Sources of data
We extensively searched all English studies with “abdominal”, “aorta”, “injury” and similar words in title / abstract. Then, we screened manually for the studies those included sufficient patient samples whose cause of injury was non-penetrating force, to reduce the risk of missing available ones. But the resulting articles including data on the primary outcome measure of interest were still few, explaining the rarity of relevant studies. In addition to the 5 studies finally included, 4 studies contained potentially adequate patients attracted our attention. Of them, two studies [32, 33] did not report the number of BAAI patients, one [34] included 20 patients without reported mortality, and one [15] reported 12 patients with a mortality of 92% containing pre-hospital deaths. We attempted to request the original data available to the authors of these articles, but all failed to be responded. It is unfortunate that the sample size of the study could not be increased in this way.
Five studies [1, 5, 35–37] were excluded due to complete duplication of data. However, three [4, 29, 31] of the included studies may have some degree of duplication of their data, because the centers might have been duplicated across these multicenter studies. We included all of them considering that the data duplication was not serious by comparing their inclusion criteria of cases (year, age, etc.). Even so, we think that this will still increase the error of this study to some extent.
In addition, the 5 included studies were all from USA; And except for the excluded article that exclusively included juveniles and had a sample size of only 16 [28], 2 of the remaining 4 exclusively included adults [4, 31]. These may reduce the fitting effect of the results of our study for populations overall.
Analyses and outcomes
In the process of extracting data, we tried to find out all the variables with data that could be displayed or analyzed. Then we found that in addition to the primary outcome measure, i.e., OHM, the HMO and HMNO of BAAI in the studies could also be calculated [4] or extracted directly, so they were both taken as the secondary outcome measures. However, due to the lack of other available data on BAAI patients in these articles, it was difficult to try other quantitative analyses, such as comparison of mortalities from various treatment modalities and screening of predictors for death.
Because this study was a single-arm meta-analysis involving no control groups, we considered all included studies for quality assessment as simple observational studies without intervention. And the result that no low-quality studies were found is satisfactory.
Since the statistical performance is higher and it is especially applicable to the samples with a very low value of rate [38], we performed the F-T double arcsine transformation of data before meta-analysis. This can reduce the risk that studies described above will be blindly excluded. Unfortunately, we still excluded the study reporting an OHM of 0 due to too high heterogeneity (I2 = 78.8%, P = 0.001 for Q test). It is worth noting that, this is a single center study with a small sample size that only enrolled pediatric individuals < 18 years of age.
Roughly we cannot find out uniqueness of the excluded study on each of the three variables including year of cases, gender, and injury severity score [39]. To explore the reasons for the extremely low mortality in this study, we must first understand which factors or variables may affect the risk of death in BAAI patients. Unfortunately, after an extensive search, it appears that only one study [31] has performed this work. The result of this study suggests indeed that in adult BAAI patients, increased age is a risk factor for death. However, it is not known whether this conclusion applies to pediatric patients. Meanwhile, due to the limited data, we have no way to evaluate the differences between the excluded study and the other 4 ones in other factors affecting mortality [31].
It is worth noting that the study [31] which attempted to explore predictors of death in BAAI, did not include the location and the grade of aortic lesions, two variables that we consider very important, in its analysis. To guide the approach to operation for BAAI, the AA was divided into 3 zones [zone I above the superior mesenteric artery (SMA), zone II from SMA to renal artery, and zone III infrarenal] in a study published in 2012 by Shalhub et al [1], and this method is widely recognized [28–30, 40, 41]. Although this study was excluded from our meta-analysis of OHM because we found that its data were completely duplicated with one [29] of the included studies, it reported mortality rates of BAAI by aortic lesion location, i.e., 60% in zone I, 100% in zone II, and 15% in zone III. In combination with the opinions of Shalhub et al. [1], we consider that the reason for the significantly lower mortality in Zone III is that, compared with those in the other two zones, the aortic injuries in this zone are fewer complicated with the injuries of other organ or aortic branches and they are easier to be exposed by open surgery or to repaired endovascularly. We note that all the locations of aortic lesion of patients in the excluded study [28] were the Zone III, whereas the other 2 available studies reported the proportion of this lesion location to be 66.4% [29] and 68.8% [30], respectively. On the other hand, we need to discuss the grade of BAAI lesion’ severity. At present, there is no consensus in this regard specifically for BAAI. The assessment methods of all previous studies were borrowed from those on BTAI, and whether they really apply to BAAI is unknown. Azizzadeh et al. [42] classified the aortic lesions of BTAI patients into 4 grades, i.e., internal tear, intramural hematoma, pseudoaneurysm, and rupture. And Rabin et al. [43] made a different classification standard, i.e., internal tear or intramural hematoma, small pseudoaneurysm (< 50% of the aortic circulation), large pseudoaneurysm (> 50% of the aortic circulation), and rupture or transection. In the excluded study [28], an aortic rupture rate of 12.5% (2 cases) was reported, whereas one included study [30] with a 31.2% OHM reported 6.2% (1 case). Due to the paucity of studies and data available, we could not draw relationships between the location and severity of aortic lesions, respectively, and the risk of mortality in BAAI patients by quantitative calculations. But we believe that patients who have aortic lesions that are more easily repaired or of lesser severity will have lower mortalities.
In addition, we note that the mechanisms of injury for all patients in the excluded study [28] were all seat belt related motor vehicle accidents. The rate of seat belt related mechanism could not be extracted directly from the 3 included studies [4, 29, 31] that reported relevant data because of nonuniformity in classification standards and missing data, but we think it should be very different from 100%. Whether this difference is one of the reasons for the large difference in OHM reported by them is unknown. The confusion in the classification standard for the mechanisms of injury in BAAI represents insufficient knowledge among investigators in this regard. BAAI and simultaneous lumbar spine fracture had been initially described in 1962 as “seat belt syndrome” [1]. BAAIs with seat belt injury mechanism were not uncommon in case reports [44–47]. Whether there are essential differences between the injury mechanisms of seat belt, those of motor vehicle accidents without seat belt, and even those of non motor vehicle induced trauma, as to cause different death risks from BAAI, still requires considerable works to explore.
In summary, a lower mean age, extremely high proportions of seat belt related injury mechanisms and of injuries in zone III may be part of the cause of that the excluded study [28] reported an unusual mortality as low as 0. But we believe that it is not all, the OHM will rise to some value if the sample size increases.
Of the 4 included studies, two included only adults (≥ 16 years [4], ≥ 18 years [31]) without reasons given. In addition, one study [30] had a sample size of only 16 cases, which appeared small compared with the others. And their other differences between the data available for extraction were not significant. So, it is likely that the above two situations are partly responsible for the heterogeneity between them.
To report a true pooled ES [48], after the heterogeneity among the finally included studies was demonstrated to be low (I2 = 47.6%, P = 0.126 for Q test), we adopted the result derived with the fixed effects model, that is, the OHM of BAAI was 28.8% (95% CI, 26.5%-31.1%). This carefully calculated OHM of nearly 30% of BAAI is very different from that of about 10% of BTAI, confirming our hypothesis that they are two different diseases.
This result was generally consistent with that obtained by the random effects model [30.1% (95% CI, 26.3%-33.9%)], and no any ES was found to be outside the previous 95%CI after omitting studies one by one, which both prove the good stability of the final model. Since the number of included studies was less than 5, it was difficult to judge whether the funnel chart was symmetrical through visual observation, so we did not draw it. The P value obtained by Egger’s test was 0.339, which was far higher than 0.05, which suggests that the model had lower publication bias.
In addition, through the statistical analyses by same method, we also obtained the HMO of 13.5% and the HMNO of 28.4%. Although the included studies were slightly different, the significantly lower hospital mortality (13.5%) still seems to mean that the operation has great benefits for BAAI patients compared with simple observation. But, for a certain BAAI patient, it is still necessary to look hierarchically at which treatment modality is more beneficial. This is back to the previous topic. After knowing that BAAI is indeed a fatal disease and that operation is a good treatment choice, we must understand which patients are at higher risk of life and need prompt operation. That is, we need to find the real predictors of hospital death of BAAI patients, and thus they can be treated more rationally. It is our next research plan in the future.