Publication

• Title: Timing of Initiation of Renal-Replacement Therapy in Acute Kidney Injury
• Acronym: STARRT-AKI
• Year: 2020
• Journal published in: New England Journal of Medicine
• Citation: STARRT-AKI Investigators. Timing of initiation of renal-replacement therapy in acute kidney injury. N Engl J Med. 2020;383(3):240-251.

Context & Rationale

• Background

  In ICU patients with severe AKI, the decision to initiate kidney replacement therapy (KRT) before “classic” indications is common but historically evidence-poor.
  Earlier RCTs produced conflicting signals (single-centre, biomarker-enriched “early” approaches vs pragmatic “watchful waiting”), while practice varied widely across sites and countries.
  Clinicians lacked high-certainty evidence on whether an accelerated strategy improves patient-centred outcomes (mortality, kidney recovery) or simply increases exposure to KRT-related harms.
• Research Question/Hypothesis

  Does an accelerated KRT initiation strategy (within 12 hours of severe AKI meeting eligibility) reduce 90-day all-cause mortality compared with a standard strategy (KRT discouraged unless conventional indications develop or AKI persists)?
  Hypothesis: accelerated initiation would lower mortality by preventing complications of severe AKI and multi-organ dysfunction.
• Why This Matters

  KRT is invasive, resource-intensive, and can plausibly worsen haemodynamics or delay renal recovery; conversely, delayed KRT risks untreated metabolic or fluid complications.
  A large, multinational, pragmatic trial was needed to resolve timing equipoise and inform guidelines, resource allocation, and shared decision-making in ICU AKI.

Design & Methods

• Research Question: In critically ill adults with severe AKI and clinical equipoise regarding KRT initiation, does an accelerated strategy (start KRT within 12 hours) reduce 90-day mortality versus a standard strategy (defer unless conventional indications or persistent AKI)?
• Study Type: Multicentre, international, pragmatic, randomised, allocation-concealed, open-label superiority trial conducted in ICUs (168 hospitals across 15 countries).
• Population:
  • Adults (≥18 years) in ICU with severe AKI (KDIGO stage 2 or 3) and clinician-assessed eligibility for KRT randomisation.
  • Key design feature: clinician declaration of equipoise (KRT neither mandated immediately nor mandated to be deferred) for trial entry ¹.
  • Exclusions included “urgent indication” features and scenarios where KRT timing was not genuinely discretionary (e.g., imminent need, competing priorities, or specific contraindications) ¹.
• Intervention:
  • Accelerated strategy: initiate KRT as soon as possible and within 12 hours after eligibility criteria were met.
  • KRT modality, prescription, and anticoagulation were determined by treating clinicians/local practice (pragmatic delivery).
• Comparison:
  • Standard strategy: KRT discouraged unless conventional indications developed (e.g., severe electrolyte/acid–base disturbance, fluid overload compromising oxygenation) or AKI persisted beyond a protocolised timeframe (watchful waiting with close monitoring) ¹.
• Blinding: Open-label (KRT timing could not be blinded); primary outcome (mortality) was objective.
• Statistics: Planned sample size 2,866 to detect a 15% relative (6% absolute) reduction in 90-day mortality (assumed 40% control mortality) with 90% power at a two-sided 5% significance level, allowing for interim analyses and ~3% attrition; primary analysis used a modified intention-to-treat approach (excluding those with consent withdrawal/loss to follow-up) ¹.
• Follow-Up Period: 90 days for primary outcome and key secondary outcomes (kidney recovery, resource use, adverse events); longer-term follow-up planned in the protocol but not reported in the index manuscript ¹.

Key Results

This trial was not stopped early. 3,019 patients were randomised; 2,927 were included in the modified intention-to-treat population (accelerated n=1,465; standard n=1,462).

• Neutral primary endpoint: 90-day mortality was essentially identical (43.9% vs 43.7%).
• Substantial “KRT avoidance” with watchful waiting: 559/1,462 (38.2%) in the standard arm never received KRT, while mortality remained unchanged.
• Trade-off signal: accelerated strategy increased KRT dependence among survivors (10.4% vs 6.0%) and increased adverse events (23.0% vs 16.5%), including hypotension (8.7% vs 5.3%) and hypophosphataemia (7.5% vs 4.2%).

Internal Validity

• Randomisation and allocation: central, web-based randomisation with stratification by site and variable block sizes (2 and 4); allocation concealment was maintained until assignment.
• Analysis population and exclusions: 3,019 randomised; primary analysis used modified intention-to-treat (2,927 analysed: 1,465 vs 1,462), limiting post-randomisation exclusions to a small proportion (3.0%).
• Performance/detection bias: open-label design; primary outcome (mortality) is robust to lack of blinding, but co-interventions and clinician decisions (e.g., fluid strategy, diuretics, vasopressors) could plausibly be influenced.
• Protocol adherence / separation: strong separation in exposure to KRT (96.8% vs 61.8%).
• Timing separation (process): time from randomisation to KRT initiation was 1.6 hours (IQR 0.9–2.8) vs 25.6 hours (IQR 18.7–47.1); time from eligibility to KRT was 6.0 hours (IQR 4.0–8.8) vs 31.8 hours (IQR 18.9–71.8).
• Baseline comparability / illness severity: groups were closely matched (e.g., SAPS II 59.6±16.3 vs 59.5±16.6; SOFA 11.7±3.4 vs 11.9±3.4; mechanically ventilated 75.3% vs 78.5%; vasopressor use 68.8% vs 72.0%).
• Heterogeneity: pragmatic multinational conduct increases clinical heterogeneity, but prespecified subgroup analyses were consistent with the primary result (no clear effect modification across key strata, including baseline KDIGO stage, sepsis, and surgical vs medical admission).
• Dose/modality: modality and prescription were pragmatic; initial modality was predominantly CRRT (68.4% vs 70.6%), limiting confounding by systematic modality imbalance.
• Outcome assessment: mortality and KRT dependence are objective; patient-reported quality-of-life measures at 90 days were included but more vulnerable to missingness and non-blinded assessment effects.
• Statistical rigour: prespecified analysis plan with interim monitoring; primary comparison used risk ratios with 95% CIs; sensitivity analyses (including as-treated analyses) were directionally consistent with the primary result.

Conclusion on Internal Validity: Strong—randomisation and allocation concealment were appropriate, outcome measurement was largely objective, and exposure separation was substantial; the main residual threats are open-label co-intervention bias and pragmatically variable “standard” management.

External Validity

• Population representativeness: enrolled a broad, multinational ICU population with high acuity (≈44% 90-day mortality), improving relevance to routine critical care.
• Key boundary of applicability: conclusions apply to patients with severe AKI without urgent, life-threatening indications requiring immediate KRT and where clinicians can reasonably choose to defer with close monitoring.
• Resource and system context: pragmatic design accommodates different ICU and nephrology models; however, settings with limited monitoring capacity or delayed KRT access may not reproduce the “safe watchful waiting” conditions.
• Subpopulations: kidney recovery and dependence signals are particularly relevant in patients with baseline CKD and haemodynamic fragility, where dialysis-induced hypotension may be consequential.

Conclusion on External Validity: High for well-resourced ICUs managing severe AKI without emergent indications; more limited for contexts where delayed access to KRT and/or monitoring constraints undermine a “standard” watchful waiting strategy.

Strengths & Limitations

• Strengths: Large sample (n=3,019), multinational ICU trial; pragmatic delivery enhances real-world relevance; meaningful separation in timing and exposure; objective primary endpoint; careful adjudication of key secondary kidney outcomes.
• Limitations: Open-label design; eligibility hinged on clinician equipoise (may select a particular AKI phenotype and introduce between-site heterogeneity) ¹; secondary outcomes were not multiplicity-adjusted; “standard strategy” includes clinician discretion that can dilute contrasts; increased KRT exposure created greater opportunity for KRT-related adverse events.

Interpretation & Why It Matters

• Clinical practice

  For severe AKI in ICU without urgent indications, an accelerated “start within 12 hours” strategy does not improve survival and increases exposure to KRT and its complications.
  A monitored watchful waiting strategy can avoid KRT in a large minority (38.2%) without worsening mortality, reframing timing as a question of selecting who truly needs KRT.
• Patient-centred trade-offs

  Accelerated initiation increased KRT dependence among 90-day survivors (10.4% vs 6.0%) and increased adverse events (23.0% vs 16.5%), implying potential harm for some patients despite neutral mortality.
  These data support shared decision-making and conservative initiation thresholds when metabolic and fluid complications are controlled.
• Trialists/methodologists

  STARRT-AKI operationalised a pragmatic “strategy” question rather than a mechanistic “dose-response” question, making it highly policy-relevant but dependent on the fidelity of the standard-arm monitoring and rescue criteria.

Controversies & Subsequent Evidence

• Equipoise-based enrolment and phenotype selection: the protocol’s reliance on clinician declaration of equipoise was explicitly recognised as a potential driver of population heterogeneity and selection (excluding patients at the extremes of “must dialyse now” and “will clearly recover”) ¹.
• “Did some patients die without KRT who might have benefitted?” NEJM correspondence highlighted deaths attributed to hyperkalaemia in the standard group and asked whether competing risk (death before KRT) could obscure benefit in a subset, and whether overlap in initiation timing diluted contrasts ².
• Kidney recovery signal and baseline CKD: correspondence proposed dialysis-associated hypotension as a plausible mechanism for higher KRT dependence among survivors, especially in patients with chronic kidney disease, and called for mediator analyses ².
• Conceptual critique (“why start?”): correspondence argued that timing framed purely by KDIGO stage can be clinically blunt, given that 38.2% of standard-strategy patients never required KRT; it emphasised enrichment for persistent AKI and actionable physiologic triggers rather than time-based initiation ².
• Editorial synthesis: an AJKD editorial positioned STARRT-AKI as resolving much of the “early vs delayed” debate in favour of monitored delay, while underscoring the remaining uncertainty around identifying persistent, non-recovering AKI in real time ³.
• Meta-analytic convergence: an individual patient data meta-analysis (pre-STARRT-AKI era) found no survival benefit for early initiation and that a substantial proportion in delayed strategies never received KRT, aligning with STARRT-AKI’s “avoidance” signal ⁴.
• Updated systematic reviews: contemporary meta-analyses incorporating STARRT-AKI generally report no mortality benefit and increased KRT-related adverse events/dialysis dependence with earlier initiation, reinforcing a selective and trigger-based approach ⁵⁶⁷.
• Bayesian reanalysis: a Bayesian reanalysis of STARRT-AKI reported a very low probability of clinically important 90-day mortality benefit with accelerated initiation and suggested higher probability of worse kidney-related outcomes and rehospitalisation, supporting conservative adoption ⁸.
• Guideline implications: KDIGO guidance emphasises emergent initiation for life-threatening derangements rather than fixed “early” timing, broadly consistent with STARRT-AKI’s neutral mortality and harm signals ⁹.
  Surviving Sepsis Campaign guidance similarly supports KRT when indicated, not solely for biochemical thresholds without clinical context ¹⁰.
  Recent ADQI consensus work emphasises phenotyping, persistence prediction, and indication-based decision-making over time-based triggers ¹¹.

Summary

• In ICU severe AKI without urgent indications, accelerated KRT initiation (≤12 hours) did not reduce 90-day mortality compared with a standard (watchful waiting) strategy (43.9% vs 43.7%).
• Watchful waiting avoided KRT in 38.2% of standard-strategy patients, without an apparent survival penalty.
• Accelerated initiation increased KRT dependence among survivors at 90 days (10.4% vs 6.0%) and reduced KRT-free days (73 vs 74; difference −0.48 days).
• Accelerated initiation increased adverse events (23.0% vs 16.5%), including hypotension (8.7% vs 5.3%) and hypophosphataemia (7.5% vs 4.2%).
• Overall, STARRT-AKI supports selective, trigger-based KRT initiation with close monitoring rather than routine early initiation for severe AKI alone.

Overall Takeaway

STARRT-AKI is landmark because it decisively tested a pragmatic strategy question at scale: routine accelerated KRT initiation for severe ICU AKI without urgent indications does not save lives and increases treatment exposure and harms. The trial shifts modern practice towards monitored deferral with clear rescue triggers, while refocusing research on identifying patients with persistent AKI who truly stand to benefit from timely KRT.

Bibliography

• 1STARRT-AKI Investigators. STandard versus Accelerated initiation of Renal Replacement Therapy in Acute Kidney Injury: study protocol for a multinational, multicentre, randomised controlled trial. Can J Kidney Health Dis. 2019;6:2054358119852937.
• 2Bouchard J, Cutter G, Mehta RL, et al. Timing of initiation of renal-replacement therapy in acute kidney injury. N Engl J Med. 2020;383(18):1796-1798.
• 3Sohaney R, Yessayan LT, Heung M. Towards consensus in timing of kidney replacement therapy for acute kidney injury? Am J Kidney Dis. 2021;77(4):542-545.
• 4Gaudry S, Hajage D, Benichou N, et al. Delayed versus early initiation of renal replacement therapy for severe acute kidney injury: a systematic review and individual patient data meta-analysis of randomised clinical trials. Lancet. 2020;395(10235):1506-1515.
• 5Castro I, Wieringa J, Crippa IA, et al. The impact of early versus late initiation of renal replacement therapy in critically ill patients with acute kidney injury: a meta-analysis. Clin Kidney J. 2022;15(10):1932-1942.
• 6Li X, Yang J, Wang J, et al. Timing of renal replacement therapy initiation for acute kidney injury in critically ill patients: a systematic review and meta-analysis. Crit Care. 2021;25(1):15.
• 7Pan HC, Chen YY, Tsai IJ, et al. Accelerated versus standard initiation of renal replacement therapy for critically ill patients with acute kidney injury: a systematic review and meta-analysis of randomised controlled trials. Crit Care. 2021;25(1):5.
• 8Zampieri FG, da Costa BR, Vaara ST, et al. A Bayesian reanalysis of the Standard versus Accelerated Initiation of Renal-Replacement Therapy in Acute Kidney Injury (STARRT-AKI) trial. Crit Care. 2022;26(1):255.
• 9KDIGO Acute Kidney Injury Work Group. KDIGO clinical practice guideline for acute kidney injury. Kidney Int Suppl. 2012;2(1):1-138.
• 10Evans L, Rhodes A, Alhazzani W, et al. Surviving Sepsis Campaign: International guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47:1181-1247.
• 11Zarbock A, Nadim MK, Pickkers P, et al. Sepsis-associated acute kidney injury: consensus report of the 28th Acute Disease Quality Initiative workgroup. Nat Rev Nephrol. 2023;19(6):401-417.