Background Hepatorenal syndrome in cirrhosis with ascites is a well-defined entity with significant morbidity and mortality. It is unclear whether milder degrees of acute kidney injury (AKI), defined as a serum creatinine increase of over 26.4 μmol/l (0.3 mg/dl) or by 50% from baseline, also has a negative impact on patient outcomes.
Objectives To determine the prevalence of AKI in cirrhosis with ascites and the impact of AKI on patient outcomes.
Design Patients with cirrhosis with ascites and baseline serum creatinine less than 110 μmol/l, and no evidence of structural renal disease, prospectively underwent 4–6-weekly blood work-up for full blood count, biochemistry and liver function. Clinical assessments occurred every 4 months for the development of AKI and other complications.
Results 90 patients (mean age 55.8±0.8 years) with a mean follow-up of 14.05±1.07 months were enrolled. 82 episodes of AKI occurred in 49 patients, with the majority of episodes precipitated by a disturbance in systemic haemodynamics. The mean peak serum creatinine of the AKI episodes was within the laboratory's normal range. 73 episodes of AKI resolved; nine did not. There was no clear clinical predictor for the development or resolution of AKI. Despite resolution of most AKI episodes, a gradual and significant increase in serum creatinine and a gradual reduction in mean arterial pressure were observed during follow-up, associated with a significant reduction in survival compared with non-AKI patients.
Conclusion Minor increases in serum creatinine are clinically relevant and can adversely affect survival. Every effort should be made to avoid precipitation of AKI in cirrhosis and ascites.
- Acute kidney injury
- cardiovascular complications
- health service research
- hepatic haemodynamics
- hepatorenal syndrome
- liver cirrhosis
- portal hypertension
- renal dysfunction
Statistics from Altmetric.com
- Acute kidney injury
- cardiovascular complications
- health service research
- hepatic haemodynamics
- hepatorenal syndrome
- liver cirrhosis
- portal hypertension
- renal dysfunction
Significance of this study
What is already known on this subject?
HRS is a severe complication of advanced cirrhosis with significant morbidity and mortality if left untreated.
The diagnosis of HRS requires the patient to fulfil a rigid set of diagnostic criteria.
Smaller increases in serum creatinine levels, insufficient for the diagnosis of HRS in cirrhosis, have a negative impact on survival in other disease states.
What are the new findings?
AKI, as diagnosed by an increase in serum creatinine of over 50% from baseline or a rise in serum creatinine of 26.4 μmol/l or greater (≥0.3 mg/dl), occurs in more than half of patients with cirrhosis and ascites.
Most episodes of AKI are precipitated by the same clinical events that precipitate HRS.
Small increases in serum creatinine can have a negative impact on patient outcome, even if these increases resolve and the peak level is within the laboratory's normal range. A gradual increase in serum creatinine occurs with time, associated with a gradual reduction in mean arterial pressure.
The development of AKI is associated with a reduction in patient survival.
Some episodes of AKI can progress on to HRS.
How might it impact on clinical practice in the foreseeable future?
The recognition that mild cases of renal dysfunction may not fulfil the traditional definition of HRS, but are still clinically relevant, permits clinicians to treat these events more effectively.
Clinicians will be more vigilant for the development of AKI in the presence of potential precipitating events. This will allow for the institution of preventive measures.
Larger studies to assess AKI in cirrhosis further will help to identify predictive factors for the development and for the non-resolution of AKI.
Renal dysfunction is a common complication in liver cirrhosis and ascites, occurring in 20% of patients with cirrhosis admitted to hospital.1 Most cases of renal dysfunction in cirrhosis are functional in nature, ie, there is no underlying renal structural abnormality. Many episodes of functional renal dysfunction are responsive to volume challenge. Those that do not respond are defined as cases of hepatorenal syndrome (HRS), which is associated with significant morbidity and mortality.2
Traditionally, renal impairment in cirrhosis has been identified using serum creatinine as an indicator of renal function. However, serum creatinine is a poor marker of glomerular filtration rate (GFR) in patients with decompensated cirrhosis and ascites.3 This is due to a reduced production of creatinine from creatine4 in the liver from significant muscle wasting, resulting in lower serum creatinine levels relative to GFR.5 Various formulae used to determine the GFR are also imprecise, as they all incorporate serum creatinine into their calculations.6–8 It is thus well recognised that in decompensated cirrhosis, the serum creatinine can be within the laboratory's normal range despite significant renal dysfunction.5 However, until simple and accurate measures of GFR are available clinically, serum creatinine remains an indicator of renal function in cirrhosis.
Because serum creatinine may underestimate the degree of renal impairment in cirrhosis, particularly in patients with ascites, the use of rigidly defined serum creatinine levels to diagnose HRS may exclude many cases of potentially clinically significant renal dysfunction—especially if the change in serum creatinine does not reach the HRS diagnostic criteria. Nephrologists define acute kidney injury (AKI) as a percentage, in addition to an absolute increase in serum creatinine.9 Similar approaches to define AKI have also been proposed for patients with cirrhosis.10 ,11 In addition, smaller rises in serum creatinine of 0.3 mg/dl (26.4 μmol/l) have been reported to have a negative impact on survival in patients undergoing open heart surgery;12 these same small increases in serum creatinine may also have a negative impact on the clinical outcome in patients with cirrhosis and ascites. Therefore, the aims of this study are prospectively to assess the natural history of cirrhotic patients with ascites but normal serum creatinine as defined by our laboratory, and to determine whether episodes of AKI, as defined in table 1, have an adverse effect on patient morbidity and mortality.
Patients and methods
The study was approved by the Research Ethics Board of University Health Network. Informed consent was obtained from all patients in the study.
Patients with cirrhosis and ascites, aged 18–75 years, were recruited from the hepatology outpatient clinic at Toronto General Hospital, University Health Network. The diagnosis of cirrhosis was either biopsy confirmed or based on a combination of laboratory, endoscopic and imaging findings. Ascites was confirmed on abdominal ultrasound taken within the previous 3 months. Patients were included if they had a laboratory normal serum creatinine of less than 110 μmol/l taken on the day of enrolment. Exclusion criteria were active malignancy including hepatocellular carcinoma, past history of transjugular intrahepatic portosystemic stent shunt (TIPS) insertion or liver transplantation, estimated survival of less than 3 months, and the presence of organic renal disease as indicated by an abnormal urinalysis, small kidneys on abdominal ultrasound, or proteinuria of more than 500 mg/day. Table 2 summarises baseline patient characteristics.
Out of approximately 30 patients with cirrhosis and ascites who attended the hepatology clinic per month, consecutive eligible patients who gave informed consent were enrolled between January 2008 and December 2009. At baseline visit, all patients underwent a review of their medical history and a physical examination including an assessment for ascites to confirm eligibility. Baseline collected blood work-up included liver enzymes, liver function tests, complete blood count, electrolytes and serum creatinine. Thereafter, patients received the same blood work-up at 4–6-week intervals. Patients were evaluated at the clinic every 4 months to review significant clinical events, including any hospital admissions or the development of AKI. AKI was defined as a rise in serum creatinine of over 0.3 mg/dl or 26.4 μmol/l within 48 h if two serum creatinine readings were available within that time period, or as a 50% increase in serum creatinine from baseline.1 ,11 Both the model for end-stage liver disease13 and the Child–Pugh scores14 ,15 were calculated at each clinical encounter. Other recorded clinical outcomes include TIPS insertion, liver transplantation, or death. If AKI were noted, patients would receive volume expansion of albumin 1 g/kg for two consecutive days up to a maximum of 100 g/day.
Results were expressed as means±SEM. All statistical analyses were made using PRISM 5.0a (Graphpad Software). Descriptive statistics were applied to all parameters collected. Because the intent of the study was to assess the clinical outcomes of those with and without AKI, the patients were divided into these respective groups. The differences between the AKI and the non-AKI groups were analysed using an unpaired t test for continuous variables and the χ2 test for nominal variables. For parameters that were repeatedly measured over the study period, the difference between various time intervals versus baseline for each parameter was compared using analysis of variance. Correlation between two variables was assessed using linear regression. The probability of survival was estimated using the Kaplan–Meier method and compared with the log-rank test. A p value of less than 0.05 was considered statistically significant.
A total of 90 patients was enrolled. There were 64 men and 26 women, with a mean age of 55.8±0.8 years. During a mean follow-up period of 14.05±1.07 months, there were 82 episodes of AKI in 49 patients, yielding a mean 1.67 episodes of AKI per patient (range one to four episodes). Twenty patients had two episodes of AKI, 10 patients had three episodes, and three patients had four episodes. There was no difference in patient demographics between those who developed AKI and those who did not, nor was there any difference in terms of baseline liver enzymes, liver or renal function, or systemic haemodynamics (table 2). The mean peak serum creatinine values of the AKI episodes were all below 133 μmol/l (1.5 mg/dl), the value required for the diagnosis of HRS2 (figure 1). The peak serum creatinine reached was 374 μmol/l (4.25 mg/dl). For all the AKI episodes, there was a significant negative correlation between the peak serum creatinine and the simultaneous mean arterial pressure (figure 2). In fact, the development of AKI was accompanied by a transient fall in the mean arterial pressure from 89±3 mm Hg to 76±3 mm Hg, which returned to baseline upon resolution of the AKI episodes.
In the AKI group, most of the AKI events had an identifiable precipitating factor (85%); only 12 episodes of AKI did not. The most common precipitants for the episodes of AKI were bacterial infections (including spontaneous bacterial peritonitis), followed by large volume paracentesis despite concomitant use of appropriate doses of albumin of 6–8 g/l of ascites removed and increase in diuretic doses (table 3). In the AKI episodes that resolved, the serum creatinine returned to baseline pre-AKI levels on resolution (figure 1). At the end of the follow-up period, nine episodes of AKI remained unresolved, two of which reached a serum creatinine level high enough for the diagnosis of HRS. None of the remaining unresolved AKI episodes had elevated levels of serum creatinine persisting for long enough for the diagnosis of type 2 HRS. In terms of patient demographics, laboratory parameters and in particular the mean serum creatinine levels before the AKI episode, there was no difference between subjects whose AKI episodes resolved (mean pre-AKI serum creatinine 74±2 μmol/l) versus those whose episodes did not resolve (mean pre-AKI serum creatinine 75±5 μmol/l). Despite the fact that most of the AKI episodes resolved, there was a gradual and significant increase in the mean serum creatinine levels in the AKI group at 8 and 12 months of follow-up, albeit still within the normal laboratory range (figure 3A). The same serum creatinine levels were also significantly higher compared with the non-AKI group at the same time intervals (figure 3A). Concomitant with the rise in serum creatinine, there was a gradual and significant fall of the mean arterial pressure in the AKI group during follow-up, with the mean arterial pressure at 8 and 12 months being significantly lower than that of the non-AKI group (figure 3B). Table 4 shows the remainder of the laboratory data in the AKI and non-AKI groups during follow-up, which demonstrates a steady and gradual deterioration in liver function in the AKI group, but not in the non-AKI group. Because of the decline in both liver and renal function in the AKI group, the model for end-stage liver disease score of the AKI group was significantly different between the AKI and non-AKI groups at 12 months of follow-up (p<0.05).
There was no difference in the ascites characteristics between the AKI and non-AKI groups. Eight of the 49 patients (16.3%) in the AKI group had regular 2-weekly paracentesis, while five of 41 patients (12.2%) in the non-AKI group underwent regular 2-weekly paracentesis (p>0.05). The volume removed was usually 6–8 l per session and albumin would be infused at the rate of 6–8 g/l of ascites removed. The mean doses of diuretic used in the AKI group decreased over time, but the doses at the end of the follow-up period were not significantly different compared with baseline. The mean diuretic doses in the non-AKI group did not change throughout the study period.
Multiple clinically significant events occurred in both the AKI and non-AKI groups during the follow-up period. There were 17 episodes of variceal bleeding, 26 episodes of hepatic encephalopathy and 16 episodes of spontaneous bacterial peritonitis in the AKI group. The corresponding numbers of episodes were 12, 17 and seven in the non-AKI group (p>0.05). Therefore, during the follow-up period, there were significantly more patients who went on prophylactic antibiotics in the AKI group (n=12) versus the non-AKI group (n=2). Three patients in the AKI and two patients in the non-AKI group developed a hepatoma during the follow-up period.
Seven patients in the AKI group and two patients in the non-AKI group underwent liver transplantation during the follow-up period, six of which were living related liver transplantations. There were eight deaths in the AKI group and one death in the non-AKI group. Causes of death in the AKI group were related to sepsis in three cases, multi-organ failure in two cases, cardiac disease in one case, uncontrolled variceal bleed in one case, and the development of a hepatoma in one case. The mean±SEM time from enrolment to death in the AKI group was 211±76 days (range 13–546 days). The death in the non-AKI group was related to intraperitoneal bleeding following a large volume paracentesis. No patient received a TIPS shunt during the follow-up period. Figure 4 shows the survival of the two groups of patients, which was significantly lower in the AKI group (p=0.049).
This study demonstrates that the natural history of ambulatory patients with decompensated cirrhosis is relatively good despite the presence of ascites, unless AKI supervenes. Many of these episodes of renal impairment were transient, and they might escape the clinician's attention, as the serum creatinine remained within the laboratory's normal range. Despite this, the development of these ‘minor’ episodes of AKI can lead to a gradual deterioration of renal function. In addition, there is evidence of worsening liver dysfunction over the course of 12 months, as indicated by indices of liver function.
There has been a great deal of discussion recently regarding the rigidity of the diagnostic criteria for HRS,11 which is an extreme form of renal failure in cirrhosis.2 As the current treatment for HRS, which includes vasoconstrictors such as terlipressin, only reverses renal failure in approximately 30% of patients,16 ,17 some have advocated starting treatment in the earlier stages of renal failure, such as those episodes of AKI that do not resolve with volume expansion, but have not yet reached the required serum creatinine levels for the diagnosis of HRS. However, the only currently accepted diagnostic criteria for renal failure in cirrhosis are those for the diagnosis of HRS. Therefore, clinicians have no other guidelines as to how to diagnose renal impairment in cirrhosis and when to start treatment. It is clear that normal serum creatinine may not represent normal renal function. In a cohort of patients on the liver transplant waiting list, a serum creatinine of less than 88 μmol/l (1 mg/dl) had a true GFR that ranged from 34 to 163 ml/min per 1.73 m2.8 Therefore, to delay treatment until patients reach a serum creatinine of 133 μmol/l or 1.5 mg/dl in order to diagnose renal dysfunction may be detrimental to their outcome. In 2008, Garcia-Tsao and colleagues1 suggested borrowing the term AKI from the nephrology literature. The International Ascites Club and the Acute Dialysis Quality Initiative Group together proposed that AKI should be recognised as an entity in cirrhosis, and that it should be defined in terms similar to those used in the nephrology literature11 (table 1). This is because smaller rises in serum creatinine have been shown to have a negative impact on survival in critically ill patients with cirrhosis in intensive care units, irrespective of the baseline serum creatinine.18–20 This is the first study that demonstrates that similar episodes of AKI can also adversely affect the outcome of otherwise stable patients with cirrhosis with ascites.
Most of the episodes of AKI were precipitated by events that have also been implicated as precipitants for HRS. These include bacterial infections, intravascular volume disturbances such as large volume paracentesis despite albumin infusion, and excess diuretics (table 3). All these events have the potential to disturb the systemic haemodynamics of these patients. This would suggest that these episodes of AKI most likely have the same pathophysiological basis as that of HRS,21 ,22 but perhaps to a lesser degree of severity. The fact that patients in the AKI group had a gradual reduction in their mean arterial pressure during their follow-up, suggesting progressive further vasodilatation would support this contention. Therefore, these patients were haemodynamically less stable, thus making them more susceptible to the development of renal dysfunction. Patients with cirrhosis and ascites also have an abnormal renal autoregulatory mechanism, the result of sympathetic hyperactivity in cirrhosis.23 As the patient with cirrhosis advances from pre-ascites through the stages of diuretic responsive ascites, diuretic refractory ascites, and finally to HRS, for every given renal perfusion pressure, the renal blood flow progressively falls.23 In other words, the renal blood flow becomes progressively compromised as cirrhosis advances. The patient is therefore increasingly susceptible to the development of renal dysfunction if there is a further disturbance in systemic haemodynamics, such as in the case of a bacterial infection. We thus see bacterial infections being a common precipitant of AKI in these patients. However, this study was not designed to assess the pathophysiological basis of AKI, and therefore we can only speculate on the mechanisms that led to the development of these AKI episodes.
The fact that most episodes of AKI in these patients were reversible would suggest that there was some renal reserve, and that correction of the precipitating factor led to a restoration of systemic haemodynamics and a return of serum creatinine to baseline levels. As we did not measure the true GFR in these patients with a clearance technique, it is unclear whether the GRF also returned to baseline levels on resolution of the precipitating factor.
Despite the fact that serum creatinine returned to the baseline level with each episode of AKI, there was a gradual and significant increase in the serum creatinine level over time by 12 months of follow-up, associated with a progressive reduction in mean arterial pressure. This would suggest that the slow deterioration of renal function probably had a haemodynamic basis to its pathogenesis. Whether this was additionally related to a cumulative loss of renal function with repeated insults to the kidney, or to the gradual deterioration of liver function with the progression of cirrhosis is unclear. The corollary from this observation is that patients with cirrhosis and ascites who develop AKI will need to be monitored more closely even if the episodes of AKI are completely reversed, so as to allow for timely referral for liver transplant assessment.
There were no clear predictors for the development of AKI, nor were there any predictors for the reversal of the AKI episodes in this study. The design of the study only allowed the collection of routine clinical and laboratory parameters, and therefore we could have missed subtle predictors of AKI development. It is possible that the patients who developed AKI were the patients who had some degree of cirrhotic cardiomyopathy, a condition recently described in patients with cirrhosis.24 These patients tend to have further deterioration of their circulatory dysfunction with the presence of bacterial infections, predisposing them to the development of renal dysfunction.25 Although conjectural, and no cardiac investigations were routinely performed in these patients, it is reasonable to expect that patients whose cardiac status cannot meet the demands of an increased cardiac output in the face of infection are at risk of the development of renal dysfunction, especially if the baseline renal blood flow is subnormal as a result of altered autoregulation in cirrhosis with ascites. It is interesting to note that the infusion of albumin post-large volume paracentesis (mean volume removed 8.67 l) did not prevent the development of AKI episodes. This is consistent with the previous findings of Ginès et al,26 which demonstrated an 18% incidence of post-paracentesis circulatory dysfunction despite patients receiving adequate doses of albumin infusion. In that particular study, 8% of their patients who received albumin developed renal dysfunction with a serum creatinine of more than 1.5 mg/dl, sufficient for the diagnosis of HRS. This is further supported by a more recent study that demonstrated that renal dysfunction can occur following large volume paracentesis despite the absence of post-paracentesis circulatory dysfunction with albumin infusions.27 What our particular cohort of patients tells us is that despite adequate volume replacement with albumin with large volume paracentesis, there must have been sufficient disturbance of the systemic haemodynamics to allow these small rises in serum creatinine to occur.
Most of the deaths in the AKI group were related to sepsis or multi-organ failure, conditions associated with haemodynamic instability, whereas the only death in the non-AKI group was associated with a procedural misadventure. There are no clear clinical or laboratory indicators to suggest that the patients in the AKI group were more susceptible to infections. It is possible that it was a susceptibility to sepsis in the AKI group that contributed to the increased mortality, with AKI occurring as a manifestation of the overall sepsis syndrome.
Irrespective of the causes of the AKI episodes, it appears that minor increases in serum creatinine, even when the peak serum creatinine values are still within the laboratory's normal range, are clinically relevant and are associated with progressively worsening renal function and increased mortality. Therefore, every effort should be made to avoid and promptly treat renal dysfunction with volume expansion, even if mild, in patients with cirrhosis and ascites. Accepting that minor increases in serum creatinine are clinically relevant, the next step would be to investigate whether earlier treatment of unresolved AKI episodes would lead to improved patient outcomes.
The authors wish to acknowledge the help provided by Ms Azra Shijiv for data entry, and that provided by Dr Vivekanandan Thayalasuthan for statistical analysis.
Competing interests None.
Patient consent Obtained.
Ethics approval Ethics approval was provided by Research Ethics Committee, University Health Network.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.