Every procedure has a learning curve. What makes REBOA different is that the cases that teach it are rare, unpredictable, and unforgiving. A clinician can complete a course, perform well in simulation, and then wait months for a real patient, who arrives at 3 a.m. with a falling blood pressure and no time to look anything up.
In September 2026, a systematic review in the Journal of Clinical Medicine gathered the clinical evidence on how REBOA proficiency actually develops in practice. Its conclusions are modest and carefully qualified, as they should be. But they point toward something we have believed since the COBRA-OS® was first designed: the easier a procedure is to do well, the sooner teams do it well, and the longer they keep doing it well. In REBOA, simplicity is not a convenience. It is a safety feature.
What the evidence actually tells us
The review identified only seven clinical studies that examined how REBOA performance changes with experience. Three tracked outcomes against case numbers directly. The largest of these, a prospective study across eight Level I trauma centres in the United States and Canada, followed 377 procedures and found that technical challenges clustered early. Individual surgeons reached a plateau after roughly five cases; institutions took about ten.
A Korean study of five regional trauma centres told a less tidy story. Risk-adjusted performance improved after somewhere between 5 and 34 cases at three centres, while the other two showed no consistent improvement at all. Registry data from Japan and the American AORTA registry showed outcomes improving year over year, with in-hospital survival in Japan rising from 39% to 60% across 2004 to 2015., But calendar time also brings better devices, better patient selection, and better trauma systems, so those trends cannot be credited to experience alone.
The authors are clear that none of these numbers is a validated competency threshold, and that the evidence cannot support a specific case-number requirement. We agree. What the data do suggest is that experience matters, that it matters at the team level as well as the individual level, and that it does not accumulate evenly everywhere.
Five cases is a lot when cases are rare
“Five cases to proficiency” sounds encouraging, and at a busy urban trauma centre it may be. But many hospitals that receive bleeding patients see only a handful of REBOA candidates in a year. Earlier work from Philadelphia found that REBOA training begins to fade at around six months in the absence of real cases. Put those two observations together and the arithmetic becomes uncomfortable: at a low-volume centre, an operator may never reach the plateau, or may reach it and quietly slide back down before the next patient arrives.
This matters because severe hemorrhage does not respect hospital volume. Patients bleed in rural hospitals, community centres, and austere military settings, often far from the teams who perform REBOA every month. If the benefits of aortic occlusion depend on climbing a steep curve that only high-volume centres can climb, then the technology will keep failing the patients who are furthest from definitive care.
Where the curve is steepest
One detail from the North American study deserves attention. The early technical challenges were not mainly about the balloon. They were about femoral access, sheath management, and catheter removal: the steps surrounding the occlusion rather than the occlusion itself. Catheter removal alone accounted for more than a third of individual challenges. The only randomized trial of REBOA in trauma, UK-REBOA, tells a related story; balloon inflation was achieved in fewer than half of the patients assigned to the REBOA strategy.
Every step a device adds is a step that must be learned, rehearsed, and remembered under stress. That is why we built the COBRA-OS® around a 4 French profile, with no routine sheath upsizing, no guidewire required, fewer procedural steps, and the accessories included in the kit. A smaller sheath is easier to place and easier to remove. A shorter sequence leaves less to forget between cases.
We want to be careful here. No study has yet measured whether a lower-profile, simpler device shortens the clinical learning curve, and we would welcome that research. But the logic is straightforward: the fewer things a team must master, the fewer things can go wrong on the first real case, and the less there is to lose during the quiet months in between.
The team learns too
The institutional curve was longer than the individual one, and in the Korean study two of five centres never showed clear improvement. That should not surprise anyone who has worked a massive hemorrhage. REBOA is a bridge, not a destination, and its value depends on what happens around it: who makes the call, whether the occlusion zone was chosen before the balloon went up, who is watching the clock, and how quickly the patient reaches surgical or endovascular control.
Those are team skills, and they can be practised without waiting for a patient. Simulation that includes the whole team rather than only the operator, standardized workflows, structured debriefing after every case, and simple performance monitoring are all reasonable ways for a centre to climb the curve deliberately instead of by accident. A simpler device helps here as well. When the technical steps demand less attention, the team has more attention left for the decisions that actually determine outcome.
Lowering the barrier to proficiency
Our mission has always been to lower the barriers to hemorrhage control. We usually describe those barriers as access, complexity, training, and cost. The learning-curve evidence adds a fifth: proficiency, and the difficulty of keeping it when real cases are rare.
The best REBOA device is not the one that performs well in the hands of a surgeon who places a balloon every week. It is the one a team can use well on its first real case, and still use well after six quiet months. That is the standard we hold the COBRA-OS® to. Because in catastrophic hemorrhage, what REBOA buys is time, and no patient should lose that time to a learning curve.


