Study designs

We conducted a retrospective cohort study using registered data from the Japan Trauma Data Bank (JTDB) to compared characteristics and outcomes between REBOA and ACC.

Data collection

Data were obtained from the JTDB, a nationwide trauma registry established in 2003 and authorized and maintained by the Japanese Association for the Surgery of Trauma and the Japanese Association for Acute Medicine to improve and assure the quality of trauma care in Japan [7]. During the study period, a total of 234 hospitals including 95% of tertiary emergency medical centers in Japan participated in the JTDB [7]. The JTDB collected variables about patients and hospitals such as patient demographics, comorbidities, injury type, mechanism, vital signs, Abbreviated Injury Scale (AIS) score, Injury Severity Score (ISS), pre-hospital treatment, in-hospital procedures, and in-hospital and emergency department (ED) mortality [7].

REBOA has recently found use as a general technique across major emergency centers in Japan. Due to the limitations of the JTDB as being a general, total trauma registry and not a REBOA-specific database, we were unable to glean information on imaging, access, and balloon zone placement specifics. Although details of REBOA in those three areas depend mainly on local facilities and expertise, in our specific cases REBOA access is typically accomplished through a common femoral artery and the balloon insertion follows a blind approach.

Patient selection

The study inclusion criteria were the presence of critical trauma and reception of either REBOA or ACC. We excluded patients who had received both REBOA and ACC. Also, we excluded subjects younger than 14 years old or those with age data missing. Patients with cardiopulmonary arrest on arrival at the ED (systolic blood pressure of 0 mm Hg or data missing on arrival) or with an AIS score of 6 (i.e., non-survivable injury) for any region were also excluded. Figure 1 shows participant selection data from this study.

Study endpoints

The primary intervention was either REBOA or ACC. Intervention strategies were solely dependent on the individual decisions of ED physicians. The primary outcome of this study was in-hospital mortality and the secondary outcome was mortality in the ED.

Statistical analysis

To report the characteristics of participants, firstly we used the Shapiro-Wilk test to identify normality of each variable. However, there was some skew in almost all the variables. We then decided whether to use the mean ± standard deviation (SD) or median (interquartile range (IQR)) for plot figures of each variable, depending on the previous reporting style. To assess the independent effects of REBOA compared with those of ACC, outcomes were evaluated by analytical models—standard logistic regression and a logistic regression incorporating the results of propensity score matching. Since the number of survivors was very small in this study population, we chose a few covariates for the standard logistic regression analyses. Within the standard logistic regression analyses, we conducted three adjustment models: a revised trauma score (RTS)-adjusted model, an ISS-adjusted model, and a trauma and injury severity score (TRISS)-adjusted model. Together, these three standard logistic regression models, coupled with a propensity score-matched model, provide a robust method for statistically reliable analysis.

Propensity score

Because the use of REBOA or ACC was not randomly assigned, a formal causal inference is not possible. Therefore, a logistic regression analysis was used to estimate propensity scores (PSs) to predict usage of REBOA or ACC from the available predictors. These variables were age, gender, mechanism of injury, cause of injury, transport type, pre-hospital treatment, vital signs at ED, and ISS, which would reflect patient wound severity. Treatments after ED arrival (e.g., blood transfusion, cardiopulmonary resuscitation (CPR), operations) were not included in the PS derivation process because they were performed after usage of REBOA or ACC. Severity scoring systems without ISS (e.g., RTS, TRISS) were not included in the PS because in them were many of the same components that had already been included in the PS, such as vital signs. Propensity score matching extracted 1:1 matched pairs of subjects who received REBOA or ACC based on an averaged PS. The absolute standardized difference of variables for the PS estimation was used to assess the match balance. An absolute standardized difference of less than 0.2 was generally considered as an acceptable match balance between the groups.

The two-sided significance level for all tests was P < 0.05. All analyses were performed using SPSS software, version 21.0 (IBM, Armonk, NY, USA).

Brief summary

This study investigated the current usage of REBOA and ACC using a large, nationwide trauma database in Japan. Mortality rates in patients requiring AO was discovered to be very high but this is attributed to the usage of ACC on patients who cannot be saved, skewing mortality out of favor with ACC. We also analyzed outcomes for patients after receiving either REBOA or ACC after adjusting for patient trauma severity. Robust analyses of the adjusted data showed that REBOA was associated with significantly reduced in-hospital mortality compared with ACC. However, due to differences in associated procedures between REBOA (e.g., increased need for angiography) and ACC (e.g., thoracotomy), there should be some consideration given to choosing either intervention.

Comparison with previous studies

To our knowledge, our current study is one of the largest cohort studies describing the use of REBOA [8]. REBOA has recently found use as a general technique across major emergency centers in Japan. The highest density of potential REBOA patients is also seen at major trauma centers in England and Wales, although the number of patients in whom REBOA was utilized is small [9]. In fact, a review of the potential use of REBOA in exsanguinating hemorrhage cases in the US suggested that this new technique should be thoroughly evaluated for broad use, but the literature currently suffers from a dearth of human studies on REBOA [10]. Although our observational study admittedly had some selection bias, we feel that our results will nonetheless become an important part of the foundation of literature supporting the evaluation of global REBOA use.

Previous studies regarding REBOA usage have been limited in size and scope, and have shown ambiguous results. For example, although previous single-center cohort studies mentioned the utility of REBOA for massive pelvic bleeding cases that could still be imaged by angiography [5, 6], another large, retrospective cohort study cautioned against REBOA usage for patients who had emergency surgery or transcatheter embolization [7, 11]. Yet another single-center cohort study also reported on the feasibility and safety of REBOA for a non-compressive torso injury (pelvic fracture or hemoperitoneum) [12], but contrasting studies also reported that REBOA usage was associated with a higher mortality compared with non-REBOA usage in JTDB [7, 11]. However, to objectively evaluate these reports, a thorough knowledge of the Japanese trauma care system is required. For example, most Japanese emergency departments see few in-house trauma surgeries, see fewer trauma cases overall, and mostly deal with older patients and age-related maladies [13]. REBOA usage, in this context, may signal “last ditch” efforts [11]. However, our results from the same database show an incongruent outcome even though our study population and comparisons are different. Previous reports have indicated REBOA usage as a last resort in the most severe trauma cases, but only one multicenter, prospective observational study (Aortic Occlusion for Resuscitation in Trauma and Acute Care Surgery (AORTA) registry) [1] has looked at direct comparisons with ACC, which is also used in the most severe cases, and found REBOA to be beneficial. As the severity criteria for both REBOA and ACC are similar, it is reasonable to compare those outcomes directly and our reports findings strengthen the conclusion of DuBose and colleagues [1], and show a promising consistency in results.

Possible explanations and implications

AO was used on 2032 patients in our database. However, 799 patients with ACC were excluded from analysis because of pre-hospital cardiopulmonary arrest. Although we controlled for this in our study, the differences between REBOA and ACC in general need more context for accurate interpretation. In Japan, ACC currently seems to be a preferred intervention tactic in non-survivable injury cases and this differs from other countries, possibly making Japanese ACC-related mortality rates non-indicative of actual outcomes [1, 9, 13]. In addition, other patient characteristics such as better Glasgow Coma Scale (GCS) scores in cases where REBOA was used versus ACC cases (with more severe GCS scores) may also skew results. This raises the issue of snap decisions by ED physicians to choose rapid thoracotomy over REBOA because ACC would be more frequently chosen in cases with worse GCS scores. A key point to keep in mind, however is that although the probability of survival (TRISS) for REBOA was better than that of ACC, it is still no guarantee of success in severe cases (0.43 ± 0.36). Furthermore, REBOA patients who die might count as preventable, but ACC deaths with TRISS scores indicating unsurvivable injuries (0.27 ± 0.30) would be counted as non-preventable. This might not be seen as a negative even though the mortality of ACC patients was very high (90%) and might be related to more severe complications in the thorax. These issues highlight the nuances necessary to objectively interpret the data, as both REBOA and ACC have complicating factors. Survival rates at ED of 78% and 51% in REBOA and ACC, respectively, indicate that both can serve a role in trauma treatment. However, it is important to keep in mind that these procedures are not panaceas; only 14% (83/607) of REBOA patients and 2.0% (12/233) of ACC patients could leave the hospital and go home. This emphasizes the difficulty of AO in clinical practice. This is especially evident in Japan as our mortality was slightly higher than other countries [1, 9, 13]. As Japan’s prevalence of penetrating trauma is quite low (6.4%), survival probabilities may follow suit. However, taking into consideration the variability in study populations and institutional skill, a general trend in the same direction can be seen with our results versus those of other countries.

Trauma severity in ACC cases versus REBOA cases were controlled for with sensitivity analyses as seen in Fig. 2, but REBOA usage showed a clear survival benefit. PS matching was also used to control for insufficiency of adjustment and the tendency of results was the same among PS-matched patients. We found that PS matching was one of the best methods to control confounders in this prevalence and mortality. Again, direct comparisons between REBOA and ACC were conducted after PS matching because of current interest in the possibility of shifting the ACC paradigm to REBOA [10]. Table 3 shows the precision of our PS matching methodology. However, covariate differences where we did not use PS (although we did properly control for physiological severity and backgrounds) unavoidably resulted in an inability to match anatomical severity. This might be classified as an indication bias, but we feel that our analysis shows the real utility of AO.

Taken together, we feel that these results should be made part of the body of knowledge that physicians consult in the decision tree of AO. Accordingly, REBOA would conceivably be used more often as a solo abdominal trauma option even though there are no formal criteria for AO treatment utility. Still, this does not necessarily mean that all AO cases would shift from ACC to REBOA. In reality, choosing ACC for severe abdominal trauma patients who present no chest trauma is a difficult choice for ED physicians who may not have enough thoracotomy experience. This is especially important because of recent reports on poor outcomes of emergency thoracotomies after abdominal exsanguination, adding to the reputation of ACC as a “last ditch” effort [13, 14]. However, to find the best position of an occlusion balloon with a blind approach is next to impossible when patients present with thoracic complications. This explains the higher incidence of abdominal and pelvic angiography in REBOA groups which we find to be an acceptable >tradeoff for accuracy in occlusion balloon positioning. We do not doubt that REBOA will be applicable as a bridge to definitive treatment in the ED, but indications and contraindications in the light of ACC must be further refined. Finally, the most critical point to remember is that any method inducing long-lasting ischemia to at least half or more of the body has serious potential to harm the patient. To this end, the decision to use REBOA or ACC should be part of a robust clinical governance framework in order to ensure high quality patient care and maximal survival chance [9].

Limitations

Potential limitations of this study should be acknowledged. First, there remained some indication bias as previously discussed, indicating caution when interpreting results for clinical standpoints. However, we controlled for patient background using logistic regression and PS-matched analysis, when possible, and found two key points in this study. First, PS-matched analysis was one of the best methods for comparison because there was a relatively small sample size of survivors. Second, there was institutional bias although covariates were carefully selected on the basis of the assumption that none were affected directly by the intervention. This assumption could be a potential weakness and requires further study. With regard to mortality rates, a population-based study in England and Wales showed only major trauma centers had a high density of REBOA use and their rate was smaller than ours [9]. We, on the other hand, did not have institutional-level data, and therefore we could not control for it and this might account for our higher mortality rate. Although a potential weakness could be variability between physicians and institutions, AORTA registry data reveals that the general tendency of outcomes is the same [1]. Although selection bias may skew towards REBOA more than ACC in both AORTA and this study, we feel that our results are worth consideration to add to the scarce body of knowledge regarding this topic. Moreover, we did not have detailed data on REBOA or ACC such as the clamping time, the ballooning time, and the tactics of that therapy. Since the patients had the issue of ischemia/reperfusion injury, their outcome may have been influenced by time. A general assumption, however, is that clamping and ballooning times were kept as short as possible by the physicians because of the common knowledge that occlusion times should be kept to a minimum.