Introduction

Acute kidney injury (AKI) refers to an abrupt and sustained decrease in the glomerular filtration rate, resulting in the accumulation of toxic end products of metabolic pathways and disturbances in fluid volume and electrolyte and acid-base balance. A multinational cross-sectional study found that AKI occurred in more than half of patients treated in the intensive care unit (ICU) and that increasing AKI severity was associated with increased mortality [1]. An estimated 8–12% of critically ill patients require renal replacement therapy (RRT) for severe AKI (AKI-D). The utilization of RRT in the ICU is rapidly growing, likely in response to an aging population with a high burden of comorbid diseases and severe illness acuity [2]. AKI has become prominent as a major contributor influencing both short-term (in-hospital global mortality 40–50%) as well long-term outcomes [2, 3].

RRT represents the cornerstone of the supportive management of severe AKI. The goals of acute RRT are the following: to replace the important functions of volume control and clearance of electrolytes or metabolic by-products and to enable functional recovery of the kidneys and other vital organs with as little disruption as possible. However, clinical decision-making related to RRT weaning for ICU patients with AKI-D is not standardized, and conflicting recommendations exist for the timing of RRT discontinuation in ICU patients who appear to exhibit improving renal function [4].

The aim of this review is to offer practitioners a comprehensive update on factors predictive of renal functional recovery and to address questions that arise in daily practice regarding RRT weaning.

Current status of weaning of acute RRT in ICU patients with AKI

Most critically ill patients surviving an acute episode of AKI sufficiently severe to require RRT will recover sufficient kidney function to be independent of RRT at ICU or hospital discharge.

While weaning from mechanical ventilation is a crucial aspect of pulmonary support that has been extensively investigated, unfortunately, no prospective randomized trial exists for the cessation of RRT in general, particularly intermittent RRT or continuous RRT [5].

The KDIGO guideline criteria for discontinuation of acute RRT suggest weaning of RRT when it is no longer required after sufficient renal recovery has occurred and meets a patient’s needs or because RRT is no longer consistent with the goals of care. In addition, diuretics are discouraged to enhance kidney function recovery or to reduce the duration or frequency of RRT [6]. The lack of specificity of these statements indicates that clear clinical factors or functional markers that may help physicians predict successful cessation of acute RRT are still lacking.

Currently, the decision to stop a course of acute RRT or to change the RRT modality is strongly influenced by a variety of clinical factors such as patient characteristics (hemodynamic stability, fluid balance/volume overload, electrolyte disturbance), renal parameters (urine output, urine chemistry, and GFR surrogates), and logistic characteristics (staff availability, catheter malfunction, circuit clotting).

The lack of evidence has translated into wide variations in the practice of RRT weaning in ICU patients with AKI-D, ranging from a “wait and see” strategy to a “fast track” procedure. The effects of inappropriate timing of RRT weaning on patient outcomes have not been specifically addressed, but both unnecessary RRT as well as re-institution of RRT likely have negative impacts on the clinical course of a patient and on economic outcomes.

Early re-institution of RRT in critically ill patients (fast track) may contribute to higher mortality among patients with failed weaning. Admittedly, the increased mortality of this subgroup of ICU patients with AKI-D likely reflects greater severity of the underlying illness rather than indicating that early weaning is harmful. However, a period of severe AKI without renal support may cause fluid overload, persisting uremic syndrome and electrolyte abnormalities and expose unwell patients to the systemic effects of ongoing AKI. Therefore, avoiding one or more failed attempts of discontinuation of RRT may be better.

By contrast, the wait and see approach may prolong renal support and expose ICU patients to hazardous effects of unnecessary RRT, such as repeated hypotension, blood stream infections, bleeding, bio-incompatibility reactions, and inadequate drug dosing [7].

Predictors of renal functional recovery and successful discontinuation of acute RRT

Most ICU patients who survive an episode of AKI-D regain sufficient kidney function to avoid long-term dialysis, especially unwell patients with normal renal function prior to the renal insult. However, wide variation in the mean duration of RRT exists. Therefore, clinical practice dictates daily assessment of both intrinsic kidney function as well as the ongoing need for acute RRT. Assessment of kidney function during RRT, particularly during CRRT, is not easy. Recovery of native kidney function can be assessed during RRT by serial determinations of serum creatinine, close attention to urine output, measurements of urinary chemistry (creatinine and urea excretion rates) and calculation of surrogates of the glomerular filtration rate (creatinine clearance).

Predictive ability of urine volume alone or in combination with clinical factors

Urinary volume over time is regarded as the most important predictor of successful weaning of both CRRT and IHD by most physicians caring for ICU patients with AKI-D. A survey of current practice regarding critically ill AKI patients managed in UK intensive care units showed that the most common reasons for discontinuing RRT were increasing urine output (74%), normalization of pH (70%), and achievement of adequate volume status (55%). Normalization of serum urea and creatinine as an indication to stop RRT was cited by 39% of respondents [8].

Wu et al. [9] conducted a retrospective observational case–control study to identify clinical and renal risk factors of early re-dialysis after weaning from post-surgery AKI-D therapy. Successful discontinuation of RRT was defined as cessation of IRRT or CRRT for at least 30 days. The authors weaned 94 ICU patients from acute RRT for more than 5 days, and 64 of these patients were successfully weaned for at least 30 days. The independent predictors for resuming dialysis treatment within 30 days after original RRT cessation were a longer duration of dialysis, a higher sepsis-related organ failure assessment (SOFA) score, oliguria (urine output less than 100 ml/8 h) on day 1 after cessation, pre-existing chronic kidney disease and age over 65 years.

Utilizing the database of the prospective Best kidney study, Uchino et al. [10] sought to investigate the practice of CRRT discontinuation in a multinational setting to identify which factors present at the time of discontinuation may assist physicians in predicting successful cessation of CRRT. These authors analyzed five hundred twenty-nine ICU patients with AKI-D who survived initial therapy among 1006 patients treated with CRRT. Three hundred thirteen patients were successfully weaned from CRRT and did not require any RRT for at least 7 days and were classified as the success group; the other 216 patients were classified as the repeat RRT group. The patients in the success group had significantly lower in-hospital mortality (28.5 vs. 42. 7%). They also had lower creatinine or urea serum concentrations and higher urine output (1500 vs. 199 ml/day) at discontinuation of CCRT. The multivariable model for successful discontinuation of CRRT showed that the most significant variable was urine output in the 24 h before stopping CRRT. Serum creatinine at discontinuation of CRRT was also found to be a significant variable in this study. The predictive ability of urine output was negatively affected by diuretic use. In the ROC curve, a urine output of 436 ml/day without diuretics and a urine output of 2330 ml/day with diuretics had the highest sensitivity, specificity and positive predictive values (80.9 and 87.9%, respectively). The authors found that pre-existing CKD was a strong (not statistically significant) negative predictor of successful discontinuation of CRRT.

A retrospective cohort study divided 18 patients with Type 1 cardiorenal syndrome into two groups according to successful weaning. The mean urine volume and fluid balance values 2 days after RRT cessation were 1350 ml/day and − 350 ml/24 h in the RRT-independent subgroup and differed significantly from the corresponding mean values (265 ml/day and + 850 ml/24 h, respectively) in the RRT-dependent subgroup. The authors of this small analysis concluded that daily urine volume and fluid balance should carefully monitored due to their importance in the clinical decision concerning cessation of RRT for type 1 CRS patients with AKI-D.

The multicentre, retrospective observational study by Katayama et al. aimed to determine the optimal timing for CRRT weaning in Japanese ICU patients with AKI-D by evaluating factors predictive of successful discontinuation of RRT. Patients were divided into two groups according to success or failure of CRRT discontinuation. In the multivariate logistic regression analysis, higher urine output, lower serum creatinine, and shorter CRRT duration were significant factors predicting successful discontinuation of CRRT [11].

Tsukamoto et al. [12] evaluated the time to terminate CRRT in patients with post-cardiovascular surgery AKI with or without a mechanical ventilation support. When CRRT was stopped, 32 patients were supported with a mechanical ventilator and 41 patients were not mechanically ventilated. The patients supported with a ventilator had a higher mean SOFA score, greater increases in body weight after cardiac surgery, higher mean fluid therapy volumes, and higher mean central venous pressure despite higher mean urine volumes per day. The rate of re-introduction of CRRT was higher in the patients supported with a ventilator (47 vs. 12%). Multivariate regression analysis revealed that successful cessation of CRRT was dependent on mechanical ventilation (odds ratio 5.2).

The retrospective observational study performed by Raurich et al. [13] on 86 ICU patients found that 6-h urine output and sex were the main risk factors of successful CRRT weaning. Univariate analysis revealed that patients with failed weaning tests had fluid overload and a longer CRRT duration. The authors found that more than 70% of the weaning tests failed in patients with a urine output of less than 0.3 ml/kg/h.

Clinical trials using the urine output criterion for discontinuation of RRT

In the small prospective randomized clinical trial by Bouman et al., CRRT was discontinued when the urine output returned to or was stable at greater than 60 ml/h [14]. In the AKIKI study [15], for both the early and late groups, discontinuation of RRT was considered when the spontaneous urine output (UO) was higher than 500 ml/24 h, highly recommended when the spontaneous UO was higher than 1000 ml/24 h in the absence of diuretic treatment (higher than 2000 ml/24 h in patients receiving diuretics), and mandatory if diuresis was sufficient to allow a spontaneous decrease in the serum creatinine concentration. In the Elain study [16], renal recovery was defined by a urine output greater than 400 ml/24 h without diuretic treatment (2100 ml/24 h with diuretic treatment) and creatinine clearance greater than 20 ml/min. None of these clinical trials reported success rates of weaning.

Prediction of RRT cessation by urine chemistry markers

Daily urinary creatinine excretion

Viallet et al. [17] conducted a retrospective study to investigate and identify parameters, particularly the 24-h urinary creatinine excretion rate, present at the time of RRT weaning that may be associated with successful cessation of RRT in ICU patients with AKI-D. The criteria required for a weaning attempt were the following: a urine output of at least 20 ml/h without diuretics, restored and stable hemodynamic and respiratory conditions, no impeding renal insult, and no ongoing need to continue RRT for medical reasons. Successful weaning was defined as cessation of RRT for at least 15 days. Twenty-six weaning attempts succeeded while 28 failed. The 24-h creatinine excretion rate (more than 5.2 mmol/24-h) was the strongest predictor of successful weaning (AUC 0.86, positive predictive value 84%, independent of age, weight and diuresis). Notably, the success group had a significantly higher urine output (2.3 vs. 1.9 l/day) and higher creatinine clearance (31 vs. 18 ml/min) on the day of the weaning attempt. None of the clinical parameters (age, pre-existing chronic kidney disease, SOFA score, or length of RRT) appeared to be an efficient and independent marker of successful weaning of RRT in this study. Moreover, blood parameters including serum urea and creatinine and their kinetics were not useful in guiding successful termination of RRT.

Daily urinary urea excretion

A retrospective single-centre cohort study analyzed 67 ICU patients with AKI-D treated with intermittent haemodialysis (IHD) for at least 7 days and four dialysis sessions [18].

Blood and urinary markers were recorded on the day of the last IHD session for 37 unweaned patients or 2 days after the last IHD session for 29 weaned patients. Urine output and urinary urea concentration were both significantly associated with IHD weaning. The optimal diagnostic thresholds for IHD weaning were a urine output greater than 8.5 ml/kg/24 h, a urinary urea concentration greater than 148 mmol/l, and a daily urea excretion greater than 1.35 mmol/kg/24 h. The AUROC of urinary urea excretion (0.96) was greater than the AUROC of urine output (0.86) or the AUROC of daily urinary urea excretion rate (0.83). These authors did not observe any difference in daily urinary creatinine excretion rates between the weaned and unweaned patients. However, approximately one-third of these patients had a baseline GFR less than 30 ml/min. By contrast, the study of Viallet et al. included only patients with a baseline GFR greater than 60 ml/min, partly explaining the discrepant urinary chemistry results [17].

Creatinine clearance and discontinuation of RRT in AKI patients

Two-hour creatinine clearance

The retrospective single-centre study by Fröhlich et al. [19] aimed to investigate whether a 2-h creatinine clearance (2-h CrCl) measurement could more accurately predict successful cessation of CRRT than serum creatinine or urine output alone. Of 85 ICU patients with AKI-D who had a 2-h CrCl performed in the 12 h preceding CRRT cessation, 53 critically ill patients (success group 62, 4%) remained dialysis-free for 7 days after CRRT cessation. The 2-h CrCl was a better predictor of remaining CRRT-free than urine output, serum creatinine or age. A 2-h CrCl value of 23 ml/min had a positive predictive value of 88.8%.

Six-hour creatinine clearance

The VA/NH Acute Renal failure trial network study defined renal recovery on the basis of creatinine clearance measured using 6-h timed urine collections [20]. When urine flow increased to more than 30 ml/h or when a spontaneous decrease in the serum creatinine level was observed, renal replacement therapy was continued if creatinine clearance was less than 12 ml/min and discontinued if creatinine clearance was greater than 20 ml. Decisions regarding discontinuation of RRT in the context of intermediate values of creatinine clearance were left to the treating physicians. The success rate of this weaning approach remains unknown.

The primary purpose of our retrospective analysis of the Acute Renal Support Registry of the University of Munich was to examine the association between severe hypophosphatemia and the outcomes of critically ill patients with AKI-D [21]. The study included 289 ICU patients, all of whom had normal glomerular filtration rates prior the renal insult(s). Intermittent haemodialysis was performed daily, on alternate days, or based on an individual patient’s need. IHD was terminated according to a predefined protocol when the 6-h creatinine clearance was greater than 15–20 ml/min (mean 22 + 2 ml/min), serum creatinine concentrations decreased spontaneously and increased urine output was associated with a balanced fluid status. All patients who recovered sufficient renal function were successfully weaned for at least 30 days.

24-h Creatinine clearance

Wheeler and Tolwani retrospectively evaluated the utility of a 24-h creatinine clearance measurement obtained while on CCRT in a small group of ICU patients with oliguric AKI at RRT initiation [22]. Of the 23 study patients, 9 had a 24-h CrCl greater than 15 ml/min prior to CRRT cessation, whereas 14 had a 24-h CrCl less than 15 ml/min. No significant differences in the mean serum creatinine or the mean urine output were identified. Eight of the nine patients (89%) in the creatinine clearance greater than 15 ml/min group successfully remained off RRT for at least 2 weeks following CRRT cessation, whereas only 4 of the 14 patients (29%) in the creatinine clearance less than 15 ml/min group successfully remained off RRT. Curve characteristics demonstrated that a threshold CrCl > 15 ml/min predicted successful discontinuation.

Novel AKI biomarkers and plasma-NT-pro BNP

Few studies in critically ill patients with AKI-D have shown that patients with lower urinary levels of biomarkers of inflammation or tissue or kidney injury or decreasing levels of these novel biomarkers over time are more likely to recover kidney function [23, 24]. Han et al. [25] found that plasma NT-pro-BNP, but not plasma neutrophil gelatinase-associated lipocalin (NGAL), was a weaning-related factor. Importantly, these studies focused more on renal recovery and survival than on successful prediction of RRT cessation.

Limitations of published data

Few studies have evaluated the process of RRT weaning or ideal markers to predict sufficient recovery of renal function to avoid re-initiation of RRT in critically ill patients with AKI-D.

The available evidence from observational studies has considerable limitations for translation to clinical practice due to the retrospective, post hoc secondary design of these studies, inherent risk of residual confounding and bias, small sample sizes and single-centre design, heterogeneity across study populations (severity of underlying illness, cause and severity of AKI, burden of comorbid diseases), discrepant thresholds of markers for RRT weaning and arbitrary definitions of successful cessation. Second, RRT weaning was not based on standardized criteria; the decision was made by attending physicians who clearly followed empirical rules. The markers of renal recovery, including urine output and urinary creatinine and urea concentrations, are significantly affected by extrarenal factors. Fourth, the data on both clinical factors and renal recovery markers are conflicting.

Algorithm for weaning a critically ill AKI patient from renal replacement therapy (Fig. 1)

Fig. 1
figure 1

Discontinuation of RRT in ICU patients with AKI-D: readiness testing, predictors of renal recovery and weaning

Full size image

If acute RRT has been initiated as a replacement therapy for severe AKI, for control of fluid volume overload or electrolyte disturbances, or for the treatment of acute uraemia, then adequate control of variables such as cardiovascular haemodynamics, respiratory function and fluid balance must be achieved prior to weaning. If these clinical factors have not been achieved, then successful discontinuation is unlikely.

Nonetheless, the main indicator of renal excretory function recovery is recovery of urine output. However, increased urine volume may not be apparent in patients with non-oliguric AKI and requires daily monitoring. Increases in urine output alone may not always indicate recovery of renal function following AKI. In the presence of oliguria, the patient’s ability to clear the solute load is significantly impaired and weaning should be postponed. The post hoc analysis of a prospective multicentre observational study found that a urine output of more than 400 ml was a reasonable cut-off value, resulting in correct classification in 79% of patients stopping CRRT [10]. A retrospective single-centre cohort study [18] found that the optimal diagnostic threshold for IHD weaning was a urine output greater than 8.5 ml/kg/24 h (1420 ml/day for a 70-kg patient). In two recent randomized trials, renal recovery was defined by spontaneous urine output rates of more than 400 ml/24 h [16] or more than 500 ml/24 h [15]. However, in the AKIKI trial discontinuation of RRT was highly recommended when urine output was higher than 1000 ml/ 24 h in the absence of diuretic therapy [15]. A urine output of 2 l/24 h or more may be necessary for a weaning test in critically ill patients receiving RRT and diuretics.

Increasing creatinine clearance and decreasing serum creatinine levels in the context of a fixed RRT prescription and a clinically stable patient suggest recovery of excretory renal function. The level of endogenous creatinine clearance needed to allow the successful discontinuation of RRT in critically ill patients is assumed to be 15–25 ml/min. In the VA/ATN study [20] and the Elain study [16], RRT was discontinued if the creatinine clearance was greater than 20 ml/min. 24-h urine collection periods are the most accurate measurement to determine creatinine clearance reliably. However, comparative studies in large cohorts of critically ill patients described a close correlation between 24-h creatinine clearance and the 2-h creatinine clearance or the 4-h creatinine clearance [19]. The VA/ATN trial used a 6-h urine collection period. In patients receiving intensive treatment, serum creatinine can be maintained in the slightly increased or normal range, and detecting a spontaneous decrease in serum creatinine may not be possible. Whether urine chemistry markers serve as the best markers for successful weaning remains speculative. The two studies exploring the role of urinary creatinine excretion in predicting successful weaning of RRT in AKI patients reported conflicting data, probably due to differences in patient characteristics.

Conclusions

Currently, the decision regarding the appropriate timing for RRT cessation is naturally complex, integrating numerous clinical variables and renal functional parameters, and should be individualized. Current evidence favors measurements of urine output and serum creatinine levels in the context of a constant CRRT regimen or prior to the next IRRT session and calculation of endogenous creatinine clearance. Decisions to delay or cancel the next RRT session may be easier for IRRT. Future clinical trials are needed to identify better parameters for individual decisions regarding discontinuation of RRT in ICU patients with AKI-D. Future work should integrate evolving data on novel kidney damage and repair biomarkers, bed-side measurements of real-time GFRs, or the furosemide stress test to provide additional support for the decision regarding the optimal timing to attempt RRT discontinuation in ICU patients. However, large randomized multicentre trials are required to definitively guide appropriate cessation.