The Institutional Review Board approved this retrospective study and informed consent was waived. Eleven consecutive patients with delayed (≥ 3 months) GI bleeding from jejunal varix caused by PV occlusion after hepatobiliary-pancreatic surgery who were treated with PV stent placement between April 2009 and December 2015 were included in the study. Of these, eight were male and three were female patients, aged from 2 to 79 years. PV occlusion and jejunal varix were diagnosed based on portal phase images of contrast-enhanced computed tomography (CT). PV occlusion was defined as segmental discontinuation of the PV and development of collateral channels. Jejunal varix was demonstrated by clustered, tortuous, and dilated veins in the wall of the jejunum around the choledochojejunostomy or hepaticojejunostomy. Treatment was determined based on laboratory findings and the clinical situation: decreased serum hemoglobin level (7.7–11.5 g/dL) and GI bleeding that was recurrent and worsening in intensity and duration.

The access route for PV stenting was determined based on the anatomic location, shape, and extent of PV occlusion depicted on preprocedural contrast-enhanced CT. The presence of a tapering end in the patent PV or superior mesenteric vein (SMV) or splenic vein (SPV) was a main determinant of the access route. Access was initially performed via right or left transhepatic and/or transsplenic approach. Percutaneous transhepatic and/or transsplenic approach was performed with a 22-G Chiba needle (Cook Incorporated, Bloomington, IN, USA) under ultrasound guidance. After insertion of a 6 Fr or 7 Fr vascular sheath into the intrahepatic PV or SPV, 4 Fr or 5 Fr angiographic catheter (either Torcon NB Advantage catheter [Cook Incorporated] or Glidecath Angled Taper [Terumo Corporation, Tokyo, Japan]) was advanced over a 0.035-inch hydrophilic guidewire. Navigation of the guidewire through the occluded PV was performed under the landmark of surgical clips seen on fluoroscopy or imaginary PV course. When the guidewire passage was difficult, a braided sheath (Super Arrow_Flex, Arrow International, Inc., Reading, PA, USA) was used to achieve better mechanical support. A 2 Fr microcatheter (Progreat, Terumo Corporation) and a microguidewire (0.016 or 0.014 inch in diameter) were also used for difficult cases. After a wire passed through an occluded segment, a catheter was advanced into the SMV or SPV; subsequently, venography was obtained to identify the length of the occluded segment, the diameter of SMV or PV, and any collateral veins including jejunal varix. Along the exchanged stiffer guidewire (Rosen Curved Wire Guide, Cook Incorporated), preballoon dilatation was performed with a 4- to 6-mm diameter balloon and a self-expandable nitinol stent (SMART control, Cordis Corporation, Miami Lakes, FL, USA) was placed to cover the entire occluded segment. Diameter of the stent was 1 to 2 mm larger than the maximum diameter of the remnant PV or SMV demonstrated on angiography. Post-dilatation was performed with an 8- or 10-mm diameter balloon. Pressure gradient was not routinely measured; however, two patients had post-procedural pressure gradient measurement. Completion angiography was performed with the tip of a catheter in the SMV or SPV. One patient underwent aspiration thrombectomy with a 6 Fr guiding catheter (Envoy, Codman & Shurtleff, Inc., Raynham, MA, USA) for a partially protruded thrombus into the stent lumen, which was identified on completion angiogram. Percutaneous access tract in the liver and/or spleen was embolized with coils and/or Gelfoam (Cutanplast, Mascia Brunelli S.P.A, Milano, Italy). Anticoagulation was performed at least 3 months after the procedure with acetylsalicylic acid 100 mg and clopidogrel 75 mg.

Review of electronic medical records was performed for underlying disease, type of surgery, and postoperative treatment. Symptom onset period was between the date of surgery and the date of symptom occurrence. The extent of PV occlusion, presence of tapering end in patent portal or SMV, and the location of ectopic varix were analyzed based on contrast-enhanced CT. The cause of PV occlusion was evaluated with postoperative CT and positron emission tomography. Technical success was defined as complete restoration of the hepatopetal portal venous flow after stent placement on completion angiography. Treatment efficacy was defined as disappearance of jejunal varix on the first follow-up CT after PV stenting. Clinical success was defined as disappearance of GI bleeding without procedure-related complications. Follow-up period was defined as the time-period between the date of interventional treatment and the date of the last hospital visit. Contrast-enhanced CT or Doppler ultrasonography was performed to evaluate stent patency in all patients, except one who was transferred after procedure. Primary patency of the PV stent was calculated with Kaplan-Meier analysis using statistical software (IBM SPSS for Windows, version 20.0, IBM Corp., Armonk, NY, USA). Complication and mortality were evaluated based on the Society of Interventional Radiology guideline (20).

All patients underwent hepatobiliary-pancreatic surgery for pancreatic cancer except two children with LT for biliary atresia. Surgery included pylorus-preserving pancreaticoduodenectomy (n = 7), total pancreatectomy (n = 1), and Whipple's operation (n = 1). In two patients, SMV resection and anastomosis was performed at the time of surgery. Between surgery and PV stenting, two patients received both radiation treatment and adjuvant chemotherapy; and the other three patients received adjuvant chemotherapy.

Gastrointestinal bleeding occurred, on average, 27 months (4 to 72 months) after surgery. GI bleeding was manifested as melena (n = 6) and hematochezia (n = 5). Initially, hemorrhage was intermittent and small in amount in all patients and progressed in only three patients. Endoscopic examination was performed in nine patients: endoscopic hemostasis was attempted but failed in three patients.

The cause of PV occlusion was benign postoperative change in six patients and local tumor recurrence in five patients. All PV occlusion was extrahepatic. In eight out of 11 patients, occlusion extended to the splenoportal junction or SMV. On CT images, the mean length of PV occlusion was 3.7 cm (range, 1.1 to 6.3 cm). Tapering appearance of the patent PV or SPV was seen in eight patients on CT and 10 patients on angiogram (Fig. 1). Varix was clearly depicted on CT in the afferent jejunal loop in all patients. Percutaneous transhepatic PV stenting was performed, on average, 37 months (4 to 121 months) after surgery. Jejunal varix embolization was not performed in all patients. Transhepatic and transsplenic approaches were used in 10 patients (right in eight patients and left in two patients) and in one patient, respectively. The approach was determined from the presence of tapering appearance in patent vein on CT (Supplementary Movie 1–4 in the online-only Data Supplement); otherwise, transhepatic approach was attempted first. Post-stent pressure gradient was measured in two patients and was less than 3 mm Hg between SMV and the intrahepatic PV.

Technical success rate was 90.9% (10 out of 11). PV stenting was successful at first trial in nine patients and at second trial in two patients. PV stenting failed in one patient who had undergone radiation therapy for local cancer recurrence. Mean follow-up period was 9 months (range, 1–27 months). Follow-up CT in seven patients and Doppler ultrasonography in three patients were performed at mean 7 months (range, 1–14 months) and 12 months (range, 4–26 months) after PV stenting, respectively. Treatment efficacy was 100%. Jejunal varix disappeared on post-stenting CT available in six patients with successful PV stenting and remained in one patient with technical failure. Clinical success rate was 81.8% (9 out of 11). One patient visited the emergency room 40 days after PV stenting because of hematochezia. This patient had a long segmental occlusion from the PV to SMV and a history of radiation therapy before procedure. Although portal venous flow was restored after placement of two nitinol stents, completion angiogram showed partial thrombi within the PV. Despite the successful aspiration thrombectomy, good portal venous flow on final angiogram, and the dual antiplatelet medication after procedure, the patient visited the emergency room due to recurred hematochezia. His follow-up CT revealed complete obstruction of the PV stent and recurred jejunal varix. Primary patency rate of the PV stent during the mean follow-up period of 9 months was 88.9% (Fig. 2). Neither procedure-related complication nor mortality occurred during the follow-up period. Cases are summarized in Table 1.