required a documented history of parenchymal renal disease in hospital records or serum creatinine values that did not improve after volume expansion and the exclusion of other etiologies of renal disease. Diagnosis of acute tubular necrosis required the presence of granular casts (epithelial cell casts) in urinary sediment, fractional excretion of sodium greater than 2%, or a urinary sodium concentration greater than 40 mmol/L, in the clinical setting of prolonged hypotension, tachycardia, sepsis, or bleeding. Nephrotic syndrome was defined by proteinuria greater than 3.5 g/d, hypoalbuminemia, hyperlipidemia, and the presence of peripheral edema.21,22Diagnosis of AIN required clinical exposure to a known nephrotoxin and either a diagnosis of AIN in the medical record or the presence of leukocytes, leukocyte casts, and eosinophils in uninfected urine.23
Type 1 hepatorenal syndrome was defined according to the International Ascites Club criteria and included rapidly progressive renal failure and a doubling in serum creatinine concentrations to a level greater than 2.5 mg/dL in less than 2 weeks.18,19Serum creatinine also did not improve (decreased to a level ⩽1.5 mg/dL), despite at least 2 days of diuretic cessation and optimization of intravascular blood volume with albumin and/or saline (of note, we considered optimization of intravascular volume with saline to be suitable because the current International Ascites Club criteria that includes albumin was published after the time frame of the study). Patient medical records and laboratory results were carefully scrutinized to confirm the diagnosis using these established criteria.
Type 2 hepatorenal syndrome was defined according the International Ascites Club criteria.19Diagnostic criteria are the same as for HRS-1; however, HRS-2 is defined by moderate rather than rapidly progressive renal failure with serum creatinine concentrations typically between 1.5 and 2.5 mg/dL.
Postrenal renal dysfunction was confirmed by imaging demonstrating hydronephrosis in the setting of oliguria or anuria.
Statistical Analysis
The primary outcome measure for this study was all-cause hospital mortality. Demographics, and clinical and laboratory characteristics of the patients were summarized using descriptive statistics. Comparisons in mortality rates between renal etiology and renal dysfunction types were determined using exact χ2 test for differences in proportions; 95% confidence intervals (CIs) on mortality rate were computed using exact binomial distribution. To identify which clinical and laboratory markers predict an increase in risk of mortality in patients, we performed logistic regression analyses. Univariate logistic regression analysis was performed to examine the risk of mortality based on renal-related markers including serum creatinine, blood urea nitrogen, renal etiology (prerenal, intrinsic renal, and HRS-1), renal injury type (acute vs nonacute), and other laboratory variables including serum bilirubin, international normalized ratio, model for end-stage liver disease (MELD) score, Child-Turcotte Pugh score, Child Pugh class, mean arterial pressure, upper gastrointestinal bleeding, and bacteremia. Multivariable logistic regression analysis was done by constructing a full stepwise sequence.24The final multivariable model was selected based on the Akaike information criterion to identify risk factors that independently predict mortality.25 P values less than 2-sided 0.05 level are considered statistically significant. Statistical analyses were performed using SPSS Version 11.5 (Chicago, IL) and SAS Version 9.2 (SAS Institute Inc., Cary, NC).
RESULTS
A total of 935 unique patients with cirrhosis were admitted to the hospital from 2003 to 2005. Of those individuals, 170 (18%) were found to have a creatinine level of 1.5 mg/dL or greater. Eighteen patients were excluded because complete clinical and/or laboratory data were not available, leaving a cohort of 152 patients for analysis.
Demographic and clinical characteristics of the cohort are shown in Table 1. Men outnumbered women approximately 3:1. The majority of patients were African American (39%), followed by a nearly equal proportion of whites and Hispanics (30% and 29%, respectively). The most common cause of cirrhosis was HCV infection (45%), followed by alcohol (38%). Some patients with HCV also had a history of alcohol consumption, and these patients were included in the HCV cirrhosis group. The cohort had substantial liver dysfunction as evidenced by the mean Child-Turcotte-Pugh and MELD scores, which were 10 ± 2 and 25 ± 9 (mean ± SD), respectively.
Key laboratory characteristics for the study cohort are displayed in Tables 1 and 2. Of note, the mean albumin was 2.7 g/dL, and the mean international normalized ratio was 1.7 in this population. Serum creatinine had a mean value of 3.0 mg/dL, and fractional excretion of sodium was less than 1% for the majority of patients. Additional features at the time of hospital admission included a mean hematocrit of 33% and leukocyte count of 11,000/mm3 as well as an average mean arterial pressure of 85 mmHg and a heart rate of 95 beats per minute.
The most common classification of renal dysfunction was AKI, identified in 107 patients (70%; Table 3). Stages 2 and 3 AKI accounted for 79% of these cases. The remaining patients had CKD, including 17% with superimposed AKI. Of note, no patient was identified with HRS-2. Eighty-five patients (56%) were admitted with prerenal azotemia, and the precise clinical circumstances are displayed in Table 4. The most common cause of prerenal azotemia was gastrointestinal hemorrhage, comprising 36% of all cases. The most common etiologies of intrinsic renal disease were nephrotic syndrome and acute tubular necrosis (Table 5). Among patients with CKD, nephrotic syndromes (86%) were most prevalent, usually a result of diabetes or long-standing hypertension.
Overall mortality in the cohort during hospitalization was 31% (47/152; 95% CI, 24%–39%). Mortality rates by cause of kidney injury are displayed in Table 6 and Figure 2. The highest mortality rate was seen in those with HRS-1 (79%). One third of the individuals with AKI and one third with AKI plus CKD died during hospitalization, whereas mortality for those with CKD was 11%. In patients with prerenal azotemia, mortality was higher in the AKI group when compared with those with AKI plus CKD. A total of 11 patients required dialysis during hospitalization, including 4 patients with CKD (21%), 5 patients with AKI on CKD (19%), and 2 patients with AKI alone (2%). Mean length of hospital stay was 11 ± 14 days, with a median length of 6 days.
We performed a logistic regression analysis in an attempt to identify clinical variables that might have prognostic value for predicting in-hospital mortality of patients with elevated creatinine (renal insufficiency) as in our cohort. Univariate analysis identified 11 variables with prognostic significance (Table 7). Whereas serum creatinine and blood urea nitrogen as well as HRS-1 predicted hospital mortality, AKI alone did not. A number of other variables predicted mortality, including those shown in Table 7. These were used in a multivariate analysis, which revealed that HRS-1 (odds ratio [OR], 9.4; P = 0.012), upper gastrointestinal bleed (OR, 6.0; P = 0.001), and bacteremia (OR, 12.8; P < 0.001) predicted in-hospital mortality (Fig. 3). In addition, MELD score was a significant predictor of mortality (OR, 1.1; P = 0.002).
DISCUSSION
In this study, renal dysfunction, defined by a creatinine level of at least 1.5 mg/dL at time of hospital admission was identified in 18% of hospitalized patients with cirrhosis over a 3-year study period. The most common type of renal dysfunction was AKI, which accounted for 70% of cases and was usually the result of prerenal azotemia (69%). In turn, the most common cause of prerenal azotemia was gastrointestinal bleeding. We also found that CKD occurred in 30% of hospitalized patients with cirrhosis and renal dysfunction.
The overall hospital mortality rate in patients with cirrhosis and elevated creatinine in this investigation was 31%, consistent with previous reports in patients with cirrhosis and renal failure.1,26–28The highest mortality rate was seen in patients with HRS-1. As expected, patients with AKI had higher mortality than those with CKD; however, mortality rates were similar in patients with AKI and preexisting CKD compared with those with AKI alone. The predominant type of renal dysfunction in patients with AKI and no history of renal disease was prerenal azotemia, whereas prerenal azotemia and exacerbation of intrinsic renal dysfunction were equally prevalent in those with AKI plus CKD. Among all patients admitted with prerenal azotemia, mortality was substantially lower in those with AKI and preexisting CKD compared with AKI alone. This difference cannot be explained by covariates as there were no significant differences between groups with regard to MELD, Childs-Turcotte-Pugh score, age, serum creatinine, bacteremia, or gastrointestinal bleeding. Instead, this observation is likely most consistent with the selection of patients with less severe AKI when occurring in the presence of preexisting renal disease.
Type 1 hepatorenal syndrome has been studied extensively and is widely accepted as associated with a very poor outcome.14,29Approximately 11% of patients with AKI in our study had HRS-1, with an in-hospital mortality of 79%, highly consistent with previous data. All deaths occurred within 3 weeks of admission. Although these results are typical for those who develop HRS-1, far less is known regarding HRS-2.
We were unable to identify any patient in our cohort meeting diagnostic criteria for HRS-2 (highlighted in the “Methods” section). This is in contrast to previously published work, in which HRS-2 has been shown to account for approximately 5% to 7% of AKI.14This discrepancy may be a result of imprecision in definitions of HRS-2 or inherent differences between previous and current diagnostic criteria.18,19Alternatively, it is possible that, simply by chance, our study cohort contained no patient with HRS-2, although this would seem unlikely, given the size of our cohort. In either case, our work has several implications. First, we believe that HRS-2 is likely inadequately defined, and diagnostic criteria for this syndrome require reevaluation. Second, our data suggest that HRS-2 may not be as prevalent as previously reported. Finally, it is possible that HRS-2, as currently defined, is not a unique functional kidney disorder. This syndrome may simply be a reflection of previously undiagnosed intrinsic renal pathology13or simply because of prerenal AKI. Future studies are clearly needed to better define (and refine) diagnostic criteria of HRS-2.
We found that the presence of upper gastrointestinal bleeding was a significant independent predictor of hospital mortality in patients with kidney injury and cirrhosis. In-hospital mortality for those with upper gastrointestinal bleeding and kidney dysfunction was 46%, with median time to death from hospital admission of 21 days (95% CI, 7–24 days). Although this finding is consistent with other data suggesting that upper gastrointestinal bleeding is an important cause of morbidity and mortality in patients with cirrhosis, the mortality rate in patients with renal dysfunction and bleeding in our study seemed to be higher than that in those with bleeding alone.30–33For example, in a retrospective review of 403 patients with cirrhosis admitted for upper gastrointestinal bleeding, 6-week mortality was 24%.30Thus, our data suggest that upper gastrointestinal hemorrhage in the setting of kidney injury merits special clinical attention. In addition, the presence of bacteremia in the setting of renal dysfunction infers a particularly poor prognosis. This is consistent with recent literature showing a 31% mortality rate at 3 months for renal failure associated with bacterial infections.29
Utilization of the MELD score has become increasingly common for the assessment of prognosis in patients with advanced hepatic cirrhosis, particularly in those preparing for liver transplantation.34Besides its ability to accurately predict outcome in this population, MELD also is readily available in all settings and provides objective and reproducible information that is superior to other prognostic indices. It has been widely postulated that the reason for this superiority is due in large part to the inclusion of serum creatinine in the model, as it is now evident that renal function is an important prognostic indicator in patients with cirrhosis.34
Few studies have examined the value of MELD in predicting prognosis in hospitalized patients with renal dysfunction and cirrhosis.26–28Model for end-stage liver disease was originally developed in those undergoing the transjugular intrahepatic portosystemic shunt procedure35and then validated in hospitalized patients with cirrhosis without renal dysfunction.36In the current investigation, MELD score was a significant predictor of hospital mortality in patients with cirrhosis and renal dysfunction. This is not surprising because creatinine is an important component of MELD. However, it is likely that a better predictor of outcome in hospitalized patients with cirrhosis exists, as these patients typically have medical comorbidities that extend beyond renal function. For example, in the current investigation, both bacteremia and upper gastrointestinal hemorrhage are important prognostic factors in patients with renal injury (Fig. 3). We speculate that inclusion of such clinical factors would improve prediction models of outcome in hospitalized patients with cirrhosis.
It is important to recognize several limitations of this study. First, we used a serum creatinine concentration of 1.5 mg/dL or greater as a sign of renal dysfunction, although it is apparent that serum creatinine may not accurately reflect renal function in cirrhosis,6,13and many patients may have renal dysfunction at even lower creatinine levels. Thus, this cutoff for serum creatinine may have led us to underestimate the prevalence of renal disease, in particular CKD. Nonetheless, serum creatinine remains the most clinically useful marker available for assessing kidney function in cirrhosis. Another potential limitation of this investigation was the exclusion of 18 patients without complete clinical or laboratory data because they could not be adequately defined as AKI alone, AKI and CKD, or CKD. In particular, it is possible that one or more of these patients may have had HRS-2; however, we would emphasize that none carried a diagnosis of HRS-2 in their medical record. Additionally, the retrospective nature of this study may have confounded the precise determination of the etiologies of renal dysfunction and, theoretically, could be a source of selection bias. To mitigate this potential limitation, we used as precise definitions for renal dysfunction as possible and, moreover, excluded patients that were unable to be adequately categorized according to predetermined definitions. Finally, data collection for this investigation ended at the time of hospital discharge, and hence, there was no long-term follow-up for determination of long-term patient outcomes. A prospective study capturing real-time laboratory and clinical characteristics may be beneficial and, in addition, might allow a better understanding of the natural history of renal dysfunction in cirrhosis.
In summary, both acute and chronic renal dysfunction are common in hospitalized patients with cirrhosis, and in-hospital mortality rate for those admitted with elevated creatinine was greater than 30%. The development of AKI on preexisting CKD did not infer worse prognosis than AKI alone. Gastrointestinal bleeding or bacteremia in the setting of kidney injury carried an especially poor prognosis (as did the presence of HRS-1). Additionally, we were not able to definitively identify patients with HRS-2, raising questions about the uniqueness of this kidney disorder and/or suggesting that diagnostic criteria for this syndrome may require reevaluation. Finally, further investigation is clearly warranted to determine the clinical significance of CKD in patients with cirrhosis.