Delayed post-pancreaticoduodenectomy hemorrhage: is endovascular treatment a strategic necessity?

Abstract

Background

Late post-pancreaticoduodenectomy hemorrhage (PPH) is a rare but potentially fatal complication associated with high morbidity and mortality. Early recognition and prompt intervention are critical for improving patient outcomes. This study aimed to evaluate the clinical features, treatment modalities, and outcomes of late-onset PPH cases managed in our institution.

Materials and methods

Between January 2016 and January 2025, a total of 220 patients underwent pancreaticoduodenectomy (PD) at the Department of General Surgery, XXX University Faculty of Medicine. Patients who developed late arterial hemorrhage, defined by the International Study Group of Pancreatic Surgery (ISGPS) as occurring more than 24 h after the index operation, were included. Demographic data, histopathological diagnoses, timing and source of bleeding, radiological and angiographic findings, treatment modalities, rebleeding events, morbidity, and mortality rates were analyzed. In patients undergoing endovascular treatment (EVT), technical success and complications were recorded.

Results

Late PPH occurred in 5.9% (13/220) of patients, with a mean onset of 10.3 days postoperatively. The most frequent bleeding source was the gastroduodenal artery (61.5%), and 61.5% of patients had an anastomotic leak. Sentinel bleeding was present in 76.9% of cases. EVT was performed in eight patients, achieving technical success in 75% after the first attempt and 100% after repeat procedures; rebleeding occurred in 25%. Surgical intervention was performed in five patients, with remnant pancreatectomy required in 60% of cases. Mortality was 60% in the surgical group and 25% in the EVT group. No cases of hepatic ischemia or abscess were observed following hepatic artery embolization.

Conclusion

Endovascular therapy appears to be a feasible first-line approach for the management of late PPH and may achieve high technical success. Surgical intervention should be reserved for selected cases in which EVT is unavailable or unsuccessful.

Introduction

Pancreatic surgery, particularly when performed for malignancies requiring advanced surgical procedures, is associated with a high risk of complications. Although the mortality rate related to pancreaticoduodenectomy (PD) has shown a marked decline to below 5% due to improvements in conditions of care, morbidity remains substantially high, ranging between 30% and 50% [1–4]. The most common major complications following PD include postoperative pancreatic fistula (POPF), delayed gastric emptying, intra-abdominal abscess, and hemorrhage.

Postoperative hemorrhage after pancreatic surgery, although less frequent than other complications, requires prompt intervention and can have a profound impact on patient prognosis. The reported incidence varies between 2.5% and 20.2%, with mortality rates ranging from 11.0% to 56.2% in various studies [5, 6]. These rates vary depending on patient-related factors, surgical technique, the type of procedure performed, and postoperative management. Post-pancreatectomy hemorrhage (PPH) may occur in either the early or late postoperative period. The International Study Group of Pancreatic Surgery (ISGPS) has proposed a consensus definition for PPH, categorizing hemorrhage occurring within 24 h of surgery as early PPH and that occurring thereafter as late PPH [7]. Early PPH is most often related to technical issues, whereas late PPH is usually associated with vascular complications arising from pancreatic fistula or infection. While severe early PPH generally warrants surgical intervention, mild cases may be managed conservatively [8]. The management of late PPH is more variable and depends on the severity and site of bleeding; however, angiographic interventions are generally recommended over surgery [6, 7, 9]. Despite advances such as computed tomography (CT) angiography and transarterial techniques for the diagnosis and management of PPH, its prognosis remains poor. No universally validated preventive measures are available; Therefore, prevention of PPH after PD continues to represent a significant clinical challenge.

In this context, PPH should not be regarded as an isolated bleeding event but rather as part of a complex postoperative period. The present study retrospectively analyzed cases of late PPH in patients who underwent PD at our institution. We evaluated the treatment modalities employed for this complication and sought to identify strategies that may improve management. The goal was to develop practical clinical recommendations that could contribute to the effective treatment of this serious complication.

Materials and methods

Study design and patients

Between January 2016 and January 2025, a total of 220 patients underwent PD in our hospital. According to the definition proposed by the International Study Group of Pancreatic Surgery (ISGPS), patients who developed late hemorrhage occuring more than 24 h after the index surgery were included in this study. Thirteen patients who met this criterion were included in the study population, and their data were reviewed retrospectively.

Demographic data (age, sex), final histopathological diagnosis, timing of hemorrhage, radiological findings, bleeding site, decrease in hemoglobin level, number of transfused blood units, diagnostic methods, management strategy, occurrence of rebleeding, need for re-intervention, as well as morbidity and mortality were retrieved from patient medical records.

Postoperative hemorrhage was defined as the presence of fresh blood in an abdominal drain, active bleeding from the surgical wound, hematemesis or melena, and/or decrease in hemoglobin level of more than 2 g/dL. The diagnosis and management of bleeding, including the use of CT angiography, surgical re-exploration, mesenteric angiography, and stent placement or embolization, were documented. Endovascular therapy was prioritized whenever possible; however, emergency surgery was performed in cases where 24-hour interventional radiology services were unavailable.

Patient comorbidity profiles were assessed using the American Society of Anesthesiologists (ASA) physical status classification as recorded in the preoperative anesthesia evaluation. For analysis, ASA scores were categorized as ASA ≤ II and ASA ≥ III. Hemodynamic status during the bleeding episode and before any surgical or endovascular intervention was evaluated based on mean arterial pressure (MAP) and the requirement for vasopressor support. Hemodynamic instability was defined as a MAP < 65 mmHg and/or the need for vasopressor support; patients not meeting these criteria were considered hemodynamically stable. Pancreatic fistula/leak was defined as an amylase value greater than three times the normal laboratory value, measured from a drain with any amount of output, after the third postoperative day. Bile leak was defined as the presence of bile in the output from a drain placed adjacent to the hepaticojejunostomy. Hepatic necrosis was defined as a serum alanine aminotransferase (ALT) level exceeding 1,000 IU/L. Liver function tests (LFTs) were recorded both before and after the endovascular intervention. Mortality was defined as any death occurring within 90 days after PD.

Surgical technique

PD was performed either as a pylorus-preserving modification or via the classical Whipple procedure, depending on the surgeon preference. Pancreatic-enteric continuity was restored with a duct-to-mucosa pancreaticojejunostomy in all cases, most commonly using a Blumgart-type technique. Portal or superior mesenteric vein resection and reconstruction were performed when indicated. In patients with a narrow pancreatic duct, a pancreatic duct stent was routinely used. Routine intra-abdominal drains were placed to drain the anastomotic region and the GDA stump, and somatostatin analog therapy was administered in patients who developed clinically relevant POPF.

Endovascular treatment protocol

In cases of post-PD hemorrhage, all endovascular interventions were performed under sterile conditions via the femoral artery approach. After insertion of a 5 F angiographic sheath, an abdominal aortogram was obtained using a 5 F pigtail catheter to assess visceral arterial anatomy. Selective catheterization of the celiac trunk and superior mesenteric artery was performed with a 5 F Cobra or Simmons-1 catheter. When digital subtraction angiography (DSA) demonstrated active extravasation, a 2.7 F microcatheter (Progreat^®, Terumo, Japan) was advanced to the bleeding site, and embolization was carried out. Coil selection was based on lesion location and vessel diameter, using Concerto^® (Medtronic, USA) or Interlock™ (Boston Scientific, USA) coils of various sizes. When appropriate, a polytetrafluoroethylene-covered stent graft was implanted. Complete vessel occlusion was confirmed by post-procedure angiography. The sheath and catheter system were then removed, and femoral hemostasis was achieved. Technical success was defined as angiographic cessation of contrast extravasation and/or successful exclusion of the culprit pseudoaneurysm at the end of the procedure.

Statistical analyses

All statistical analyses were performed using IBM SPSS Statistics (version 26.0; IBM Corp., Armonk, NY, USA). Continuous variables with a normal distribution were expressed as mean ± standard deviation, whereas non-normally distributed variables were presented as median (minimum–maximum). Categorical variables were expressed as counts and percentages. Comparisons between groups were performed using the Mann–Whitney U test for continuous variables and the Fisher’s exact test for categorical variables. A p value < 0.05 was considered statistically significant.

Results

Between 2016 and 2025, a total of 220 patients underwent PD at our institution. The incidence of clinically relevant POPF (grade B and C) was 17.7%, while PPH occurred in 7.2% of patients (n = 16). Of these, three patients (1.3%) experienced early PPH (< 24 h), and thirteen patients (5.9%) developed late PPH (> 24 h).

All three patients with early PPH underwent emergency surgical exploration. The bleeding source was identified at the site of portal vein reconstruction in one patient, the splenic vein in another, and the spleen in the third. Splenectomy was performed in two patients, and portal vein reconstruction was revised in one. The patient with portal vein bleeding died in the early postoperative period.

Among the thirteen patients with late PPH, nine were male and four were female, with a median age of 61 years (48–77). PD was performed for pancreatic head tumors in ten cases and ampullary tumors in three cases. Histopathology revealed adenocarcinoma in all patients. The median tumor size was 2.75(1.6–7.0.6.0) cm, and the mean number of metastatic lymph nodes was 3.14. Anastomotic leakage from the pancreatojejunostomy (n = 7) or hepaticojejunostomy (n = 1) was detected in 61.5% of patients (n = 8). The most frequent initial sign of bleeding was hemorrhagic output from intra-abdominal drains, meeting the criteria for sentinel bleeding in 76.9% of cases. CT angiography demonstrated extravasation in 61.5% of the cases (Fig. 1), while in the remaining 38.5%, hematoma was observed as an indirect sign of bleeding. The mean time to bleeding onset was 10.3 days (3–23 days). Late PPH after discharge occurred in one patient, on postoperative day 23. The bleeding site was identified as the gastroduodenal artery (GDA) in eight patients, the common hepatic artery (CHA) in four patients, and the left gastric artery (LGA) in one patient (Table 1).

Fig. 1.

Contrast-enhanced computed tomography angiography (CTA) demonstrating a peripancreatic hematoma (yellow arrow) consistent with postoperative hemorrhage

Table 1.

Demographic and clinical features of patients according to treatment modality

	All(n, %)	EVT(n, %)	Surgery(n, %)	p value
Number of patients	13(100,0)	8(61,5)	5(38,5)	-
Age				0,71
Mean±SD	63,5±7,8	64,1±8,2	62,6±7,4
Median (Min-Max)	61 (48-77)	62 (50-77)	60 (48-71)
Sex				0,62
Male	9 (69,2)	6 (75,0)	3 (60,0)
Female	4 (30,8)	2 (25,0)	2 (40,0)
Tumor Size				0,84
Mean ± SD	3,15±1,48	3,2±1,5	3,0±1,4
Median (Min-Max)	2,75 (1.6–7.0.6.0)	2,9(1.6–7.0.6.0)	2.6(1.8–6.5.8.5)
Tumor Location				1,0
Pancreas	10(76,9)	6 (75,0)	4 (80,0)
Ampulla	3(23,1)	2 (25,0)	1 (20,0)
Histologic Type				-
Adenocarcinoma	13(%100)	8 (100)	5 (100)
Number of Metastaticlymph Nodes				0,78
Mean ± SD	3,14 ±2,57	3,0 ± 2,4	3,4 ±2,9
Median (Min-Max)	3,50 (0–8)	3 (0–7)	4 (0–8)
ASA Score				0,22
ASA ≤ II	9 (69,2)	7 (87,5)	2 (40,0)
ASA ≥ III	4 (30,8)	1 (12,5)	3 (60,0)
Hemodynamic Status				0,29
Stable	8 (61,5)	6 (75,0)	2 (40,0)
Unstable	5 (38,5)	2 (25,0)	3 (60,0)
Decrease in Hb (g/dL)				0,67
Mean ± SD	2,77±0,95	2,71±0,65	2,86±1,39
Median (min-max)	2,9 (1.3–4.7.3.7)	2,9 (1.3–3.2.3.2)	2,7 (1,5–4,7,5,7)
Units of Blood Transfused				0,41
Mean ± SD	3,77±0,78	3,6±0,7	4,0±0,9
Median (min-max)	2,9(2,0–4,7,0,7)	2,8 (2,0–4,5,0,5)	3,2 (2,5–4,7,5,7)
POPF and Bile Leak				0,29
Yes	8(61,5)	6 (75,0)	2 (40,0)
No	5(38,5)	2 (25,0)	3 (60,0)
Bleeding Onset Day (POD)				0,39
Mean ± SD	10,3±5,1	11,2 ±5,4	8,9 ±4,6
Median (min-max)	9 (3–23)	10 (3–23)	8 (4–18)
Initial Sign				1,00
Drain Bleeding	10 (76,9)	6 (75,0)	4 (80,0)
Wound Bleeding	2 (15,4)	1 (12,5)	1 (20,0)
Syncope	1 (7,7)	1 (12,5)	0 (0,0)
Bleeding Side				0,84
GDA	8 (61,5)	4 (50,0)	4 (80,0)
CHA	4 (30,8)	3 (37,5)	1 (20,0)
LGA	1 (7,7)	1 (12,5)	0 (0,0)

ASA American Society of Anesthesiologists, POPF Postoperative pancreatic fistula, GDA Gastroduodenal Artery, CHA Common Hepatic Artery, LGA Left Gastric Artery, Hb Hemoglobin, POD Postoperative Day, EVT Endovascular Therapy

Five patients underwent emergency laparotomy due to the unavailability of 24-hour endovascular therapy services. In these cases, GDA ligation was performed in four patients and CHA ligation in one patient, as a mandatory salvage procedure. Pancreatic anastomotic exploration and remnant pancreatectomy were required in three of these patients (60%) (Table 2).

Table 2.

Detailed clinical characteristics and outcomes of patients

	POPF and BileLeak	BleedingTime	Initial Sign	Sentinelbleeding	CTA Findings	BleedingSite	Treatment	Re-bleeding	Outcome	Cause OfDeath
1	PJ Leak	20	Drain Bleeding	Yes	Extravasation	GDA	Surgery;GDA Ligation	Yes	Dead	DIC, MOFS
2	PJ Leak	10	Drain Bleeding	Yes	Extravasation	GDA	Surgery;GDALigation+Subtotal PD	No	Dead	MOFS
3	No	3	Drain Bleeding	Yes	Hematoma	CHA	Surgery;HALigation+Subtotal PD	Yes	Dead	DIC, MOFS
4	PJ Leak	7	Drain Bleeding	Yes	Extravasation	GDA	Surgery;GdaLigation+Subtotal PD	No	Alive
5	No	8	WoundBleeding	No	Extravasation	GDA	Surgery;GDA Ligation	No	Alive
6	PJ Leak	23	Drain Bleeding	Yes	Extravasation	GDA	EVT; Stent+CHA Coil	Yes	Alive
7	PJ Leak	9	Drain Bleeding	Yes	Extravasation	GDA	EVT; Stent+CHA Coil	No	Alive
8	No	8	Drain Bleeding	Yes	Hematoma	CHA	EVT;CHA Coil	No	Alive
9	HJ Leak	10	Drain Bleeding	Yes	Hematoma	CHA	EVT;CHA Coil	No	Dead	PulmonaryAspiration
10	No	10	Drain Bleeding	Yes	Hematoma	CHA	EVT;CHA Coil	No	Alive
11	No	9	Syncope	No	Extravasation	GDA	EVT;CHA Coil	No	Alive
12	PJ Leak	9	Drain Bleeding	Yes	Extravasation	GDA	EVT;CHA Coil	Yes	Alive
13	PJ Leak	10	WoundBleeding	No	Hematoma	LGA	EVT; LGA Stent	No	Dead	Sepsis

POPF Postoperative pancreatic fistula, CTA Computed Tomography Angiography, PJ Pancreaticojejunostomy, HJ Hepaticojejunostomy, GDA Gastroduodenal Artery, CHA Common Hepatic Artery, LGA Left Gastric Artery, DIC Disseminated Intravascular Coagulation, MOFS Multiple Organ Failure Syndrome

Endovascular therapy was performed in eight patients. Angiography revealed GDA extravasation in four cases (Fig. 2A), CHA pseudoaneurysm in three, and LGA pseudoaneurysm in one. Coil embolization of the CHA was performed in five patients (Fig. 2B), as stent placement was not feasible due to anatomical factors such as a short GDA stump, unfavorable arterial angulation, or a narrow pseudoaneurysm neck. Covered stent placement was carried out in three cases (two GDA, one LGA). In the patient with LGA bleeding, the covered stent achieved successful hemostasis. In one patient with GDA bleeding, the stent failed to achieve hemostasis, simultaneously necessitating CHA embolization. In another GDA case, rebleeding occurred two days later due to stent malrotation, requiring repeat CHA embolization (Table 2).

Fig. 2.

A Digital subtraction angiography (DSA) demonstrating the bleeding source arising from the gastroduodenal artery (yellow arrow) prior to endovascular intervention. B Post-procedural DSA image showing successful coil embolization of the gastroduodenal artery with no residual contrast extravasation (yellow arrow)

Rebleeding after endovascular therapy occurred in two patients (25%). In one case, stent malrotation after GDA bleeding led to CHA embolization, while in the other, recurrent hemorrhage occurred after initial CHA embolization, also necessitating repeat embolization of the CHA. In total, CHA coil embolization was performed in seven patients, and a covered stent was placed in the LGA in one patient. Hemostasis was achieved after the first intervention in six patients (75%) and after a second intervention in two patients (25%) (Table 3). Transient and mild elevations in ALT levels were observed in three of seven patients undergoing CHA embolization, with normalization within one week. None of the patients developed hepatic ischemia or abscess (Table 3).

Table 3.

Detailed angiographic findings and clinical outcomes of patients undergoing endovascular treatment

1	Angiographic Findings	EVT	Re-bleeding	Second Line Management	Hemostasis	Liver Ischemia	Survival
	GDA Extravasation	GDA Stent	Yes	CHA Coil	Complete	Transient elevation of LFTs	Alive
2	GDA Extravasation	GDA Stent + CHA Coil	No		Complete	Transient elevation of LFTs	Alive
3	CHA Pseudoaneurysm	CHA Coil	No		Complete	No	Alive
4	CHA Pseudoaneurysm	CHA Coil	No		Complete	Transient elevation of LFTs	Died
5	CHA Pseudoaneurysm	CHA Coil	No		Complete	No	Alive
6	GDA Extravasation	CHA Coil	No		Complete	No	Alive
7	GDA Extravasation	CHA Coil	Yes	CHA Coil	Complete	No	Alive
8	LGA Pseudoaneurysm	LGA Stent	No		Complete	No	Died

GDA Gastroduodenal Artery, CHA Common Hepatic Artery, LGA Left Gastric Artery, EVT Endovascular Therapy, LFT Liver Function Test

Among surgically treated patients, three of five (60%) died due to multiple organ failure syndrome (MOFS) and disseminated intravascular coagulation (DIC). In the endovascular group, two of eight patients (25%) died, one due to POPF-related sepsis and the other due to pulmonary aspiration. The overall mortality rate among patients with late PPH was 38.4% (5/13), with rates of 60% (3/5) in the surgical group and 25% (2/8) in the endovascular group. The mean hospital stay for patients with late PPH was 20.8 days (11–37) (Table 4).

Table 4.

Comparison of 90-day survival outcomes between endovascular treatment and surgical management

Survival	EVT	Surgery
Died	2 (%25)	3 (%60)
Alive	6 (%75)	2 (%40)
Total	8 (%100)	5 (%100)

EVT Endovascular Therapy

Discussion

Late-onset hemorrhage following pancreaticoduodenectomy is a rare but life-threatening complication, associated with high rates of morbidity and mortality, and requiring urgent intervention. Although no universally accepted treatment guidelines exist for PPH, early diagnosis is emphasized as a determinant of successful management [10]. The classification proposed by the International Study Group of Pancreatic Surgery (ISGPS), differentiating early (< 24 h) from late (> 24 h) PPH, has been instrumental in guiding treatment strategies [11]. Endovascular approaches are now frequently favored over surgery, owing to their minimally invasive nature and lower complication rates. In this study, we compared the efficacy of surgical treatment and EVT in the management of late PPH. Early PPH is typically caused by technical inadequacies or insufficient hemostasis, whereas late PPH most often results from vascular complications secondary to pancreatic anastomotic leakage or surgical trauma during lymphadenectomy [12–14]. The incidence of late PPH in the literature ranges between 2.3% and 10%, and the rate observed in our series (5.9%) is consistent with this range [1, 3, 10, 15].

In our cohort, pancreaticojejunostomy or hepaticojejunostomy leakage was detected in 61.5% of late PPH cases, emphasizing its role as a significant risk factor for bleeding. Sentinel bleeding, observed in 76.9% of our patients as hemorrhagic output from an intra-abdominal drain, was found to be a critical early diagnostic sign and an important predictor for prompt endovascular evaluation. In contrast, the intraluminal bleeding manifestations commonly described in the literature, such as hematemesis and melena [1], were not observed at the clinical onset of delayed PPH in our series. In hemodynamically stable patients, CT angiography can help localize the bleeding source by detecting pseudoaneurysms or extravasations, thereby potentially increasing the success rate of endovascular therapy. In our cohort, CT angiography demonstrated extravasation in 61.5% of the cases, while in the remaining 38.5%, hematoma was observed as an indirect sign of bleeding. Similarly, Blanc et al. reported that sentinel bleeding serves as a valuable indication for urgent CT angiography [5]. The mean onset time of late PPH in our study was 10.3 days (3–23 days), with one patient experiencing bleeding after hospital discharge on the 23rd postoperative day. This finding highlights the importance of post-discharge surveillance protocols [1]. Beyond sentinel bleeding, early warning signs may include an unexplained drop in hemoglobin, tachycardia, hypotension, abdominal pain, fever or signs of sepsis, jaundice, and gastrointestinal bleeding manifestations such as melena or hematemesis. Routine outpatient follow-up after discharge, particularly in patients with POPF or bile leakage, may help detect late PPH earlier by prompting careful assessment of these warning signs. The most common bleeding source was identified as the GDA, which is consistent with literature data indicating that the GDA is the most frequently involved artery in pseudoaneurysm formation [4, 7, 16].

For hemodynamically stable patients, EVT is generally considered the first-line treatment. In our series, EVT was performed in eight cases, with technical success achieved in 75% after the initial procedure and 100% after repeat intervention. Reported technical success rates for EVT in the literature range from 72% to 100%, with rebleeding rates between 7% and 30% [6, 7, 10, 17, 18], which are closely consistent with our results.

One of the most concerning complications of EVT is hepatic and biliary ischemia, potentially leading to recurrent liver abscesses, particularly following hepatic artery embolization [19–21]. Reported rates of hepatic ischemia and abscess formation after hepatic artery embolization vary widely, from 0% to 83% [6, 7, 22–26]. Such variation is likely due to multiple factors, including the level of embolization, portal vein patency, the presence of arterial collaterals, and comorbid conditions such as sepsis or fistula. Proximal embolization in the setting of portal vein obstruction or inadequate arterial collaterals is particularly associated with markedly increased complication rates. Although both stent placement and coil embolization achieve high rates of hemostasis, early stent occlusion and rebleeding remain notable concerns, and embolization of major hepatic arteries has been reported to be not universally well tolerated [6, 7, 25]. The literature indicates that hepatic artery occlusion is tolerated only in selected clinical subgroups. In the hepatic artery stent series by Finch et al., the initial hemostasis rate was 72%, but thrombotic occlusion developed in 71% of stents within a median of 10 days. The authors highlight that stent placement can be a valuable option in cases of short GDA stump bleeding or direct hepatic artery hemorrhage, particularly when embolization may risk complete loss of hepatic arterial inflow. However, it carries a substantial risk of stent occlusion [6]. Nozawa et al. demonstrated that hepatic artery occlusion is safe only in cases with preserved portal vein patency and well-developed extrahepatic collateral circulation, whereas the risk of hepatic ischemia increases significantly in the presence of portal vein stenosis [23]. Similarly, Zhang et al. reported high success and low complication rates for hepatic artery embolization in late PPH, but mainly in well-collateralized patients [26]. In our series, transient ALT elevation was observed in three patients following hepatic artery embolization, with liver function tests returning to normal within 10 days. The absence of hepatic ischemia after CHA embolization in our series may be related to the fact that patients were in the late-stage PPH group, preserved of portal vein patency, and adequate collateralization. Although our findings suggest that hepatic artery embolization can preserve hepatic perfusion in most patients, the safety of this approach is closely dependent on patient-specific anatomical and hemodynamic conditions. Therefore, CHA occlusion should not be interpreted as universally safe for all patients undergoing pancreaticoduodenectomy, and the potential long-term biliary complications must also be taken into consideration.

Stent placement offers a potential advantage by preserving arterial flow and reducing ischemic complications [6, 7, 27, 28]. However, the use of covered stents carries additional disadvantages, including the risk of vessel injury, recurrent bleeding, and stent thrombosis [29, 30]. In particular, the need for antiplatelet therapy to prevent thrombosis may a significant drawback in patients who remain at risk for rebleeding. Therefore, in patients who stent placement is not possible or who are not suitable for antiplatelet therapy, coil embolization may a safer alternative [25]. The choice between stent placement and coil embolization in endovascular therapy depends on several factors, including the anatomical location of the bleeding, the need to preserve hepatic perfusion, arterial variations, and overall technical feasibility. The presence of a replaced hepatic artery may complicate the procedure, as sharp arterial angulations, tortuous vessel courses, and short landing zones can make stent deployment technically difficult or even unfeasible. In addition, portal vein patency and the presence of extrahepatic collateral circulation are critical determinants of treadment success [23, 25]. Therefore, the endovascular approach should be individualized by considering arterial anatomical variations, portal vein status, and collateral circulation. Both the need for antiplatelet therapy following stent placement and the potential loss of perfusion associated with coil embolization should be carefully considered to determine the optimal strategy for each patient.

Surgical intervention is generally necessary in cases where endovascular treatment is unavailable, the patient cannot be stabilized for the procedure, or embolization is unsuccessful. One of the main limitations of surgery is that localization of the bleeding site often requires opening anastomoses. This may lead to disruption of the anastomosis and, in some cases, complete removal of the remnant pancreas (remnant pancreatectomy). Although bleeding control may be achieved, the systemic inflammatory response triggered by surgical trauma can further compromise hemodynamic status, particularly in unstable patients, thereby limiting their ability to tolerate emergency surgery. Consequently, mortality rates remain high despite successful hemostasis. Surgical mortality rates as high as 30%–50% have been reported [26, 31–34], whereas EVT generally achieves technical success rates above 80%–90% with mortality rates between 10% and 25% [2, 8, 10, 17, 18, 25].

In our series, mortality in the emergency relaparotomy group was 60%, primarily due to MOFS and DIC, compared to 25% in the EVT group. In the EVT group, deaths were mainly related to non-bleeding causes. However, these findings should be interpreted cautiously due to the limited sample size and treatment selection. Treatment selection may have been influenced by real-world factors such as hemodynamic status at the time of bleeding and access to interventional radiology. Importantly, timely access to interventional radiology may vary across different healthcare systems and institutions, which can directly influence treatment selection in late PPH. In addition, mortality may be affected by non-bleeding-related factors such as POPF/leakage-related sepsis or systemic deterioration; therefore, 90-day mortality should not be interpreted solely as a measure of hemostatic strategy effectiveness.

This study has several limitations. It was a retrospective, single-center study with a relatively small sample size. Treatment allocation was not randomized and may have been influenced by real-world factors such as hemodynamic status at the time of bleeding and institutional availability of interventional radiology, limiting direct comparability between groups. In addition, non-bleeding-related factors such as POPF/leakage-related sepsis and systemic deterioration may have contributed to mortality outcomes. Therefore, mortality may not solely reflect the effectiveness of the hemorrhage control strategy. As our analysis was limited to patients who developed late PPH, we were unable to evaluate the impact of leakage-mitigation strategies on the overall incidence of late PPH. Finally, the long study period may have introduced time-related effects, including changes in perioperative management or procedural techniques. Further studies are needed to better define optimal management strategies for late PPH after pancreaticoduodenectomy.

Our findings support EVT as the preferred first-line approach for the management of late PPH after pancreaticoduodenectomy. In our cohort, EVT achieved high technical success, even in cases of recurrent bleeding requiring repeat intervention. In our cohort, mortality was lower in the EVT group than in the emergency relaparotomy group. Timely access to interventional radiology remains an important component of care in late PPH, and centers performing pancreatic surgery should ensure appropriate access to endovascular expertise. Surgical intervention should be reserved for selected cases in which EVT is unavailable or unsuccessful.

Conclusıon

Our study suggests that endovascular therapy is a feasible first-line approach in the management of late post-pancreaticoduodenectomy hemorrhage and achieves high technical success. Surgical intervention should be reserved for selected cases in which endovascular treatment is unavailable or unsuccessful. Ensuring timely access to interventional radiology in centers performing pancreatic surgery may contribute to improved patient safety by reducing potentially preventable mortality.

Abbreviations

POPF

Postoperative pancreatic fistula

GDA

Gastroduodenal Artery

CHA

Common Hepatic Artery

LGA

Left Gastric Artery

Hb

Hemoglobin

CTA

Computed Tomography Angiography

PJ

Pancreaticojejunostomy

HJ

Hepaticojejunostomy

DIC

Disseminated İntravascular Coagulation

MOFS

Multiple Organ Failure Syndrome

EVT

Endovascular Therapy

LFT

Liver Function Test

Authors' contributions

Conceptualization, O.A; methodology, O.A.,E.B.Ç.; formal analysis, O.A.,S.A.; investigation, O.A.,S.A.,A.E.,A.T.; resources, O.A.,E.B.Ç.,S.A.,A.E.,A.T.,A.E.S.; writing—original draft preparation, O.A.; writing—review and editing, O.A.; visualization, O.A.,E.B.Ç,A.E.; supervision, O.A.,A.T.; project administration, O.A. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data availability

The data that support the findings of this study are available from the corresponding authorupon reasonable request.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Ethics Committee of Aydın Adnan Menderes University Medical Faculty, Turkey (approval date: 24 july 2025; approval number: 2025/242). Informed consent for participation is not required, due to the retrospective design of the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.