Dialysis Status Becomes the Critical Modifier in Planned Pantoprazole Analysis
A new secondary analysis plan from the REVISE trial turns attention to a group that sits at the intersection of bleeding risk and organ failure: critically ill patients who are receiving renal replacement therapy when they enter the study, or who were already on maintenance dialysis before admission. The question is not whether pantoprazole works in general. A prior international randomized trial already found that the proton pump inhibitor reduced clinically important upper gastrointestinal bleeding in invasively ventilated ICU patients, without changing mortality. The question now is whether kidney failure requiring dialysis shifts the magnitude of benefit, or alters the balance of benefit and harm.
The protocol is built around a biologically plausible tension. Kidney dysfunction can promote bleeding through platelet dysfunction, uremic effects on hemostasis, anemia, and the complex coagulation changes that accompany critical illness. At the same time, proton pump inhibitors have been scrutinized for possible kidney-related adverse effects in broader clinical settings, which makes subgroup exploration more than an academic exercise. In a population already exposed to renal failure and critical illness, the margin between preventing a stress-related bleed and introducing another problem may be narrow.
One sentence captures the clinical stakes plainly: dialysis status could change the size of pantoprazole’s benefit without changing the direction of effect.
Why the REVISE Dataset Is Being Reopened for This Question
REVISE enrolled invasively ventilated ICU patients and compared pantoprazole with placebo for stress ulcer prophylaxis. This planned post hoc analysis will use that randomized framework to test effect modification by kidney replacement therapy status at baseline. The investigators have defined two related exposure questions. The primary one asks whether any RRT on study day 1 modifies pantoprazole’s effect. That category includes maintenance dialysis already being delivered before hospitalization and hospital-initiated therapies such as intermittent hemodialysis, continuous renal replacement therapy, sustained low efficiency dialysis, and peritoneal dialysis. The secondary question narrows the lens further to outpatient maintenance RRT before the index admission, specifically intermittent hemodialysis or peritoneal dialysis.
This split is clinically sensible. Patients who arrive already dependent on chronic dialysis are not the same as patients who develop acute kidney failure after ICU admission and then begin extracorporeal support. Their bleeding risks, inflammatory burden, and baseline physiology differ. Chronic dialysis patients also arrive with a different set of medication exposures, vascular access issues, and competing causes of anemia. A single subgroup defined as “kidney failure” would blur those differences. The protocol instead separates broad RRT exposure from preexisting maintenance dialysis, which should help distinguish persistent chronic risk from acute critical illness physiology.
Outcomes Focused on Bleeding First, Mortality Second
The primary outcome for both planned analyses is clinically important upper GI bleeding. Ninety-day mortality is the co-primary clinical concern in the overall framing of the project, while patient-important upper GI bleeding, ICU mortality, and hospital mortality are listed as secondary outcomes. The statistical plan is straightforward: Cox proportional-hazard models will include randomized treatment group, the RRT exposure of interest, and the interaction term between them. Results will be reported as hazard ratios with 95% confidence intervals, along with p-values for interaction.
That interaction term is the key. If pantoprazole’s relative benefit is stronger in dialysis patients, the analysis would suggest a subgroup with heightened absolute need for stress ulcer prophylaxis. If the effect is similar across strata, dialysis status may matter less than clinical intuition suggests. If the direction differs, the finding would force a more careful reconsideration of how renal failure reshapes both mucosal bleeding risk and the response to acid suppression.
The protocol does not claim an answer yet. It lays out the test.
The Statistical Design Relies on an Interaction Test, Not a New Trial
Because this is a secondary analysis, the investigators are working within the limits of an already completed randomized trial. That design has strengths and constraints. Randomization preserves the core comparison between pantoprazole and placebo, but subgroup analyses are inherently more vulnerable to reduced power and chance findings than the parent trial itself. The planned use of prespecified exposure definitions and Cox modeling helps keep the analysis disciplined, yet the central inference will still depend on whether the interaction is large enough to be clinically credible.
The protocol also reflects an important ICU reality: bleeding prevention is not a one-size-fits-all intervention. Stress-related mucosal disease develops in the setting of shock, ventilation, coagulopathy, and multi-organ failure. Dialysis can amplify several of those hazards at once. Circuit anticoagulation, platelet dysfunction, and fluctuating hemodynamics may all contribute to the risk environment. A proton pump inhibitor may therefore produce its greatest value in the patients who are already most fragile from a hemostatic standpoint.
What Clinicians Will Need If the Analysis Is Positive or Null
If the planned analysis finds stronger benefit in patients on RRT, the result would support targeted stress ulcer prophylaxis in a subgroup often treated as uniformly high risk but rarely examined with enough granularity. If the interaction is null, pantoprazole’s benefit would appear more consistent across dialysis and non-dialysis patients, which would simplify bedside decision-making. Either way, the most immediate clinical issue is whether baseline kidney failure meaningfully changes the tradeoff between preventing upper GI bleeding and exposing patients to another routine ICU medication.
A second sentence matters here: the study is designed to answer a subgroup question, not to overturn the parent REVISE trial.
For now, the analysis remains a protocol and statistical plan, not a results paper. Its value lies in forcing a precise question onto a common ICU practice. Patients receiving dialysis are frequent, clinically complex, and often underrepresented in subgroup-specific critical care decision-making. This planned REVISE analysis will determine whether pantoprazole behaves differently in them, or whether the same bleeding-prevention effect seen in the full trial persists regardless of renal replacement therapy status.