🧠🩸 Too Early… or Too Late? When Dialysis Is a Decision, Not a Reflex. - By Dr Bram Rochwerg🩸🧠

👨‍⚕️ Author of the Month

 

Dr Bram Rochwerg

Intensivist & Clinical Trialist,
McMaster University, Canada
Chair – Internal Medicine Section,
Society of Critical Care Medicine
Chair-Elect – Canadian Critical Care Trials Group Vice-Chair – Surviving Sepsis Campaign
Associate Editor – Critical Care Medicine
Editorial Board – CHEST

Dr. Rochwerg doesn’t just interpret evidence—he builds it.This blog reflects how a trialist thinks when evidence meets reality.

“Critical Care / Renal Replacement Therapy

A 74-year-old woman with hypertension and coronary artery disease is admitted to the ICU with septic shock from intra-abdominal sepsis. She requires norepinephrine at 0.15 µg/kg/min and mechanical ventilation. Over 48 hours, she develops worsening acute kidney injury with creatinine rising from 1.1 mg/dL to 3.4 mg/dL, oliguria (<200 mL/day), metabolic acidosis (pH 7.26), and rising urea levels.

She has no life-threatening hyperkalemia or pulmonary edema. The ICU team debates whether to initiate early renal replacement therapy (RRT) preemptively or to delay RRT until standard indications develop.

One intensivist argues that early RRT may prevent complications and improve outcomes. Another cites large randomized trials suggesting no mortality benefit to early initiation and potential harms, including catheter-related complications and dialysis dependence.

The patient’s family asks whether starting dialysis now will improve her chances of survival or recovery.

Key Questions at the Bedside

(Click / Tap on Questions to Reveal Content)

Population (P): Critically ill adult with severe AKI, without life-threatening complications that mandate immediate RRT (no refractory hyperkalemia, no severe pulmonary edema causing hypoxemia, no refractory severe acidosis, no overt uremic complications).

Intervention (I): Early/accelerated/immediate initiation of RRT/CRRT

Comparator (C): Delayed initiation of RRT/CRRT (watchful waiting; start only if prespecified complications develop eg. refractory acidosis or hyperkalemia or AKI persists).

Outcomes (O):

  • Patient-important: short-term mortality (e.g., 28–90 days), long-term mortality (e.g., 1 year if available), quality of life/function.
  • Kidney-important: dialysis dependence at 90 days, renal recovery, MAKE (major adverse kidney events).
  • Others: ventilator-free days, ICU/hospital LOS, fluid balance/organ support.
  • Harms: catheter-related complications, hypotension, bleeding, infection, electrolyte derangements (e.g., hypophosphatemia), delivered dose issues

Randomized Controlled Trial

Strengths

  • Minimizes confounding by indication (big issue in timing questions).
  • Allows prespecified, operational definitions of “early” vs “delayed.”
  • Best for causal inference on mortality and dialysis dependence.

Limitations

  • Crossover is common (delayed arm may start early for evolving indications; early arm may not receive RRT if instability/logistics intervene).
  • “Early” and “delayed” are not uniform across trials, creating interpretability issues.
  • Expensive and time intensive – ++ resources

Observational cohort

Strengths

  • Large samples, real-world practice, can explore subgroups.
  • Faster/cheaper; good for hypothesis generation.

Limitations

  • Highly vulnerable to confounding by severity, clinician judgment, and immortal-time/lead-time biases (sicker patients may get “early” RRT, or patients must survive long enough to become “late”).

Patient-important: short-term mortality (e.g., 28–90 days), long-term mortality (e.g., 1 year if available), quality of life/function, return to work, return to independent living.

Key issues that make this topic tricky:

  • Definition heterogeneity: “Early” and “delayed” vary (AKI stage-based triggers vs time-based triggers vs complication-based triggers).
  • Population heterogeneity: septic shock vs mixed ICU; surgical vs medical; baseline CKD; varying illness severity.
  • Intervention heterogeneity: modality (CRRT vs intermittent), dose, anticoagulation, catheter strategy.
  • Crossover and protocol adherence: common and can dilute effects; important to extract adherence and perform sensitivity analyses.
  • Competing risk structure: death competes with “dialysis dependence” and “kidney recovery.” Analyses must handle this carefully.
  • Outcome timepoints: 28 vs 60 vs 90 days; MAKE definitions differ.
  • Risk of bias: open-label designs are common; mortality is less prone to measurement bias, but “kidney recovery” and decisions to start RRT can be influenced. Important to assess certainty in evidence using framework such as GRADE.
  • Balancing publication bias vs including unpublished sources

A good recommendation must integrate:

  • Net benefit: Large trials show no consistent mortality benefit to routine early/accelerated RRT in the absence of urgent indications; early strategies may increase exposure to RRT and some harms.
  • Avoidability: Delayed strategies can avoid RRT in a meaningful fraction of patients (important for resource use and avoiding unnecessary invasive therapy).
  • Operational triggers: Recommendations must specify what “standard indications” are, and define “persistent AKI” thresholds that should prompt starting (e.g., refractory acidosis, hyperkalemia, pulmonary edema, uremic complications, prolonged oliguria/anuria, high urea).
  • Feasibility & resources: staffing, CRRT availability, anticoagulation practices, ICU nurse ratios.
  • Patient values/preferences: some patients prioritize avoiding long-term dialysis; others prioritize “doing everything early.” That affects how strongly you recommend watchful waiting.

The patient from the vignette (septic shock, ventilated, on norepinephrine, severe AKI without emergent indications) is quite similar to patients enrolled in several key trials:

  • STARRT-AKI (NEJM 2020): broad ICU population with severe AKI; compared accelerated vs standard initiation; good match for “no urgent indications at randomization.”
  • IDEAL-ICU / AKI in sepsis: specifically addressed AKI in septic shock populations (helpful for your case framing).
  • AKIKI-2: explored “how far can we delay,” with structured thresholds (e.g., prolonged oliguria or urea triggers) — relevant if you want to debate how long to wait.

External validity limits (common in practice):

  • More severe hemodynamic instability than trial medians (e.g., very high vasopressor dose).
  • Profound fluid overload/ARDS needing aggressive net negative balance.
  • Chronic advanced CKD, transplant, cirrhosis, or unique toxic ingestion scenarios.
  • Local practice differences: earlier RRT for fluid management, or limited CRRT capacity.
  • Different thresholds for “urgent indication” (especially acidosis or uremia).

A trialist would have to manage:

  • Selection bias: enrolling only “stable enough” patients can limit applicability; need broad screening logs and clear eligibility.
  • Performance bias (open-label): clinicians know allocation; might change co-interventions (diuretics, fluids, nutrition/protein targets).
  • Contamination/crossover: delayed-arm patients getting early RRT for soft indications; early-arm not receiving RRT due to logistics. Predefine criteria and monitor adherence.
  • Co-intervention bias: differences in modality, dose, anticoagulation, ultrafiltration goals.
  • Outcome ascertainment: mortality is robust; kidney recovery/dialysis dependence is sensitive to follow-up completeness and practice patterns; standardize definitions and follow-up timepoints.
  • Competing risks: death affects dialysis dependence and recovery endpoints; prespecify appropriate statistical handling.
  • Center effects: CRRT expertise varies; consider stratification/random effects or minimum standards.

Here’s a bedside-style explanation:

“Your family member’s kidneys are under severe stress from the infection and shock. Dialysis is a supportive treatment — it can correct dangerous potassium levels, severe acid build-up, fluid overload, or symptoms of toxin build-up. Right now, we’re not seeing those emergency problems yet.

Large studies in ICU patients like her compared starting dialysis early versus waiting and starting only if specific problems develop. Overall, starting early did not improve survival, and it exposed some patients to dialysis who later would have recovered without it. Some studies also suggest early dialysis can lead to more complications from the catheter or the treatment itself, and in some groups a higher chance of still needing dialysis later.

So our plan is active watchful waiting: we’ll monitor her labs, acid level, urine output, fluid balance, and breathing closely. If she develops a clear reason that dialysis will help — like worsening fluid overload affecting oxygen levels, dangerous potassium, severe acid that isn’t responding, or signs of toxin build-up — we will start dialysis promptly. If those don’t happen and her kidneys begin to recover, we may be able to avoid dialysis altogether.”

Loading

Leave Your Comment