From our prospectively maintained database, we retrospectively reviewed the records of all patients who had undergone EUS-HGS at our institution between January 1, 2016, and December 31, 2021. Only patients who had received PCSEMS with 5-mm or 20-mm uncovered portions on the liver side were included, whereas those who had received fully covered SEMS or PCSEMS with other uncovered lengths were excluded. Patients in whom the procedure was terminated before stent placement were also excluded. Written informed consent was obtained from all patients before SEMS placement.

The procedure time was defined as the time from the first liver puncture to completion of stent deployment. Technical success was defined as the successful placement of a stent extending from the intended intrahepatic branch (B2 or B3) to the stomach. Clinical success was defined as at least 50% improvement or normalization of the total bilirubin level without the need for further biliary intervention within 14 days after the procedure. Adverse events were defined based on the American Society of Gastrointestinal Endoscopy lexicon for endoscopic adverse events.9 Only early adverse events that occurred within 14 days from the procedure were considered for the analyses.

RBO was defined according to the TOKYO criteria 2014.10 However, only RBO involving the HGS route was included. Time to RBO was defined as the time between EUS-HGS and the next intervention involving the HGS route or intrahepatic branches drained by the HGS stent.

In our institution, we performed ERCP in the first session, where it was possible to achieve biliary drainage, remove any transpapillary plastic stents or SEMS, and place a nasobiliary tube in one of the left intrahepatic ducts (generally B3). After resolution of cholangitis or jaundice, EUS-HGS was performed in the second session by injecting saline from the nasobiliary tube to dilate the target intrahepatic duct, as required. EUS-HGS was occasionally preceded by percutaneous transhepatic biliary drainage in patients who were hemodynamically unstable and/or for whom ERCP was challenging. Although EUS-HGS was initially reserved for patients with end-stage disease, we gradually expanded its indications in line with our institutional learning curve.

All procedures were performed by expert endoscopists with >5 years of experience in therapeutic pancreatobiliary endoscopy. All EUS-HGS procedures were performed using a convex echoendoscope (GF-UCT260; Olympus Corp.). Reinterventions through the HGS route were generally performed using duodenoscopes with 4.2-mm working channels (TJF-260V or TJF-Q290V; Olympus Corp.).

From January 2016 to June 2019, we used the Niti-S Biliary S-type Stent (Niti-S S-type Stent; Taewoong Medical Co., Ltd.), a PCSEMS with a 5-mm uncovered portion on the liver side and a short, 2-mm flare on the stomach side (Fig. 1A). From July 2019 to June 2020, we used a modified EGIS Biliary Stent, Double Covered (EGIS Stent; Biotech Inc.) (Fig. 1B). The stent was initially designed with a short, uncovered portion on the liver side,11 but was modified via an extension to 20-mm to reduce the risk of migration towards the stomach during EUS-HGS. From July 2020 onward, we used the Niti-S S-type Spring Stopper Stent (Spring Stopper Stent; Taewoong Medical Co., Ltd.), which has an uncovered portion on the liver side and a flange on the stomach side to prevent migration from the stomach towards the peritoneum/liver (Fig. 1C).12 While stents with uncovered portions measuring 15 and 20 mm are available, only cases using the 20-mm type were included in this study. We placed the 5-mm uncovered portion of the Niti-S S-type Stent in the punctured branch (B2 or B3) immediately proximal to the B2/3 bifurcation, to avoid covering the bifurcation. However, we placed the 20-mm uncovered portion of the other two stents in the B2/3 branch (beyond the bifurcation) due to the reduced risk of side branch obstruction associated with the EGIS Stent and the Spring Stopper Stent.

Chemotherapy was initiated or resumed after EUS-HGS at the oncologist’s discretion. The response to chemotherapy was evaluated based on the Response Evaluation Criteria in Solid Tumors guidelines (ver. 1.1).13 OS was defined as the time from EUS-HGS to death from any cause or the last follow-up date. Progression-free survival (PFS) was defined as the time from EUS-HGS to death from any cause, disease progression confirmed by imaging studies, or the last follow-up date. Follow-up data were confirmed up to June 30, 2022.

Categorical variables are presented as absolute numbers (percentages), whereas continuous variables are presented as medians (range). Denominators of ratios were adjusted for missing data. Statistical analyses were conducted using Pearson’s chi-square test or Fisher exact test for categorical variables and the Mann-Whitney U test for continuous variables. Survival analysis using Kaplan-Meier curves and log-rank analyses were conducted to evaluate OS and time to RBO. Cox regression analyses were performed to evaluate factors affecting OS. Multivariate analysis was performed on variables with p<0.1 in univariate analysis. Statistical significance was set at p<0.05. All statistical analyses were conducted using IBM SPSS Statistics ver. 28.0 (IBM Corp.).

This study was approved by the Institutional Review Board of Cancer Institute Hospital of Japanese Foundation for Cancer Research (2021-GB-110). Patient consent for inclusion in the study was waived owing to its retrospective study design. The study was publicized on the hospital website and patients were permitted to opt out without any impact on their care.

Sixty-two patients underwent EUS-HGS during the study period. Thirty-eight patients had pancreatic cancer, 19 had biliary tract cancer, and the remaining five had stomach, liver, and duodenal cancer. Thirty-two patients received PCSEMS with 5-mm uncovered portions on the liver side (Niti-S S-type Stent: short uncovered group), while 30 received PCSEMS with 20-mm uncovered portions on the liver side (EGIS Stent: 9/Spring Stopper Stent: 21; long uncovered group).

The short uncovered group was more likely than the long uncovered group to have received previous chemotherapy (78.1% vs. 53.3%, p=0.039) and to have been undergoing chemotherapy at the time of EUS-HGS (68.8% vs. 40.0 %, p=0.023). No other baseline characteristics differed significantly between the two groups were observed (Table 1).

EUS-HGS was caused by duodenal tumor invasion in 35 patients (56.5%) and surgically-altered anatomy in 12 patients (19.4%). Other reasons included other causes of ERCP failure, repeated RBO after ERCP drainage, and inability to perform additional interventions through a previously placed SEMS in hilar malignant biliary obstruction, peribiliary abscess, and concurrent pancreatitis. EUS-HGS was performed after the resolution of cholangitis via endoscopic or percutaneous biliary drainage in 32 patients (51.6%). Twenty-nine patients underwent endoscopic nasobiliary drainage and three underwent percutaneous transhepatic biliary drainage prior to EUS-HGS during the same admission.

There were no significant procedural differences between the short uncovered and long uncovered groups, although patients in the former tended to be treated more commonly with 10-cm stents (75% vs. 53.3%, p=0.086) (Table 2). The technical success rate was 100% and clinical success was achieved in 51 patients (82.3%). Adverse events were observed in 11 patients (short/long uncovered groups: 6/5 patients), most of which resolved spontaneously. Two patients required admission to the intensive care unit (due to biliary hemorrhage [long uncovered group] and sepsis [short uncovered group]), and one patient died (due to bleeding and peritonitis secondary to complete stent migration towards the liver [short uncovered group]).

This fatal case involved an 80-year-old man with malignant biliary obstruction due to pancreatic head cancer. EUS-HGS was performed on the B2 intrahepatic branch. After tract dilation with a 6-mm balloon dilator, a 6 mm×10 cm Niti-S S-type Stent was successfully deployed (Supplementary Fig. 1A). Bleeding was observed during the session, but hemostasis was confirmed endoscopically (Supplementary Fig. 1B). However, the patient experienced cardiopulmonary arrest before dawn the following day. Plain computed tomography after successful resuscitation revealed complete stent migration, with the proximal end inside the peritoneal cavity, accompanied by peritoneal free air and new-onset ascites (Supplementary Fig. 1C). Blood was observed in the stomach; however, no active bleeding was confirmed, and the puncture site could not be visualized on emergency esophagogastroduodenoscopy (Supplementary Fig. 1D). The patient was unfit for surgery and died several hours later, despite multiple blood transfusions, and administration of catecholamines and antibiotics.

Stent migration (three cases, including the previously described fatal case, one patient with incomplete migration towards the liver requiring additional SEMS placement four days later and one patient with incomplete migration towards the stomach requiring the addition of a second SEMS during the same session) occurred only in the short uncovered group. While both groups exhibited similar postprocedural courses, more patients in the long uncovered group underwent chemotherapy (28.1% vs. 56.7%, p=0.029). Among those who received chemotherapy after EUS-HGS, the long uncovered group exhibited no association with longer median OS (5.0 vs. 5.5 months, p=0.319).

RBO was observed in 21 patients (9 [28.1%] in the short uncovered group vs. 12 [40.0%] in the long uncovered group, p=0.213) (Table 3). Mucosal hyperplasia of the uncovered portion occurred in six (18.8%) and seven patients (23.3%) in the short and long uncovered groups, respectively (p=0.628). The causes of RBO did not differ significantly between the two groups.

All patients with RBO were endoscopically treated with 100% clinical success. Most patients were treated with balloon cleaning (57.1%), followed by stent-in-stent placement (76.2%). Stent-in-stent placement was performed more frequently in the long uncovered group (55.6% vs. 91.7%, p=0.055). Plastic stents were placed more often in the long uncovered group (22.2% vs. 91.7%, p<0.001), whereas metal stents were placed more often in the short uncovered group (33.3% vs. 0%, p=0.031); multiple RBOs occurred in two (22.2%) and three (25.0%) patients (p=0.633), respectively, all of whom were also treated endoscopically.

RBO was generally observed in patients with less advanced disease who had lived long enough to experience it. Specifically, patients experiencing RBO were less likely than those not experiencing RBO to have peritoneal dissemination (4.8% vs. 41.5%, p=0.003) or to receive chemotherapy before EUS-HGS (42.9% vs. 78.0%, p=0.010), but were more likely to receive chemotherapy after EUS-HGS (66.7% vs. 29.3%, p=0.007) or to have longer median OS (7.3 vs. 2.4 months, p<0.001). The length of the uncovered portion had no effect on the RBO rate (long uncovered group: 57.1% vs. short uncovered group: 48.8%, p=0.423).

Forty-one patients received chemotherapy before EUS-HGS, of which 34 continued to receive chemotherapy up to less than 4 weeks before EUS-HGS. Of these, chemotherapy resumed in 14 patients after EUS-HGS, while the remaining 20 patients received the best supportive care. Twelve patients received chemotherapy for the first time after EUS-HGS, while nine never received chemotherapy before or after EUS-HGS; therefore, 26 patients (16 with pancreatic cancer, seven with biliary tract cancer, and three with other cancers) received chemotherapy after EUS-HGS (Supplementary Material 1, Supplementary Table 1).

Patients who received chemotherapy after EUS-HGS were more likely to be in the long uncovered group (17 patients [65.4%] vs. 13 patients [36.1%], p=0.023), to achieve clinical success (100% vs. 75.0%, p=0.008), to have shorter median lengths of stay after EUS-HGS (8 vs. 11 days, p=0.015), to subsequently experience RBO (57.7% vs. 25.0%, p=0.009), and to have longer median OS (6.1 vs. 2.7 months, p<0.001) than those of patients who did not resume chemotherapy (Supplementary Table 1). These patients also tended to have better Eastern Cooperative Oncology Group performance status (all had a performance status of 0 or 1; p=0.059), lower preprocedural total bilirubin (median 1.2 vs. 1.8 mg/dL, p=0.065), and lower likelihood to have a history of previous chemotherapy (53.8% vs. 75.0%, p=0.082). The median PFS of patients who had resumed chemotherapy was 3.4 months. Nine and three patients remained progression-free after 6 and 12 months, respectively.

The short uncovered group had significantly shorter median OS than that of the long uncovered group (3.1 vs. 4.9 months, p=0.037). Five patients in the short uncovered group and three in the long uncovered group died during the admission for the EUS-HGS procedure (p=0.509). However, all patients were on best supportive care before the procedure, and only one died due to an adverse event. The median time to RBO (6.8 vs. 7.1 months, p=0.760; Fig. 2), 3-month patency rate (82.4% vs. 84.2%, p=0.881), and 6-month patency rate (62.5% vs. 75.0%, p=0.551) did not differ significantly between the two groups.

According to Cox regression analysis, the predictors of longer OS included performance status of 0 or 1 (hazard ratio [HR], 0.33; p=0.032), absence of hepatic metastases (HR for the presence of hepatic metastases, 3.30; p<0.001), use of stents with long uncovered portions (HR, 0.58; p=0.040), and chemotherapy after EUS-HGS (HR, 0.36; p<0.001) (Table 4). While performance status, hepatic metastases, and chemotherapy after EUS-HGS remained significant in multivariate analysis (HR for presence of hepatic metastases: 3.63, 95% confidence interval [CI], 2.03–6.51; p<0.001; HR for performance status of 0 or 1, 0.22; 95% CI, 0.09–0.57; p=0.002; HR for chemotherapy after EUS-HGS, 0.34; 95% CI, 0.18–0.63; p<0.001), the use of stents with long uncovered portions was not an independent predictor of OS (HR, 0.64; 95% CI, 0.36–1.14; p=0.130).

Regarding the time to RBO, peritoneal dissemination was an independent predictor for a longer time to RBO (HR, 0.11; 95% CI, 0.01–0.78), although only one patient with peritoneal dissemination survived long enough to experience RBO (Supplementary Table 2). Hilar or diffuse biliary obstruction also tended to predict longer time to RBO, but this was not significant in multivariate analysis.

A review of eight studies (including ours) with at least 20 patients who had undergone EUS-HGS with PCSEMS with sufficient data revealed weighted average technical and clinical success rates of 98.6% and 88.5%, respectively (Table 5),4,12,14-19 which are consistent with data from a review of 27 studies (96% and 90%, respectively).20 Interestingly, no two studies used the same PCSEMS or same uncovered length, highlighting the lack of consensus on the optimal uncovered length. Early adverse events were observed in a weighted average of 19.4% of the cases, but most resolved with conservative treatment alone. Our RBO rate of 33.9% was comparable to the weighted average of 33.4%, although the different follow-up periods across studies preclude a direct comparison. Only a few studies provided data on the median time to RBO (weighted average, 5.9 months) and median OS (weighted average, 3.7 months).

In our study, early adverse events occurred in 17.7% of the cases, with no significant difference between groups. However, all three cases of stent migration occurred in the short uncovered group. The case of stent migration towards the stomach triggered our transition to PCSEMS with longer uncovered portions, which serve as anchors and permit deeper stent insertion with less concern about side branch obstruction. In addition, the anti-migration flange of the Spring Stopper Stent may have reduced the risk of migration towards the liver.12 The two groups did not differ in terms of the distance between the stomach and liver, which has been reported as a risk factor for stent migration.21

One case of incomplete stent migration towards the stomach occurred in the short uncovered group. A systematic review of 1,192 EUS-guided biliary drainage procedures revealed stent migration in 2.68% of cases, although the study was not limited to EUS-HGS cases.22 Our review of all 42 studies included in the systematic review revealed seven cases of outward stent migration towards the stomach, one of which was fatal.23-29 Although it is rarer than stent migration towards the liver, care is required to avoid this potentially catastrophic event.

Consistent with past reports, the majority of RBO cases resulted from hyperplasia of the uncovered portion.8 The length of the uncovered portion did not significantly affect the likelihood of RBO due to hyperplasia in our study. Although differences in patient backgrounds preclude direct comparisons, PCSEMS with various uncovered lengths had similar RBO rates (Table 5), and longer uncovered portions may not necessarily be associated with increased risk of hyperplasia. Reintervention was successful in all cases through the HGS route, requiring stent-in-stent placement in 75% of the cases and only balloon cleaning or dilatation in the remaining cases. Plastic stents were primarily used in stent-in-stent placement, particularly when biliary access was only possible through the mesh of the EUS-HGS stent. This was especially the case in the long uncovered group, in which 91.7% of the RBO cases were treated with stent-in-stent placement of plastic stents. This difference between the groups reflects our increased understanding over time that hyperplasia eventually develops in the uncovered portion, even when it is not the primary cause of RBO. Thus, when we transitioned to the EGIS Stent, stent-in-stent placement of plastic stents became our main strategy for RBO.

The effect of chemotherapy on stent patency and OS in patients undergoing EUS-HGS has not been studied in depth. One study with 51 participants found that patients on chemotherapy before EUS-HGS had longer patency than those who did not undergo chemotherapy, with a HR of 3.02 (95% CI, 1.45–6.30).20 However, the study did not clarify the number of patients who were able to resume chemotherapy after EUS-HGS. Another study with 21 participants reported that only one received chemotherapy (concurrent chemoradiotherapy) after EUS-HGS.15 We found that longer uncovered portion, clinical success, and shorter length of stay after EUS-HGS were associated with the ability to commence or resume chemotherapy after EUS-HGS. Median PFS was mediocre, at 3.4 months. Resuming chemotherapy after EUS-HGS was also an independent predictor for longer OS and, therefore, led to higher likelihood of survival until RBO.

In addition to chemotherapy after EUS-HGS, good performance status and absence of hepatic metastases were independent predictors of longer OS in patients who had undergone EUS-HGS. Hepatic metastases not only signify advanced disease, but may also limit the available EUS-HGS puncture routes or cause segmental biliary obstruction. The use of stents with 20-mm uncovered portions was associated with longer OS, although this association was not significant based on multivariate analysis. No notable factors predicting the time to RBO were identified in this study, as one significant factor in the multivariate Cox analysis reflected a single patient with peritoneal dissemination who happened to enjoy long stent patency.

This study had several limitations. This was a retrospective study conducted at a single institution with a limited sample size, making selection bias inevitable. Patients who had received two different types of PCSEMS were included in the long uncovered group. As the choice of PCSEMS was dictated by the period during which EUS-HGS was conducted, institutional learning curves and advances in chemotherapy or supportive treatment might have worked in favor of the long uncovered group. Characteristics of SEMS other than the length of the uncovered portion, such as proximal flares and flanges, wires used, stent type (e.g., braided or laser-cut), and membranes for the covered portion, which were not analyzed in this study, might also have affected stent migration and other causes of RBO. The stents used in this study were not available in some regions, limiting the generalizability of our results.

In conclusions, PCSEMSs with 20-mm uncovered portions on the liver side performed at least as well as those with 5-mm uncovered portions in all material respects. RBO due to hyperplasia and the time to RBO did not increase, when the uncovered portion was extended. The ability to reduce potentially fatal stent migration to the stomach may be an advantage of PCSEMSs with longer uncovered portions. Additional studies are warranted to determine the optimal uncovered length and to investigate whether certain patient subgroups may benefit from longer or shorter uncovered lengths.