We conducted a combined retrospective–prospective observational cohort study in a 1,000‑bed tertiary referral trauma center serving northeastern Thailand. The study population comprised all trauma patients undergoing REBOA for NCTH between January 2012 and December 2024. The retrospective phase encompassed January 2012 to December 2022; the prospective phase extended from January 2023 to December 2024. The Ethics Committee approved this study; informed consent requirements were waived for retrospectively collected data.

This was a complete enumeration study including all consecutive patients meeting the inclusion criteria during the 12‑year study period.

Inclusion criteria: adult trauma patients (age > 18 years); clinical diagnosis of NCTH (abdominal, pelvic, or thoracic hemorrhage); REBOA performed as a component of resuscitation in non‑responder hemorrhagic shock patients.

Exclusion criteria: age < 18 years; REBOA performed for non‑traumatic indications; incomplete medical records (retrospective cohort only).

No formal sample size calculation was performed a priori. With 23 patients (18 from the retrospective phase and five from the prospective phase) and an observed survival rate of 21.7%, the 95% confidence interval for survival ranges from 7.5% to 43.7%, providing reasonable precision for this rare intervention in a resource‑limited setting.

All patients underwent the primary survey with simultaneous resuscitation following the ATLS^®‑protocol. Extended Focused Assessment Sonography in Trauma (eFAST) and portable chest X‑ray were performed as the adjunct to the primary survey to exclude significant intrathoracic injuries. All of the results were reviewed immediately by the trauma surgeon staff on duty before deciding to undergo the REBOA procedure. However, owing to limitations in capabilities (such as hybrid emergency room, emergency computed tomography (CT) scan, or availability of interventional radiology), most decisions regarding the procedure are the responsibility of the trauma surgeon staff on duty (Figure 1).

The procedure can be performed in the emergency department (ED), operating room, or the intensive care unit (ICU). This decision depends on the hemodynamics of the patient and the availability of the operating room. In other words, if the patient’s condition is stable enough and the operating room is available, the patient should undergo the procedure in the operating room. However, if the operating room is unavailable at that time, the procedure can buy some time for resource management.

All REBOA procedures adhered to our standardized institutional protocol:

All of the patients received the treatment in the same protocol and same devices.

The attending trauma surgeon determined REBOA deployment based on clinical presentation, hemodynamic instability despite resuscitation, and suspected NCTH.

All data, such as electronic medical records, trauma registry databases, and operative documentation, were extracted from the internal hospital database. Variables collected encompassed:

Demographic characteristics: Age, sex, weight, height, body mass index; mechanism of injury; prehospital cardiopulmonary resuscitation.

Presentation variables: Vital signs on arrival (systolic blood pressure measured by cuff in the upper extremity; heart rate); Glasgow Coma Scale (GCS) score on arrival (prior sedation); injury severity scores.

Procedural variables: REBOA deployment location (ED, operating room, ICU); vascular access method (open versus percutaneous); balloon Zone (1 or 3); time metrics (door‑to‑procedure, door‑to‑occlusion, prehospital‑balloon‑inflation, emergency department‑to‑operating room transfer, aortic occlusion duration, total operative time).

Physiologic response: Change in systolic blood pressure post‑REBOA; return of spontaneous circulation; cardiopulmonary resuscitation requirement during REBOA. If the patient survived the first operation, the invasive monitoring is performed in the ICU to measure the physiologic response.

Resource utilization: Total blood products transfused (first 24 hours); ICU length of stay; hospital length of stay.

NCTH was defined as suspected or confirmed hemorrhage from thoracic, abdominal, or pelvic sources not amenable to external compression. Non‑responder hemorrhagic shock was defined as systolic blood pressure < 90 mmHg with clinical signs of hypoperfusion after 1‑L of intravenous (IV) crystalloid resuscitation. Acute kidney injury was defined according to Kidney Disease: Improving Global Outcomes (KDIGO) criteria (a serum creatinine increase of >0.3 mg/dL within 48 hours, or serum creatinine increase 1.5–1.9 times baseline within 7 days, or urine output < 0.5 mL/kg/h for 6–12 hours). Lower limb ischemia was defined by clinical manifestations: pain, pallor, pulselessness, paresthesia, and paralysis, which is assessed by the trauma team physicians using clinical assessment and Doppler ultrasonography immediately post‑operative and every 6–8 hours until the hemodynamic is stable.

Continuous variables were expressed as mean ± standard deviation (SD) for normally distributed data or median with interquartile range (IQR) for non‑normally distributed data. The Shapiro–Wilk test assessed normality. Categorical variables were presented as frequencies and percentages.

Univariate comparisons between survivors and non‑survivors used Student’s t‑test or the Mann–Whitney U‑test for continuous variables, and chi‑square or Fisher’s exact test for categorical variables. Crude odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. Statistical significance was defined as P < 0.05 (two‑tailed). Given the small sample size (N = 23) and limited number of outcome events, multivariable regression models were restricted to avoid overfitting. All analyses were conducted using SPSS version 26.0 (IBM Corp., Armonk, NY, USA).

Complete case analysis was performed for all variables. The retrospective cohort (N = 18, 2012–2022) had complete data for all primary and secondary outcomes. Missing data occurred for the following variables: door‑to‑procedure time, door‑to‑aortic‑occlusion time, ED‑to‑operating room time, operative time, and aortic occlusion time. Patients with missing data for specific variables were excluded only from analyses involving those variables. No imputation methods were employed, given the small sample size.

The study protocol was approved by the Khon Kaen University Ethics Committee in Human Research (HE641418). The study was conducted according to the Declaration of Helsinki. For the retrospective component, informed consent was waived by the ethics committee. For the prospective component, written informed consent was obtained from all participants or their legal representatives. Patient data were de‑identified prior to analysis.

During the 12‑year study period, from a total of 659 patients with torso hemorrhage (492 abdominal hemorrhages, 60 pelvic hemorrhages, and 107 thoracic hemorrhages), 28 trauma patients underwent REBOA at our institution. Five patients were excluded: three were < 18 years of age, and two had incomplete medical records (retrospective cohort). The final study cohort comprised 23 patients (Figure 2).

During the study period, 23 patients underwent REBOA for NCTH. Baseline characteristics are summarized in Table 1. Mean age was 38.4 ± 19.2 years; 20 patients (87%) were male. Mean body mass index was 23.1 ± 3.7 kg/m². Ten patients (43.5%) received prehospital cardiopulmonary resuscitation.

On arrival, patients demonstrated non‑responsive hemorrhagic shock after 1‑L of IV crystalloid resuscitation: median systolic blood pressure 60 mmHg (IQR 40–68). Five patients (21.7%) presented with unmeasurable blood pressure; only three patients (13.0%) had systolic blood pressure ≥81 mmHg. Median GCS was 7 (IQR 3–11); 15 patients (65.2%) exhibited severe neurological impairment (GCS 3–8).

The REBOA was decided upon and initiated by a trauma surgeon staff on duty at the emergency room based on the clinical judgement of non‑responsive hemorrhagic shock from NCTH after 1‑L of IV resuscitation without significant neck or chest injury. The procedure was performed in the operating room (13 patients, 56.5%), ED (nine patients, 39.1%), or ICU (one patient, 4.3%). The principle of the procedures in the operating room is to slow down bleeding and increase blood flow to vital areas. The procedures performed in the emergency room serve as a bridge to the operating room. Percutaneous access was achieved in 13 cases (56.5%), and open cutdown was required in 10 cases (43.5%). Zone 1 occlusion was employed in 22 patients (95.7%), and one patient received Zone 3 occlusion. However, the REBOA procedure did not influence the principle of damage control resuscitation.

Seven patients (30.4%) required cardiopulmonary resuscitation during REBOA deployment. Following balloon inflation, 19 patients (82.6%) achieved return of spontaneous circulation.

Only one type of balloon, the Stent graft aortic balloon, was used throughout the study period.

Median door‑to‑procedure time was 33 minutes (IQR 18–54), and median door‑to‑occlusion time was 39 minutes (IQR 22–62). Mean time to balloon inflation was 6.7 ± 3.4 minutes. For patients undergoing ED REBOA, the median operating room transfer time was 40 minutes (IQR 15–73). Median aortic occlusion time was 53 minutes (IQR 30–93), and mean operative time was 215.4 ± 99.8 minutes.

REBOA produced a mean systolic blood pressure increase of 39.8 ± 20.7 mmHg. Patients required massive transfusion: mean 38.7 ± 25.6 units of blood products within the first 24 hours.

Our cohort represented exceptionally critically ill patients. Median systolic blood pressure on arrival was 60 mmHg; 21.7% presented with unmeasurable blood pressure. Nearly half (43.5%) required prehospital cardiopulmonary resuscitation, and two‑thirds demonstrated severe neurological impairment (GCS 3–8). These characteristics indicate that our patients were moribund, representing a population that would traditionally undergo resuscitative thoracotomy with anticipated survival rates of only 2–8.5% [4,5].

Compared with these historical outcomes, REBOA appears to confer superior survival in this critically ill population. The American Association for the Surgery of Trauma (AAST) Aortic Occlusion for Resuscitation in Trauma and Acute Care Surgery (AORTA) registry reported overall REBOA survival of 54.8%; however, their cohort demonstrated a median systolic blood pressure of 80 mmHg with only 23% receiving cardiopulmonary resuscitation [11]. When examining exclusively their most severely injured patients (systolic blood pressure < 70 mmHg), survival approximated 30%, which is more comparable with our results [11].

The most striking finding was the robust association between presenting GCS and survival. Patients with GCS 9–12 had 23‑fold higher odds of 24‑hour survival than those with GCS 3–8, with important implications for patient selection.

GCS likely serves as a surrogate marker for multiple physiologic derangements. Lower GCS may reflect: prolonged hypoperfusion with cerebral injury; more severe hemorrhagic shock; concomitant traumatic brain injury; advanced acidosis and coagulopathy, which can improve after resuscitation. This finding aligns with previous investigations suggesting that REBOA achieves optimal efficacy when deployed before complete cardiovascular collapse [12,13]. The Japanese DIRECT‑IABO (Diagnostic and Interventional Radiology in Emergency, Critical care and Trauma‑Intra‑Aortic Balloon Occlusion registry) study similarly identified consciousness level as a significant REBOA survival predictor [14].

We did not perform formal sensitivity analyses given the small sample size (N = 23). However, the consistency of GCS as a predictor across both 24‑hour and 28‑day mortality on univariate analysis, and the magnitude of effect (crude OR 23.33 for 24‑hour, crude OR 14.00 for 28‑day) strengthens confidence in this finding. The finding remained robust when analyzed as both a continuous variable (OR 1.76 per point) and a categorical variable (9–12 vs 3–8). Nevertheless, GCS alone cannot be the sole deciding factor in initiating or terminating a procedure. We recommend using it as part of a decision‑making process, along with other factors.

Our median door‑to‑occlusion time of 39 minutes exceeds that reported by many Western centers (20–30 minutes) [15,16]. However, this must be contextualized by our deployment locations. Only 39.1% of REBOA procedures were performed in the ED; the majority (56.5%) occurred in the operating room, necessitating patient transport. For patients undergoing emergency department REBOA, the median door‑to‑occlusion time was 37 minutes, which is more comparable to the published literature. Upon reviewing their medical history, we found that some patients reported a transient response to resuscitation for a short period, which might have been the reason for the delay in deciding to proceed with the procedure. But once the decision was made to proceed with the procedure, it only took about 6 minutes to the aortic occlusion time.

Interestingly, shorter prehospital‑occlusion showed a trend toward improved survival, but this did not reach statistical significance (P = 0.438). This might reflect competing priorities in trauma resuscitation. Some patients might benefit from rapid ED REBOA deployment, whereas others might be better served by immediate operative intervention with REBOA performed in the operating room concurrent with definitive hemorrhage control.

REBOA achieved a mean systolic blood pressure increase of 39.8 mmHg, consistent with the physiologic objective: temporarily augmenting afterload and maintaining vital organ perfusion proximal to the balloon. The 82.6% return‑of‑spontaneous‑circulation rate demonstrates REBOA’s effectiveness in reversing cardiovascular collapse.

Nevertheless, hemodynamic improvement must be weighted against the risk of prolonged aortic occlusion. Our median occlusion time of 53 minutes falls within recommended limits; however, 25% of patients experienced occlusion times exceeding 93 minutes. Prolonged occlusion is associated with reperfusion injury, metabolic acidosis, and spontaneous‑circulation dysfunction [17].

Our complication profile was favorable compared with some published series. Most notably, no cases of lower limb ischemia requiring intervention were observed, in contrast to some studies reporting limb ischemia rates of 10–20% [18,19]. Several factors could explain this finding: short occlusion times (median 53 minutes), potentially preventing severe distal ischemia; Zone 1 occlusion predominance (95.7%), possibly providing superior collateral flow versus Zone 3 occlusion; percutaneous access (56.5%), potentially reducing vascular injury compared with open cutdown; and early mortality before limb ischemia manifestation.

Acute kidney injury occurred in 21.7% of patients, exclusively among survivors, representing an acceptable complication rate for a life‑saving intervention and lower than the 30–40% acute kidney injury rate reported in some REBOA series [17]. The true rate might be higher, as many patients died before kidney injury.

Although our study lacked a direct comparison with a resuscitative thoracotomy group, our 21.7% survival rate compares favorably with historical resuscitative thoracotomy survival rates of 2–8.5% in similar patient populations [4,5]. REBOA offers several theoretical advantages: less invasive (percutaneous capability); faster deployment (thoracotomy not required); maintainability during operating room transport; reversibility (controlled reperfusion via gradual balloon deflation) and lower technical skill threshold than resuscitative thoracotomy.

However, the REBOA cannot replace the resuscitative thoracotomy in the event of witnessed cardiac arrest, or the need of intrathoracic intervention. The procedure serves only if it is certain that the patient has abdominopelvic bleeding and no significant neck or chest injury.

The resuscitative thoracotomy retains advantages in specific clinical scenarios: cardiac tamponade requiring pericardial window; cardiac injury requiring direct repair; massive air embolism requiring cardiac venting; and inability to obtain femoral access. The optimal algorithm likely incorporates both techniques, with REBOA serving as the initial approach for most NCTH patients without significant neck or chest injury, and resuscitative thoracotomy reserved for specific thoracic injuries or failed REBOA [20].

Our experience demonstrates the feasibility of REBOA in resource‑limited settings. Thailand is classified as an upper‑middle‑income country; while our center maintains adequate basic equipment and blood supply, we lack certain resources available at Western trauma centers, such as hybrid operating rooms, hybrid emergency room, 24/7 interventional radiology, or a fluoroscopic machine in the emergency room. Despite these limitations, reasonable outcomes were achieved.

Key factors enabling REBOA success in our setting included: institutional commitment (a dedicated trauma team and standardized protocols); training (ongoing education for surgeons and staff); equipment availability (maintenance of REBOA catheter inventory); and a multidisciplinary approach (surgery, anesthesia, and nursing collaboration). Our findings suggest that REBOA can be successfully implemented in similar healthcare environments, without delaying the diagnosis and management, potentially saving lives in regions where resuscitative thoracotomy is not available.

However, this procedure only slows down bleeding in the abdominopelvic cavity but does not achieve hemostasis. Therefore, there is an ongoing hemorrhage. A hospital with an insufficient blood supply or massive transfusion protocol may be another factor contributing to the mortality, which should be further studied.

Based on our findings, we propose the following clinical recommendations:

Patient selection: REBOA should be considered for NCTH patients in hemorrhagic shock; patients with higher GCS (≥9) are most likely to benefit; spontaneous‑circulation cardiopulmonary resuscitation associates with poor outcomes but should not constitute an absolute contraindication and patients in complete cardiovascular collapse might derive limited benefit.

Deployment strategy: ED deployment may be appropriate for patients requiring immediate intervention; operating room deployment is reasonable for patients tolerating brief transport; goal door‑to‑occlusion time < 40 minutes and percutaneous access should be attempted when feasible.

The decision to transfer a patient depends on several factors, such as the patient’s hemodynamic status at the time and the availability of the team and equipment, as surgical procedures are usually performed in an operating room with more advanced facilities. Furthermore, moving patients with so much equipment is not easy and could lead to complications.

Occlusion management: Target occlusion time < 60 minutes; consider partial deflation or intermittent occlusion if prolonged hemorrhage control is required and coordinate with the surgical team for definitive hemorrhage control.

Complication prevention: Monitor for limb ischemia, though risk appears low with appropriate technique; anticipate acute kidney injury in survivors and optimize renal protection and employ gradual balloon deflation to minimize reperfusion injury.

Several important limitations exist: small sample size (23 patients) limiting statistical power and generalizability; single‑center experience with potentially limited external applicability; absence of comparison group precluding direct comparison with resuscitative thoracotomy or alternative strategies; retrospective component with potential for missing data such as the duration of CPR en‑route and selection bias; heterogeneous population mixing injury mechanisms and anatomic locations; lack of long‑term outcomes (functional status, quality of life); variable operator experience as the technique evolved over the 12‑year period.

The small sample size (N = 23) with limited outcome events (six survivors at 24 hours, five at 28 days) severely constrains our statistical analyses. With an events‑per‑variable ratio of 5–6:1, well below the recommended 10:1 threshold, the multivariable models are at high risk of overfitting and should be interpreted as exploratory rather than confirmatory.

Because some patients did not respond to the procedure and resuscitation, ongoing data collection, such as lactate trends and base‑deficit, was incomplete. Therefore, the analysis of complications related to the procedure was lacking.

Of the 17 fatalities, 10 were consistent with clinical signs of severe traumatic brain injury, but only six were confirmed by CT scans or autopsies. Owing to religious and belief reasons, the patients’ relatives did not consent to the examination. Some of this information is therefore missing.

Similarly, the complication data (such as acute kidney injury, lower limb ischemia, or multiorgan failure) of the early deaths cannot be collected.

Several areas warrant further investigation: randomized controlled trials comparing REBOA with resuscitative thoracotomy in appropriate patient populations; partial REBOA evaluating partial occlusion strategies to reduce ischemic complications; multiorgan REBOA assessing prehospital deployment feasibility and safety; predictive models developing algorithms to identify ideal REBOA candidates; long‑term outcomes examining quality of life and functional status in survivors; cost‑effectiveness analyses of REBOA implementation; training methodologies determining optimal education strategies for REBOA proficiency and zone selection comparing Zone 1 versus Zone 3 occlusion effectiveness.