Study design and population

This retrospective cohort study included consecutive adult patients (≥18 y) admitted with refractory or early recurrent acute EVB who underwent SEMS placement after failure of standard endoscopic therapy. The patients were treated at a tertiary referral center between January 2011 and December 2022, and all available cases meeting the inclusion criteria were analyzed. The study was conducted in accordance with the Declaration of Helsinki and approved by the institutional Ethics Committee at the Medical University of Warsaw (AKBE/238/2026). Due to the retrospective nature of the study, the requirement for obtaining individual informed consent was waived. A total of 64 consecutive eligible individuals were identified and comprised the analytic cohort.

The patients were considered eligible if they had: 1) endoscopically confirmed esophageal varices associated with portal hypertension; 2) failure of standard therapy, defined by ongoing hematemesis, aspiration of over 100 ml of fresh blood via a nasogastric tube more than 2 hours after endoscopy, or a drop in hemoglobin level equal to or greater than 3 g/dl without transfusion; and 3) SEMS placement for control of acute refractory or early recurrent variceal bleeding. Individuals with nonvariceal bleeding or isolated gastric varices without esophageal involvement were excluded. Data completeness varied for some laboratory variables, but it was not a reason for exclusion of any patient from the study; survival analyses included the full cohort. The missing values were mitigated by variable‑specific complete‑case analysis rather than excluding cases from the examined cohort.

All patients received standardized initial therapy, including intravenous vasoactive drugs (terlipressin or somatostatin analogues), antibiotic prophylaxis (ceftriaxone 1 g/day), and urgent endoscopic evaluation with variceal ligation performed within 12 hours of admission following hemodynamic stabilization.

Rescue and definitive therapies

Outcome measures

Primary outcomes were all‑cause mortality within 5 days, 6 weeks, and 1 year of SEMS placement, bleeding‑related mortality within the same time windows, and rebleeding after SEMS placement. Short‑term mortality was defined as death within 5 days and within 6 weeks of the index bleeding episode, whereas long‑term mortality was defined as death within 1 year. Secondary outcomes included technical success of SEMS placement and TIPS, feasibility of definitive therapies, and procedure‑related complications, categorized as stent migration, postremoval ulceration or necrosis, SEMS‑related ulcer bleeding, and post‑TIPS rebleeding.

Statistical analysis

All statistical analyses were performed using R statistical software, version 4.3.2 (R Foundation for Statistical Computing). Continuous variables were assessed for distribution using the Shapiro–Wilk test, and are presented as mean with SD or median with interquartile range (IQR), as appropriate. Categorical variables are presented as counts and percentages. Survival was estimated using the Kaplan–Meier methods and compared with the log‑rank test. Comparisons between the SEMS alone and SEMS + TIPS groups were considered descriptive, because treatment allocation was nonrandom and TIPS represented a postbaseline sequential intervention.

Univariable logistic regression analyses were initially performed to evaluate associations between individual variables and study outcomes. Exploratory multivariable logistic regression was then used to examine baseline predictors of 6‑week and 1‑year mortality and rebleeding. Before the multivariable model construction, potential collinearity between candidate predictors was assessed using the Spearman rank correlation coefficient, and highly correlated variables were not entered simultaneously into the same model. The variables were selected based on clinical relevance and univariable analysis results. A backward elimination approach was applied, with sequential removal of nonsignificant variables to obtain the final model.

Model performance was assessed using receiver operating characteristic curve analysis, and discriminative ability was expressed as the area under the curve. Results of logistic regression analyses are reported as odds ratios (ORs) with 95% CIs and corresponding P values. Cox proportional hazards models were not fitted, and TIPS was not modeled as a time‑dependent exposure. A 2‑sided P value below 0.05 was considered significant.

Baseline characteristics

A total of 64 consecutive patients with recurrent acute EVB refractory to standard therapy were analyzed. Mean (SD) age of the participants was 46.8 (12) years, and 59.4% of them were men. Alcohol‑related liver disease was the main cause of portal hypertension (53.1%), followed by metabolic (14.1%) and viral cirrhosis (9.4%). Advanced liver disease prevailed: 61.7% of the patients were graded as Child–Pugh class C, with a median (IQR) Model for End Stage Liver Disease (MELD) score of 24 (14–30) points. Large esophageal varices (F3) were present in 71.9%, and portal vein thrombosis in 16.4% of the participants (Table 1). Laboratory parameters confirmed severe hepatic dysfunction, with median bilirubin concentration of 5.9 mg/dl, international normalized ratio (INR) of 1.7, albumin level of 2.5 g/dl, and platelet count of 72.5 × 10⁹/l (Table 1).

Outcomes of self‑expanding metal stent placement

SEMS implantation was performed in 64 patients as rescue therapy after failed pharmacologic and endoscopic treatment. Technical success rate was 96.9%, with immediate hemostasis achieved in 93.8%. Rebleeding occurred in 61.4%, and failure to control bleeding within 5 days in 34.4% of the study cohort. Stent migration occurred in 43.6% of the patients, with complete migration accounting for nearly two‑thirds of these events. Among the individuals undergoing SEMS removal, median dwell time was 12 (9–14) days. Postremoval mucosal ulceration or necrosis occurred in 29.4%, and SEMS‑related ulcer bleeding in 17.6% of the patients (Table 1).

Feasibility and outcomes of transjugular intrahepatic portosystemic shunt

When retrospectively assessed according to the Baveno VII criteria,¹ 55 patients met the indications for TIPS; however, due to contraindications, including severe hepatic failure, advanced encephalopathy, portal vein thrombosis, and active sepsis, only 13 (23.6%) underwent the procedure. All TIPS procedures were technically successful, and rebleeding after TIPS occurred in 25% of the cases (Table 1).

Comparison of outcomes between self‑expanding metal stent + transjugular intrahepatic portosystemic shunt and self‑expanding metal stent alone

Nearly half of the patients (46.9%) died with a SEMS in situ, which likely reflected the severity of underlying liver disease rather than procedural failure. In the patients treated with SEMS alone, bleeding‑related mortality accounted for 22.2% of the deaths within 5 days and 64.7% of those occurring within 6 weeks (Table 1). In descriptive comparisons, the patients who subsequently underwent TIPS had lower 6‑week mortality (46.2% vs 66.7%) and 1‑year mortality (50% vs 80.4%) than the patients managed with SEMS placement alone; early bleeding‑related death (≤5 days) was also less frequent (8.3% vs 25.5%; Supplementary material, Table S1). Given the nonrandom allocation to TIPS and the variable timing of the procedure, these intergroup differences should be interpreted with caution.

Liver transplant

A total of 29 patients were listed for LT at the time of SEMS placement. After expedited evaluation, 21 were found eligible; however, only 10 underwent the procedure, while the remainder died awaiting graft availability. LT markedly improved survival, eliminating bleeding‑related mortality and significantly reducing 6‑week and 1‑year mortality, as compared with the nontransplanted patients (Supplementary material, Table S2).

Survival analysis

Kaplan–Meier survival analysis was performed in the entire cohort stratified by treatment strategy: SEMS alone (n = 51) vs SEMS + TIPS (n = 13). No patients were excluded because of incomplete time‑to‑event data (Supplementary material, Figure S1). The curves are presented as unadjusted survival estimates. The patients treated with SEMS alone had poor outcomes, with median overall survival of 14 days and 6-, 12-, and 24‑month survival rates of 29.4%, 27.5%, and 25.5%, respectively (38 deaths within 24 months). In contrast, 24‑month survival in the SEMS + TIPS subgroup was 53.8%, with 6 deaths. Because of the small size of the TIPS subgroup and nonrandom treatment allocation, these estimates should be interpreted cautiously. The Kaplan–Meier curves showed early and sustained separation favoring SEMS + TIPS (log‑rank P = 0.04). At 400 days, estimated survival was 25.5% (95% CI, 13.5–37.5) for SEMS alone and 53.8% (95% CI, 26.7–80.9) for SEMS + TIPS (Supplementary material, Figure S1).

Exploratory regression analysis

In exploratory multivariable logistic regression analysis, INR independently predicted 6‑week mortality (OR, 14.3; 95% CI, 3.13–115; Supplementary material, Table S3). One‑year mortality was associated with MELD score and INR, whereas no independent predictors of rebleeding were identified. These models address fixed follow‑up windows rather than time‑to‑event risk, and should therefore be interpreted as supportive analyses.

Limitations

First, our findings apply only to endoscopically confirmed refractory EVB and should not be extrapolated to nonvariceal causes of gastrointestinal bleeding. Rare tumor‑related etiologies, including gastrointestinal bleeding associated with intussusception, may present with overlapping clinical symptoms but require a distinct diagnostic and therapeutic approach.²¹

Most patients were treated before publication of the Baveno VII consensus, which endorsed a more individualized approach to TIPS despite relative contraindications; thus, the low TIPS rate in our cohort likely reflects historical practice patterns. This was a retrospective, single‑center study, and only a small subgroup received TIPS. Because treatment allocation was nonrandom, comparisons between the SEMS alone and SEMS + TIPS groups remained vulnerable to confounding by indication, selection bias, and immortal time bias. In addition, mortality was analyzed with logistic regression for fixed follow‑up windows rather than Cox proportional hazards models, and TIPS was not modeled as a time‑dependent exposure. These issues substantially limit statistical precision, subgroup robustness, generalizability, and causal interpretation.

Conclusions

SEMS placement achieved effective temporary hemostasis in individuals with refractory acute EVB, whereas overall prognosis remained closely linked to the severity of underlying liver disease. In this real‑world cohort, the patients who subsequently underwent TIPS or LT had longer observed survival, which supports the clinical role of SEMS in a sequential rescue strategy. These findings reinforce the value of SEMS placement as a bridging intervention in the management of refractory EVB, while the observed survival differences should be interpreted cautiously given the retrospective and nonrandomized design of the study.