Initiation Strategies for Renal-Replacement
Therapy in the Intensive Care Unit

Gaudry. N Engl J Med 2016; 375:122-133. doi: 10.1056/NEJMoa1603017

Clinical Question

  • In critically ill patients with acute kidney injury does delayed compared with early initiation of renal replacement therapy (RRT) reduce mortality at 60 days?


  • Randomised controlled trial
    • Computer-generated
    • Variable block size
    • Stratified by centre
  • Non-blinded
  • Follow up for 60 days post randomisation
  • Sample size calculation: sample size of 546 patients would have 90% power to demonstrate a 15% absolute risk reduction from a baseline mortality of 55%, with a false positive rate of 5%.


  • 31 ICUs in France
  • September 2013 – January 2016


  • Inclusion criteria: Critically ill adult patients with severe acute kidney injury
    • Age ≥18
    • Receiving invasive ventilation and/or catecholamine infusion (adrenaline/noradrenaline)
    • Admitted to the ICU with acute kidney injury that was compatible with diagnosis of acute tubular necrosis in the context of ischaemic or toxic injury
    • Within 5 hours of validation of KDIGO stage 3 kidney injury, defined by at least 1 of the following
      • Creatinine >354 µmol/litre or greater than 3* baseline
      • Anuria (urine output <100ml/day) for >12 hours
      • Oliguria (urine output <0.3ml/kg/h or <500ml/day) for >24 hours
  • Exclusion criteria:
    • Pre-existing severe chronic renal failure (Creatinine clearance <30ml/min)
    • Cardiac arrest without awakening
    • Severe laboratory abnormalities
      • Urea >40mmol/l, K >6mmol/l (or >5.5 despite medical treatment), pH <7.15 in context in context of either pure metabolic acidosis or mixed acidosis
    • Acute pulmonary oedema due to fluid overload causing severe hypoxaemia
      • Requirement of >5l/min oxygen to maintain SpO2 >95%, or requiring FiO2 >50% in ventilated patients despite diuresis
  • 620 patients randomised out of 5528 patients assessed

Comparing intervention vs control group, well matched at baseline apart from prothrombin ratio

  • Age: 64.8 vs. 67.4
  • Serum creatinine before ICU admission: 0.95 mg/dl vs. 0.97 mg/dl
  • Chronic renal failure: 7% vs. 12%
  • SOFA score at enrollment: 10.9 vs. 10.8
  • Exposure to nephrotoxic agent in past 2 days: 62% vs. 63%
  • Invasive mechanical ventilation: 86% vs. 87%
  • Vasopressor support with adrenaline/noradrenaline: 85% vs. 85%
  • Septic shock: 67% vs. 66%
  • Oliguria/anuria: 65% vs. 62%
  • Prothrombin ratio: 57.1% vs. 53.5%


  • Early renal replacement therapy (n=312)
    • Aimed to start within 6 hours of documentation of stage 3 KDIGO acute kidney injury
    • Started at median 4.3 hours (IQR 2.7-5.9) post documentation of stage 3 injury and fulfillment of other inclusion criteria
    • 6 patients did not receive RRT


  • Delayed renal replacement therapy (n=308)
    • Renal replacement therapy initiated if patient developed any of the following
      • Severe laboratory abnormalities: Urea >40mmol/l, K >6mmol/l (or >5.5 despite medical treatment), pH <7.15 in context in context of either pure metabolic acidosis or mixed acidosis
      • Acute pulmonary oedema due to fluid overload causing severe hypoxaemia (as defined in exclusion criteria)
      • Oliguria/anuria lasted for >72 hours post randomisation
    • 51% of patients received RRT at a median of 57 hours (IQR 25-83) post randomisation
    • Median interval between occurrence of at least 1 criteria mandating RRT and its initiation was 4.7 hours (IQR 1.7-10.0)
    • Five patients received RRT without meeting criteria

Comparing early vs. delayed group

  • Medical treatment of acute kidney injury related metabolic complication prior to receiving RRT
    • Diuretics: 1.3% vs. 36.5%, p<0.001
    • Medical treatment of acidosis: 6.8% vs. 16.7%, p<0.001

In both groups

  • Choice of method of RRT, duration, interval between sessions, anticoagulation, was at discretion of physicians
    • Over 50% received intermittent RRT
    • 30% received continuous renal replacement therapy (CRRT) as sole method
  • Discontinuation of RRT
    • Considered if spontaneous urine output >500ml/24 hours
    • Highly recommended if spontaneous urine output >1000ml/24 hours  (>2000ml/24 hours in patients receiving diuretics)
    • Mandatory if diuresis was sufficient to allow for spontaneous decrease in serum creatinine concentration
  • RRT resumed if diuresis insufficient to result in spontaneous decrease in creatinine level or if urine output <1000ml/24 hours (or <2000ml/24 hours in patients receiving diuretics)


  • Primary outcome: mortality at 60 days – no significant difference
    • 48.5% in early vs. 49.7% in delayed, P=0.79, Hazard ratio 1.03 (95% C.I. 0.82-1.29)
    • Fragility index: -21 patients (Number of patients lost to follow up = 6)
  • Secondary outcome: comparing early vs. delayed
    • Mortality at 28 days – no significant difference
      • 41.6% vs. 43.5%
    • Received renal replacement therapy – significantly higher in early group
      • 98% vs. 51%, p<0.001
    • Median length of ICU stay – no significant difference
      • Survivors: 13 days vs. 13 days, p=0.87
      • Non-survivors: 6 days vs. 6 days, p=0.92
    • Median length of hospital stay – no significant difference
      • Survivors: 29 vs. 32 days, p=0.58
      • Non-survivors: 6 vs. 6 days, p=0.85
    • Patients with catheter related nosocomial infection – significantly higher in early group
      • 10% vs. 5%, p=0.03
    • Dependency on RRT – no significant difference
      • At day 28: 12% vs. 10%, p=0.51
      • At day 60: 2% vs. 5%, p=0.12
    • Proportion of patients with diuresis of >1000ml urine /24 hours (or >2000ml if had diuretics) significantly higher in delayed group, p=0.003
  • Post-hoc analysis
    • Mortality at 60 days
      • Delayed RRT group but never received RRT: 37.1%
      • Early RRT: 48.5%
      • Delayed RRT group and received RRT: 61.8%
      • p<0.0001
    • SOFA score at baseline, median (IQR)
      • Delayed RRT group but never received RRT: 10 (8-12)
      • Early RRT group: 11 (9-13)
      • Delayed RRT group and received RRT: 12 (9-14)
      • p<0.0001

Authors’ Conclusions

  • In critically ill patients with severe acute kidney injury, an early compared with a delayed initiation of renal replacement therapy showed no difference in mortality.


  • Randomised controlled trial
  • Multi-centre
  • Adequately powered
  • Defined criteria for initiation of RRT


  • For over 50% of patients the initial modality of RRT was intermittent RRT. This was despite the fact that 85% of patients required vasopressor support upon initiation of RRT. This may limit the external validity as this is not standard practice in the UK.
  • A limitation that the authors state is that the study only included patients with advanced kidney injury, and therefore the results may not be generalisable to patients with less severe kidney injuries.
  • Non-blinded – although unavoidable the unblinded clinician also directed care which may introduce bias.

The Bottom Line

  • In critically ill patients with severe acute kidney injury, an early approach to RRT did not provide a mortality benefit compared to a delayed approach; when the RRT modality was predominately intermittent RRT.
  • ~50% of the patients in the delayed group did not receive RRT
  • There was a significantly higher rate of catheter related nosocmial infections in the early group

External Links


Summary author: @davidslessor
Summary date: 01.06.16.
Peer-review editor: @DuncanChambler


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.