Association of AKI With mortality and complications in hospitalized patients with cirrhosis§


  • Justin M. Belcher,

    1. Program of Applied Translational Research, Yale University School of Medicine, New Haven, CT
    2. Section of Nephrology, Yale University School of Medicine, New Haven, CT
    3. Clinical Epidemiology Research Center, VAMC, West Haven, CT
    Search for more papers by this author
  • Guadalupe Garcia-Tsao,

    1. Clinical Epidemiology Research Center, VAMC, West Haven, CT
    2. Section of Digestive Diseases, Yale University School of Medicine, New Haven, CT
    3. VA-Connecticut Healthcare System, West Haven, CT
    Search for more papers by this author
  • Arun J. Sanyal,

    1. Division of Gastroenterology, Department of Internal Medicine, Virginia Commonwealth University School of Medicine, Richmond, VA
    Search for more papers by this author
  • Harjit Bhogal,

    1. Division of Gastroenterology, Department of Internal Medicine, Virginia Commonwealth University School of Medicine, Richmond, VA
    Search for more papers by this author
  • Joseph K. Lim,

    1. Section of Digestive Diseases, Yale University School of Medicine, New Haven, CT
    Search for more papers by this author
  • Naheed Ansari,

    1. Division of Nephrology, Department of Internal Medicine, Jacobi Medical Center, South Bronx, NY
    Search for more papers by this author
  • Steven G. Coca,

    1. Program of Applied Translational Research, Yale University School of Medicine, New Haven, CT
    2. Section of Nephrology, Yale University School of Medicine, New Haven, CT
    3. Clinical Epidemiology Research Center, VAMC, West Haven, CT
    Search for more papers by this author
  • Chirag R. Parikh,

    Corresponding author
    1. Program of Applied Translational Research, Yale University School of Medicine, New Haven, CT
    2. Section of Nephrology, Yale University School of Medicine, New Haven, CT
    3. Clinical Epidemiology Research Center, VAMC, West Haven, CT
    • Section of Nephrology, Yale University and VAMC, 950 Campbell Avenue, Mail Code 151B, Building 35A, Room 219, West Haven, CT 06515
    Search for more papers by this author
    • fax: 203-937-4932

  • for the TRIBE-AKI Consortium

    1. Program of Applied Translational Research, Yale University School of Medicine, New Haven, CT
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

  • Supported by a grant from the National Institutes of Health (NIH) R21-DK078714 to Dr. Parikh. Dr. Belcher was supported by an institutional fellowship training grant from NIH.

  • §

    Collaborators: Isabel Butrymowicz, Aldo J. Peixoto, Mark A. Perazella.

  • See Editorial on Page 435


Acute kidney injury (AKI) is a common and devastating complication in patients with cirrhosis. However, the definitions of AKI employed in studies involving patients with cirrhosis have not been standardized, lack sensitivity, and are often limited to narrow clinical settings. We conducted a multicenter, prospective observational cohort study of patients with cirrhosis and AKI, drawn from multiple hospital wards, utilizing the modern acute kidney injury network (AKIN) definition and assessed the association between AKI severity and progression with in-hospital mortality. Of the 192 patients who were enrolled and included in the study, 85 (44%) progressed to a higher AKIN stage after initially fulfilling AKI criteria. Patients achieved a peak severity of AKIN stage 1, 26%, stage 2, 24%, and stage 3, 49%. The incidence of mortality, general medical events (bacteremia, pneumonia, urinary tract infection), and cirrhosis-specific complications (ascites, encephalopathy, spontaneous bacterial peritonitis) increased with severity of AKI. Progression was significantly more common and peak AKI stage higher in nonsurvivors than survivors (P < 0.0001). After adjusting for baseline renal function, demographics, and critical hospital- and cirrhosis-associated variables, progression of AKI was independently associated with mortality (adjusted odds ratio = 3.8, 95% confidence interval 1.3-11.1). Conclusion: AKI, as defined by AKIN criteria, in patients with cirrhosis is frequently progressive and severe and is independently associated with mortality in a stage-dependent fashion. Methods for earlier diagnosis of AKI and its progression may result in improved outcomes by facilitating targeted and timely treatment of AKI. (HEPATOLOGY 2013)

Acute kidney injury (AKI) is one of the most severe complications of cirrhosis and portends an ominous prognosis.1 The development of AKI is often linked with the onset of other complications of cirrhosis such as variceal bleeding and spontaneous bacterial peritonitis (SBP) and occurs in up to 19% of hospitalized patients with cirrhosis.2 Although the hepatorenal syndrome (HRS) has long been associated with prodigious mortality,3 more recent recognition of the general hazard associated with AKI in cirrhosis has led to the incorporation of serum creatinine as one of the three variables comprising the model of endstage liver disease (MELD) score. This model has met with marked success in predicting short-term mortality and is used to determine allocation priority for orthotropic liver transplantation.4

Although recognition of the primacy of AKI in determining outcomes in hospitalized patients with cirrhosis has been a welcome advance, studies attempting to quantify and expand on AKI's impact have been limited by several flaws. Most significantly, studies of AKI in cirrhosis suffer from a lack of standardization in AKI definitions. Moreover, they have often used elevated creatinine thresholds that are outdated and lack sensitivity.5-8 Creatinine, a suboptimal marker for renal function under any circumstance, is especially insensitive to a decline in glomerular filtration rate (GFR) in the setting of cirrhosis.9 The reliance on elevated thresholds leads to an overselection of the most severe cases, limiting the ability to evaluate factors associated with disease progression and the bearing of AKI severity on outcomes. Many studies examining the impact of AKI in cirrhosis have treated the presence of AKI as a dichotomous variable and assessed outcomes relative to patients with stable kidney function. Although much attention has been devoted to eliciting risk factors for the development of AKI, few studies have explored variables associated with disease progression and outcomes only in the subset of patients with AKI.10 Those few studies that have employed more current definitions such as the acute kidney injury network (AKIN) or the risk, injury, failure, loss, and endstage kidney disease (RIFLE) criteria have primarily focused on patients admitted to an intensive care unit (ICU) and are thus not generalizable to all hospitalized patients.11, 12 In addition, rigorous attempts to establish an accurate baseline creatinine based on outpatient values were not performed. Finally, due to enrollment difficulties inherent to this extremely ill population, studies of AKI in cirrhosis have often been retrospective or restricted to small numbers of patients.

The aim of this study was to prospectively assess a large cohort of hospitalized cirrhosis patients with AKI to understand the natural history, trajectory, and recovery patterns of the disease. Specifically, we sought to evaluate the impact of AKI severity and disease progression on in-hospital death. The AKIN criteria for the diagnosis of AKI were used to detect a decline in renal function at the earliest possible instance.13


ADQI, Acute Dialysis Quality Initiative; AKI, acute kidney injury; AKIN, acute kidney injury network; AUC, area under the curve; CKD, chronic kidney disease; GFR, glomerular filtration rate; HRS, hepatorenal syndrome; IAC, International Ascites Club; ICU, intensive care unit; IQR, interquartile range; MELD, model of endstage liver disease; OR, odds ratio; RIFLE, risk, injury, failure, loss and endstage kidney disease; SBP, spontaneous bacterial peritonitis; UTI, urinary tract infection.

Patients and Methods

Study Design

This prospective, multicenter observational cohort study was carried out at four tertiary care academic centers in the U.S. Potential participants were identified by a daily screening of patients on medical ICUs, transplant floors, and on each hospital's hepatology service. Laboratory tests of all patients with cirrhosis were reviewed daily for the presence of AKI (see “Variables”). Patients were eligible if they presented for admission with AKI or developed it during the course of hospitalization. Inclusion criteria included a known diagnosis of cirrhosis (see “Variables”), age ≥18 years, presence of AKI, and the availability of a documented serum creatinine within 1 year prior to AKI. Exclusion criteria included prior kidney or liver transplant, advanced chronic kidney disease (baseline creatinine >4.0 mg/dL), acute or chronic renal replacement therapy at the time of enrollment, estimated life expectancy less than 3 days, confirmed pregnancy, other known causes of renal insufficiency such as glomerulonephritis or hydronephrosis, and previous participation in the study. If a patient was unable to provide consent, a surrogate decision maker was sought. All patients were enrolled within 5 days of meeting AKI criteria. The study was approved by the Institutional Review Board or Human Investigations Committee at each institution.


Independent Variables


Patients were eligible with an existing diagnosis of cirrhosis obtained from medical records and which was based on liver biopsy, when available, or on a combination of clinical, biochemical, ultrasonographic, and endoscopic findings.


The AKIN criteria (Table 1) were applied for the diagnosis of AKI. As urine collection and output documentation can be inconsistent, only the definition of an increase in serum creatinine of 0.3 mg/dL or a 50% rise from baseline was used. Renal failure in the setting of cirrhosis has previously been defined as a serum creatinine greater than 1.5 mg/dL.6-8 However, in light of recent evidence that much smaller decrements in renal function are associated with adverse outcomes,14 a working group composed of members of the International Ascites Club (IAC) and the Acute Dialysis Quality Initiative (ADQI) have proposed employing the AKIN definition in the setting of cirrhosis.15 Patients were considered to have worsening of AKI if they progressed to a higher AKIN stage or, if they presented in stage 3, if they subsequently required renal replacement therapy. Death was not considered to represent progression of AKI.

Table 1. Classification/Staging System for Acute Kidney Injury According to AKIN13
AKI StageSerum Creatinine CriteriaUrine Output Criteria
  1. AKIN, acute kidney injury network; AKI, acute kidney injury.

AKI Stage 1Increase in serum creatinine ≥ 0.3 mg/dL or increase to ≥ 150-200% from baselineUrine output < 0.5 ml/kg/hr for > 6 hr
AKI Stage 2Increase of serum creatinine to > 200-300% from baselineUrine output < 0.5 ml/kg/hr for > 12hr
AKI Stage 3Increase of serum creatinine to > 300% from baseline or serum creatinine ≥ 4.0 mg/dL after a rise of at least 0.5 mg/dL or treatment with renal replacement therapyUrine output < 0.3 mL/kg/hr for 24hr or anuria for 12 hr
Baseline creatinine

Baseline creatinine was defined as the most recent stable measurement prior to admission for the index hospitalization. When possible, outpatient measurements were selected, although values were also used from previous admissions not complicated by AKI. In rare cases, patients without an outpatient measurement were included in the analytic cohort if, prior to a rise in creatinine fulfilling the above definition of AKI, they manifested at least 5 days of stable values within the normal creatinine range following admission. In these instances, the creatinine at admission was considered the baseline.

Other variables

A preexisting decrease in GFR was defined as a GFR <90 mL/min as calculated with the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation using the baseline creatinine value.16 We have avoided using the term “chronic kidney disease,” as this classically implies structural damage. Many patients with cirrhosis have a chronically depressed GFR due instead to persistent hypoperfusion and their renal function may thus be partially reconstituted with restitution of perfusion. As the granular data required to distinguish true structural kidney injury, such as chronically active urine sediment or proteinuria, from a preexisting decrease in GFR due to hypoperfusion was not consistently available on patients prior to admission, we have instead termed such baseline dysfunction as chronic renal impairment. When controlling for baseline renal dysfunction in our model predicting in-hospital mortality, GFR >90 mL/min was used as the reference range. Baseline proteinuria was defined as an outpatient value of 1+ or greater on dipstick or 30 mg/dL when quantitated by clinical laboratory. HRS therapy refers to the use of midodrine and octreotide. Although such therapy is often paired with albumin, albumin use was nearly ubiquitous in our cohort and thus not considered indicative of dedicated therapy for HRS. Urinary tract infections (UTIs) and bacteremia were defined by positive cultures. The diagnosis of pneumonia required either a positive sputum culture or findings on radiography. SBP was defined by a fluid polymorphonuclear leukocyte count ≥250 cells/mm3. The presence of hepatic encephalopathy was determined by clinical diagnosis reflected in the patient's medical chart. MELD and Child-Pugh scores were calculated on the day of first sample collection.


Our primary outcome was in-hospital mortality during the index hospitalization.


Categorical variables were expressed as proportions and compared using the chi-square and Fisher's Exact test, as appropriate. Normally or near-normally distributed variables were reported as means with standard deviations (SDs) and compared by Student's t test. Nonnormally distributed continuous variables were reported as medians with interquartile ranges (IQR) and compared by the Kruskal-Wallis test. Normality was assessed using the Kolmogorov-Smirnov test. Variables thought likely to associate with the primary outcomes were assessed by univariate analysis. Those found to have a P-value of less than 0.2 on univariate analysis were included in a multiple logistic regression model evaluating the impact of worsening of AKI on mortality. Stepwise selection of variables was performed to build the model. Model accuracy was assessed by calculating the area under the receiver operating characteristic curve (AUC) and goodness-of-fit verified with the Hosmer-Lemeshow test. A two-sided P < 0.05 was considered significant for all analysis. Statistical analysis was performed using SAS, v. 9.2 (SAS Institute, Cary, NC).


Cohort Characteristics.

A total of 219 patients with cirrhosis and AKI were enrolled into the study over a period of 29 months. Twenty-seven patients were subsequently excluded, with reasons being an excessive interval between the onset of AKI and time of first sample collection (n = 15), lack of documented baseline creatinine level (n = 4), recent treatment with nephrotoxins (n = 3), a diagnosis of acute hepatitis rather than cirrhosis (n = 2), and other causes (n = 3). Baseline demographic, clinical, and laboratory data of the 192 patients included in the study are shown in Table 2. The mean patient age was 55.1 ± 9.3 and 136 (71%) were male. Fifty (26%) patients died during their hospitalization. The primary etiologies of cirrhosis were alcohol (29%), alcohol and hepatitis C virus (HCV) (27%), and HCV alone (17%). There was no difference in etiologies between survivors and nonsurvivors. The majority of patients had previously suffered complications of cirrhosis including ascites, 76%, hepatic encephalopathy, 67%, variceal bleeding, 23%, and SBP, 16%. Reasons for admission were similar between the two groups. The median Child-Pugh score was 10.5 and MELD 26.3 at the time of enrollment. Not unexpectedly, both Child-Pugh (12 versus 10, P < 0.0001) and MELD scores (34.1 versus 23.6, P < 0.0001) were higher in nonsurvivors than in survivors. However, there was no difference in median serum sodium levels or presence of hyponatremia at the time of enrollment.

Table 2. Baseline and Clinical Characteristics of All Patients and Those for Nonsurvivors and Survivors
 Total N=192Nonsurvivors N=50Survivors N=142P
  • N, number; SD, standard deviation; BMI, body mass index; IQR, inter-quartile range; HCV, hepatitis C virus; NASH, nonalcoholic steatohepatitis; SBP, spontaneous bacterial peritonitis; AKI, acute kidney injury; GI, gastrointestinal; MELD, model for endstage liver disease.

  • a

    Child-Pugh Class is at time of enrollment.

  • Jonckheere-Terpstra trend test.

  • Serum sodium <130 mEq/L.

Age in years - mean ± SD55.1 ± 9.354 ± 8.955.5 ± 9.50.35
Male sex – n (%)136 (71)35 (70)101 (71)0.88
BMI – median (IQR)31.3 ± 8.932 (26.5-34.6)31 (25-35.2)0.28
Race – n (%)    
 White137 (71)27 (54)110 (77)0.002
 Black27 (14)10 (20)17 (12)0.16
 Hispanic24 (13)11 (22)13 (9)0.02
Diabetes – n (%)51 (27)11 (22)40 (28)0.40
Active Cancer – n (%)21 (11)6 (12)15 (11)0.78
Cirrhosis etiology – n (%)    
 Alcohol56 (29)16 (32)40 (28)0.61
 Alcohol and HCV52 (27)13 (26)39 (27)0.84
 HCV33 (17)9 (18)24 (17)0.86
 NASH17 (9)2 (4)15 (11)0.25
 Cryptogenic12 (6)2 (4)10 (7)0.73
 Autoimmune11 (6)5 (10)6 (4)0.16
 Other11 (6)4 (8)7 (5)0.48
Previous complications of cirrhosis – n (%)    
 Ascites146 (76)37 (74)109 (77)0.63
 Hepatic encephalopathy129 (67)35 (70)94 (66)0.62
 Variceal bleed45 (23)9 (18)36 (25)0.29
 SBP31 (16)10 (20)21 (15)0.33
Reason for admission – n (%)    
 Hepatic encephalopathy52 (27)13 (26)39 (27)0.84
 Refractory ascites/edema23 (12)5 (10)18 (13)0.62
 AKI22 (11)3 (6)19 (13)0.95
 GI bleed15 (8)4 (8)11 (8)0.42
 Abdominal pain14 (7)4 (8)10 (7)0.76
 Jaundice10 (5)4 (8)6 (4)0.29
 Transplant work-up6 (3)1 (2)5 (4)1
 SBP6 (3)3 (6)3 (2)0.18
 Infection other than SBP6 (3)5 (10)1 (1)0.005
 Other38 (20)10 (20)28 (20)1
Child-Pugh Classa - n (%)    
 A4 (2)04 (3)<0.0001
 B60 (31)4 (8)56 (39) 
 C125 (65)46 (92)79 (56) 
Child-Pugh score – median (IQR)10.5 (9-12)12 (11-13)10 (8-11)<0.0001
MELD score – mean ± SD26.3 ± 9.534.1 ± 8.623.6 ± 8.2<0.0001
Bilirubin – median (IQR)4.1 (1.8-10.6)12.7 (6.3-23)3 (1.6-5.8)<0.0001
INR – median (IQR)1.7 (1.3-2.2)2.2 (1.7-2.6)1.5 (1.3-1.9)<0.0001
Sodium – mean ± SD133 ± 6.4134 ± 8133 ± 60.57
Hyponatremia at enrollment – n (%)64 (33)16 (32)48 (34)0.82
Length of hospitalization – median (IQR)12 (6-19)15 (8-40)10 (6-17)0.008

Kidney Variables and Mortality.

The impact of renal variables on survival is shown in Table 3. A majority of patients, 91%, had a documented outpatient creatinine, whereas 17 (9%) had creatinine values from their admission used as a baseline level. Overall, 119 (62%) patients had evidence of a preexisting decrease in GFR <90 mL/min, whereas 53 (28%) had a baseline GFR <60 mL/min. Chronic renal impairment was more prevalent and median GFR was lower (73 mL/min versus 91 mL/min, P = 0.048) in survivors than nonsurvivors. Proteinuria was present at baseline in 12% of patients and did not differ between the two groups. Remarkably, AKI was present in 116 (60%) patients at admission, whereas an additional 17 (9%) developed AKI within 48 hours of hospitalization. The remaining 59 (31%) patients experienced AKI later in the course of their hospital stay at a median of 7 days postadmission, IQR 4-10 days. Mortality was significantly higher in those patients who developed AKI subsequent to admission than in those who presented with AKI, 36 versus 21%, respectively (P = 0.01). At the time of first fulfilling AKIN criteria, 48% of patients had stage 1 AKI, 29% stage 2, and 23% stage 3. A decrease in serum creatinine occurred in 70 (37%) patients within 48 hours of first meeting AKIN criteria. Such early evidence of improvement was significantly more common in survivors than in nonsurvivors, 59 (42%) versus 11 (22%), respectively (P = 0.01). Conversely, the severity of AKI worsened following the initial fulfillment of AKIN criteria in 85 (44%) of patients. Progression of AKI was significantly more common among those patients who developed AKI in the hospital, 59%, than in those who presented already experiencing AKI, 35% (P = 0.001). Critically, worsening of AKI was markedly more common among nonsurvivors, 80%, than among survivors, 32% (P < 0.0001). A strong stepwise association was noted between degree of progression and mortality (Fig. 1).

Table 3. Renal Variables and Associations with Survival
 Total N=192Nonsurvivors N=50Survivors N=142P
  • N, number; eGFR, estimated glomerular filtration rate; IQR, inter-quartile range; CKI, chronic kidney impairment; CVVH, continuous venovenous hemofiltration; HD, hemodialysis; AKIN, acute kidney injury network; AKI, acute kidney injury; CKI, chronic kidney impairment.

  • a

    eGFR at baseline by CKD-EPI equation: GFR = 141 × min(Scr/κ, 1)α × max(Scr/κ, 1)-1.209 × 0.993Age × 1.018 [if female] or 1.159 [if black], where Scr is serum creatinine, κ is 0.7 for females and 0.9 for males, α is -0.329 for females and -0.411 for males, min indicates the minimum of Scr/κ or 1, and max indicates the maximum of Scr/κ or 1.

  • Microalbuminuria (30mg/dL) or greater on dipstick or quantitative measurement prior to admission.

  • Jonckheere-Terpstra trend test.

Baseline eGFR – median (IQR)a76 (58-101)91 (60-110)73 (56-98)0.048
CKI stages    
 GFR 60-89 ml/min/m266 (35)12 (24)54 (39)a0.04
 GFR 30-59 ml/min/m246 (24)11 (22)35 (25) 
 GFR 29-15 ml/min/m27 (4)1 (2)6 (4) 
Proteinuria – n (%)23 (12)5 (10)18 (13)0.62
Creatinine on admission – median (IQR)1.8 (1.2-2.55)1.6 (1-2.5)1.9 (1.3-2.6)0.13
Creatinine at enrollment – median (IQR)2.2 (1.6-3.4)3 (2.1-3.7)2 (1.4-3.1)0.0008
Peak creatinine – median (IQR)2.7 (1.9-4.2)3.8 (2.7-5.2)2.4 (1.8-3.9)<0.0001
Timing of AKI relative to admission    
 Outpatient116 (60)23 (46)93 (66)0.01
 Inpatient76 (40)27 (54)49 (34) 
Any creatinine decrease within 48 hrs    
 Yes70 (37)11 (22)59 (42)0.01
 No120 (63)39 (78)81 (58) 
AKIN stage at first meeting criteria    
 191 (48)20 (40)71 (51)0.06
 256 (29)13 (26)43 (31) 
 343 (23)17 (34)26 (19) 
AKIN stage progressed    
 Yes85 (44)40 (80)45 (32)<0.0001
 No107 (56)10 (20)97 (68) 
Peak AKIN stage    
 150 (26)1 (2)49 (35)<0.0001
 247 (24)7 (14)40 (28) 
 395 (49)42 (84)53 (37) 
Dialysis – n (%)46 (24)29 (58)17 (12)<0.0001
 CVVH21 (46)18 (36)3 (2)<0.0001
 HD12 (26)5 (10)7 (5)0.2
 Both13 (28)6 (12)7 (5)0.09
Figure 1.

Degree of AKI progression and mortality. Progression is defined by an increase in AKIN stage after initially fulfilling AKIN criteria. Progression to dialysis refers to any patient who presented as nondialysis-dependent but subsequently developed the requirement for dialysis.

AKI in the setting of cirrhosis was ultimately severe with peak AKIN stages 1, 2, and 3 attained in 26%, 24%, and 49% of patients, respectively. Mortality increased in a stage-response manner with severity of AKI. The likelihood and degree of progression, along with subsequent mortality, is presented by initial AKIN stage in Fig. 2. For patients with peak stages of 2 and 3, those who progressed to that degree had higher mortality than those who presented with that level of dysfunction but did not progress (Fig. 3). Remarkably, patients with a peak severity of AKIN stage 1 did extremely well, with only 1 (2%) death. Nonsurvivors ultimately experienced significantly more severe AKI, with 84% reaching a peak of stage 3 versus 38% of survivors (P < 0.0001). Dialysis was required for 46 (24%) patients and was used more frequently among nonsurvivors, 58%, than among survivors, 12%, (P < 0.0001). Of those patients requiring dialysis, 57% died during the index hospitalization still requiring renal replacement therapy, 17% were discharged on dialysis, and 26% recovered renal function by the time of discharge.

Figure 2.

Incidence and extent of AKI progression and subsequent mortality by initial AKIN stage. Patients were categorized by their stage upon first meeting AKIN criteria. Progression refers to worsening to a higher AKIN stage, with patients who are initially in stage 3 by creatinine criteria but not requiring dialysis counted as progressing if dialysis was subsequently initiated. *Three patients were initiated on dialysis on the day of enrollment.

Figure 3.

Impact of AKI progression versus peak severity on mortality. Patients are categorized by the peak AKIN stage they attained. “Progressors” refers to patients who worsened from their initial stage to achieve this peak, whereas “Non-progressors” were already in their peak stage at the time of first meeting AKIN criteria.

Multivariate logistic regression was employed to evaluate the independent association between worsening of AKI and death. On univariate analysis, progression of AKI was associated with death with an odds ratio (OR) of 8.62 (95% confidence interval [CI] 3.96-18.77, P < 0.0001). After adjustment for baseline renal function, demographics, hospital events, and variables related to severity of cirrhosis, the adjusted OR was attenuated but remained strongly significant, OR 3.8 (95% CI 1.31-11.08). On ROC curve analysis, worsening of AKI alone was able to predict death with an AUC of 0.74.

Other Complications.

The associations between severity of AKI and general medical and cirrhosis specific hospital complications are listed in Table 4. The rate of both general medical and cirrhosis-specific complications was higher with worsening severity of AKI. HRS-specific therapy was often employed, with 45% of patients receiving midodrine and 46% octreotide, whereas albumin use, 82%, was nearly ubiquitous. The use of midodrine, octreotide, and albumin increased significantly with severity of AKI. Patients with higher-stage AKI were more likely to be admitted to ICUs and less likely to be transferred out alive. The use of mechanical ventilation and vasopressors was higher with worsening peak AKIN stage. Of patients surviving to discharge, median length of hospital stay increased with AKIN stage from 9 to 10 to 14 days, respectively (P = 0.01).

Table 4. Hospital Events by Severity of AKI
 Total N=192AKIN Stage 1 N=50AKIN Stage 2 N=47AKIN Stage 3 N=95P
  1. N, number; ICU, intensive care unit; SICU, surgical intensive care unit; CCU, cardiac care unit; GI, gastrointestinal, UTI, urinary tract infection; SBP, spontaneous bacterial peritonitis; HRS, hepatorenal syndrome; IQR, interquartile range.

Medical complications     
 Bacteremia – n (%)34 (18)5 (10)4 (9)25 (26)0.005
 GI bleed – n (%)45 (24)10 (20)10 (21)25 (27)0.34
 Pneumonia – n (%)35 (18)6 (12)6 (13)23 (24)0.04
 UTI – n (%)55 (29)7 (14)8 (17)40 (42)<0.0001
Hepatic complications     
 Ascites160 (83)34 (68)39 (83)87 (92)0.0004
 Hepatic encephalopathy120 (63)23 (46)32 (68)65 (68)0.02
 SBP38 (20)1 (2)11 (23)26 (27)0.0008
 Variceal bleed16 (8)4 (8)3 (6)9 (9)0.66
HRS specific therapy     
 Albumin158 (82)33 (66)40 (85)85 (89)0.001
 Midodrine86 (45)11 (22)15 (32)60 (63)<0.0001
 Octreotide89 (46)8 (16)18 (38)63 (66)<0.0001
Admitted to ICU – n (%)94 (50)15 (30)17 (36)62 (65)<0.0001
 Transferred out alive – n (%)60 (64)14 (93)15 (88)31 (50)0.002
 Survived to discharge – n (%)47 (50)14 (93)11 (65)22 (35)<0.0001
Mechanical ventilation62 (32)6 (12)8 (17)48 (51)<0.0001
Vasopressor usage49 (26)3 (6)6 (13)40 (42)<0.0001
Days from admission to discharge – median (IQR)10 (6-17)9 (6-13)10 (6-17)14 (8-35)0.01


The development of AKI in the setting of cirrhosis has long been recognized to confer a grim prognosis and is known to be independently predictive of death in patients with SBP and variceal hemorrhage.8, 17 Unfortunately, estimates of the incidence of AKI in cirrhosis and attempts to quantify AKI's impact on mortality have suffered from a lack of standardization in the definition of AKI. Utilizing markedly elevated creatinine thresholds ranging from 1.56-8 to 3.518 mg/dL, AKI in cirrhosis has been associated with a striking mortality of 55%-91%. However, such stringent cutoffs ensure selection bias wherein only the most severe cases of AKI qualify. The lack of sensitivity inherent in these AKI definitions is particularly problematic in patients with cirrhosis where significant muscle atrophy and reduced hepatic conversion of creatine to creatinine results in potentially significant renal dysfunction being masked by an ostensibly normal creatinine value. This danger is compounded by cirrhosis patients' unique vulnerability to AKI. In addition to inducing what is functionally a state of constant diminished renal bloodflow,19 progression of cirrhosis is associated with a loss of ability to maintain renal perfusion by way of tubuloglomerular feedback.20, 21 In this setting, frequent volume shifts accompanying titration of lactulose and altered oral intake due to encephalopathy will precipitate numerous episodes of AKI not captured by such rigid definitions.

Several recent studies have attempted to rectify this shortcoming by investigating the impact of AKI on mortality in the setting of cirrhosis using the modern RIFLE criteria, whose stages of “R,” “I,” and “F” are analogous to AKIN stages 1, 2, and 3.22 Jenq et al.11 studied 134 patients with cirrhosis admitted to the ICU and found a mortality of 32.1% in those without AKI, 68.8% for RIFLE-R, 71.4% for RIFLE-I, and 94.8% for RIFLE-F. AKI was diagnosed based on creatinine at the time of ICU admission and the association of mortality with peak RIFLE stage or AKI progression was not assessed. Cholongitas et al.12 followed a large cohort of 412 cirrhosis patients also admitted to the ICU, evaluating the impact of AKI on mortality during ICU stay or within 6 weeks of unit discharge. The authors noted a similar stage-dependent association between AKI and mortality, with rates increasing from 42.5% in those without AKI to 71% for RIFLE-R and 88% for RIFLE-I/F. The significant increase in mortality in those patients with only mild AKI (RIFLE-R) speaks to the value of these sensitive criteria for ICU prognosis. However, both studies only included patients in the ICU, where AKI is often associated with multisystem organ failure and severe sepsis. Carvalho et al.23 studied 91 patients with cirrhosis and AKI by AKIN criteria at hospital admission including 83 with stage 1, 5 stage 2, and 3 with stage 3. Patients were staged by comparing creatinine values drawn within 48 hours of admission. Any patient with a change of 0.3 mg/dL or greater, in either direction, was classified as having AKI. The magnitude of this change determined the AKIN stage, as no baseline values were considered and no assessment was made of progression or peak stage. Presence of AKI conferred an overall OR of 2.6 for hospital mortality but quantifying the risk by stage was limited by the small number of patients with more advanced disease.

In our study we investigated the impact of AKI, using the AKIN definition, on mortality in hospitalized patients with cirrhosis, independent of admission ward (Table 5). The overall mortality was 26%, significantly lower than in studies using less sensitive definitions or confined ICUs. However, a pronounced stage-dependent response was again seen, with mortality for peak AKIN stages 1, 2, and 3 of 2%, 15%, and 44%, respectively. Similarly, advancing stages of AKI were associated with a higher incidence of medical complications, including bacteremia, pneumonia, and UTI, and with cirrhosis-specific complications such as ascites, encephalopathy, and SBP. Paradoxically, the presence of a lower baseline GFR conferred a survival advantage. Although the CKD-EPI equation correlates best with measured GFR,24 it can significantly overestimate renal function in patients with cirrhosis.25 It may be that more advanced cirrhotics with lower muscle mass and decreased hepatic creatinine production were falsely estimated to have higher GFRs. Additionally, chronic kidney disease is a strong risk factor for AKI and may be underappreciated in cirrhosis.7, 26 It is possible that the development of AKI without preexisting renal dysfunction requires a stronger renal insult and greater systemic illness, thus placing this group at higher risk for death from nonrenal causes.

Table 5. Association of AKIN Stage Progression and In-Hospital Mortality Adjusted for Multiple Variables
ModelAdjusted Odds Ratio95% CIP
  • a

    GFR >90 mL/min taken as no CKD and serves as reference.

  • Race, age and sex.

  • Pressor use, UTI, pneumonia, GI bleed, bacteremia.

  • §

    Albumin therapy, HRS therapy, HE, SBP, MELD.

  • AKIN, acute kidney injury network; CI, confidence interval; CKD, chronic kidney disease; GFR, glomerular filtration rate; UTI, urinary tract infection; GI, gastrointestinal; HRS, hepatorenal syndrome; HE; hepatic encephalopathy; SBP, spontaneous bacterial peritonitis; MELD model for endstage liver disease.

AKIN stage progression8.623.96-18.77<0.0001
AKIN stage progression + CKD stagea10.044.43-22.74<0.0001
AKIN stage progression + CKD stage + demographics9.564.23-21.61<0.0001
AKIN stage progression + CKD stage + demographics + hospital events7.132.42-21.060.0004
AKIN stage progression + CKD stage + demographics + hospital events + cirrhosis variables§3.801.31-11.080.01

Although AKI is typically thought an inpatient syndrome, the majority (60%) of our patients presented to the hospital already in AKI. However, nearly half of the patients (48%) were still in stage 1 AKI at the time of first meeting AKIN criteria. A critical result, then, of our study is the identification of progression of AKI as a powerful independent risk factor for mortality. Indeed, as seen in Figs. 2 and 3, assessment of AKI progression has the potential to add granularity to the association between AKI severity and mortality. The accuracy of AKI progression for predicting death, evidenced by an AUC of 0.74, is remarkable given the high mortality in the cohort. The striking, nearly 4-fold, increase in mortality in those patients whose AKI progressed, and contrasting decrease in mortality in those who showed early improvement, is vital due of the presence, rare in AKI, of disease-specific therapies in the setting of cirrhosis. The etiology of AKI in cirrhosis is estimated to be 68% hypoperfusion and 32% intrarenal, primarily acute tubular necrosis (ATN).2 Rapid and aggressive intervention early in the course of AKI to optimize volume status and restore renal perfusion may prevent progression of AKI and subsequent development of ATN. Those patients unresponsive to volume and lacking evidence of frank kidney injury have classically been diagnosed with HRS, long the most dreaded of cirrhosis complications.

HRS, however, is undergoing a revolution, as improved understanding of its physiology has facilitated targeted treatments.28 Terlipressin, a nonselective V1 vasopressin agonist, has been successfully employed along with albumin to mitigate splanchnic and systemic vasodilatation and restore effective circulating volume and renal perfusion in patients with HRS.29, 30 Critically, application of terlipressin improves renal hemodynamics and GFR even in patients who do not yet meet HRS diagnostic criteria.31 However, the use of such early interventions has until recently been hampered by a consensus that the diagnosis of AKI in cirrhosis requires a serum creatinine of at least 1.5 mg/dL.5, 32 The risk inherent in such stringent criteria is evidenced by recent data demonstrating that response to terlipressin declines with increasing creatinine at treatment initiation.33

Seeking to modernize this perception, a recent working group comprised of members of the IAC and ADQI proposed adopting the AKIN criteria within the spectrum of what they term “hepatorenal dysfunction” for all acute deteriorations in renal function in cirrhosis patients, irrespective of etiology.15 However, for HRS Type 1, distinguished as a specific form of AKI, the new proposal retains a diagnostic creatinine threshold of 2.5 mg/dL. Given the well-known limitations of serum creatinine as an accurate marker of renal function in patients with cirrhosis,34 patients must suffer a marked decrease in GFR before their creatinine rises to this level. With the availability of effective interventions and our demonstration of poor outcomes associated with worsening, we believe that this threshold should be lowered. When there is no clear evidence of ATN or other intrinsic disease, vasoconstrictor therapy should be initiated in all patients when they progress to a higher stage of AKI. The potential impact of this approach is apparent in our study, where 56 patients had a creatinine <2.5 mg/dL upon first meeting AKIN criteria but ultimately rose to >2.5. Of these, 31 (55%) progressed to a higher AKIN stage prior to reaching 2.5 and thus would have been treated sooner under this strategy. The effectiveness of such an approach could be studied in a trial enrolling patients with AKI and fulfilling the IAC criteria of ascites, lack of response to 48 hours of volume resuscitation, and absence of shock, nephrotoxic exposure, or evidence of structural injury but without a creatinine cutoff. Patients would be randomized to receive vasoconstrictors either upon first stage progression or when creatine reaches 2.5.

More fundamentally, there is no reason to think that, even under the threshold of 2.5, higher creatinine at the initiation of vasoconstrictor therapy will not associate with decreased response rates. HRS must be conceptualized as the terminal end of a physiologic spectrum; guidelines for treatment should hinge upon phenotyping patients on this spectrum and need not invoke a given degree of renal dysfunction. Ideally, patients at high risk of progression would be treated immediately upon meeting criteria for AKI. We have shown development of AKI as an inpatient to be a risk factor for progression but prognosis could be enhanced further through novel biomarkers capable of distinguishing structural from functional AKI etiologies and, if functional, quantifying the intensity of renal vasoconstriction. Ultimately, trials of vasoconstrictors could be optimized by using such biomarkers as entrance criteria, thereby selecting only those patients with predominately functional disease.

Our study has several important strengths. We prospectively enrolled one of the largest cohorts of patients with cirrhosis and AKI in the literature. Baseline creatinine levels were rigorously assessed, with >90% of patients' determined by stable values drawn within a year prior to admission. The critical importance of this approach toward ascertaining baseline is underscored by 60% of our patients presenting with AKI, where the use of admission creatinine as the baseline value would have obscured the severity or even the presence of AKI. In using the sensitive AKIN definition, we included patients with a broad spectrum of AKI severity. Patients were enrolled throughout the hospital, including those who presented with AKI and those who subsequently developed it, rendering our findings regarding the impact of AKI severity and progression on mortality broadly generalizable. However, our study is not without limitations. As an observational study, we were unable to assess the impact of volume expansion and HRS-specific therapy on AKI progression. The onset of AKI in cirrhotics appears to be frequently in the outpatient setting but we were unable to evaluate patterns of progression and recovery prior to admission.

In conclusion, the results of the current study confirm that AKI, as defined by AKIN criteria, is associated with in-hospital mortality in the setting of cirrhosis in a stage-dependent manner. Although those patients who exhibit early recovery from AKI do well, worsening of AKI is independently associated with mortality. Further studies are required to investigate the implementation of more sensitive criteria for AKI and the development of earlier and more discriminating diagnostic tests. It is possible that early recognition of AKI and prompt, aggressive treatment to mitigate disease progression may improve outcomes in this complex clinical setting.