A critical review of the first randomised trial of prehospital REBOA in non-traumatic out-of-hospital cardiac arrest.

Mario Rugna

A technique JUST for TRAUMA

REBOA was not designed for cardiac arrest. It grew out of trauma care, as a way to buy time in patients bleeding to death from injuries no tourniquet can reach non-compressible torso haemorrhage. The idea is mechanical and direct: thread a balloon catheter up the femoral artery, inflate it in the aorta, and physically dam the flow above the bleeding source. Occlude in Zone 1 (the descending thoracic aorta) or Zone 3 (below the renal arteries) depending on where the haemorrhage sits, and you both slow the exsanguination and prop up the pressure reaching the heart and brain.

That is the world REBOA has lived in for most of its history. The case series, the registries, the device approvals all sit within traumatic haemorrhage and traumatic cardiac arrest. In that setting the balloon does two jobs at once: it controls bleeding and augments proximal perfusion pressure.

REBOARREST is different! That’s WHY

Non-traumatic cardiac arrest is a fundamentally different problem, and that is what makes REBOARREST unusual. The patient isn’t bleeding; the heart has simply stopped. Occluding the thoracic aorta during chest compressions redistributes the little flow that CPR generates toward the coronary and cerebral circulations, raising coronary perfusion pressure in the way adrenaline is meant to. It is REBOA repurposed as “mechanical adrenaline,” stripped of the haemorrhage-control rationale that justified it everywhere else.

Until now, that idea rested almost entirely on physiological reasoning, animal data, case reports and small uncontrolled series. REBOARREST is the first randomised controlled trial to test it the first time REBOA has been asked, under proper experimental conditions, to prove itself outside of bleeding. That context is worth holding onto while reading what follows: a technique borrowed from one domain and applied to another carries its assumptions with it, and not all of them travel.

So WHAT!

The headline from REBOARREST is easy to summarise and easy to misread. Among 179 patients with non-traumatic out-of-hospital cardiac arrest (OHCA), adding resuscitative endovascular balloon occlusion of the aorta (REBOA) to advanced life support did not improve sustained ROSC: 28% in the intervention arm versus 26% in controls, an adjusted risk difference of 1.8% (95% CI −11 to 15, p=0.78). The graphical abstract states it plainly the strategy “did not significantly improve rates of sustained ROSC.”

A busy clinician skimming the abstract will file REBOA-for-OHCA under “tried, didn’t work.” The trial itself doesn’t support that conclusion, and the gap between what was found and what it means is the most interesting thing about this paper.

This is, first and foremost, a genuine achievement. It is the first RCT of REBOA in non-traumatic OHCA, run pragmatically across 12 sites in three countries, with concealed allocation, a blinded statistician, a prespecified analysis plan, an active data monitoring committee, and independent review of ALS quality. Those are not small things in prehospital research, where trials of this kind barely exist. But it’s worth being clear about what it can and cannot tell us.

The trial was built to detect an effect almost no intervention produces (not only in cardiac arrest)

REBOARREST was powered to detect a doubling of sustained ROSC from a baseline of 18% to 36%. The authors, to their credit, call this goal “optimistic.” It is more than that: powering a trial for a doubling means designing it to be blind to anything smaller. And in cardiac arrest, an absolute increase of even 5–10% in ROSC would be clinically meaningful and worth chasing.

This is the difference between absence of evidence and evidence of absence, and it matters. A non-significant p-value in an underpowered trial tells you the study couldn’t see an effect, not that no effect exists. REBOARREST cannot distinguish “REBOA doesn’t help” from “REBOA helps by a real but sub-doubling amount that this trial was never equipped to detect.”

Most of the intervention arm patients never got the intervention

The second structural issue compounds the first. Of 88 patients randomised to REBOA, only 51 (58%) actually underwent aortic occlusion. The rest didn’t: 19% achieved ROSC before the balloon could be inflated, 16% had an unsuccessful procedure, and 7% were aborted.

The primary intention-to-treat (ITT) analysis therefore compares a group in which four out of ten patients never received the treatment against a control group. This is a defensible and honest way to answer a strategy question “should an EMS system deploy REBOA in OHCA?” but it is not a clean test of whether aortic occlusion works physiologically. Dilution of this magnitude pushes the result mechanically toward the null before biology gets a say.

That’s why the as-treated (AT) signal deserves a mention, with heavy caveats. When patients who actually received occlusion were compared with controls, the first AT sensitivity analysis showed a significant benefit: a 16% absolute increase in sustained ROSC (95% CI 2 to 29, p=0.02). Read alone, that looks like the intervention doing exactly what it was supposed to. But this analysis conditions on a post randomisation event (whether the balloon went in and trims controls who died or achieved ROSC early) wide open to selection and time bias. It cannot confirm efficacy. What it can do is tell us ITT analysis is that the null result is hiding a signal worth taking seriously, not closing the book.

By the time the balloon inflated, the physiology may already have been lost

The rationale for REBOA in arrest is time critical: occlude the aorta, raise proximal and coronary perfusion pressure, mimic the haemodynamic effect of adrenaline, and buy a shot at ROSC. That logic depends entirely on getting there early.

In REBOARREST, the median interval from arrest to occlusion was 47 minutes. The authors name this “the major limitation,” and they’re right. By three-quarters of an hour, most of these patients were in prolonged, poor prognosis arrest, and the window in which augmented perfusion pressure might have mattered had largely closed. The trial reflects a mixed urban rural, largely helicopter dispatched system where reaching the patient takes time so what it really tested was late prehospital REBOA. Whether early occlusion (a metropolitan short-transport service, or an in-hospital arrest) would behave differently is a question this trial leaves wide open.

The one number that would explain the result was never recorded

Here is the quiet problem at the centre of the paper. The entire hypothesis rests on REBOA raising aortic and coronary perfusion pressure, and blood pressure was never measured. The Prytime catheter used in Italy lacked the equipment; the Reboa Medical catheter used elsewhere couldn’t measure pressure without exceeding its CE approval. Intra-aortic pressure data simply don’t exist.

So when the primary outcome comes back null, we can’t tell which of three very different stories is true: the balloon failed to raise coronary perfusion pressure, or it raised pressure but too late, or it raised pressure and ROSC still didn’t follow. EtCO₂ did rise significantly after occlusion but that was measured only within the occlusion subgroup, with no contemporaneous control comparison, so it’s a within group observation consistent with the mechanism rather than between group proof of it.

The endpoint that moved isn’t the endpoint that matters

Sustained ROSC ≥20 minutes is a surrogate. What patients and families care about is survival with an intact brain, and on those measures the arms were flat and consistent: 30-day survival was 7% in both groups, and good neurological outcome (mRS 0–3) was 6% versus 3%, not significant.

Even if the ROSC trend had been real, it didn’t carry through to survival or neurology. But the mirror image is also true: a trial powered on ROSC is hopelessly underpowered for these harder outcomes, so it can’t rule a survival difference in or out either.

So what does REBOARREST actually tell us?

On feasibility and safety, the trial is convincing. A two-person prehospital team can perform this procedure with a short procedure time (median 14 minutes from randomisation to occlusion), acceptable cannulation success, and no excess of adverse events. That’s a real, well-supported result, and it opens the door to other endovascular interventions in the field.

On efficacy, the honest verdict is: unproven, not disproven. What REBOARREST demonstrates is that a strategy of late prehospital REBOA, in a mixed urbanrural, expert-staffed, ECPR adjacent system, did not improve a surrogate outcome in a trial that could only ever have detected a very large effect, that never delivered the intervention to 42% of the treatment arm, and that never measured the pressure it was built around.

That is not “REBOA doesn’t work in cardiac arrest.” It’s “we still don’t know, and here’s exactly why we don’t.” The right response to this trial isn’t to abandon the question. It’s to design the study REBOARREST couldn’t be: earlier occlusion, invasive pressure monitoring, a realistic effect size, and enough events to see it. The as treated signal is reason enough to build it.

Reference: Brede et al. Prehospital resuscitative endovascular balloon occlusion of the aorta in non-traumatic out-of-hospital cardiac arrest (REBOARREST): an international, multicentre, open label, pragmatic, randomised, controlled trial. Critical Care 2026;30:324.

Tags: cardiac arrest, emergency medicine, Out of Hospital Cardiac Arrest, prehospital emergency medicine, REBOA