Chapter 3 — Indications & The Timing Debate

This is the most consequential and most misread chapter in CRRT. The timing literature is not ambiguous — it is specific, and the specificity is the whole point.

3.1 The Absolute (Emergent) Indications

The classic mnemonic AEIOU:

💡 The most important word in the mnemonic is "refractory." AEIOU describes complications of renal failure, not degrees of renal dysfunction. A creatinine value is not on this list.

Dialysable toxins favour low molecular weight, low volume of distribution, low protein binding, and water solubility — classically toxic alcohols, salicylates, lithium, metformin-associated lactic acidosis, and valproate. Note that CRRT's low clearance makes it a poor choice for rapid toxin removal when intermittent HD is feasible; CRRT is used for these when haemodynamics forbid IHD or to prevent rebound.

3.2 Guideline Framing

KDIGO 2012 AKI:

• Initiate RRT emergently when life-threatening changes in fluid, electrolyte, and acid–base balance exist. (Not graded)

• Consider the broader clinical context, the presence of conditions that can be modified with RRT, and trends of laboratory tests — rather than single BUN and creatinine thresholds alone — when making the decision to start RRT. (Not graded)

Both statements are ungraded, which is itself the finding: KDIGO declined to define a numeric trigger because the evidence did not support one.

3.3 The Landmark Timing Trials

The single greatest source of confusion is treating these five trials as if they asked the same question. They did not. Each defined "early" and "delayed" differently, in different populations, against different primary outcomes.

AKIKI (Gaudry et al., NEJM 2016)

• Population: 620 patients with KDIGO stage 3 AKI, mechanically ventilated and/or receiving vasopressors, with no potentially life-threatening complication directly related to renal failure.
• Early arm: RRT initiated immediately after randomisation (within 6 hours of documenting stage 3).
• Delayed arm: RRT withheld until at least one of — severe hyperkalaemia, severe metabolic acidosis, pulmonary oedema from fluid overload causing severe hypoxaemia, BUN above 112 mg/dL (urea >40 mmol/L), or oliguria persisting beyond 72 hours.
• Primary outcome: 60-day overall survival.
• Result: No significant difference — mortality 48.5% early vs 49.7% delayed (p = 0.79).
• Key secondary finding: Approximately half of the delayed arm never received RRT at all. Catheter-related bloodstream infections were more frequent in the early group.

ELAIN (Zarbock et al., JAMA 2016)

• Population: 231 patients, single centre (Münster), predominantly surgical — with a large cardiac surgery component. KDIGO stage 2 AKI plus plasma NGAL above 150 ng/mL, plus at least one of: severe sepsis, vasopressor requirement, refractory fluid overload, or progressive non-renal organ dysfunction.
• Early arm: RRT within 8 hours of reaching stage 2.
• Delayed arm: RRT within 12 hours of reaching stage 3, or an absolute indication.
• Primary outcome: 90-day mortality.
• Result: Favoured early — 39.3% vs 54.7%, HR 0.66 (p = 0.03).
• Critical caveats: single centre; a highly selected surgical population; a biomarker-enriched entry criterion used in no other trial; and over 90% of the delayed arm did ultimately receive RRT — so ELAIN compared earlier with later, not start with maybe never.

IDEAL-ICU (Barbar et al., NEJM 2018)

• Population: 488 patients with early septic shock and RIFLE failure-stage AKI. Trial stopped early for futility.
• Early arm: RRT within 12 hours of documenting failure-stage AKI.
• Delayed arm: RRT deferred for 48 hours unless an emergent indication arose.
• Primary outcome: 90-day mortality.
• Result: No significant difference — 58% early vs 54% delayed (p = 0.38).
• Key secondary finding: Roughly 38% of the delayed group never required RRT.

STARRT-AKI (STARRT-AKI Investigators, NEJM 2020)

• Population: 3,019 patients — by far the largest — with severe AKI. Notably included a run-in period in which clinicians could exclude patients they felt strongly should or should not start RRT, which is essential to interpreting generalisability.
• Accelerated arm: RRT within 12 hours of eligibility.
• Standard arm: RRT discouraged unless a conventional indication developed or severe AKI persisted beyond 72 hours.
• Primary outcome: 90-day all-cause mortality.
• Result: No difference — 43.9% accelerated vs 43.7% standard (RR 1.00, p = 0.92).
• Key secondary findings — these are the ones that changed practice: among 90-day survivors, RRT dependence was higher in the accelerated arm (10.4% vs 6.0%), and adverse events were more frequent in the accelerated arm (23.0% vs 16.5%). About 62% of the standard arm received RRT.

AKIKI 2 (Gaudry et al., Lancet 2021)

• The question is different. Having established that delay is safe, AKIKI 2 asked whether delaying even further is better.
• Population: 278 patients with severe AKI who had already reached the AKIKI "delayed" trigger — oliguria beyond 72 hours or BUN 112–140 mg/dL — without an emergent indication.
• Delayed arm: RRT started at that point.
• More-delayed arm: RRT withheld until BUN exceeded 140 mg/dL or another emergent indication appeared.
• Primary outcome: Number of days alive and free of RRT at day 28 — not mortality.
• Result: No significant difference in RRT-free days. In adjusted analysis, the more-delayed strategy was associated with higher 60-day mortality (HR 1.65, 95% CI 1.09–2.50).
• Interpretation: there is a floor. Watchful waiting is safe; indefinite waiting is not.

3.4 Synthesis Table

⚠️ Note the outcome timepoints. AKIKI and AKIKI 2 report 60-day mortality; ELAIN, IDEAL-ICU and STARRT-AKI report 90-day. Comparing headline mortality percentages across these trials without noting the timepoint is a common and serious error.

3.5 Reconciling ELAIN

ELAIN is the lone positive trial and deserves a precise explanation rather than dismissal:

1. Single centre, 231 patients — the smallest of the five and the most susceptible to centre effects.
2. Biomarker-enriched entry (NGAL >150 ng/mL) selected a population with established tubular injury, unlike the others.
3. Predominantly surgical/cardiac surgical, a different AKI phenotype from septic shock.
4. Both arms overwhelmingly received RRT. ELAIN's real question was "sooner or later," while AKIKI, IDEAL-ICU and STARRT-AKI asked "now or possibly never."

This is not proof that ELAIN was wrong — it is the reason its result should not be generalised to the mixed medical ICU population studied by the larger trials.

3.6 What the Evidence Actually Supports

💡 The three defensible conclusions:

1. In the absence of an emergent indication, a watchful-waiting strategy is safe. No mortality penalty across the three large trials, and a substantial proportion of patients recover without ever needing RRT.

2. Starting early carries real costs — STARRT-AKI showed more adverse events and greater 90-day RRT dependence in the accelerated arm.

3. Waiting has a floor. AKIKI 2 found no benefit and a signal of harm from pushing delay beyond the conventional threshold.

⚠️ Declared gaps:

• No trial has validated a biomarker-guided initiation strategy against a clinical strategy. ELAIN used NGAL as an entry criterion, not as a randomised comparator. Biomarker-triggered initiation is unproven.

• No trial defines an optimal trigger for the individual patient. All five tested strategies in populations, not thresholds in individuals.

• Furosemide stress testing has been studied for predicting AKI progression, not for timing RRT initiation. Do not present it as a validated initiation trigger.

3.7 A Practical Bedside Framework

1. Emergent indication present? → Start now. No trial contradicts this; all five excluded such patients.
2. No emergent indication? → Watchful waiting is evidence-supported. Reassess frequently — clinical trajectory, urine output trend, fluid balance, acid–base, potassium.
3. Trajectory improving? → Keep waiting. Roughly a third to a half of such patients never require RRT.
4. Trajectory static or worsening beyond the conventional threshold (persistent oliguria beyond ~72 hours, rising urea, accumulating fluid) → Start. This is where AKIKI 2 draws the line.
5. Never let "the trials showed delay is safe" become "delay indefinitely." That is the precise error AKIKI 2 was designed to detect, and it found harm.