Splenic artery embolization: technically feasible but not necessarily advantageous

Abdominal trauma with hemodynamic instability is an absolute indication for operative management in most protocols. However, in cases of splenic injury, SAE and NOM is increasingly being performed even in patients that are hemodynamically unstable. The use of NOM or SAE in hemodynamically unstable patients is highly controversial and should be further studied. Delays in diagnostic testing and inappropriate selection of hemodynamically unstable patients for this treatment protocol could increase morbidity and mortality rates, as some authors have suggested [14‐16]. Importantly, there has been little research on managing patients with transient hemodynamic stability, although Hagiwara et al. did report successful SAE in 15 patients with transient responses to fluid administration [17].

Most studies in the literature are retrospective case series that describe the immediate technical success of SAE without comparisons with surgical treatment or a purely conservative approach. Operative intervention by means of splenectomy is thought to increase morbidity and mortality rates, but this is biased by higher injury severity scores in patients that are surgically managed. A large study of 11,793 patients by Zarzaur et al. (2011) could not confirm a correlation between splenectomy and in-hospital mortality when the data were corrected for demographic, physiologic, and injury-related factors [18]. The risk of overwhelming post-splenectomy sepsis has been advocated as a concern and relative contraindication for splenectomy; however, this rare complication occurs in only approximately 0.9 % of splenectomy cases and is preventable by immunization [19]. A recent, large, population-based study classified 9719 patients with splenic injuries into two groups, splenectomy or non-splenectomy, and compared these against a control group of 30,413 patients with other abdominal injuries. This study warned of a two-fold increased risk of developing type II diabetes post-splenectomy [20]. In contrast, some authors have reported higher morbidity after SAE than after surgical intervention [21]. In addition, there is evidence of splenic functional alterations after SAE, although research suggests that no immunizations are necessary after NOM combined with SAE [22]. In a 2010 paper, Shih and colleagues demonstrated a cytokine hyporesponse after SAE [23] therefore, SAE should not be considered definitively safer than splenectomy.

SAE, as opposed to pure conservative management, is even more difficult to advocate in hemodynamically stable patients. In a 2015 study, Olthoff et al. looked at the management of 253 patients with splenic injury in Dutch trauma centers taking into account hemodynamic instability on admission, high-grade injury, and ISS [24]. Although the rate of NOM was comparable between the five trauma centers reviewed, the use of SAE for splenic injuries was highly variable between the centers, illustrating the lack of consensus that exists regarding the management of blunt splenic injury.

In hemodynamically stable patients, retrospective studies have shown that NOM has a reduced failure rate when combined with SAE [12, 25‐30], but prospective studies and certain retrospective studies have failed to confirm these findings [11, 15]. Several studies have been published that show increased numbers of both minor and major complications, leading to an increase in time spent in the hospital, with the use of SAE [15, 31‐33]. Moreover, there is no consensus on the follow-up management and imaging of patients with splenic injuries. Given the lack of evidence in the literature, Olthof et al. published a study in 2013 using the Delphi method to reach an expert opinion on the optimal management and follow-up protocol for blunt splenic injury [28]. In several questionnaires, trauma surgeons and interventional radiologists have tried to reach consensus on the indications for, and optimal follow-up management and imaging to use with, NOM combined with SAE. Almost all of the experts used the American Association for the Surgery of Trauma (AAST) scoring system to grade splenic injuries (Table 3). They all agreed on the use of NOM with or without SAE in small graded injuries (I–II) with no or only a small hemoperitoneum. In high-grade injuries (III–V) with a large hemoperitoneum and active contrast extravasation half of the experts would still attempt NOM + SAE, but most of them would not make a second attempt if the initial SAE failed. Most importantly, they all agreed that rapid intervention is needed for SAE to succeed. Current recommendations are that intervention should be performed within 60 min of admission in stable patients with active contrast extravasation and within 15 to 30 min when a large hemoperitoneum is present. No consensus was reached on management after failure of the initial NOM, or on the appropriate length of stay (LOS) in the hospital.

In search of better criteria for determining the need for intervention, Koca et al. correlated the results of NOM with the American Association for the Surgery of Trauma (AAST) splenic injury grade (see below) [34]. They found that hemodynamically stable patients can safely be treated with NOM, including patients with multi-organ injuries and higher grade splenic injuries. Most injuries are lower grade injuries (I–III), and not surprisingly transfusion is correlated with the injury grade. A 2008 article by Gonzalez et al. added the degree of hemoperitoneum and associated injuries as predictive factors for NOM failure [35]. In 2004, Wahl and colleagues retrospectively analyzed 164 patients with blunt splenic injuries; [36] after univariate analysis they found that lower blood pressure, higher ISS, lower pH, and more packed RBC transfusions are the best indicators of the need for operative intervention. Age, heart rate, splenic abbreviated injury scale score, and GCS did not significantly correlate with the need for operative management. Similarly, Tugnoli et al. (2014) presented the algorithm used in the Bologna-Maggiore Hospital; in 293 patients with splenic injuries admitted between 2009 and 2013 [37] they reported a NOM success rate of 95.8 %. All hemodynamically stable cases with active contrast extravasation or grade V injuries were embolized, preferably proximally within the splenic artery.

In a 2015 AAST prospective observational study, Zarzaur et al. investigated the risk of delayed splenectomy after NOM with or without SAE [38]. They found that the only risk factor for delayed splenectomy was the finding of active extravasation at presentation; thus calling for closer observation of patients with this finding irrespective of whether SAE was performed. Patients with grade I injury had no risk for delayed splenectomy. In cases with higher grade (II to V) injuries, the authors advocated close observation for 10–14 days. In the 2005 study of Haan et al. the presence of arteriovenous fistula on CT imaging was predictive for 40 % of failure rates in nonoperatively managed patients [9]. The presence of vascular lesions (arteriovenous fistulas, pseudoaneurysms) and impact on NOM success rates was further investigated by several authors. In a 2014 study, the Wake Forest University investigated failure rates after angiography with subsequent embolization, ignoring presence or absence of any vascular lesion, for all grade III to V splenic injuries in stable patients in a protocolized manner and comparing this to a historic control group [39]. From 168 patients 67 % were managed nonoperatively with an overall NOM failure rate of 5 %. In the patient group where protocol was violated the failure rate was 25 % compared to 3 % in the protocol group (P = 0.03). Comparing this to a historic control group, 52 % patients underwent NOM with a significant higher failure rate of 15 % (P = 0.04). For each injury grade they reported lower failure rates in the study group compared to the control group and when protocol was followed. However, they did not reach statistical significance when comparing patients treated according to protocol or patients deviating from protocol in injury grades IV and V. The authors did not report on complications or on splenic function after embolization in their study group, no data is available whether proximal or distal embolization was used.

Another prospective study from Parihar et al. [40] found a shorter LOS in SAE patients (5.4 days compared with 6.6 days for NOM patients without SAE), as well as higher hemoglobin concentrations and systolic blood pressures. A recent study, which used propensity score analysis, found no statistically significant difference between observation and embolization in the subsequent splenectomy rate [41].

Many scoring systems have been developed to grade splenic injuries, among which the AAST grading system for splenic injury (Table 3) is the most popular. Two separate studies have correlated CT findings with the AAST injury grade and subsequent need for intervention. In the study by Barquist et al. [42], four radiologists retrospectively reviewed 200 CT images of patients who had undergone splenectomy; the operative grading of the splenic injuries was used as a gold standard. The weighted kappa score for intrarater reproducibility was 0.15–0.77 and the interrater kappa score was 0–0.84 (mean 0.23). In the study by Cohn et al. [43], five radiologists reviewed CT scans of 300 patients with liver and splenic injuries using the AAST grading system. Twenty-one percent of the patients with splenic injuries visible on CT images required intervention. Cohn et al. reached the same conclusion as Barquist et al.; the sensitivity of the AAST injury grade for predicting the need for intervention is poor and interrater variability, even among experienced radiologists, is high. Both thus concluded that the AAST scoring system is unreliable because even experienced radiologists often underestimate the magnitude of the injury, and other factors should also be considered when evaluating indications for surgery or angiography. The authors also evaluated kappa scores for the reporting of contrast blush on CT imaging and found that they were similarly variable, although contrast blush on CT imaging is considered a major indication that intervention is needed.

In 2001, Protetch et al. reported that the incidence of contrast blush was related to the grade of injury, with 3.2 % of grade I/II injuries and up to 37.6 % of grade IV and V injuries showing a contrast blush in CT images [44]. Interventions were more frequent when a contrast blush was present, but multivariate regression analysis showed no correlation between finding a contrast blush and splenic intervention. This was confirmed by both Thompson et al. and Michailidou et al.; both also reported that the size of the contrast blush was correlated with the need for intervention [45, 46]. They both found that similar cut-off values (1 cm and 1.5 cm) indicated the need for intervention. Other factors that also correlated with the need for intervention in splenic injury were injury grade, hypotension on admission to the emergency department, and age. The findings of these studies support the need for better protocols and parameters to allow proper assessment of the need for intervention in cases of splenic trauma.

In 2007, Marmery et al. suggested a new grading system for splenic injury that takes into account active bleeding or splenic vascular injury (including splenic pseudoaneurysms and arteriovenous fistulas) [47]. Their new grading system was statistically significantly better than the AAST grading system in terms of its ability to predict the need for splenic arteriography (P = 0.0036) or the combination of arteriography and surgery (P = 0.0006). Table 4 presents the proposed new grading system.

There has been much research published, and work is still ongoing, to understand the complications after SAE compared with those after surgery and NOM. All of the treatment modalities for splenic injury are associated with high morbidity rates. Specific morbidities associated with SAE include pancreatitis, splenic infarction, post-embolization syndrome, and intestinal perforation.

Two retrospective studies examined the complications that occur after SAE [32, 33]. Ekeh et al. reviewed 1383 patients with blunt splenic injury over an 11-year period; 78.5 % were treated nonoperatively, and of this group 8.1 % underwent SAE. Major complications including splenic infarction, splenic abscess, contrast-induced renal insufficiency, and splenic cysts occurred in 14 % of the patients who underwent SAE. Wu et al. reported that 28.5 % of patients who underwent SAE experienced major complications. Distal embolization was associated with more major complications than proximal SAE. Minor complications (pleural effusion, coil migration, and fever) occurred in 34 % and 61.9 % of the patients in the SAE groups in the studies by Ekeh et al. and Wu et al., respectively.

In 2004, Haan et al. reported SAE complication rates in 140 patients and reported a splenic salvage rate of 87 % [31]. This rate indirectly correlated with injury severity scores. More than 80 % of the grade IV and V injuries in their study were successfully managed nonoperatively. No significant difference was noted in patients older than 55 years of age. The presence of a significant hemoperitoneum did not alter success rates; however, the presence of arteriovenous fistulas did. They concluded that complications after SAE are common, but do not seem to influence outcomes.

In 2008, Wei et al. published a retrospective study showing lower complication rates after SAE even with higher splenic Abbreviated Injury Scores in this group compared with the operatively managed group (6 % vs. 36 %, P < 0.01) [30]. They found that the introduction of SAE reduced operative interventions by 16 % between 2000 and 2006 in their Level 1 center. A French prospective multicenter study investigated complication rates in 91 patients [21]. Twenty percent underwent splenectomy with a post-surgical morbidity of 15 %. Fifteen patients (16 %) underwent embolization and 67 were initially managed nonoperatively. In the embolization group with severe injury (grade III–V), the morbidity was 73 %. The splenectomy group with severe injury had a total morbidity of 70 %. The NOM group had a morbidity of 58 %. Specific morbidities after NOM, surgery, and SAE were 10, 15, and 47 % (P = 0.02), respectively. The study concluded that embolization should not be used as a prophylactic measure, but should only be used in cases with active bleeding. We could not find well-designed studies comparing complication rates between the three treatment modalities.

In their 2015 study of 253 patients, Olthoff et al. found a very large variation but no significant difference in the need for transfusion between patients treated with SAE and those treated with NOM [24]. The mean number of transfused blood products was 5.5 units (±9.9) in the NOM group versus 9.1 units (±17.2) in embolized patients (P = 0.75). Rosati et al. reported on the need for transfusion in patients requiring immediate splenectomy (70 %), embolization (46.5 %), and NOM (25.9 %), but failed to adjust these data for injury severity and confounders [12]. They found that the need for transfusion was a major risk factor for mortality (adjusted OR = 2.63; CI 1.27 – 5.42, P = 0.009). Dent et al. compared transfusion rates between NOM, early embolization, and operative management [25]. They reported a 2.1-unit difference in the transfusion rate (P < 0.01) of the NOM group versus the operative management group, and found no statistically significant difference between transfusion rates in the early embolization and operative management groups without correcting for confounders. They also found no difference in transfusion rates when they compared patients who underwent early embolization versus late embolization.

A retrospective review in 2011 investigated the impact of direct supervision (DS) or indirect supervision (IS) on the management of splenic injuries [48]. Using data from 506 cases the authors found significant differences in compliance with protocols (DS, 95 % vs. IS, 82 %, P < 0.001), operation rates (DS 16 %, vs. IS, 8 %, P = 0.016), ICU use (IS, 84.1 % vs. DS, 73.0 %, P = 0.029), hospital costs (IS, $142.956 ± $36.219 vs. DS, $62.981 ± $8.784, P = 0.048), and use of SAE without indication (IS, 8 vs. DS, 0). Interestingly, there were no reported differences in mortality or splenectomy rates.

In 2014, Olthof et al. retrospectively examined the time to intervention in patients undergoing SAE and patients undergoing surgery in a cohort of 96 adults [49]. In hemodynamically stable patients, the median time to intervention for patients in the SAE group was 117 min compared with 105 min for patients in the surgery group. The differences in time to intervention or re-intervention, and the complication rates were not significantly different between the two patient groups. There was a higher transfusion rate for patients in the splenic surgery group; the median number of transfused units of packed RBCs was eight for hemodynamically unstable patients undergoing SAE versus 24 for patients undergoing splenic surgery (P = 0.09). A report published in 2015 reviewed 10,405 records in the National Trauma Data Bank to find a correlation between high and low angio centers and to investigate any relationship between angiography use and its timing or splenectomy after angiography [50]. They concluded that early angiography is associated with higher splenic salvage rates but found no statistically significant difference in splenectomy rates between the high and low angiography volume centers.

We could not find any articles comparing the relationship between time to intervention and success or complication rates between patients undergoing NOM, SAE, or splenic surgery.

Different authors have investigated the costs of SAE versus surgical management. Wahl et al. reported no significant cost differences between these two treatment modalities (SAE, $49.300 ± $40.460 vs. OM, $54.590 ± $34.760) [36]. This was confirmed by Wei et al., who found the costs associated with SAE to be $47,000 ± $31,000 and the costs associated with OM to be $40,000 ± $34,000 [30]. Both authors assessed the LOS and differences in LOS in patients who received different treatments. Wahl et al. found an average LOS of 45 ± 26 days for patients who went directly to the operating theatre, compared with 14 ± 15 days for patients who underwent SAE (P < 0.02). Wei et al. found that the average LOS in the surgical group was 14 ± 10 days and in the SAE group was 12 ± 12 days (P > 0.05). Bruce et al. also came to the same conclusion in their 2011 study; [51] in that study the median total hospital charges were $41,269 (± $31,128) for the NOM + SAE group compared with $46,356 (± $11,334) for the OM group (P = 0.545). They did report higher procedure-related charges for OM patients compared with patients undergoing NOM + SAE ($28,709 ± $6941 vs. $19,062 ± $14,025; P = 0.16); however, this was offset by more charges for a greater number of radiological evaluations in the other group.