Our study showed that oronasal packing combined with TAE successfully achieved hemostasis in 47 patients (96%), and 36 patients (73%) survived by adhering to the treatment protocol at our institution. Only two patients (4%) died of persistent bleeding from inaccessible lesions by angiography in the present study. Protocol-guided management of maxillofacial trauma-related oronasal bleeding was effective in achieving hemostasis in most patients. To the best of our knowledge, this is the largest single-center report on the management of maxillofacial trauma with oronasal bleeding (Table 4).

Hypoxic encephalopathy was the main cause of death in patients with massive oronasal bleeding. The current study showed that 92% of deaths were caused by hypoxic encephalopathy, and 85% of deaths occurred because of cardiac arrest related to a compromised airway. Securing the airway is the top priority for saving maxillofacial trauma patients with active oronasal bleeding, especially those who are unconscious [6, 7, 13, 14]. Additionally, timely management may reduce the consequences of hypoxic injury, such as death, prolonged coma, permanent cognitive deficits, seizures, and other neurological abnormalities. In the face of active oronasal bleeding, trauma surgeons should not hesitate to establish a surgical airway when the first attempt of orotracheal intubation fails. Moreover, prehospital intubation with an intent to prevent blood from obscuring the airway may decrease the incidence of hypoxia and aspiration, which may be associated with better neurological outcomes and reduced in-hospital mortality [15, 16].

Anterior and posterior nasal packing, balloon tamponade, and conservative measures have been proposed as the first attempts to manage oronasal hemorrhage [6,7,8,9,10,11]. Anterior and posterior nasal packing were used to apply pressure on the Kiesselbach's plexus and the Woodruff plexus, respectively. Cogbill et al. reported that anterior, posterior, or both packing controlled the bleeding in only 29% of patients and slowed the rate in 44% of patients [6]. This finding is comparable with that of the current study, in which hemostasis was achieved in 15 patients (30%) by oronasal packing alone. However, disruption of facial buttresses and extensive collateral circulation between the external and internal carotid artery systems may render packing and tamponade less effective in hemostasis [17]. External carotid artery (ECA) ligation has been recommended in the literature if oronasal bleeding continues despite packing. ECA ligation is rarely effective in controlling such hemorrhage because of the abundant collateral circulation of the face [18]. TAE can precisely identify and stop bleeding with few complications. TAE has emerged as the primary choice for hemostasis in patients with persistent bleeding after packing [3, 6, 10, 12, 13]. In this study, 34 patients (69%) proceeded to angiography for persistent bleeding after packing; among them, 22 patients (65%) were salvaged by TAE.

The advantages of TAE over ligation include shorter procedure time, more precise hemorrhage localization, and the ability to embolize a possible concomitant abdominal or other bleeding during the same session [10, 19]. Based on a review of 205 cases, the efficacy of TAE was 79.4%–100%, while the rate of major complications was approximately 2%–4% [20]. In our study, 34 patients underwent TAE, and definitive control of hemorrhage was achieved in 32 patients (94%), which is consistent with the finding reported in the literature. The internal maxillary artery and its branches are most frequently associated with traumatic maxillofacial hemorrhage [3, 6, 8, 10, 12].

The proper choice of embolic agents is crucial for interventional radiologists to avoid serious neurological complications, such as cerebrovascular accidents, cranial nerve palsy, and brain infarction. Microcoils and gelfoam pledgets are the embolic agents most frequently used for TAE in maxillofacial trauma [20]. Embolic agents such as gelfoam may pass through the extracranial–intracranial anastomoses, consequently causing embolic complications [21]. No major procedure-related complications were observed in our study.

Embolization can be used to prevent bleeding. However, the approach to craniofacial bleeding has some limitations. Bleeding from the ICA and its branches should be treated carefully with TAE in the case of passage of embolic agents into the brain, leading to neurological injury. Four bleeding sources in the ICA or its branches were noted in our study: ruptured ACA pseudoaneurysm, unruptured ICA pseudoaneurysm, bilateral ICA laceration, and ethmoidal artery.

Coils were used in the ruptured ACA pseudoaneurysm, but they only partially reduced the bleeding because of the difficulty in accessing the lesion. No embolization was performed in the unruptured ICA pseudoaneurysm because of the absence of contrast extravasation. Active bleeding from the bilateral ICA laceration was identified; therefore, coils were used, and bleeding was successfully arrested. A low flow rate of contrast extravasation was identified in the ethmoidal artery; therefore, embolization was not performed.

It is worth noting that arterial or venous bleeding from oral and nasal mucosa might continue after packing and arterial embolization. Therefore, our protocol suggests that packing should be performed for 48–72 h until hemostasis is confirmed. A second angiography can be performed if delayed or persistent bleeding occurs. Seven patients received a PRBC transfusion of more than 6 units 24 h after angiography. Only one of those patients underwent a second angiography after receiving 20 units of PRBC transfusion. No contrast extravasation in the maxillofacial area was identified in the second angiography. Owing to persistent bleeding from the oronasal mucosa found on examination, prolonged packing and suture ligation were performed, and the patient survived after successful hemostasis. Four patients achieved hemostasis with prolonged packing and survived. One patient died of persistent bleeding from the skull base AV fistula, which was not embolized owing to its inaccessibility. One patient achieved hemostasis by packing but died of hypoxic encephalopathy.

Matsumoto et al. selected cases of Le Fort III fractures with blood loss greater than 20% from Japan trauma data bank and reported that the in-hospital mortality of the TAE group was significantly lower than that of the non-TAE group (23.1% vs. 44.6%; P = 0.048) [13]. Our study found a similar result that TAE can effectively arrest bleeding. However, the angiography group showed a significantly higher mortality rate (35%). Our study showed that the angiography group had significantly lower GCS and higher ISS scores. Therefore, the higher mortality rate was more likely to result from the patient-associated injury than TAE.

Intracranial injury or hemorrhage is the most common type of concomitant injury in patients with panfacial fractures [1, 2]. In this study, 49% of the included patients had a head AIS score of > 3. Bromberg et al. [22] reported that blunt trauma patients with arterial hemorrhage from neck/nose/mouth and midface fractures (Le Fort II or III) are at an increased risk of blunt cerebrovascular injury (BCVI), which might lead to permanent severe neurologic deficits in survivors. This study identified five cerebrovascular injuries after trauma, including carotid-cavernous fistula (two patients), ACA pseudoaneurysm (one patient), ICA pseudoaneurysm (one patient), and skull base AV fistula (one patient) in the angiography group. The incidence of BCVI was 15% in the angiography group. Two patients with carotid-cavernous fistulas survived with conservative management and did not develop severe neurologic deficits. Two patients with a ruptured ACA pseudoaneurysm and skull base AV fistula died due to persistent bleeding. The endovascular intervention failed to access the lesion and arrest the bleeding. Although no active bleeding was identified in the ICA pseudoaneurysm, the patient died of hypoxic encephalopathy.

This study has several limitations. This study was a retrospective analysis of the data. First, the number of patients identified in this single-center study was small. The timing of TAE depended on clinical judgment, which may have influenced the outcome. Moreover, the selective or nonselective embolization techniques and the choice of embolic agents depended on the clinical assessment of the interventional radiologists. This variation could also affect the outcomes. During the study period, some changes were made in the treatment of bleeding trauma patients. In 2013, 1 g tranexamic acid was adopted in the emergency department and the massive transfusion protocol was established in 2015 as standard care for bleeding trauma patients. Both may affect clinical outcomes. However, because of the small number of patients, no survival benefit was observed with these two changes. Multi-institutional prospective studies are needed to further this research.