Introduction

Bleeding is a severe complication in pancreatic surgery. Mortality after re-laparotomy for bleeding following pancreatectomy is up to 11–38% in the last 5 years [1–6].

The International Study Group of Pancreatic Surgery (ISGPS) developed the classification of bleeding after pancreatic surgery in 2007. This grading system determines treatment strategy. Depending on the period after surgery, bleedings are divided into early (within 24 hours) and delayed (after 24 hours) [7]. Early postoperative bleeding is a consequence of inadequate intraoperative hemostasis or homeostatic disorders. Delayed bleeding is almost always arrosive and follows postoperative pancreatitis complicated by pancreatic or biliary fistula [8].

Bleedings are divided into 3 grades depending on blood loss (A, B and C) [9]. Hemorrhage may also be mild and severe. Bleeding grade A is early and mild, decreases hemoglobin level by < 30 g/l, does not significantly affect patient condition and requires conservative treatment. Bleeding grade B may be early or late and mild or severe. Severe bleeding is accompanied by hemoglobin level decrease by more than 30 g/l. This event is characterized by significant clinical deterioration, need for transfusions, re-laparotomy or transcatheter arterial embolization (TAE). Bleeding grade C is late and accompanied by hemoglobin level decrease by more than 30 g/l. It significantly worsens clinical status and should be considered as potentially life-threatening event requiring immediate interventions [8].

The purpose of this study was to evaluate technical aspects and clinical results of TAE for delayed postoperative arterial bleeding after pancreatic surgery.

Material and methods

Study design

A retrospective observational single-center study was carried out at the Vishnevsky National Medical Research Center of Surgery between January 2012 and September 2020. There were 821 pancreatectomies for pancreatic cancer (ICD-10 C25.0) and complicated forms of chronic pancreatitis (K86.1). Postoperative bleeding occurred in 106 (12.9%) patients.

Inclusion criteria: late (arrosive) bleeding ISGPS grade B or C (2007).

Exclusion criteria: early bleeding after pancreatic surgery (ISGPS (2007) grade A or B); stress-related gastrointestinal ulcerative bleeding. In addition, we excluded patients with chronic pancreatitis and previous preventive endovascular interventions for false aneurysms of celiac trunk pool.

Finally, the study included 74 patients. There were 48 (64.9%) men and 26 (35.1%) women. Median age of men was 57 years (IQR 50 — 62 years), women — 50.5 years (IQR 42 — 64 years). Pancreatic cancer was diagnosed in 53 (71.6%) patients, complications of chronic pancreatitis with indications for surgical treatment — in 21 (28.4%) cases. Postoperative bleeding followed pancreatic fistula or fluid accumulation around the vessel. Late bleeding grade B was observed in 7 (9.5%) patients, grade C — in 67 (90.5%) cases. The majority of interventions (75.7%) were performed on the pancreatic head; pancreatic body resection was required in 17.6% of patients, pancreatic tail resection — in 6.8% of patients. We analyzed the incidence of bleedings from various arteries of celiac trunk and superior mesenteric artery pools.

Embolization agents for TAE, effectiveness of this procedure and specific complications were critically assessed.

Planning primary digital subtraction angiography (DSA), we applied transfemoral approach in 56 (75.7%) cases, transbrachial approach — in 18 (24.3%) cases. Data of CT angiography before open and endovascular surgery underlied the choice of certain vascular approach. Mesenteric artery discharge angle was very important. For example, we preferred transfemoral approach in patients vascular angle > 45° and transbrachial approach for more acute angle (<45°) (Fig. 1).

Fig. 1. CT angiography.

a — discharge of mesenteric arteries at an angle of more than 45° (indication for femoral approach); b — discharge of mesenteric arteries at an angle of less than 45° (indication for brachial approach).

Endovascular hemostasis was performed under hemodynamic support with saline, colloidal solutions and, in particular, blood derivatives (red blood cells, fresh frozen plasma and platelets). Introducers 5F or 6F were used for DSA regardless arterial access. We performed angiography using Simmons 1/2 5F, Cobra 1/2 5F, Vertebral 5F and 5F multipurpose catheter. Coaxial systems of microcatheter 1.8 — 2.8F were used for superselective angiography. The following embolization agents were used for TAE: uncalibrated PVA (polyvinyl alcohol) emboli, hydrogel cylindrical emboli, N-butyl cyanoacrylate (NBCA, Histoacryl) diluted with Lipiodol, metal coils and occluders. Stent-grafts were also used for hemostasis. In embolization with coils exiting into great arteries, we additionally used stents for renal or coronary arteries to prevent thrombosis of the artery.

TAE was followed by control angiography for assessment of hemostasis. Primary endpoint was technical success of TAE defined as hemostasis or no extravasation of contrast agent in the target vessel after TAE. Secondary endpoints included major complications of visceral artery embolization (non-targeted organ embolization followed by ischemia, liver and renal failure), recurrent bleeding followed by repeated angiography and TAE.

Statistical analysis

Statistical analysis was performed using IBM SPSS Statistics software version 26. Chi-square test, Fisher's exact test and Mann-Whitney test were used to determine the relationship between bleeding, clinical and laboratory characteristics of patients. Differences were significant at p-value <0.05.

Results

Angiography followed by endovascular hemostasis in celiac and mesenteric arterial pools was performed in 74 patients.

Extravasation of contrast agent from the damaged vessel into abdominal cavity and/or bowel was the most common sign of bleeding. Contrast agent leakage outside the vessel in the form of a "cloud" or "cone" (in case of severe bleeding) is observed during CT and digital subtraction angiography (Fig. 2). These symptoms were found in 50 (67.6%) patients.

Fig. 2. Extravasation of contrast agent.

a — CT angiography of abdominal aorta (MPR), the arrows indicate extravasation of contrast agent into abdominal cavity (source of bleeding — superior mesenteric artery); b — selective DSA of superior mesenteric artery, the arrow indicates extravasation of contrast agent into abdominal cavity and postoperative outcome (stent-graft).

False aneurysm is a cavity communicating with a damaged artery and surrounded by adjacent tissues and/or organs. This aneurysm can be asymptomatic for a long time. CT or digital subtraction angiography visualize aneurysm like a round or oval formation with contrast enhancement identical to the artery (Fig. 3). Severe bleeding from a false aneurysm follows damage to the aneurysm wall. We encountered such a situation in 13 (17.5%) patients.

Fig. 3. False aneurysm of common hepatic artery.

a — CT angiography of abdominal aorta, the arrow indicates false aneurysm of common hepatic artery; b — selective DSA of common hepatic artery, the arrows indicate false aneurysm and post-TAE result (coils + coronary stent).

Interpretation of indirect signs of bleeding is the most difficult. This situation occurred in 11 (14.9%) patients without ongoing bleeding during examination. In these cases, diagnosis was based on CT data and angiographic signs of bleeding. Indirect signs of previous bleeding were uneven vascular wall ("cankered vessel"), sharp break of contrast enhancement ("vascular stump"), hypervascularity (Fig. 4—6).

Fig. 4. MSCT angiography of the abdominal aorta, arrows indicate a false aneurysm of the common hepatic artery

a — CT angiography of the abdominal aorta, the arrows indicate intra-abdominal hematoma at the level of gastroduodenal artery; b — selective DSA of common hepatic artery with imaging of "vascular hemp" and post-TAE result (embolization with a spiral, its distal part is located in abdominal cavity).

Fig. 5. Selective digital subtraction angiograms of the common hepatic artery.

a — CT angiography, the arrow indicates intra-abdominal hematoma; b — selective DSA of jejunal artery with visualization of a «corroded vessel».

Fig. 6. MSCT angiograms of the abdominal aorta: arrows indicate hematoma of the abdominal cavity at the level of hastroduodenal artery.

a — selective DSA of posterior arterial arcade of pancreatic head with imaging of hypervascularity; b — after TAE (spiral).

The main signs of vascular lesions and their incidence are summarized in Table 1.

Table 1. Angiographic signs of bleeding and their incidence (n=74)

Digital subtraction angiography identified 84 sources of bleeding in 74 patients. In most cases, bleeding arose from gastroduodenal artery, superior mesenteric artery and liver arteries (Table 2).

Table 2. Sources of bleeding identified during DSA in 74 patients

Empirical hemostasis was performed in patients with obvious clinical signs of bleeding, CT angiography data and indirect angiographic signs of vascular destruction. High risk of recurrence is a feature of empirical hemostasis (repeated bleeding occurred in 4 (36.7%) out of 11 patients with indirect angiographic signs of bleeding).

In 36 (48.6%) cases, we used a combination of metal coils with uncalibrated PVA particles (“sandwich”). This technique was effective for bleeding from gastroduodenal, splenic and liver arteries. In case of insufficient hemostasis (n=11, 14.9%), Histoacryl (NBCA) was applied. Additional stenting was required after coil embolization in 2 (2.7%) patients (renal tent and large-diameter coronary stent, respectively). In 15 (20.3%) patients, stent-graft was used for damage to superior mesenteric artery and common hepatic artery in the area of gastroduodenal artery stump or in case of its short stump (Fig. 7).

Fig. 7. Techniques of embolization.

If stent-graft implantation was technically impossible (vascular convolution, small vessel, no appropriate stent-graft), we performed spiral embolization with exit of coils into the lumen of the parent artery with subsequent implantation of renal or coronary stent (Fig. 8). This approach carries certain risk of recurrent bleeding, since dual antiplatelet therapy will be required to prevent thrombosis of stent-graft or balloon-expandable peripheral stent.

Fig. 8. Coil embolization combined with stenting.

Time of TAE via transfemoral and transbrachial approaches was similar (Mann-Whitney test, p = 0.523). Pooled median time of TAE was 65 min (IQR 40 — 80). Transfemoral procedure required 65 min (IQR 40 — 77.5), transbrachial approach — 60 min (IQR 50 — 90).

Technical success of surgery defined as the absence of contrast agent extravasation and/or stasis in the target vessel was achieved in all patients, including those with only indirect angiographic signs of bleeding. Abnormalities of vascular anatomy (excessive tortuosity, branched arterial collaterals) were found in 17 (16.4%) out of 106 patients with clinical and laboratory signs of postoperative bleeding who underwent DSA. If source of bleeding was technically inaccessible for endovascular hemostasis, TAE procedure was finished at the diagnostic stage to minimize vascular and organ complications.

Bleeding recurrence developed in 13 (17.6%) patients that required redo endovascular intervention. The period between primary endovascular intervention and redo TAE was 9.5 days (IQR 6.25 — 12.0). Recurrent bleeding followed damage to another vessel in 8 (61.5%) patients, destruction of the same vessel in 3 (23.8%) patients. In 2 (15.4%) patients, the cause of repeated bleeding could not be established.

Subsequently, bleeding recurred in 3 (4.05%) patients that required DSA. Mean period until recurrence was 4 days (IQR 2.25 — 6.5). Bleeding was the result of vascular wall destruction outside the area of primary endovascular hemostasis.

Specific TAE-related postoperative complications occurred in 4 patients. In the first case, bowel wall infarction developed because of inappropriate embolization of superior mesenteric artery. In the second case, acute liver failure occurred due to thrombosis of common hepatic artery and required hemodiafiltration. In the third patient, deliberate embolization of splenic artery resulted splenic infarction with subsequent abscess. The fourth patient developed brachial artery thrombosis after surgery and subsequent local phlegmon.

Complications not associated with TAE were observed in 3 (42.9%) cases. One patient with paroxysmal atrial fibrillation developed acute coronary syndrome at the end of TAE procedure. This event was intraoperatively diagnosed. Emergency coronary angiography revealed LAD occlusion, and a bare-metal stent was implanted. Stenting required intake of P2Y12 receptor inhibitors. Nevertheless, there was no recurrence of intra-abdominal bleeding. In another case, pulmonary thromboembolism developed in 11 days after TAE. This complication required high doses of unfractionated heparin without recurrent bleeding. The third patient developed floating thrombosis of common femoral vein in 9 days after TAE that required cava filter implantation and injections of low-molecular-weight heparins without recurrent bleeding.

In-hospital mortality after TAE was 12.2% (n=9), median period until death — 17 days (IQR 10 — 23). Median length of hospital-stay was 32 days (IQR 21 — 42), in case of lethal outcome — 28 days (IQR 17 — 28).

Fig. 9. Selective digital subtraction angiograms of the posterior arterial arcade of the head of the pancreas noah gland.

Fig. 10. Methods for performing embolization (description in the text).

Fig. 11. Embolization with a coil in combination with a stent.

Discussion

Postoperative bleeding after pancreatectomy is still a scourge of surgical pancreatology. Postoperative pancreatitis and pancreatic fistula underlie this complication. These events lead to local spread of proteolytic and lipolytic enzymes contributing to tissue and vascular wall destruction [10, 11]. In our study, the majority of patients with delayed postoperative bleeding had pancreatic fistula or fluid accumulations. Adequate drainage of fluid accumulations is necessary to reduce the likelihood of vascular destruction. However, it is not always possible. In case of re-laparotomy for postoperative bleeding after pancreatoduodenectomy, disconnection of anastomoses or pancreatic stump extirpation are usually required. Endovascular intervention is an alternative option for hemostasis.

There are few Russian-language studies devoted to endovascular treatment of arrosive bleeding after pancreatectomy. Most reports describe endovascular hemostasis for false aneurysm as a complication after pancreatic surgery [1]. There are only a few studies devoted to arrosive bleeding, but they have descriptive design [12, 13]. Only one study contains data on TAE as the first stage of treatment of acute arrosive postoperative bleeding [13].

The gold standard in the diagnosis of delayed postoperative bleeding is CT angiography (sensitivity 79–92%, specificity 92–95% at blood loss > 0.3 ml/min) [14–16]. Moreover, CT angiography makes it possible to identify not only the source of bleeding (arterial or venous), but also to detect local hematoma [17]. The greatest diagnostic difficulties arise in interpretation of indirect signs of vascular destruction.

Abdominal organs have well-developed collateral vasculature and multiple anastomoses. Therefore, the type/variant of abdominal blood supply should be considered before TAE to minimize ischemic complications. According to the Michels classification of hepatic arteries, 10 types of celiac-mesenteric vascular anatomy are distinguished [18]. This system is used in abdominal surgery for surgery of stomach, liver, biliary tract and pancreas. However, this classification does not currently meet the modern requirements of endovascular surgery. Balakhin P.V. and Tarazov P.G. analyzed data of 3756 angiographies and identified 114 variants of celiac-mesenteric vascular anatomy. The authors combined these types into 5 groups: central (common hepatic), celiac, celiac-mesenteric, mesenteric and aortic [19].

Embolization agents/substances also carry certain risk associated with inappropriate embolization. For example, the choice of small PVA particles significantly increases this risk. Characteristics of patients with unfavorable outcomes are summarized in Table 3. In 1 patient, inappropriate arterial embolization was followed by bowel necrosis, redo surgeries and subsequent death.

Table 3. Characteristics of patients with unfavorable outcome (n=9)

Note. * — autopsy was not carried out at the request of relatives. DPDE – distal pancreatoduodenectomy, DPDE+S – distal pancreatoduodenectomy + splenectomy; HPDE – pancreatic head resection; GE+PDE – pancreatoduodenectomy + gastrectomy; lPDE – laparoscopic pancreatoduodenectomy. CT – celiac trunk, LGA – left gastric artery, IPDA – inferior pancreatoduodenal artery, GDA – gastroduodenal artery, CHA – common hepatic artery, PPA – posterior pancreatic artery; C – cylindrical emboli, H – histoacryl, UP – uncalibrated particles, MOF – multiple organ failure.

Conclusions

1. The main angiographic signs of delayed arterial bleeding after pancreatectomy were contrast agent extravasation (67.6%), arterial false aneurysms (17.5%) and indirect signs of arrosive bleeding (14.9%).

2. Overall mortality in patients with delayed arrosive bleeding after TAE was 12.2% and occurred only in patients with angiographic signs of contrast agent extravasation.

3. CT angiography is necessary to verify late arrosive arterial bleeding after pancreatectomy in patients with stable hemodynamics.

4. If a source of bleeding is identified, CT angiography should be immediately followed by digital subtraction angiography of celiac-mesenteric arteries and endovascular hemostasis.

5. Empirical endovascular hemostasis is advisable in patients with indirect signs of arterial destruction and no signs of active bleeding.

6. Recurrent bleeding in patients with stable hemodynamics is an indication for immediate CT angiography followed by endovascular hemostasis.

The authors declare no conflicts of interest.