Renal Replacement Therapy in Critically ill Patients: Navigating the Timing Debate

Abstract

Acute kidney injury (AKI) is a frequent complication in critically ill patients and is associated with high mortality. While renal replacement therapy (RRT) remains a cornerstone of supportive care for severe AKI, the optimal circumstances for RRT initiation in critically ill patients is a longstanding subject of debate. This narrative review aims to provide an up to date summary of the evidence regarding timing of RRT initiation in the intensive care unit (ICU) and its impact on clinical outcomes. Additionally, management strategies for patients while RRT is delayed are suggested and means to identify patients who have a high probability of receiving RRT are explored.

Introduction

Acute kidney injury (AKI) is a common complication of critical illness and an independent risk factor for mortality.^1,2 When AKI complicates an episode of sepsis, for example, mortality is reported to be as high as 60%^3–6. AKI rarely occurs in isolation and is often a consequence of multiple concurrent insults including shock, sepsis/inflammation, major surgery, trauma, burns or intoxications.^4–9 In such cases, the kidneys are the victims of a systemic process and treatment efforts are therefore aimed at resolving the underlying disease process and providing supportive care, as there is no targeted treatment for established AKI. ¹⁰ Renal replacement therapy (RRT), in its different configurations, is an important tool for the achievement of metabolic and volume stabilization during acute illness, serving as a bridge to kidney function recovery. The optimal circumstances for RRT initiation in critically ill patients is a longstanding subject of debate. While there are unambiguous indications for the initiation of RRT in the setting of AKI (eg, medically refractory hyperkalemia, severe acidosis and diuretic-refractory volume overload), RRT initiation is often started in the absence of such indications and with a great deal of inter-clinician variability.^11–13 The fundamental clinical dilemma, which has been the topic of several clinical trials over the past decade, is whether RRT initiation in the setting of severe AKI, but in the absence of an overt emergency, is beneficial for patients. ¹⁴

In this narrative review, we aim to summarize the evidence regarding the timing of RRT initiation in the ICU, in the absence of immediate and urgent medically refractory complications. Specifically, we will challenge the definitions of “early” and “late”, analyse the impact of timing on clinical outcomes, suggest management strategies for patients while RRT is deferred and explore means to identify patients who have a high probability of receiving RRT.

How Early, how Late?

In the absence of traditional urgent indications, the ability of the clinician to identify the ideal point in time to start RRT in an ICU setting remains somewhat of an art. Numerous studies on this topic used various definitions of “early” and “late” initiation of RRT^15–17 (Table 1). The reference point of time was also variable, with studies looking at time from hospital or ICU admission, time from development of AKI or its complications, duration of oliguria or anuria and time from enrolment.^18,19 Some studies included biochemical thresholds of serum urea^13,20,21 or creatinine concentrations.^22,23 Others used urine output alone^24,25 or in combination with urea and/or creatinine concentrations, which were defined either separately^22,26 or using the Risk, Injury, Failure, Loss and End-stage renal failure criteria (RIFLE) criteria. ²⁷ More recent prospective RCTs, incorporated the Kidney Disease Improving Global Outcomes (KDIGO) criteria ²⁸ and defined the time point for “early” RRT initiation by the presence of severe AKI but without complications that would impel the urgent initiation of RRT.^29–31 The KDIGO criteria rely solely on serum creatinine and urine output and do not account for other dynamic clinical variables. They are intended to provide a standardised framework for classifying the severity of AKI, rather than to serve as stand-alone indication for real-life clinical decision-making regarding the initiation of RRT. Rather than time per se, decision-making around RRT initiation incorporates data regarding the global clinical state of the patient. But even when a predefined threshold is crossed, it may be reasonable and safe to apply close surveillance and to delay RRT. This is because a considerable number of patients will, in retrospect, recover kidney function without receipt of RRT. ³² Moreover, we must acknowledge the real-world challenges of resource allocation in busy ICUs, where dialysis equipment and personnel may be limited.

Table 1.

A summary of the various definitions of early versus delayed renal replacement therapy across major trials.

Study	Study Design	Inclusion Criteria	Definition of Early RRT	Definition of Delayed RRT
Andrade 18 (2007)	Retrospective comparison of practice	Patients with severe leptospirosis and AKI requiring RRT	As soon as possible (mean 265 ± 58 min)	after 6 h of conservative treatment (mean 1638 ± 592 min)
Iyem 19 (2009)	Observational	ARF requiring RRT post CABG	as soon as possible	>48h
Liu 20 (2006)	Observational	ARF requiring RRT in patients from the PICARD database	BUN ≤ 76 mg/dl	BUN > 76 mg/dl
Bagshaw 21 (2009)	Observational	Critically ill patients with severe AKI admitted to intensive care units (ICUs).	Urea ≤ 24.2 mmol/L or SCr < 309 μmol/L or RRT within 2 days of ICU admission	urea > 24.2 mmol/L or SCr ≥ 309 μmol/L or >2 days from ICU admission
Demirkilic 22 (2004)	Retrospective comparison of practice	ARF post cardiac surgery	SCr > 5 mg/dL or K > 5.5 mEq/L	UO < 100/8h
Elahi 23 (2004)	Observational	ARF requiring RRT post cardiac surgery	Urea ≥ 30 mmol/L, SCr ≥ 250 mmol/L or K > 6.0 mEq/L	UO < 100/8h
Sugahara 24 (2004)	RCT	AKI post CABG	UO < 30 ml/h for 3 h or <750 ml/d	UO < 20 ml/h for 2 h or <500 ml/d
Manche 25 (2008)	Retrospective comparison of practice	AKI requiring RRT post cardiac surgery	UO < 0.5 ml/kg/h	Refractory hyperkalemia
Bouman 26 (2002)	RCT	UO < 30 ml/h for 6 h and Ccr < 20 ml/min	within 12 h	urea > 40 mmol/l, K > 6.5 mmol/L or severe pulmonary edema
Shiao 27 (2009)	Observational	AKI requiring RRT post major abdominal surgery	RIFLE stage R	RIFLE stage I or F
Zarbock 29 (2016)	RCT	AKI with KDIGO stage 2	within 8 h of KDIGO stage 2	within 12 h of KDIGO stage 3 or urgent need
Gaudry 30 (2016)	RCT	ICU patients with severe AKI requiring RRT.	within 6 h of stage 3 KDIGO;	Oliguria or anuria > 72 h; BUN > 112 mg/dL (40 mmol/L); or urgent indication
Bagshaw&Wald 31	RCT	AKI with KDIGO stage 2 or 3	within 12 h of eligibility;	urgent indications or per clinician discretion if AKI > 72 h
Wald 32 (2015)	RCT		within 12 h	standard indications
Barbar 36 (2018)	RCT	RIFLE-Failure stage	within 12 h	>48 h or urgent criteria
Payen 37 (2009)	RCT	Sepsis with SAPS > 35	within 24 h	urgent indication
Combes 38 (2015)	RCT	Shock post cardiac surgery	within 24 h	AKI-AKIN stage 3, urea > 36 mmol/L, need for ECMO or life-threatening hyperkalemia
Durmaz 39 (2003)	RCT	Cr > 220 umol/L post CABG	Postoperative creatinine > 10% within 48 h	≥50% rise in SCr or exceeds 420 umol/L, UO < 400 ml/24 h or urgent indication

Abbreviations: AKI – Acute Kidney Injury; RRT – Renal Replacement Therapy; ARF – Acute Renal Failure; CABG – Coronary Artery Bypass Grafting; PICARD – Program to Improve Care in Acute Renal Disease; BUN – Blood Urea Nitrogen; SCr – Serum Creatinine; ICU – Intensive Care Unit; UO – Urine Output; RCT – Randomised Controlled Trial; Ccr – Creatinine Clearance; RIFLE – Risk, Injury, Failure, Loss, End-stage renal disease; KDIGO – Kidney Disease: Improving Global Outcomes; ECMO – Extracorporeal Membrane Oxygenation; SAPS – Simplified Acute Physiology Score; AKIN – Acute Kidney Injury Network.

Is Earlier RRT Initiation Beneficial or Harmful?

There have been many attempts to relate the timing of RRT initiation to clinical outcomes. Older data, summarized in two meta-analyses from 2008 and 2010,^33,34 pointed to a benefit of early RRT in reducing the risk of mortality and of RRT dependence at hospital discharge. However, these meta-analyses were largely comprised of observational studies and included only patients who actually started RRT. In clinical practice, it is well known that even patients with the most severe degree of AKI may die or experience kidney function recovery without receiving RRT. Hence any assessment of the timing of RRT initiation in AKI must incorporate the clinical scenarios of death and kidney recovery before RRT is even commenced.

Over the past decade, 4 randomized trials aimed to examine the effect of a strategy of earlier RRT initiation, versus a structured delay in RRT initiation, on patient-relevant outcomes. Though the eligibility criteria and specific triggers for RRT initiation differed in each study, all of the trials had a common over-arching objective: to determine whether, in patients with severe AKI but no current emergent indications for RRT initiation, a strategy of pre-emptive RRT initiation leads to a reduction in mortality.

The Early versus Late Initiation of Renal Replacement Therapy in Critically Ill Patients with Acute Kidney Injury (ELAIN) trial was a single centre RCT that recruited 231 participants with KDIGO Stage 2 AKI. The vast majority of participants experienced AKI in the post-surgical setting or following trauma. ²⁹ Participants were randomized to commence RRT immediately following enrolment (early) as opposed to commencing RRT upon reaching Stage 3 AKI or the development of a clinical urgency (delayed). ELAIN demonstrated that early RRT initiation conferred a significant reduction in mortality through 90-days (HR 0.66 [95%CI, 0.45 to 0.97]). In addition, the early arm led to a reduction in RRT duration and hospital length of stay. The benefits attributed to early RRT initiation- including mortality and the composite outcome of major adverse kidney events (sustained loss of kidney function, dialysis dependence or death)- were sustained at one year. ³⁵

The Artificial Kidney Initiation in Kidney Injury (AKIKI) trial was a multicentre trial of 620 critically ill patients with KDIGO stage 3 AKI. ³⁰ Participants were randomized to immediate RRT initiation (early) versus a strategy of deferral whereby RRT was only commenced if a clinical urgency developed, blood urea nitrogen exceeded 112 mg/dL or oligoanuria persisted for 72 h. There was no difference in 60-day mortality between both groups. Notably, nearly half of the patients in the delayed-strategy group did not receive RRT and the rate of catheter-related bloodstream infections was significantly lower than in the early-strategy group.

The Initiation of Dialysis Early Versus Delayed in the Intensive Care Unit (IDEAL-ICU) trial was a multicentre trial in France that recruited patients with severe AKI in the setting of septic shock. Participants were randomized to receive RRT within 12 h of meeting criteria for Stage 3 AKI or delayed RRT, whereby RRT was commenced in the event of a clinical urgency or after 48 h if renal recovery had not occurred. ³⁶ The trial was stopped early for futility. The primary outcome of 90-day all-cause mortality and nearly all secondary outcomes were not modified by earlier RRT initiation.

The multinational Standard versus Accelerated Initiation of Renal Replacement Therapy in Acute Kidney Injury (STARRT-AKI) trial randomized 3019 patients with severe AKI, defined as KDIGO stage 2 or 3, to accelerated versus standard RRT. ³¹ In the accelerated arm, RRT was initiated within 12 h after the patient met the eligibility criteria, whereas in the standard arm RRT was discouraged unless emergency indications for RRT developed. If AKI persisted for > 72 h, clinicians could initiate RRT at their discretion. Accelerated RRT did not confer a reduction in mortality. RRT-related adverse events and RRT dependence at 90 days were higher in the accelerated arm.

Two other studies failed to show a benefit of early RRT for patients with sepsis and either AKI or other organ dysfunction.^26,37 Combes et al enrolled patients who were within 24 h post-cardiac surgery and in severe shock requiring high-dose catecholamines. ³⁸ High-volume hemofiltration for 48 h, followed by standard volume hemodiafiltration until resolution of shock and recovery of renal function, neither lowered mortality at 30 days nor impacted on other secondary outcomes, compared to a conservative strategy. Conversely, two smaller studies that focused on patients undergoing cardiac surgery were able to demonstrate significant survival benefits with a strategy of prophylactic perioperative dialysis.^24,39

Despite the variability in patient populations, designs and methods of these studies, the bulk of the evidence suggests that a strategy of early initiation of RRT (before emergency indications are met) does not improve outcome. This has been summarized by updated meta-analyses and a Cochrane database systematic review.^17,40,41 It is possible that the potential positive effect of early RRT was attenuated by the demonstrated increased risks of dialysis, such as hemodynamic instability, ³¹ subclinical myocardial ischemia with cardiac stunning,^42,43 infection risk, ³⁰ electrolyte disturbances ³¹ and loss of residual kidney function. ⁴⁴

Remaining Questions

While the aforementioned trials generally demonstrated that an earlier strategy of RRT initiation did not confer improved clinical outcomes in a broad population of critically ill patients, several questions remain. Specifically, in patients with non-resolving AKI but with no evidence of “classic” indications for RRT initiation, how long is RRT deferral acceptable? What effect should fluid status have on the decision to initiate RRT? Moreover, how should AKI be managed when RRT is being deferred?

How Long can we Wait?

While recent evidence supports the initial deferral of RRT in severe AKI, what should clinicians do when AKI persists in the continued absence of classic indications for RRT initiation? In the aforementioned RCTs, the average duration from enrolment to RRT initiation in the delayed arm (for those who commenced RRT) ranged from 25 to 57 h.^29–31,45 It is thus unclear if the protracted deferral of RRT can be justified when AKI persists for 3 days or longer. The AKIKI-2 trial addressed this dilemma and compared two strategies for delayed RRT in patients with AKI. ⁴⁶ The trial focused on patients with unremitting Stage 3 AKI who were oliguric for more than 72 h or had a BUN > 112 mg/dL, at which point they were randomized to start RRT or to further watchful waiting (“more delayed” strategy) until an emergent indication emerged or when BUN exceeded 140 mg/dL. Although fewer patients in the more-delayed strategy received RRT (79% vs 98% in the delayed strategy, p < 0.001), the overall number of RRT-free days was comparable in both groups. Moreover, there was a trend to higher mortality with the more-delayed strategy (55% vs 44% in the delayed strategy, p = 0.071), and a multivariable analysis revealed significantly higher adjusted 60-day mortality. The more-delayed strategy was also associated with worse neurological outcomes in the subgroup of patients who were non-responsive to verbal or painful stimulation at randomization. ⁴⁷

Volume Expansion and the Decision to Start RRT

It may be reasonable to assume that a high degree of fluid accumulation and overload could be an indication to commence RRT even in the absence of florid pulmonary edema. Fluid accumulation has been independently associated with multiorgan dysfunction and death, while restrictive fluid management and de-resuscitation with fluid offloading were associated with improved patient outcomes.^48–51 This was explored in a post hoc analysis of the STARRT-AKI trial ⁵² which examined regional variations in clinical practice and outcomes. Participants from Australia and New-Zealand (ANZ) had significantly lower maximal fluid balance at 14 days (+2211 ml compared to +5641 ml in patients from Europe and +7199 ml in patients from North America, p < 0.001), and a negative fluid balance of −493 ml from randomization to 14 days, compared to +2900 ml and +2018 ml in patients from Europe and North America, respectively. Participants from ANZ were significantly more likely to be alive at ICU and hospital discharge as well as at 90 days and enjoyed significantly more ICU-, hospital-, RRT- and ventilator-free days. Moreover, in two prospective observational multicentre studies, fluid overload at the time of RRT initiation, which was defined as an increase in cumulative fluid balance adjusted for body weight of greater than 10% (%FO > 10%) from baseline, was associated with a 2-fold risk for mortality and a reduced likelihood of kidney function recovery.^53,54 For this reason, one might expect that the earlier initiation of RRT may be beneficial in those with a greater positive fluid balance. However, in the STARRT-AKI trial the effect of randomization to the accelerated strategy was not modified by the degree of fluid accumulation at randomization. ⁵⁵ It should nonetheless be noted that the STARRT-AKI was largely devoid of patients with severe fluid overload at the time of randomization (mean %FO ∼ 4%) likely because patients perceived to require immediate RRT due to volume excess were excluded from the trial. Among IDEAL-ICU trial participants, 35.5% of participants met AKI criteria by oliguria alone and had the greatest fluid balance compared to other groups. Early RRT in those patients was associated with a non-significant 10% absolute decrease in 90-day mortality. ⁴⁵

How to Manage AKI-Related Complications Without RRT

Adoption of a conservative strategy of RRT deferral in the setting of severe AKI mandates the need to manage AKI complications safely and effectively.

Diuretics play an important role in maintaining fluid balance and avoiding excessive fluid accumulation in patients with AKI. More recent data from the STARRT-AKI cohort revealed that in patients from ANZ, who demonstrated the most favourable survival rates, the use of diuretics was highest (41.9% of patients compared to 25.3% and 34.2% in patients from Europe and North America respectively, p < 0.001) and their average time between randomization and RRT was the longest. ⁵⁵ A recent pilot trial compared a strategy of active fluid restriction, coupled with diuretics to mitigate fluid excess, versus usual care, in critically ill patients with AKI. Cumulative fluid balance was approximately 1.1 L lower in the restrictive fluid management arm at 72 h following enrolment. The restrictive fluid management approach conferred a 60% lower risk for RRT initiation at 14 days from enrolment. ⁵⁶ The importance of fluid stewardship, comprising careful consideration of fluid input complemented by diuretic use, is increasingly recognized in critically ill patients and is likely to be particularly vital for patients with severe AKI. ⁵⁷

Hyperkalemia refractory to medical management is a dangerous complication of AKI. The reversal and/or prevention of severe hyperkalemia is necessary for continued deferral of RRT. The discontinuation of hyperkalemia-promoting drugs, a low potassium diet, the management of acidosis and diuretics are all crucial to medical management of hyperkalemia. Sodium zirconium cyclosilicate (SZC) and patiromer are novel potassium binders and have been shown to be effective in the management of hyperkalemia.^58–61 Though no trials have been conducted in the setting of AKI, given their mechanism of action, safety profile and proven efficacy, it seems reasonable to utilize them as an adjunct to treat or prevent hyperkalemia in selected critically ill patients with AKI and a functioning GI tract, aiming to prevent the need for RRT as a result of hyperkalemia.

Worsening metabolic acidosis can be mitigated by the infusion of sodium bicarbonate. Correction of metabolic acidosis may counterbalance the deleterious consequences of severe acidosis on myocardial contractility, systemic vasodilatation, tissue perfusion, or cellular function, thus offering better conditions to allow for kidney function recovery.^62,63 Indeed, this strategy was found to be beneficial in the BICAR-ICU multicentre RCT. ⁶⁴ The study enrolled 389 critically ill patients with severe metabolic acidosis and compared administration of bicarbonate 4.2% by infusion to no bicarbonate infusion. In the subgroup of patients who had AKI with stage 2 or 3 AKI at the time of randomization (47% of patients), sodium bicarbonate infusion decreased the need for RRT during the ICU stay and resulted in fewer deaths by day 28. A weak recommendation for the use of sodium bicarbonate in adult patients with septic shock, severe metabolic acidosis (pH ≤ 7.2) and severe AKI was incorporated in the Surviving Sepsis Campaign guidelines. ⁶⁵ Two ongoing RCTs, the BICAR-ICU 2 (NCT040106300) and the MOSAIIC (ISRCTN14027629) trials, are likely to shed more light on the role of bicarbonate in the setting of critical illness complicated by AKI.

Identifying Patients who Will Require RRT

A major challenge in the study of RRT timing in AKI is the significant rate of spontaneous kidney recovery and non-receipt of RRT. By virtue of the significant number of patients who never received RRT in the delayed arms of the AKIKI, IDEAL-ICU and STARRT-AKI trials it would seem that, in retrospect, a substantial proportion of patients who received RRT in the early arms of those trials may not have needed it.^30,31,36 In patients with severe AKI, the ability to distinguish between patients with AKI progression and a subsequent need for RRT from those who will experience kidney recovery would be a vital contribution to patient care. ⁹ AKI biomarkers can be broadly categorized as follows: markers of glomerular filtration, markers of tubular integrity and markers of inflammation. ⁶⁶ While some studies have suggested that some biomarkers have the potential to distinguish patients in whom an indication for RRT will be more likely, ⁶⁷ a more recent meta-analysis of 41 studies that examined 13 different biomarkers, could not identify a single biomarker that could reliably predict the need for RRT. Cystatin C showed the best predictive value, followed by neutrophil gelatinase-associated lipocalin (NGAL), but their performance did not significantly exceed that of serum creatinine. ⁴⁴ In the RUBY study which included critically ill patients with stage 2 or 3 AKI, elevated urinary C-C motif chemokine ligand 14 (CCL14) was the most robust predictor of persistent stage 3 AKI among a wide range of biomarkers, with more than 60% of patients with levels at the highest tertile of CCL14 concentrations ultimately requiring RRT. ⁶⁸ A plasma NGAL > 150 ng/ml was incorporated into the inclusion criteria of the ELAIN trial presumably to enrich the cohort with individuals with more severe AKI. ²⁹ Notably, only 3/604 patients with Stage 2 AKI who were screened for the trial were excluded due to low NGAL which casts doubt on the specificity of this threshold. In analysis of patients with severe community acquired pneumonia and severe AKI, plasma NGAL was able to predict failure to recover kidney function (defined as need for RRT or persistent RIFLE-F criteria at hospital discharge), but failed to outperform a clinical model that included age, serum creatinine, pneumonia severity, and nonrenal organ failure. ¹⁴ Furthermore, although it was feasible to use NGAL to triage patients with severe AKI, early initiation of RRT in this high-risk group did not affect 28-day mortality when compared to standard AKI management. ⁶⁹ The combination of urinary TIMP-2 × IGFBP-7 showed good predictive ability for the outcome of AKI requiring RRT in 3 small studies involving kidney transplant recipients, recipients of percutaneously inserted aortic valves and surgical patients, respectively.^70–72

The furosemide stress test (FST) is considered a functional test that can shed light on renal tubular functional capacity. Two studies showed that “non-response” to furosemide administration (1 mg/kg if furosemide naïve or 1.5 mg/kg in patients with prior exposure) reliably predicted progression of AKI and receipt of RRT, with the ideal threshold for “non-response” identified as urine output < 200 mL at 2 h.^73,74 A more recent multicentre pilot study incorporated the FST into a trial of early initiation of RRT. ⁷⁵ The researchers were able to demonstrate the protocol's feasibility, with 100% completion of FST in all patients, and patients who were non-responders were then randomized to early initiation of RRT within 6 h, or delayed RRT until a mandatory indication developed. In the delayed RRT groups, 75% of FST non-responders eventually met an urgent indication and received RRT, compared to only 13.6% in FST responders. These data support the ability of this simple and widely-available tool to anticipate the future receipt of RRT in patients with AKI.

In this era of prolific scientific research and fast-evolving artificial intelligence technology, one might wish to imagine a future in which novel investigations and machine-learning models of prediction could help refine RRT timing decision, however we currently need to rely on existing knowledge to navigate the timing debate. To this end, Figure 1 offers a flowchart, mapping out key decision points for RRT initiation in critically-ill patients with AKI and offering suggestions for circumstances in which the initiation of RRT might be considered.

Figure 1.

A flowchart mapping out key decision points for RRT initiation in critically-ill patients with AKI.

Summary

The preponderance of evidence has demonstrated that in critically ill patients with severe AKI and no urgent indications for RRT initiation, pre-emptive or early RRT initiation is not beneficial and may even be harmful. Recognizing that many patients with AKI will recover kidney function spontaneously, deferral of RRT with watchful waiting is a reasonable initial approach. Avoidance of nephrotoxins, medical treatment of hyperkalemia, optimization of acid-base status and mitigation of fluid overload with fluid stewardship and diuretics enable the safe deferral of RRT in the hope that endogenous kidney function will recover. In the setting of unremitting AKI, notwithstanding the absence of an AKI-related emergency, it is unclear how long RRT can be safely deferred. Preliminary evidence suggests possible harm if RRT is deferred beyond 72 h. Further research is needed to determine whether certain sub-populations benefit from earlier RRT initiation and whether novel biomarkers can better inform the care of this high-risk population.