How Late it was, How Late!

More information
The AKIKI-2 Visual Abstract
The AKIKI-2 Visual Abstract
How Late it was, How Late!
How Late it was, How Late!

This week, we will discuss the (probably) final story in the timing of RRT in AKI story. We discussed ELAIN, AKIKI and STARRT AKI. Early dialysis is not a good idea. But how late can you go? Let’s see with AKIKI-2

Introduction

Acute kidney injury (AKI) in critically ill patients is bad. Once they require dialysis, half of them die. Would starting dialysis earlier be helpful? The timing of initiation of kidney replacement therapy (KRT)  in AKI has hence been a matter of debate for several years. Observational studies are fraught with problems: lead time bias, and the fact that starting earlier doesn’t help you understand who, perhaps, didn’t even need it. While there is little doubt about the need for initiating KRT in those with life-threatening complications of AKI, the use and timing of KRT in those with severe AKI but no such clear indication for KRT remains to be answered. Are objective criteria useful in this setting?

Theories underpinning early dialysis are a postulated benefit of avoiding hypervolaemia, eliminating toxins early, maintaining electrolyte balance, and preventing other complications that may be attributed to AKI. (Check out the introduction of this NephJC blog on the STARRT-AKI trial  for a bit more history) However, this strategy might lead to initiation of KRT in patients who would have otherwise spontaneously recovered kidney function without genuine need, thus exposing them to its risks and preventing efficient use of a limited and expensive resource. Among the possible KRT complications are those associated with KRT itself (e.g. catheter insertion, infections, hypotension) as well as the fact that some of the patients initiated on KRT develop repeated hypotensive injury, and remain dialysis dependent (see Douvris et al, Int Care Med 2019). Over the last 5 years several randomised clinical trials (RCTs) have made an attempt to answer these questions. This week on NephJC we discuss the AKIKI 2 trial (Gaudry et al, Lancet, 2021) which builds on what we have learnt from these recent trials to bring us closer to answering this question.

The important trials on KRT timing in AKI to date include AKIKI (Gaudry et al, NEJM 2016), ELAIN (Zarbock et al, JAMA 2016), IDEAL-ICU (Barbar et al, NEJM, 2018)  and STARRT-AKI (Wald et al, NEJM, 2020). These studies and their findings have been summarised in Table 1. 

You can find some great visual abstracts created by the NSMC interns here and a discussion on some of these trials by the Renal Fellow Network (RFN) here. Check out the NephJC summary on STARRT-AKI here. 

A summary of the recent RCTs on timing of initiation of KRT in AKI. *All studies excluded patients with an urgent indication for KRT

KDIGO Stage 3 AKI= Serum creatinine 3x the baseline or ≥ 4.0 mg/dl accompanied by a rapid rise ≥ 0.3mg/dl OR urine output < 0.3ml/kg/h for ≥24 hours or anuria for for ≥12 hours 

RIFLE (Failure)= Serum creatinine 3x baseline or ≥ 4.0 mg/dl accompanied by a rapid increase of ≥ 0.5mg/dl OR urine output <0.3ml/kg/h for ≥24 hours or anuria for ≥12 hours 

AKI= Acute Kidney injury, KRT = Kidney Replacement Therapy, UOP= Urine output, BUN= Blood urea nitrogen, CRBSI= Catheter related blood stream infection

Apart from the small, single-centre ELAIN study, all the above RCTs suggested there was no benefit in an earlier vs a later strategy for initiating KRT. An individual patient-level meta-analysis (including all 10 major trials except STARRT-AKI) showed that the timing of KRT initiation does not affect survival in critically ill patients with severe AKI and that delaying KRT initiation, with close patient monitoring, might lead to a reduced use of KRT, thereby saving health resources.(Gaudry et al, Lancet, 2020) But how late might too late be? The time to KRT initiation in the earlier and later strategies were very variable across the trials. 

A comparison of the timing at which  KRT was initiated in the two arms of each of the landmark RCTs, AKKI, ELAIN, IDEAL-ICU and STARRT-AKI. Note in the ELAIN trial, even in the ‘delayed’arm, the plan was always to start KRT, not wait for a clinical indication. Median time refers to time from randomisation/ meeting eligibility criteria

So, from the story so far, delaying RRT in AKI is not harmful, and in fact may prevent complications related to RRT, and reduce RRT dependence in the long run. so, can delaying dialysis even further, continue to provide additional benefits? How far can we push the envelope? Enter AKIKI-2.

The Artificial Kidney initiation in acute Kidney Injury 2 (AKIKI 2) trial, a comparison of two delayed strategies for KRT initiation for severe AKI asks: how long can KRT be postponed in severe AKI without incurring risk when there are no life-threatening complications of AKI? The preceding RCTs used BUN levels and duration of oliguria as indications for KRT, but the limits were arbitrary. 

The AKIKI 2 trial was designed as a response to the first AKIKI trial. The AKIKI trial had the longest delay in the “delayed” arm (57 hours) with the lowest percentage of KRT initiation in the delayed group (51%). The AKIKI 2 trial studied whether a delayed strategy as defined in the AKIKI trial had inferior outcomes to a more-delayed strategy where KRT was only initiated for life-threatening emergencies or BUN >140 mg/dl, without any consideration of oliguria. The primary outcome was the number of KRT-free days. 

The study

Study design

A multicentre, open-label, two-arm, randomised, controlled trial done in 39 intensive care units in France

Study design from Gaudry et al, Trials, 2019

Study population

Inclusion criteria

  • Age >18 years

  • Receiving ICU care - on mechanical ventilation or catecholamine infusion

  • KDIGO Stage 3 AKI 

  • Oliguria or azotemia

    • oliguria: urine output < 0.3 ml/kg/h or < 500 ml/d) or anuria (urine output < 100 ml/d) for > 72 hours

    • azotemia: blood urea nitrogen concentration between 112 mg/dl (40 mmol/l) and 140 mg/dl (50 mmol/l)

Exclusion criteria (key criteria)

  • Patients presenting with an urgent indications for KRT

  • Blood urea nitrogen level greater than 140 mg/dL

  • Severe chronic kidney disease (creatinine clearance < 30 ml/min) 

  • Patients with inclusion criteria already present for more than 24 hours (to avoid delayed inclusions)

  • AKI caused by urinary tract obstruction or renal vessel obstruction or tumour lysis syndrome or thrombotic microangiopathy or acute glomerulopathy

  • Poisoning by a dialyzable agent

  • Child C liver cirrhosis 

  • Moribund state (patient likely to die within 24h)

  • Kidney transplant

eTable 1: Criteria for considering KRT at anytime before or after randomisation 

Randomisation and masking

The randomisation list was computer generated, balanced by blocks of variable and undisclosed size, and stratified by centre. 

Patients were randomly assigned in a 1:1 ratio to a delayed or more delayed strategy

Procedures

The delayed strategy

KRT to be initiated within 12 hours of fulfilling randomisation criteria

The more-delayed-strategy

KRT postponed until an urgent indication occurred (See e-table 1) or BUN reached 140 mg/dl for one day. (note that duration of anuria was not a criterion)

Steps to prevent dialysis disequilibrium were recommended when the BUN >112 mg/dl. These included use of slow haemodialysis initially, use of higher dialysate sodium concentration, use of high glucose concentration dialysate, and administration of hypertonic glucose in the venous line of the dialyser during dialysis. 

KRT modality, duration, intervals, anticoagulation and other management issues related to KRT were left to the discretion of each study site and were prescribed and monitored according to national guidelines. 

Discontinuation of KRT was considered if spontaneous diuresis was 500 ml/d or more, and highly recommended if it went over one liter a day (two liters if on diuretics). Discontinuation was mandatory if serum creatinine was falling spontaneously.

Patients were followed up for 60 days.

Primary outcome

The primary outcome was the number of KRT-free days between randomisation and day 28. (Each patient received one point for each day he/she remained alive and free of KRT, provided the patient remained free of KRT and survived 3 at least calendar days after having been weaned off KRT. )

Secondary outcome

There were many! See below, with some key ones highlighted

  • Vital status (eg survival) at ICU and hospital discharge, day 28 , and day 60

  • Percentage of patients receiving KRT at least once

  • Number of KRT sessions between randomisation and day 28

  • Time between inclusion in the observational stage and KRT initiation

  • Number of patients with renal recovery between randomisation and day 60

  • Number of ventilator-free and catecholamine-free days between randomisation and day 28

  • Duration from randomisation to ICU or hospital discharge

  • Reason for initiation of KRT

  • KRT modalities and duration

  • Number of dialysis catheter-free days between randomisation and day 28

  • Rate of catheter-related bloodstream infection

  • Barthels activities of Daily Living Index at day 60

  • Complications potentially related to AKI or KRT

  • Number of patients with treatment limitation

  • Hydration

  • Nutritional status

Statistical analysis

Based on an expected number of KRT free days at day 28 in the delayed arm of 17 days, as derived from the AKIKI study, a sample size of 270 (135 per group) was calculated to detect an increase of 4 days (approximately 25%) in the more-delayed arm with a power of 80% and accepting a drop-out rate of 5%. 

An interim analysis was done at day 28 for survival and complication rates. (This did not include the primary outcome).

Funding

The study was funded by a grant from the French Ministry of Health (similar to AKIKI). The funder had no role in the study design, data analysis, data interpretation or writing of the manuscript.

Results

During the trial period (May 7 2018- October 11 2019) 5,336 patients assessed, 767 met criteria, and among these 37% (278) were randomized.

Figure 1 Gaudry et al, Lancet 2021

Of the 278 undergoing randomization, 137 were assigned to delayed strategy with a mean start of 44 hours, and 141 to more-delayed strategy with a mean start of 94 hours (for 79% who started KRT). 49% patients received diuretics in the first 7 days (50% in delayed and 48% more-delayed group).

Patient characteristics at baseline

69% of the included cohort was male. 10% had a history of chronic renal failure, (10% delayed vs 12% more-delayed), 25% had diabetes (29% delayed vs 22% more-delayed). The remainder of patient characteristics are outlined in Table 1.

Patient characteristics at time of monitoring/randomization

Table 1 lists patient characteristics and lab values at time of monitoring vs time of randomization ~55%  were in septic shock, three-quarters required mechanical ventilation, which increased slightly to 82% in both groups at the time of randomization.  80% required vasopressor support at the onset of monitoring, decreased to 60-70% by the time of randomization. Creatinine was ~5-6 mg/dL at onset of randomization with a BUN ~100. Potassium was ~4.5 with a bicarb ~18. 

Table 1 from Gaudry et al, Lancet 2021

Note that the characteristics differed considerably from those at the time of KRT initiation (eTable 5).

Within the more-delayed group, 111/141 (78%) ended up ultimately receiving KRT. The most common reason for initiation was:

  • BUN >140 (59%)

  • K concentration (20%)

  • pH <7.15 (13%)

  • pulmonary edema (10%)

Primary outcome

Number of KRT free days from days 0-28 did not differ between the delayed strategy (12 days) and more delayed strategy (10 days).  Though the 60 day mortality did not reach significance, 44% in the delayed group died, vs 55% in the more delayed group (p = .071), this changes when analysing the time to event, and adding in the other covariates. 

In the multivariable analysis, risk factors associated with 60 day mortality included

  • more-delayed strategy (HR 1.65)

  • simplified Acute Physiology Score III (HR 1.03)

  • mechanical ventilation (HR 3.44)

Use of catecholamine infusions, sepsis status, or time from ICU admission and AKI did not show a significant difference.

Table 3 Gaudry et al Lancet 2021

Secondary Outcomes and Adverse Events

KRT dependence at day 60 and complications related to AKI or KRT (including hemorrhage, thrombocytopenia, thrombosis, hypo/hyperkalemia, hypo/hypernatremia, hypophosphatemia, rhythm disorders, pneumo/hemothorax, air embolism, AVF, pericarditis, unexpected cardiac arrest, hypothermia) did not differ between groups. Fluid balance at 2 and 7 days did not differ between groups. Note that this is different from AKIKI and STARRT-AKI, where the delayed arm had fewer complications. 

Other secondary outcomes did not differ between groups. Amongst patients with a BUN >112 or oligo-anuria >72 hours (ie, patients who would have been started on KRT in the STARRT-AKI trial), there was also no statistically significant difference in RRT free days and mortality at day 60. However these are smaller and smaller subgroups with limited power. 

Discussion

AKIKI-2 builds off many prior trials (ELAIN, IDEAL-ICU, AKIKI, STARRT AKI) examining the timing of KRT for severe acute kidney injury. Only one of the major trials, (ELAIN-AKI) showed a survival advantage, but it was a single centre trial of very early compared to early KRT, and both arms did get KRT, thus negating any advantage of the delay which seems to stem mostly from avoiding KRT and its complications. In contrast, AKIKI, IDEAL-ICU, and most recently, STARRT-AKI showed early dialysis for AKI did not confer any survival advantage and in fact, early KRT posed serious harms, including adverse events related to KRT initiation, and delayed kidney recovery. Refer to the NephJC summaries for a review of STARRT-AKI  and prior trials .

Building off the findings in these trials, AKIKI-2 sought to answer the logical next question: How late is too late?

Overall, AKIKI-2 employed two strategies for KRT initiation in severe AKI, the standard “delayed strategy” as employed in prior studies, and a more delayed strategy designed to postpone KRT initiation even longer. Further delay in KRT did not show significant difference in KRT-free days or 60-day mortality between the two strategies. The multivariable analysis did reveal the 60-day mortality was higher with more delayed strategies.  

Strengths

  • Real World Adjacent: One criticism of other trials was lack of applicability to the real world. The criteria for initiation in STARRT AKI (the largest and biggest study of its kind) included oligo-anuria and BUN <112. AKIKI-2 removed the “oliguria” criteria as an indicator for dialysis initiation and set the BUN threshold higher to 140 mg/dL, which may - or may not be more representative of clinical practice.  

  • Less variability: Another criticism of other trials included the wide variability on timing for KRT initiation in the delayed groups. In these studies, the longer the duration of delay of therapy initiation, the higher the percentage of patients who avoided KRT entirely. The small number of patients who avoided dialysis entirely in AKIKI-2 illustrates the “upper limit” of waiting on dialysis- though notably most were started for BUN elevation (see below).

Limitations

  • Small sample size: This was a multi-center study with 279 patients participating across ICU in France. A much larger trial (Sarah Faubel’s DreamRCT) has proposed an ideal study should likely be powered for a cohort of 2000 patients. The authors note this study may be underpowered to detect an increase of 25% of the number of KRT -free days.

  • Cohort: While the cohort was relatively representative of patients in most ICUs in high-income countries, it was not comparable to prior studies. For example, 45% of the STARRT-AKI cohort had chronic kidney disease, compared with 10% in the AKIKI-2 study. STARRT-AKI also had slightly higher proportion of diabetes and heart failure in their cohort- the AKIKI-2 cohort appeared healthier prior to their acute illness. However, patient factors at the time of enrollment were relatively similar to STARRT-AKI, including mechanical ventilation, use of vasopressors, and laboratory values.

  • The debate over BUN: the applicability of BUN as a marker of uremia has long been debated. Though some experts feel a BUN >100 absolutely mandates KRT initiation, the clinical benefit remains unclear in the absence of objective data. Notably, most of the patients in the more-delayed trial were started for a BUN >140. The authors noted that the lack of mortality benefit observed between delayed and more-delayed thus indicated the “objective credence to the limit that the concentration of BUN should not exceed”. However, given that this magic number has not yet been determined, it is unclear whether more patients in the more-delayed group would have avoided dialysis if this [high] threshold had been pushed further.

Where do we go from here? Does this mean we should use a lower BUN as a threshold for initiation? Whoever uses BUN as a criteria anyways? Other clinical criteria such as hyperkalemia and volume overload remain important. The investigators in AKIKI-2 decided to ignore oliganuria as an important criteria by itself to initiate. And though it is not a reason to start, one does know that when anuria sets in, hyperkalemia and pulmonary edema are not far behind.

Most importantly, how can we identify the point of no return? When we can be confident that recovery is not possible, and the patient needs KRT? Incorporation of an ideal biomarker (with Nephrocheck and the furosemide stress test as possible front runners) may eventually be helpful.  

Conclusions

This is the first trial to illustrate the upper limit on delaying KRT and what risk factors may be associated with poor outcomes when KRT is delayed.  Overall, the benefit of delaying KRT is avoiding harm in patients who may otherwise not need it. AKIKI-2 trial pushed out dialysis further than any trial had before. Overall, it did provide reasonable assurance that it is safe to postpone KRT in patients with stage 3 AKI, prolonged oliguria, and BUN <112 without other urgent indications.

The authors note that although the mortality endpoint was not significant, the p value of 0.071 may be a “warning signal” against the more delayed strategy. Compared to prior trials, it did not show the benefit of KRT-free days observed in STARRT AKI. Notably, only 21% of the more-delayed group ended up avoiding dialysis entirely. Overall, the measures we currently use for AKI prognosis are still lacking, and until new evidence emerges, the decision to start or postpone KRT must be done on the individual basis.  


Summary by Dilushi Wijayaratne
Nephrologist, Colombo, Sri Lanka

and Katie Rizzolo,
Nephrology Fellow, University of Colorado

NSMC Interns, Class of 2021