Clinical Question
Among critically ill adults with severe acute kidney injury and an indication for renal-replacement therapy, is intermittent hemodialysis noninferior to continuous renal-replacement therapy as the initial treatment strategy for major adverse kidney events by day 90?
TL;DR
Intermittent hemodialysis did not demonstrate noninferiority to CRRT. The primary composite of death, dialysis dependence, or persistent kidney dysfunction occurred in 78.1% versus 71.6% of patients, respectively. The adjusted difference was 6.28 percentage points against intermittent HD, with a 95% confidence interval from −1.22 to 13.79. That interval crossed the prespecified 7.5-point noninferiority margin and also included no difference.
The more concerning finding was mortality. Estimated 90-day mortality was 63.3% with intermittent HD and 53.6% with CRRT, with an adjusted hazard ratio of 1.30 (95% CI, 1.04–1.63). Enrollment stopped early after repeated safety analyses showed higher 28-day mortality with intermittent HD.
Early stopping reduced precision and may have exaggerated the size of the mortality difference. Still, I think this strengthens the case for starting with CRRT in critically ill patients like those enrolled. It does not establish which modality is preferable after stabilization or when a patient should transition to intermittent HD.
The Deets
Choosing between intermittent HD and CRRT is a familiar ICU discussion. Hemodynamic tolerance, fluid removal, staffing (especially important), equipment, and the patient’s trajectory all influence the decision. Earlier randomized trials had not established a consistent mortality advantage for either modality.
ICRAKI revisited that question in a contemporary population with a recognized indication for dialysis. These were critically ill patients: approximately 84% were invasively ventilated and 79% were receiving catecholamines like norepinephrine or epinephrine on the day of randomization.
The investigators initially planned to enroll 1,000 patients across 42 French ICUs. 38 centers enrolled patients. They randomized 554 adults to intermittent HD or CRRT as the initial modality. The main analysis included 535 patients.
Patients were supposed to remain on the assigned modality from day 0 through day 3 unless kidney recovery allowed dialysis to stop. Switching was permitted from day 4 onward. This design tests an initial treatment strategy rather than a commitment to one modality throughout the ICU stay.
The trial used a noninferiority design. Investigators wanted to determine whether intermittent HD could exclude an unacceptable increase in major adverse kidney events compared with CRRT. They set that acceptable disadvantage at 7.5 absolute percentage points.
Intermittent HD failed that test. Although the estimated disadvantage was below the margin, the upper confidence limit extended well beyond it. The primary composite also failed to establish CRRT superiority because the interval included zero.
Meanwhile, the safety monitoring board saw a sustained mortality difference favoring CRRT. After three interim reviews of 28-day mortality, enrollment stopped. The mortality difference persisted through day 90.
Study Snapshot
Investigator-initiated, multicenter, open-label, randomized noninferiority trial in France
42 participating ICUs; 38 enrolled patients
Planned enrollment: 1,000
Randomized: 554; main analysis: 535
Intermittent HD: 260
CRRT: 275
Adults with AKI, a recognized indication for renal-replacement therapy, and current or prior invasive ventilation, catecholamine infusion, or both during the same illness
Primary outcome: major adverse kidney event by day 90
Primary noninferiority margin: 7.5 absolute percentage points
Funded by the French Ministry of Health
What They Found
Primary Outcome
Death, dependence on renal-replacement therapy, or serum creatinine more than 25% above baseline by day 90
Measure | Intermittent HD | CRRT |
|---|---|---|
Primary-outcome events | 203/260 | 197/275 |
Event rate | 78.1% | 71.6% |
Adjusted absolute risk difference, intermittent HD minus CRRT: 6.28 percentage points
95% CI: −1.22 to 13.79
Adjusted risk ratio: 1.09 (95% CI, 0.98–1.20)
P=0.10 for the between-group difference; this is not the noninferiority P value
Per-protocol results were similar to the main analysis
How to Read the Noninferiority Result
The two relevant thresholds answer different questions:
Question | Required finding | ICRAKI result |
|---|---|---|
Can intermittent HD exclude an unacceptable disadvantage? | Upper confidence limit below +7.5 percentage points | No: upper limit was +13.79 |
Is CRRT superior for the primary composite? | Entire confidence interval above 0 | No: lower limit was −1.22 |
The point estimate of +6.28 being below +7.5 is insufficient. Noninferiority requires enough precision to rule out harm beyond the margin.
Early stopping made that harder by reducing the sample size. But the estimate itself also leaned toward harm and sat close to the margin. We cannot assume that completing enrollment would have demonstrated noninferiority.
Mortality
Kaplan–Meier mortality estimate | Intermittent HD | CRRT |
|---|---|---|
Day 28 | 53.6% | 44.8% |
Day 60 | 62.2% | 50.7% |
Day 90 | 63.3% | 53.6% |
The adjusted mortality hazard ratio through day 90 was 1.30 (95% CI, 1.04–1.63) for intermittent HD versus CRRT. This describes a higher hazard of death over follow-up, not a 30-percentage-point increase in mortality.
Kidney Outcomes and Safety
Outcome | Intermittent HD | CRRT |
|---|---|---|
Dialysis dependence at day 90, among patients treated as alive in the analysis | 2.1% | 2.3% |
Creatinine >25% above baseline at day 90, among patients treated as alive and off dialysis | 39.4% | 37.6% |
Hypotension | 68.8% | 66.5% |
Hemorrhage | 23.5% | 21.5% |
Severe cardiac rhythm disorders | 5.4% | 3.6% |
Overall hypotension rates were similar. Exploratory post hoc analyses found more transient catecholamine escalation and higher daily minimum and maximum catecholamine doses with intermittent HD. Those findings suggest a possible hemodynamic explanation but do not establish the mechanism of the mortality difference.
What Bothers Me
Early stopping complicates the effect estimate. Only 554 of the planned 1,000 patients were randomized. Stopping reduced power for the primary outcome and kidney-specific components, while stopping on an emerging mortality difference may overestimate its magnitude.
Subgroup results do not identify a clearly safe intermittent-HD population. The investigators found no convincing treatment-effect heterogeneity. That does not prove the effect is identical across vasopressor requirements, age groups, or other subgroups; the trial was not large enough to settle those questions.
The later treatment decision remains unresolved. Switching after the initial treatment period was allowed and common. This trial cannot tell us when a recovering patient should move from CRRT to intermittent HD.
Dashevsky's Dissection
For patients: Both approaches provide dialysis support while the kidneys are injured. In this trial, starting with CRRT was associated with better survival than starting with intermittent HD. The exact size of that difference is uncertain because the study stopped early.
The choice still depends on the patient’s condition, specific dialysis needs, available expertise, and goals of care. The trial does not suggest that someone who stabilizes on CRRT must remain on it throughout recovery.
For pulmonary and critical care physicians: I think ICRAKI raises the threshold for choosing intermittent HD as the initial modality in patients resembling this population. Most were ventilated, receiving vasopressors, and had substantial organ failure. The mortality signal makes it harder to assume the two approaches are interchangeable during that initial period.
The nonsignificant primary composite should not reassure us that outcomes were equivalent. The study could not exclude an unacceptable disadvantage with intermittent HD, and mortality separately favored CRRT.
At the same time, the trial does not justify concluding that every patient with AKI requires CRRT. It enrolled critically ill patients with recognized dialysis indications and excluded conditions favoring a particular modality. It also leaves the transition after stabilization unresolved.
If intermittent HD is selected, treatment delivery and tolerance deserve close attention: dialysis dose, fluid-removal goals, blood-pressure changes, and vasopressor escalation. The exploratory hemodynamic findings support further investigation without proving that those factors caused the survival difference.
This also fits into a broader sequence of AKI decisions. Trials of dialysis timing address when to start, while ICRAKI addresses how to start once an indication exists. Our prior BICARICU-2 trial covers a related question: whether bicarbonate can reduce or delay dialysis use in severe acidemia with AKI. Those decisions should remain distinct.
For the health system: CRRT requires equipment, consumables, trained staff, and sustained bedside support. A stronger clinical preference for initial CRRT has operational consequences, particularly in units with limited capacity.
ICRAKI did not establish cost-effectiveness or compare staffing models. Hospitals should examine whether they can reliably deliver the modality clinicians judge most appropriate, while avoiding the assumption that the observed survival difference will translate unchanged across institutions.
In summary, ICRAKI failed to demonstrate noninferiority of initial intermittent HD and found a concerning mortality difference favoring CRRT. Early stopping limits precision and may inflate the observed effect, but the findings strengthen the case for CRRT during the initial phase of severe critical illness. The best approach after stabilization remains an open question.
Gaudry S, et al. “Intermittent or Continuous Renal-Replacement Therapy in Acute Kidney Injury.” New England Journal of Medicine. Published October 14, 2026. doi:10.1056/NEJMoa2608397.
