Initiating Dialysis

By the end of this chapter you should be able to:

• Separate the emergent, effective-care indications for dialysis from the preference-sensitive questions of when, and whether, to start.
• List the emergent indications — refractory acidosis, hyperkalaemia, intoxication, overload, and uraemia — and act on them without delay.
• Summarise the timing-of-RRT trials and justify watchful waiting with prompt initiation on a developing indication.
• Explain why neither starting early nor delaying excessively improves survival.
• Choose a modality — CRRT, intermittent haemodialysis, SLED, or peritoneal dialysis — matched to haemodynamics and goals, knowing survival is similar.
• Prescribe an evidence-based dose and explain why intensive dosing does not improve survival.
• Select anticoagulation (regional citrate for CRRT) and vascular access, and adjust drug dosing on dialysis.
• Conduct a shared decision about whether to start dialysis in a poor-prognosis patient, including a time-limited trial.

• Two distinct questions hide inside 'should we dialyse?': whether to start at all, and — if so — when, with what, and how much.
• The emergent indications are effective care, not a choice: refractory acidosis, hyperkalaemia, a dialysable intoxication, diuretic-resistant overload, and uraemic complications all mandate dialysis.
• Outside those emergencies, the timing of RRT is genuinely uncertain, and large trials have resolved much of the debate.
• Accelerated or early initiation does not improve survival and leaves more patients dialysis-dependent, with more line and circuit complications.
• But excessive delay is not free either, so the standard is watchful waiting with prompt initiation once an indication develops.
• Within that watchful window, the exact threshold to start weighs the burdens of dialysis against the risks of waiting — a preference-sensitive judgement.
• Modality is matched to physiology: CRRT for the haemodynamically unstable or those needing slow fluid removal, intermittent haemodialysis for efficiency and mobility, SLED as the hybrid.
• CRRT and intermittent haemodialysis do not differ in survival, so the choice is driven by haemodynamics, goals, and logistics, not mortality.
• The dose question is settled: intensive dosing does not improve survival over standard, and adds hypophosphataemia, hypokalaemia, and drug clearance.
• Target a CRRT effluent of 20–25 mL/kg/h delivered (prescribe higher to offset downtime), and an intermittent Kt/V around 1.2–1.4 at least three times weekly.
• Regional citrate is the preferred CRRT anticoagulation — less bleeding and longer circuit life than heparin — watching for citrate accumulation in liver failure.
• Adjust drug dosing on dialysis, and beware under-dosing antibiotics in the septic patient on CRRT.
• Whether to start dialysis at all in a patient with little prospect of meaningful recovery is a goals-of-care decision, made with the patient and family.
• A time-limited trial of dialysis, with explicit goals and a review date, is a legitimate option when the prognosis is uncertain.

Dialysis in AKI is two decisions wearing one name. The first is whether to support this kidney at all, which in a dying or irreversibly failing patient is a question of values, not physiology. The second — once support is agreed — is when to start, with what modality, and at what dose, and here the evidence is unusually clear. Confusing the two is the central error: treating an emergent indication as a debatable choice, or treating a values-laden 'whether' as if a trial could answer it.

Two different questions: whether, and when

Begin by separating the questions. Some decisions about dialysis are effective care — determined by physiology and evidence, with no meaningful role for preference. When a patient has refractory hyperkalaemia, you dialyse; that is not a conversation about values. Other decisions are preference-sensitive — the exact moment to start in a non-emergent patient, and above all whether to start dialysis in someone with little prospect of recovery. These genuinely depend on what the patient wants and what dialysis can realistically achieve. The whole chapter is organised around keeping these straight, because the commonest mistakes come from blurring them.

The emergent indications: effective care, not a choice

When dialysis is urgently needed, it is needed, and the indications are easy to remember. Refractory metabolic acidosis that will not correct medically; refractory hyperkalaemia threatening the heart; a dialysable intoxication, which Chapter 17 covers; diuretic-resistant fluid overload causing pulmonary oedema; and the uraemic complications — encephalopathy, pericarditis, and bleeding. Any of these, when refractory to medical management, mandates dialysis regardless of the AKI's stage or cause, and regardless of where the patient sits on any timing algorithm. These are not entries on a preference map; they are effective care, and delaying them while debating timing is dangerous.

When to start in the non-emergent patient: the equipoise

The hard question is the patient with severe AKI but no emergent indication — oliguric, a high creatinine, but a tolerable potassium, pH, and volume state. Should you start now, to 'get ahead of it,' or wait? This was the subject of a series of major trials, and they have largely answered it. A single-centre study suggested early benefit, but the larger, multicentre trials did not: accelerated initiation produced no survival advantage over a watchful, standard approach, and left more patients still dialysis-dependent later, with more of the complications that come with a line and a circuit. A subsequent trial testing an even more delayed strategy against a standard delayed one found no benefit to waiting longer, and a hint of harm. The synthesis is a sweet spot: do not start early without an indication, but do not delay so long that complications develop — watch closely, and start promptly the moment an indication appears. Within that window, the precise threshold reflects a weighing of the burdens of dialysis against the risks of waiting, which is where patient values legitimately enter.

Choosing the modality

Modality is mostly matched to physiology. Continuous renal replacement therapy removes fluid and solute slowly and is the choice for the haemodynamically unstable patient, for cerebral oedema where rapid shifts are dangerous, and for large-volume removal that must be gentle; its costs are immobility, continuous anticoagulation, and resource intensity. Intermittent haemodialysis clears efficiently and quickly — ideal for severe hyperkalaemia or a poisoning — and frees the patient for mobility and procedures, but it can destabilise the haemodynamically fragile. SLED, the prolonged intermittent hybrid, offers much of the haemodynamic tolerance of CRRT with the logistical advantages of intermittent treatment. Peritoneal dialysis has a role in resource-limited and paediatric settings. Crucially, CRRT and intermittent haemodialysis do not differ in survival, so the choice is driven by haemodynamics, goals, and logistics — not by a belief that one saves more lives.

Getting the dose right — and why more isn't better

The dose question once seemed open and is now closed. Two large trials compared intensive with less-intensive dosing and found no survival benefit to the higher dose. More dialysis is not better dialysis: it adds hypophosphataemia and hypokalaemia, clears drugs — antibiotics above all — more aggressively, and costs more, without improving outcomes. The targets that follow are a delivered CRRT effluent of 20 to 25 mL/kg/h, prescribed somewhat higher to offset the inevitable downtime from clotting and procedures, and an intermittent Kt/V of around 1.2 to 1.4 per session at least three times weekly, intensified only for the catabolic or fluid-overloaded. The discipline is to deliver the standard dose reliably rather than chasing a higher one.

Anticoagulation and access

For CRRT, regional citrate anticoagulation is preferred over systemic heparin where there is no contraindication: it causes less bleeding and prolongs circuit life by anticoagulating only the extracorporeal blood, with the calcium returned to the patient. Its hazard is citrate accumulation, chiefly in liver failure, which produces a metabolic alkalosis and a rising ratio of total to ionised calcium — a signal to reduce or stop it. Intermittent haemodialysis uses heparin or saline flushes. Access is a temporary non-tunnelled catheter, and site matters for both function and the patient's dialysis future: the right internal jugular is first choice, the femoral and left internal jugular acceptable, and the subclavian avoided because of the stenosis it causes for future permanent access. The access detail belongs to Book 2; the principle here is to protect future options.

Whether to dialyse at all: the preference-sensitive decision

The decision this chapter shares with the capstone is whether to start dialysis at all. In a patient with advanced frailty, irreversible multi-organ failure, or a terminal illness, dialysis may not restore a life the patient would value — it may only prolong dying, at the cost of lines, immobility, and time in hospital. This is not a question physiology can answer, and it is not the clinician's to answer alone; it is a goals-of-care decision made with the patient and family, informed by an honest prognosis. A time-limited trial of dialysis — starting with explicit goals and a defined review date, agreeing in advance what improvement would justify continuing and what would justify stopping — is a legitimate and often wise option when the prognosis is genuinely uncertain. The shared-decision scripts in this chapter exist to make that conversation possible; the fuller framework is built in Chapter 18.

Where the evidence is firm, and where values decide

It is worth being explicit about which parts of this chapter are settled and which are not, because the two demand different reasoning. The timing trials, the dosing trials, the survival equivalence of modalities, and the anticoagulation preference are matters of evidence, and the evidence is strong: start on indication, dose to standard, choose modality by physiology, anticoagulate with citrate. The 'whether to start at all' decision, and the precise threshold within the watchful window, are matters of values, and no trial will ever settle them, because they turn on what this patient wants and what dialysis can realistically deliver for them. Good practice means applying the evidence firmly where it speaks and sharing the decision genuinely where it cannot.

Table 13.1 — Emergent indications for dialysis (effective care)

Table 13.2 — Timing-of-RRT trials

Table 13.3 — Modality comparison

Table 13.4 — Dose targets and the dosing evidence

Table 13.5 — Anticoagulation and access

Table 13.6 — Effective-care versus preference-sensitive decisions

Clinical Reasoning

Safety & Evidence

Panel A — Red flags

Panel B — Never do

Pitfall 1 — Debating an emergency

Pitfall 2 — Early start to 'get ahead'

Pitfall 3 — More dose is better

Pitfall 4 — Choosing modality for survival

Pitfall 5 — Defaulting to dialysis

Graded statements (by evidence type)

Patient Decisions

Natural-frequency estimates for orientation, from the major RRT trials; they vary with population and severity. They convey the size of the dialysis decisions, expressed per 100 comparable patients.

First, what is NOT on this map. Several dialysis decisions are effective care, settled by evidence and physiology, with no role for preference: the emergent indications (Table 13.1), matching modality to haemodynamics, the KDIGO dose targets, and regional-citrate anticoagulation. These should not be presented to patients as choices — they are recommendations.

What IS preference-sensitive. The decisions below genuinely depend on the patient's values and prognosis, and belong in a shared conversation.

Language for the two conversations this chapter's equipoise demands. Adapt to the patient; the teach-back confirms understanding.

Apply & Test

Paste-ready notes. Tick the boxes that apply and delete the rest; make the effective-care-versus-preference distinction explicit in the record.

Template 1 — RRT initiation decision

• AKI stage ___ ; candidate for meaningful recovery/benefit: ☐ yes ☐ no → goals-of-care/SDM (Chapter 18).
• Emergent indication: ☐ refractory acidosis ☐ hyperkalaemia ☐ intoxication ☐ overload ☐ uraemia ☐ none.
• If emergent → dialyse now (effective care).
• If none → watchful waiting; start trigger defined as ___ ; reviewed ___ .
• Preference-sensitive elements discussed (if applicable): ☐ whether to start ☐ time-limited trial ☐ modality preference.
• Decision and rationale: ___ .

Template 2 — RRT prescription

• Modality: ☐ CRRT (unstable) ☐ IHD (stable/efficiency) ☐ SLED ☐ PD — rationale (haemodynamics/goals) ___ .
• Dose: ☐ CRRT effluent 20–25 mL/kg/h delivered (prescribed ___ ) ☐ IHD Kt/V ___ , frequency ___ .
• Anticoagulation: ☐ regional citrate ☐ heparin ☐ saline flushes; citrate accumulation watch (liver failure): ☐ yes.
• Access: ☐ right IJ ☐ femoral ☐ left IJ (subclavian avoided).
• Drug dosing reviewed for RRT clearance: ☐ antibiotics (avoid under-dosing) ☐ other ___ .
• Review/reassessment plan: modality and continued-need reviewed ___ .

Genuine tensions this evidence and these values leave open; sit with them rather than resolving them too quickly.

• The timing trials tell us early dialysis doesn't help on average, but the patient in front of you is not an average. How do you apply population evidence to an individual whose trajectory you cannot see?
• Watchful waiting requires tolerating a rising creatinine and an oliguric patient without acting. How do you hold that nerve, and how do you know when watching has become waiting too long?
• A time-limited trial sounds clean in principle, but stopping dialysis once started feels different from never beginning. How do you make the review genuine rather than a formality that always continues?
• The 'whether' decision turns on what the patient would value, which you often cannot ask them directly. How much weight should family interpretation carry, and how do you guard against projecting your own values?
• Effective-care and preference-sensitive decisions look similar at the bedside — both end in 'dialyse' or 'don't.' What discipline keeps you from quietly converting a values decision into a technical one, or vice versa?