Critical Care Nephrology · Two-Week Intensive · Lecture 8 of 9

Kidney Replacement Therapy I

Access, modality and timing — three decisions made before a single litre of effluent is produced

40 minutes Nephrology Fellows Week 2

Based on Koyner, Handbook of Critical Care Nephrology (2021) · NTUH Yunlin Branch

Learning objectives

By the end of this session you will be able to…

  1. Select site, length and type of acute dialysis access from the evidence — and recognise dysfunction early.
  2. Explain diffusion, convection and ultrafiltration well enough to write a CRRT, SLED or PD prescription that matches the goal.
  3. Quote what the four timing trials actually showed and convert them into a written trigger list.
  4. Decide when to stop kidney replacement therapy using urine output and measured clearance.
  5. Run a time-limited trial of KRT with pre-agreed goals and a date to review them.

Koyner Ch 29, 31–32, 56 · Ronco Ch 144, 163, 165–167, 179 · KDIGO AKI 2012

01

Vascular access

Every dose you fail to deliver, every filter you lose and half the infections you cause were decided at the moment the catheter went in.

Koyner Ch 29 · Ronco Ch 167 · KDIGO 2012
“The access is the mother of dialysis.”
  • An acute dialysis catheter is not a central line: dual-lumen, 11–13.5 Fr, stiff enough to hold its lumen at high negative pressure, built to deliver 200–400 mL/min
  • Non-tunnelled (uncuffed) is KDIGO's first choice for acute KRT — placed at the bedside, removed the day it is not needed
  • Tunnelled cuffed catheters carry a lower infection risk and are preferred when use is expected to exceed 1–3 weeks — convert electively, on a planned list, not after the third bacteraemia
  • An arteriovenous fistula or graft can run CRRT in an established dialysis patient, but needle dislodgement in a sedated ICU patient is a lethal event — most units use a catheter instead

What the catheter has to survive

Negative arterial pressure without collapsing, 24 h a day for CRRT, in a patient who is coagulopathic, oedematous, proned and moved for imaging. Nothing about the prescription matters if the blood pump cannot be fed.

Koyner Ch 29 · KDIGO AKI 2012 Section 5 · Al Rifai A et al. Hemodial Int 2018;22:50 (AVF/AVG for CRRT)

Site selection: the order is evidence-based, not habit

SiteRankWhyCatheter dysfunction
Right internal jugular1stShortest, straightest path to the cavoatrial junction; compressible; lowest recirculation6.6%
Femoral2ndFastest in an arrest or a coagulopathic patient; no pneumothorax risk; costs mobility10.3%
Left internal jugular3rdTwo near-right-angle turns before the SVC — kinking, tip malposition, worse flow19.5%
SubclavianLastCentral vein stenosis destroys future fistula territory; not compressible if it bleeds
Parienti et al. · JAMA 2008

750 ICU patients needing acute KRT randomised to femoral vs jugular: catheter colonisation 40.8 vs 35.7 per 1000 catheter-days (HR 0.85, 95% CI 0.62–1.16) with no difference in catheter-related bloodstream infection. Femoral was worse only in the highest BMI tertile (>28.4). Femoral access is a legitimate choice, not a compromise.

Parienti JJ et al. JAMA. 2008;299:2413–2422 · Parienti JJ et al. Crit Care Med. 2010;38:1118–1125 (dysfunction rates, n=736) · KDIGO 2012 · Koyner Ch 29

Length and tip position decide the dose you actually deliver

Insertion siteCatheter lengthTarget tip
Right internal jugular15–20 cmRight atrium / cavoatrial junction
Left internal jugular20–24 cmRight atrium / cavoatrial junction
Femoral vein≥24 cmInferior vena cava, above the iliac confluence
Subclavian (avoid)As for jugularCavoatrial junction

Recirculation by site: internal jugular ≈ 10% (5–15%); femoral ≈ 20% (5–38%), worst with short catheters; reversing the lines to rescue flow pushes recirculation to 20–30%.

Longer is better — and it was randomised

100 CRRT patients, 20–24 cm vs 15–20 cm catheters in a great thoracic vein: the longer catheter prolonged filter survival by a mean 6.5 h, delivered more dialysis dose and caused fewer clotting episodes (2.3 vs 3.6 per patient).

Delivered clearance ≈ prescribed clearance × ( 1 − recirculation fraction ) A femoral catheter that is too short can quietly cost a fifth of the dose you charted — and no laboratory value will tell you.

Morgan D et al. Am J Kidney Dis. 2012;60:272–279 · Ronco Ch 167 (recirculation) · Koyner Ch 29 Table 29.1

Insertion: ultrasound, a bundle, and a check before first use

01

PlanSite, length, platelets/INR, and a pre-scan for thrombus or a collapsed vein. Mark the side you are preserving.

02

BarrierHand hygiene, cap, mask, sterile gown and gloves, full-body drape, >0.5% chlorhexidine in alcohol, allowed to dry.

03

Real-time ultrasoundNeedle tip visualised into the vein — fewer passes, fewer arterial punctures, fewer haematomas and pneumothoraces.

04

ConfirmAspirate and flush both lumens freely, then chest radiograph for a thoracic tip before the first treatment.

Key point

Real-time ultrasound guidance is the standard of care for dialysis catheters, not an option for difficult patients. Landmark insertion is now a documented deviation.

Rabindranath KS et al. Cochrane Database Syst Rev. 2011;(11):CD005279 · O'Grady NP et al. Clin Infect Dis. 2011;52:e162 · Koyner Ch 29 Table 29.2

Complications: three families, one scorecard

Mechanical — at insertion

  • Arterial cannulation, local haematoma, cannulation failure
  • Pneumothorax and haemothorax (jugular, subclavian)
  • Femoral: arteriovenous fistula, arterial insufficiency, retroperitoneal bleed
  • Guidewire arrhythmia, tip malposition

Infectious — over days

  • Exit-site infection and catheter-related bloodstream infection
  • Risk rises with every catheter-day; tunnelled cuffed catheters carry a lower risk
  • Draw paired catheter and peripheral cultures before you blame the line
  • Never exchange a suspected infected catheter over a guidewire

Thrombotic — over weeks

  • Mural and intraluminal thrombus; the catheter is a nidus
  • Host-vein thrombosis and central venous stenosis
  • Risk increases with duration of use and with subclavian placement
The number that should change your behaviour

In one cohort of temporary haemodialysis catheters, more than half had to be removed because of a complication rather than because they were no longer needed. Assume the line will cause a problem and plan its exit on the day you insert it.

Kairaitis LK, Gottlieb T. Nephrol Dial Transplant. 1999;14:1710 · Clark E et al. Can J Kidney Health Dis. 2016;3:2054358116669128 · Koyner Ch 29

Catheter dysfunction: a ladder, not a shrug

1

Define it — cannot sustain the prescribed blood flow, or arterial pressure more negative than about −250 mmHg / venous pressure above about +250 mmHg at target flow. Rising transmembrane pressure is a filter problem, not an access problem.

2

Position first — hip flexion with a femoral line, head turn, a tip that has migrated up the SVC, a kink under the dressing. Free, instant, and the commonest cause.

3

Volume — a hypovolaemic patient collapses the vein around the arterial lumen. Look at the patient before you blame the catheter.

4

Reverse the lines only as a bridge — it buys flow at the cost of 20–30% recirculation and a silent loss of dose. Document it; do not normalise it.

5

Thrombolytic lock — alteplase dwell for suspected intraluminal thrombus before considering exchange.

6

Exchange or resite — guidewire exchange with new sterile gloves and a new field if uninfected; a new site if infection is suspected or the vein is thrombosed.

Ronco Ch 167 · Koyner Ch 29 · Clark E et al. Can J Kidney Health Dis 2016

Daily catheter care — and what to lock it with

Every shift, every connection

  • Inspect the exit site for redness, discharge, tenderness; re-dress if damp, loosened or soiled
  • Scrub the hub with friction at every connect and disconnect; clamp before uncapping; never leave a hub open or unattended
  • Sterile gauze or transparent semipermeable dressing; chlorhexidine at dressing changes
  • Chart blood flow achieved, arterial/venous pressures and catheter-days at every handover
  • Ask on every ward round: does this line still need to be in?

Locking solutions — what the evidence supports

  • Heparin vs saline: minimal difference in patency in a Cochrane review — saline is a reasonable default
  • 4% sodium citrate: comparable or better patency, less bleeding and, in meta-analysis, fewer catheter infections
  • Concentrated citrate flushed systemically can cause severe hypocalcaemia and arrhythmia — aspirate the lock, never push it
  • Antibiotic locks are not for routine prophylaxis; reserve them for salvage of a needed tunnelled catheter with the infectious-diseases team

Lopez-Briz E et al. Cochrane Database Syst Rev. 2018;7:CD008462 · Zhao Y et al. Am J Kidney Dis. 2014;63:479 · Polaschegg HD, Sodemann K. Nephrol Dial Transplant. 2003;18:2688 · CDC scrub-the-hub protocol · Koyner Ch 29

Two access errors you cannot take back

Pitfall 1 — the convenient subclavian line

Subclavian catheterisation causes central venous stenosis far more often than jugular access — in the classic angiographic series, 42% vs 10%. A stenosis is silent until the day the arm swells after a fistula is created, and by then the whole limb is unusable. It is also the one site you cannot compress when a coagulopathic patient bleeds. If a subclavian is unavoidable, KDIGO advises using the dominant side.

Pitfall 2 — burning the arm before nephrology is called

Vein preservation is a whole-hospital behaviour: no PICC lines, no routine cannulation and no venepuncture in the non-dominant forearm of any patient who might need a fistula — which includes anyone with CKD 4–5, anyone dialysis-dependent, and anyone whose AKI may not recover. Put the sign on the bed on day one, not on the day you consult vascular surgery.

Schillinger F et al. Nephrol Dial Transplant. 1991;6:722 · Hernández D et al. J Am Soc Nephrol. 1998;9:1507 · KDIGO 2012 · Koyner Ch 29

02

Modalities & the physics behind them

No modality has ever been shown to save more lives than another — so choose the one whose physics matches the problem you are trying to solve.

Koyner Ch 32 · Ronco Ch 165–166, 176–179

The alphabet, decoded once

IHD
Intermittent haemodialysis — 3–4 h, blood 300–400 mL/min, dialysate 500–800 mL/min. Diffusion. Maximum clearance per hour.
SLED / PIRRT
Sustained low-efficiency dialysis, the prolonged intermittent family — 6–12 h, blood ≈200 mL/min, dialysate ≈300 mL/min. Diffusion, slowly.
SCUF
Slow continuous ultrafiltration — plasma water removal only. No meaningful solute clearance. Use it when volume is the entire problem.
CVVH
Continuous venovenous haemofiltration — convection alone; no dialysate, replacement fluid pre- or post-filter. Best middle-molecule clearance, simplest circuit, highest filtration fraction if post-dilution.
CVVHD
Continuous venovenous haemodialysis — diffusion alone; countercurrent dialysate, no replacement fluid. Small-solute control with the lowest filtration fraction and the best filter life.
CVVHDF
Continuous venovenous haemodiafiltration — dialysate and replacement fluid; split the effluent dose between the two mechanisms.
PD
Peritoneal dialysis — the peritoneum as the membrane; diffusion plus osmotic ultrafiltration, no extracorporeal circuit and no anticoagulation.
OMAKI · Crit Care 2012

Pilot RCT of CVVH vs CVVHD in critically ill AKI: convection was associated with a lower vasopressor requirement but no survival difference. A meta-analysis of 19 randomised comparisons of haemofiltration vs haemodialysis found no difference in mortality or any clinical outcome. Pick the mechanism that suits the circuit, not the survival curve.

Neri M et al. Crit Care. 2016;20:318 (ADQI standardised nomenclature) · Wald R et al. Crit Care. 2012;16:R205 · Friedrich JO et al. Crit Care. 2012;16:R146 · Koyner Ch 32 · Ronco Ch 176–177

Three mechanisms — and why the size of the molecule decides

MechanismDriving forceClearsFalls off when
DiffusionConcentration gradient across the membraneSmall solutes <500 Da — urea, creatinine, potassium, bicarbonateMolecular weight rises: clearance falls steeply above ~1 kDa
ConvectionTransmembrane pressure dragging solvent, and solute with itSmall and middle molecules 500–5000+ Da — β₂-microglobulin, cytokines, myoglobinThe membrane's cut-off is reached, or protein layering blocks pores
UltrafiltrationHydrostatic pressurePlasma water — solutes only incidentallyIt is the only mechanism running (SCUF): no solute control at all
AdsorptionBinding to the membrane surface and poresSome hydrophobic peptides and proteinsSites saturate within hours — minor in routine CRRT
Convective solute flux = JF × Cplasma × SC  ·  SC = 1 − rejection coefficient Urea SC = 1.0. A 5 kDa solute sieves at 1.0 in saline but only ≈0.4 in plasma, because adsorbed protein forms a secondary membrane that plugs the pores. The membrane you prescribed is not the membrane the patient has after an hour.

Ronco Ch 165 · Brunet S et al. Am J Kidney Dis. 1999;34:486 · Koyner Ch 32

Pre- vs post-dilution: filter life against clearance

Filtration fraction = ultrafiltration rate ÷ plasma flow into the filter Keep it ≤20–25% in post-dilution. Above that, haemoconcentration drives protein–membrane interaction, pores clog, and the filter dies.

Pre-dilution

  • Replacement fluid enters before the filter, diluting the blood it will clean
  • Costs roughly 15% of solute clearance for the same effluent volume
  • Lowers filtration fraction and haematocrit in the fibres → longer filter life, higher achievable effluent rates
  • Default choice in a high-haematocrit, clot-prone, anticoagulation-limited patient

Post-dilution

  • Replacement fluid enters after the filter — every litre of effluent is fully cleared plasma
  • Most efficient per litre, so a lower effluent volume achieves the target dose
  • Raises filtration fraction and accentuates the secondary protein membrane → shorter filter life
  • Needs a good blood flow; unforgiving if the access under-delivers

Uchino S et al. Nephron Clin Pract. 2003;94:c94 · Ronco Ch 165 · Koyner Ch 32 (Figure 32.1)

CRRT vs IHD: the comparison that actually decides cases

DimensionCRRTIHD
Haemodynamic toleranceBest — ultrafiltration spread over 24 hWorst — the whole fluid target in 3–4 h
Net ultrafiltration controlHourly, adjustable, matched to inputsIntermittent and bolus-like; balance drifts between sessions
Solute controlLow hourly clearance, near steady stateHigh hourly clearance, peak-and-trough with rebound
Intracranial pressure / osmolar stabilityStable — the reason it is preferred in brain injuryRapid urea fall risks osmotic shift and raised ICP
AnticoagulationUsually required — regional citrate preferredOften none or a short heparin course
Rapid removal of a toxin or of potassiumSlowThe correct tool
Mobility, physiotherapy, theatre and imagingTethered to the machineFree between sessions
Nursing modelICU nurse, one-to-one, 24 hDialysis nurse for the session
Nutrition and blood-product volumeUnlimited — space is created continuouslyConstrained between sessions
Drug clearanceContinuous; antibiotic dosing must be re-thought (Lecture 9)Intermittent; post-dialysis dosing
Cost and consumablesHighLower

Koyner Ch 32 Tables 32.2 and 32.4 · Ronco Ch 166 · KDIGO 2012 Section 5

The honest evidence: no modality has won on survival

Hemodiafe · Lancet 2006

360 patients with AKI and multiorgan failure randomised to CVVHDF vs IHD, with protocolised tolerance measures in the IHD arm: no difference in 60-day survival. Well-delivered IHD is safe in sick patients if you protect the blood pressure.

CONVINT · Crit Care 2014

252 patients randomised to continuous vs intermittent KRT: no difference in survival, ICU stay or kidney recovery. Randomised data have never reproduced the recovery advantage seen in observational cohorts.

How to say this on a ward round

CRRT is chosen for control — of pressure, of volume, of osmolality — not for survival. Observational studies favouring CRRT for kidney recovery are confounded by indication: the unstable patient gets CRRT and the unstable patient recovers less. And practice varies far more than case mix does — CRRT was the initial modality in 80% of KRT starts across 54 ICUs in 23 countries in BEST Kidney, with wide between-country spread. When practice varies that much and outcomes do not, defend the delivery, not the acronym.

Vinsonneau C et al. Lancet. 2006;368:379–385 · Schefold JC et al. Crit Care. 2014;18:R11 · Schneider AG et al. Intensive Care Med. 2013;39:987 · Uchino S et al. JAMA. 2005;294:813 (BEST Kidney) · Koyner Ch 32

So when does each one actually win?

Choose CRRT when…

  • Vasopressor-dependent shock — you need net ultrafiltration measured in mL per hour, not litres per session
  • Acute brain injury, cerebral oedema or fulminant liver failure — cerebral blood flow falls and ICP rises during IHD in AKI; osmolar stability is the whole argument
  • Severe fluid overload with a fixed obligatory input (nutrition, antibiotics, blood products)
  • Extreme dysnatraemia needing a controlled correction rate (Lecture 5)
  • Integration with ECMO or targeted temperature management

Choose IHD or a hybrid when…

  • Speed is the therapy — life-threatening hyperkalaemia, a dialysable intoxication, severe refractory acidaemia
  • The patient is haemodynamically stable or has stabilised on CRRT and is ready to step down
  • Mobilisation, physiotherapy, theatre or imaging matter more than another 24 h of tight control
  • Anticoagulation is contraindicated and a short heparin-free session is achievable
  • Machines, circuits or ICU nursing are the limiting resource — a legitimate clinical input, not a failure

KDIGO 2012 · Lund A et al. Acta Anaesthesiol Scand. 2019;63:493 · Regolisti G et al. Nephrol Dial Transplant. 2013;28:79 · Koyner Ch 32

SLED / PIRRT — the pragmatic middle

  • Slow diffusive clearance over 6–12 h, usually overnight: blood ≈200 mL/min, dialysate ≈300 mL/min, on a standard haemodialysis machine or a CRRT platform
  • Delivers adequate small-solute clearance and ultrafiltration without the swings of a 4-hour session
  • Frees the day for imaging, theatre and physiotherapy — and removes the need for one-to-one dialysis nursing
  • Less anticoagulant exposure than 24-hour therapy; the circuit comes down every morning
  • The honest gap: no consensus on dose or frequency, unlike IHD (Kt/V 1.3 thrice weekly) and CRRT (20–25 mL/kg/h)

Where it sits in practice

In most institutions PIRRT substitutes for CRRT, not for IHD — it is the step-down that keeps the haemodynamic profile while returning the patient's day.

Schwenger et al. · Crit Care 2012

232 critically ill patients with AKI randomised to SLED vs CVVH: 90-day mortality 49.6% vs 55.6% (not significant), with shorter mechanical ventilation, shorter ICU stay and less nursing time in the SLED arm. Single centre, and the CVVH arm was dosed near 31 mL/kg/h. A meta-analysis of seven RCTs found no mortality difference.

Schwenger V et al. Crit Care. 2012;16:R140 · Zhang L et al. Am J Kidney Dis. 2015;66:322 · Kumar VA et al. Am J Kidney Dis. 2000;36:294 · Koyner Ch 32

Acute peritoneal dialysis: not obsolete, just misplaced

Where it genuinely wins

  • No vascular access and no safe site — or a coagulopathy that makes a large-bore line dangerous
  • Resource-limited and disaster settings: manual PD needs no electricity, no water treatment and no machine
  • Paediatric AKI, where it remains a first-line therapy worldwide
  • No anticoagulation, no extracorporeal blood contact, no disequilibrium syndrome
  • ISPD endorses PD as a suitable alternative to blood-based therapy for AKI — with a tunnelled catheter, a closed Y-system and, in critical illness, bicarbonate-buffered solutions

Where it fails you

  • Unpredictable ultrafiltration and clearance — you cannot promise a net balance by morning
  • The hypercatabolic patient outruns it; there is no way to escalate quickly
  • Recent abdominal surgery, peritoneal injury, adhesions, a fresh laparotomy or an open abdomen
  • Raised intra-abdominal pressure and diaphragmatic splinting worsen respiratory mechanics — and worsen renal venous congestion
  • Peritonitis risk, hyperglycaemia from the dextrose load, and continuous peritoneal protein loss in an already catabolic patient

Chionh CY et al. Clin J Am Soc Nephrol. 2013;8:1649 (24 studies, 4 RCTs — no mortality difference vs extracorporeal KRT) · Cullis B et al. Perit Dial Int. 2014;34:494 (ISPD) · Phu NH et al. N Engl J Med. 2002;347:895 · Koyner Ch 32 · Ronco Ch 179, 183–186

Choosing a modality: the one-slide algorithm

1

Is there a life-threatening indication that needs speed? Refractory hyperkalaemia or a dialysable toxin → IHD now, whatever the blood pressure, then reassess.

2

Is the patient vasopressor-dependent, brain-injured or grossly overloaded?CRRT, with the dose and anticoagulation strategy from Lecture 9.

3

Is volume the only problem, with acceptable chemistry?SCUF, or a diuretic strategy first — do not start solute clearance you do not need.

4

Is the patient stabilising, or does the day need to be free?SLED / PIRRT overnight as the step-down.

5

Stable, mobilising, chemistry controlled between sessions?IHD thrice weekly at Kt/V 1.3; more frequent treatment has not improved outcomes and may impair recovery.

6

No safe vascular access, no machine, or no anticoagulation possible?acute PD, provided the abdomen and the catabolic rate allow it.

Key point

Modality is not a one-time decision. It is a prescription reviewed daily and expected to change — CRRT to SLED to IHD is the normal trajectory of a recovering patient, and going backwards is not a failure. Local expertise, nursing ratios and machine availability are legitimate inputs; what actually changes outcome is a working access, a delivered dose and a circuit that stays alive.

KDIGO 2012 · Vijayan A et al. Kidney Int Rep. 2018;3:456 (frequency and recovery) · Koyner Ch 32

03

Timing of initiation

Four large randomised trials asked whether starting earlier helps; the consistent answer is that it does not, and that starting later has a floor.

Koyner Ch 31 · Ronco Ch 144, 163

AEIOU — the indications nobody argues about

IndicationWorking threshold
AAcidaemiapH ≤7.20 or bicarbonate ≤12 mmol/L, refractory to ventilation and buffer — the STARRT-AKI trigger
EElectrolytesPotassium ≥6.0 mmol/L refractory to medical therapy, or any level with ECG change; also tumour lysis and severe refractory hypercalcaemia
IIntoxicationA dialysable toxin — toxic alcohols, lithium, salicylate, valproate, metformin-associated lactic acidosis (EXTRIP recommendations)
OOverloadPulmonary oedema refractory to diuretics; STARRT-AKI used PaO₂/FiO₂ ≤200 judged to be volume-related
UUraemiaEncephalopathy, pericarditis or bleeding — clinical, not a number. AKIKI's delayed arm used blood urea nitrogen >112 mg/dL as one trigger
Key point

In the presence of any of these, refractory to medical therapy, there is no debate and no trial — start, provided KRT is consistent with the patient's goals of care.

KDIGO 2012 · STARRT-AKI Investigators. N Engl J Med. 2020;383:240 · Gaudry S et al. N Engl J Med. 2016;375:122 · Koyner Ch 31 · Ronco Ch 163

…and yet AEIOU is rarely why KRT is started

  • Observational data are consistent: emergency indications are not the most common triggers for starting KRT in modern ICUs. In critically ill older patients the primary triggers were oligoanuria, fluid overload and acidaemia — trends, not thresholds
  • The real decision blends creatinine trajectory, urine output, cumulative balance, non-kidney organ failures and the trajectory of the underlying illness — which is why practice surveys show such wide disagreement between clinicians
  • Renal replacement vs renal support (Mehta): replacement waits for failure; support uses the circuit to create space for nutrition, antibiotics and blood products, and to hold acid–base and volume where you want them
  • Renal support is a legitimate rationale — but it is the rationale the trials tested, and did not vindicate

The question to ask instead

Not "does this patient meet an indication?" but: is the metabolic and volume demand exceeding this kidney's remaining capacity, and for how many more hours can medical therapy hold the line? That is answerable at the bedside every morning.

Mehta RL. Blood Purif. 2001;19:227 · Bagshaw SM et al. Clin J Am Soc Nephrol. 2019;14:496 · Koyner Ch 31 · Ronco Ch 163

The trials, part 1 — the one positive result and the one that answered it

ELAIN · JAMA 2016

Population: 231 patients, single centre in Germany, mostly surgical (about half cardiac surgery), KDIGO stage 2 plus sepsis, refractory overload, worsening SOFA or vasoactive support. Comparison: KRT within 8 h of stage 2 vs at stage 3 or a conventional indication; CRRT mandated in both arms. Result: 90-day mortality 39.3% vs 54.7% (p = 0.03), median separation only 21 h. Interpretation: a real result in a narrow surgical population at one centre — hypothesis-generating, never replicated.

AKIKI · NEJM 2016

Population: 620 patients, 31 French ICUs, KDIGO stage 3 on ventilation and/or vasopressors, no life-threatening complication; two-thirds septic shock. Comparison: KRT within 6 h of stage 3 vs deferral until oliguria >72 h, urea nitrogen >112 mg/dL, hyperkalaemia, acidosis or pulmonary oedema. Result: 60-day mortality 48.5% vs 49.7% (HR 1.03, 95% CI 0.82–1.29). 49% of the delayed arm never received KRT at all; the early arm had more catheter-related bloodstream infection and more hypophosphataemia. Interpretation: half of the patients you were about to dialyse do not need it.

Zarbock A et al. JAMA. 2016;315:2190–2199 · Gaudry S et al. N Engl J Med. 2016;375:122–133 · Koyner Ch 31 Table 31.1

The trials, part 2 — the two that settled it

IDEAL-ICU · NEJM 2018

Population: 488 patients with septic shock and stage 3 AKI, 29 French ICUs. Comparison: KRT within 12 h of stage 3 vs after 48 h of persistent AKI or an emergent indication. Result: stopped early for futility; 90-day mortality 58% vs 54% (p = 0.38). 97% of the early arm received KRT vs 62% of the delayed arm, and most of those who never started recovered spontaneously. Interpretation: the sepsis-specific "early support" hypothesis fails on its own ground.

STARRT-AKI · NEJM 2020

Population: 3019 randomised across 168 centres in 15 countries, stage 2–3 AKI, enrolled only where the attending physician confirmed genuine equipoise. Comparison: accelerated start within 12 h vs a standard strategy that discouraged KRT unless K ≥6.0, pH ≤7.20, bicarbonate ≤12 or volume-related PaO₂/FiO₂ ≤200 supervened. Result: 90-day mortality 43.9% vs 43.7% (RR 1.00, 0.93–1.09), consistent across every prespecified subgroup. Among survivors, dialysis dependence at 90 days was 10.4% vs 6.0% (RR 1.74, 1.24–2.43), and adverse events were more common with acceleration (23% vs 16.5%, mainly hypotension and hypophosphataemia).

Barbar SD et al. N Engl J Med. 2018;379:1431–1442 · STARRT-AKI Investigators. N Engl J Med. 2020;383:240–251 · Koyner Ch 31

There is a floor as well as a ceiling

AKIKI-2 · Lancet 2021

Population: 278 patients who had already reached the AKIKI delayed-strategy threshold — oliguria beyond 72 h or urea nitrogen 112–140 mg/dL — with no emergent indication. Comparison: start now ("delayed") vs wait further until urea nitrogen exceeded 140 mg/dL or an emergency appeared ("more delayed"). Result: no gain in days alive and free of KRT; 60-day mortality was numerically higher in the more-delayed arm, and in adjusted analysis the more-delayed strategy was associated with an increased hazard of death. Interpretation: deferral is not a free option that can be extended indefinitely.

The shape of the answer

Between the ELAIN ceiling and the AKIKI-2 floor lies a wide, flat therapeutic window. Within it, the timing of the start does not determine the outcome — but having a written trigger list and a daily review does, because it is what keeps you inside the window.

Gaudry S et al. Lancet. 2021;397:1293–1300 (AKIKI-2) · Koyner Ch 31

The synthesis: watchful waiting is an active prescription

Write this in the notesThreshold
Potassium≥6.0 mmol/L, or rising despite medical therapy, or ECG change
Acid–basepH ≤7.20 or bicarbonate ≤12 mmol/L without a correctable cause
OxygenationPaO₂/FiO₂ ≤200 attributable to volume, or refractory pulmonary oedema
Cumulative balance>10% of body weight and diuretic-unresponsive
UraemiaEncephalopathy, pericarditis or bleeding
DurationOliguria beyond 72 h with no sign of recovery → escalate the review, do not simply keep waiting

What makes it "watchful"

  • A named threshold list in the notes, visible to the night team
  • A daily review with a documented decision — including "not today, because…"
  • An access plan ready, so starting takes 30 minutes and not 6 hours
  • Diuretics, potassium binders and ventilation optimised while you wait

STARRT-AKI N Engl J Med 2020;383:240 · Gaudry S et al. Lancet 2021;397:1293 · KDIGO 2012 · Koyner Ch 31

Sharpening the decision: the furosemide stress test and biomarkers

Furosemide stress test

  • Furosemide 1.0 mg/kg (loop-diuretic naive) or 1.5 mg/kg (prior exposure) in a euvolaemic patient, replacing urine losses mL for mL
  • Urine output <200 mL over 2 h predicts progression to stage 3 AKI — AUC ≈0.87
  • In a head-to-head comparison it outperformed damage biomarkers for predicting progression and receipt of KRT
  • A flat response is a reason to secure access and tighten monitoring, not an automatic order to start

Biomarkers

  • Cell-cycle arrest markers (TIMP-2 · IGFBP7) and damage markers stratify risk, and stratify it well (Lecture 4)
  • A systematic review concluded the evidence is insufficient to guide KRT initiation at the individual-patient level
  • Use them to change surveillance intensity and nephrotoxin exposure
  • No biomarker has been shown to improve outcome by triggering earlier KRT
Key point

These tools move the probability that this kidney will recover. They do not answer the question the trials asked, which is whether acting on that probability earlier helps — and it did not.

Chawla LS et al. Crit Care. 2013;17:R207 · Koyner JL et al. J Am Soc Nephrol. 2015;26:2023 · Klein SJ et al. Intensive Care Med. 2018;44:323 · Koyner Ch 16, 31

Two ways watchful waiting goes wrong

Pitfall 3 — waiting while the balance climbs

Fluid accumulation at KRT initiation and at KRT cessation both track mortality; the adjusted odds of death associated with fluid overload at cessation were around 2.5 in the PICARD cohort. "We are watching" is not a defence when the cumulative balance has crossed 10% of body weight and the patient is on 60% oxygen. Diurese, restrict, and if you cannot, start.

Pitfall 4 — calling it waiting when nobody is watching

Watchful waiting without a written trigger list, a daily documented review and a ready access plan is not the strategy the trials tested — it is drift. The delayed arms of AKIKI, IDEAL-ICU and STARRT-AKI were protocolised, monitored, and had a defined point at which they acted.

Bouchard J et al. Kidney Int. 2009;76:422–427 · Ronco Ch 163 · Koyner Ch 31

04

Stopping

The literature on when to stop is thin and largely empiric — which makes the few validated signals worth knowing precisely.

Ronco Ch 144, 164 · Koyner Ch 31, 51

Predicting recovery: what is worth measuring

  • Urine output in the 24 h before stopping is the strongest simple predictor of successful cessation. In the post-hoc analysis of the BEST Kidney cohort it discriminated success from failure with an AUC of about 0.81; the useful cut-off sits around 400–500 mL/day off diuretics
  • Diuretics degrade the signal — a litre produced on a furosemide infusion is not evidence of recovered clearance. Interpret urine output off diuretics, or not at all
  • Measured creatinine clearance on a timed collection is the objective test: the ATN study stopped support above 20 mL/min, left 12–20 mL/min to clinical judgement, and treated <12 mL/min as inadequate
  • A pre-treatment creatinine that falls between sessions, and a rising urine output in an oliguric patient, both signal returning function

Why late stopping costs you

Every extra day is another day of catheter risk, anticoagulation, hypophosphataemia, drug removal and immobility. And repeated intradialytic hypotension delivers fresh ischaemic hits to a kidney that is trying to repair — the plausible mechanism behind the extra dialysis dependence seen with accelerated initiation in STARRT-AKI.

Uchino S et al. Crit Care Med. 2009;37:2576 · Palevsky PM et al. N Engl J Med. 2008;359:7 (ATN) · Ronco Ch 144 · Koyner Ch 31, 51

How to wean, in the order you should do it

01

Stop taking fluidSet net ultrafiltration to zero once the patient is at target volume and see whether balance holds on diuretics alone.

02

Step down the intensityCRRT → SLED overnight → alternate-day IHD, guided by chemistry, not by the calendar.

03

Space the treatmentsExtend the interval as urine output and pre-treatment chemistry allow; measure a timed clearance if you are unsure.

04

Hold and observeStop while the patient is still in hospital and watch for 48–72 h. Resuming is easy; restarting from a ward bed at 2 a.m. is not.

Practical rules

Keep the catheter until the hold has clearly succeeded — then remove it the same day. Space treatments closely enough that no single session has to remove a large volume. Feed the patient adequately throughout; and remember that 6–10% of survivors remain dialysis-dependent at 90 days, so every discharge needs a nephrology follow-up plan (Lecture 9).

Ronco Ch 144 · STARRT-AKI N Engl J Med 2020;383:240 · Koyner Ch 51

05

Shared decisions & time-limited trials

"Can we dialyse this patient?" is a technical question with an easy answer; "should we?" is the one that needs a structure.

Koyner Ch 56 · Ronco Ch 148 · RPA guideline 2010

Clinical momentum, and the numbers that should slow it

  • Dialysis decisions in the ICU are rarely made alone — they arrive bundled with intubation, vasopressors and tracheostomy. Clinical momentum accumulates interventions without anyone revisiting whether the direction is one the patient wanted
  • Older patients starting dialysis for AKI in hospital do badly: in one cohort of patients aged ≥65 starting dialysis, survival was 55.8% at 6 months and 45.5% at 1 year, with ADL dependence, age ≥85, inpatient initiation and ≥4 comorbidities all independently predicting death
  • AKI on established CKD frequently does not recover — a large fraction of these patients become permanently dialysis-dependent
  • Resist the technological imperative: being able to dialyse is not a reason to dialyse. A treatment whose likely harms exceed its benefits is not medically indicated, and should not be offered as though it were neutral

Ask before you consent

Baseline kidney function · comorbidity burden · functional and nutritional status · capacity and advance directives · what matters more, quality or length of life · is this patient a plausible long-term dialysis candidate?

Kruser JM et al. Ann Am Thorac Soc. 2017;14:426 · Wachterman MW et al. JAMA Intern Med. 2019;179:987 · Bagshaw SM et al. Clin J Am Soc Nephrol. 2019;14:496 · Koyner Ch 56 Box 56.1

The time-limited trial: four steps you can run this week

01

PrepareICU team, nephrology and consultants agree the prognosis, what KRT can plausibly achieve, and the milestones that would count as success.

02

CommunicateAsk–Tell–Ask. Share the estimate, elicit values, propose KRT explicitly as a trial with a defined end point and a date.

03

ConductDeliver full therapy. Review progress against the agreed milestones daily and update the family — no silent drift.

04

ConcludeMeet on the agreed date. Met → continue. Partly met → negotiate a new trial. Not met → move to comfort-focused care as agreed at step 2.

Opening
"I would like to use dialysis for the next five days and meet again on Friday to see whether her pressure is off, her lungs are clearing and she is awake."
Eliciting values
"What would he say about all this if he could see himself here? What would he not want us to do?"
Naming the worry
"I am worried that dialysis is keeping his kidneys going in a body that is not turning the corner."
Setting the frame
"A trial can succeed either way — either she improves, or we learn that this is not something dialysis can fix."

Scherer JS, Holley JL. Clin J Am Soc Nephrol. 2016;11:344 · Renal Physicians Association. Shared Decision-Making in the Appropriate Initiation of and Withdrawal from Dialysis. 2nd ed. 2010 · Koyner Ch 56

Withholding, withdrawing and conservative kidney management

When it is ethically appropriate to withhold or withdraw

  • A patient with capacity who, fully informed, refuses or asks to stop
  • A patient without capacity who previously refused dialysis in an advance directive
  • A patient without capacity whose health care proxy declines or asks to stop
  • A patient with irreversible, profound neurological impairment
  • Withholding and withdrawing are ethically equivalent — the fear of not being able to stop is a bad reason not to start a trial

Conservative and palliative kidney management

  • Active care without dialysis: volume and symptom control, potassium management, honest prognostication
  • Anticipate the course after withdrawal — uraemia and hyperkalaemia over days — and prescribe for dyspnoea, agitation and pain in advance
  • Involve palliative care early; provision of palliative care to patients with dialysis-requiring AKI is documented to be infrequent
  • If conflict persists, use a systematic due-process approach; if KRT is needed emergently and the surrogate requests it, provide it while the conflict is worked through

Renal Physicians Association guideline (2nd ed, 2010) Table 56.2 · Chong K et al. Clin J Am Soc Nephrol. 2017;12:1744 · Koyner Ch 56 · Ronco Ch 148

06

Cases & wrap-up

Three referrals, three different questions: when, which, and whether.

Koyner Ch 29, 31–32, 56
Case 1 · Septic AKI, stage 3, no emergency

66 M, community-acquired pneumonia, ICU day 3. Norepinephrine weaning, lactate normalised, ventilated on FiO₂ 0.4. Loop diuretic infusion started 12 h ago with little effect.

Cr 1.1 → 3.6 mg/dLUO 12 mL/h × 8 hK 5.2pH 7.31HCO₃ 18P/F 240Balance +3.4 LNE 0.06 µg/kg/min

Start kidney replacement therapy today, or not?

Hands up · answer on the next slide

Case 1 — what "waiting" has to look like on paper

01

Write triggersK ≥6.0 · pH ≤7.20 · HCO₃ ≤12 · P/F ≤200 from volume · refractory oedema

02

Optimise nowStop nephrotoxins, review every infused volume, escalate or stop the diuretic on evidence

03

TestFurosemide stress test if the trajectory is unclear; a flat response means prepare, not start

04

Review dailyDocument the decision — including the decision not to start, and why

Why this is the evidence-based answer

AKIKI: 49% of the delayed arm never needed KRT. STARRT-AKI: identical mortality, and 10.4% vs 6.0% dialysis dependence at 90 days favouring the standard strategy. AKIKI-2: but do not extend the wait past the point where an emergency is the only thing that will make you act.

Gaudry S et al. N Engl J Med 2016;375:122 · STARRT-AKI N Engl J Med 2020;383:240 · Gaudry S et al. Lancet 2021;397:1293 · Chawla LS et al. Crit Care 2013;17:R207

Case 2 · Severe traumatic brain injury with AKI and overload

27 M, day 4 after a motorcycle collision. Decompressive craniectomy, ICP monitored, on hypertonic saline and mannitol. Rhabdomyolysis-associated AKI, now anuric. Ventilated, noradrenaline 0.15 µg/kg/min.

ICP 21 mmHgNa 152 mmol/LCr 4.4 mg/dLK 5.8Balance +6.2 LP/F 180CK 42 000 U/L

Which modality, and what are the two numbers you must protect?

Think 45 seconds — name the modality and the two targets

Case 3 · The referral where the question is "whether"

84 M, CKD stage 4 (baseline creatinine 2.8 mg/dL), moderate dementia, dependent for washing and dressing, three admissions this year for fluid overload. Admitted with aspiration pneumonia and shock; now ventilated, on two vasopressors, oliguric for 30 h. No advance directive. The family asks you to "do everything".

Cr 2.8 → 5.1 mg/dLUO 8 mL/hK 5.4pH 7.26Balance +5 L2 vasopressorsKarnofsky ~40%

What do you offer, and how do you say it?

Quick poll

Four patients with KDIGO stage 3 AKI are on your list this morning. Which one should start KRT today?

A. pH 7.28, K 5.5, urine 0.3 mL/kg/h, balance +1 L, no oedema · B. K 6.4 with peaked T waves, unchanged after insulin–dextrose and salbutamol · C. Urea nitrogen 105 mg/dL, alert, eating, balance neutral · D. Oliguric 40 h, creatinine still rising, otherwise well and improving off vasopressors

Hands up — then name the trigger that fired

Key takeaways

  • Right internal jugular first, femoral a legitimate second (equivalent infection risk in Parienti), left internal jugular third, subclavian essentially never — and use the right length so the tip sits at the cavoatrial junction or in the IVC.
  • Protect the non-dominant arm from day one; a subclavian stenosis or a PICC line can cost a patient their fistula years before anyone plans one.
  • Diffusion clears small solutes, convection adds middle molecules, ultrafiltration moves only water — and pre- versus post-dilution trades about 15% of clearance for filter life.
  • No modality improves survival over another; choose CRRT for haemodynamic and osmolar control, IHD for speed, SLED to give the day back, PD when access or resources dictate.
  • Early initiation does not save lives: STARRT-AKI showed 43.9% vs 43.7% mortality with more dialysis dependence at 90 days when accelerated — but AKIKI-2 shows the wait has a floor.
  • Watchful waiting means a written trigger list, a daily documented review and a ready access plan; stopping is guided by urine output off diuretics and a measured clearance, and every uncertain start should be framed as a time-limited trial with a date.

References & further reading

Where to go deeper

  1. Parienti JJ, Thirion M, Mégarbane B, et al. Femoral vs jugular venous catheterization and risk of nosocomial events in adults requiring acute renal replacement therapy: a randomized controlled trial. JAMA. 2008;299:2413–2422.
  2. Morgan D, Ho K, Murray C, Davies H, Louw J. A randomized trial of catheters of different lengths to achieve right atrium versus superior vena cava placement for continuous renal replacement therapy. Am J Kidney Dis. 2012;60:272–279.
  3. Zarbock A, Kellum JA, Schmidt C, et al. Effect of early vs delayed initiation of renal replacement therapy on mortality in critically ill patients with acute kidney injury: the ELAIN randomized clinical trial. JAMA. 2016;315:2190–2199.
  4. Gaudry S, Hajage D, Schortgen F, et al. Initiation strategies for renal-replacement therapy in the intensive care unit (AKIKI). N Engl J Med. 2016;375:122–133.
  5. Barbar SD, Clere-Jehl R, Bourredjem A, et al. Timing of renal-replacement therapy in patients with acute kidney injury and sepsis (IDEAL-ICU). N Engl J Med. 2018;379:1431–1442.
  6. STARRT-AKI Investigators. Timing of initiation of renal-replacement therapy in acute kidney injury. N Engl J Med. 2020;383:240–251.
  7. Gaudry S, Hajage D, Martin-Lefevre L, et al. Comparison of two delayed strategies for renal replacement therapy initiation for severe acute kidney injury (AKIKI 2): a multicentre, open-label, randomised, controlled trial. Lancet. 2021;397:1293–1300.
  8. Schwenger V, Weigand MA, Hoffmann O, et al. Sustained low efficiency dialysis using a single-pass batch system in acute kidney injury: a randomized interventional trial. Crit Care. 2012;16:R140.
  9. Chionh CY, Soni SS, Finkelstein FO, Ronco C, Cruz DN. Use of peritoneal dialysis in AKI: a systematic review. Clin J Am Soc Nephrol. 2013;8:1649–1660.
  10. Scherer JS, Holley JL. The role of time-limited trials in dialysis decision making in critically ill patients. Clin J Am Soc Nephrol. 2016;11:344–353.
  11. Koyner JL, Topf JM, Lerma EV, eds. Handbook of Critical Care Nephrology. Wolters Kluwer; 2021 (Ch 29, 31–32, 56). · Ronco C, Bellomo R, Kellum JA, Ricci Z, eds. Critical Care Nephrology. 3rd ed. Elsevier; 2019 (Ch 144, 163, 165–167, 179).
Critical Care Nephrology · Two-Week Intensive

Thank you

Questions & discussion — tomorrow we prescribe the therapy we have just decided to give.

Next: Lecture 09 — KRT II: Dose, Anticoagulation, Complications & Recovery