Acute upper gastrointestinal bleeding caused by a peptic ulcer is a frequent and life-threatening condition [1]. The gold standard treatment is endoscopic intervention [2]. Rebleeding following treatment for a peptic ulcer bleeding increases mortality significantly [2]. Thus, it is of utmost importance to reduce the risk of rebleeding.

Despite achieving hemostasis endoscopically, 12–25% of patients experience rebleeding, depending on several factors (e.g. ulcer type, patient comorbidities, etc.) [3, 4]. Following successful endoscopic hemostasis, surgeons can consider prophylactic transarterial embolization to reduce the risk of rebleeding. Transarterial embolization decreases both the rate of rebleeding, the need for reintervention, and mortality [5]. Further, transarterial embolization has a high technical success rate, a low complication rate, and does not prolong in-hospital stay significantly [5].

In Denmark, patients do not undergo prophylactic embolization routinely. Thus, it is the surgeon performing the endoscopy, who assesses the need for prophylactic embolization. Among other factors, this evaluation is based on the Rockall score of the patient, and the Forrest classification of the ulcer (Table 1) [6, 7]. Two studies have investigated the effect of prophylactic transarterial embolization following peptic ulcer bleeding in a randomized setup [8, 9]. However, both studies performed the prophylactic arterial embolization after just one endoscopy. In our population, patients were evaluated by surgeons based on the Rockall-score and might have undergone more than one endoscopy [10].

Table 1 Rockall Score and Forrest Classification
Full size table

The aim of this study was to retrospectively analyze patients who have undergone prophylactic transarterial embolization at Rigshospitalet, Denmark, following a peptic ulcer bleeding. We aimed to evaluate the risk of rebleeding and 30-day mortality following embolization, and identify risk factors for rebleeding and mortality.

Materials and methods

The study was a retrospective, single-center cohort study. The study was approved by the local ethics committee of the Capital Region of Denmark (R-22007493). The study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies [11].

The study was conducted at Rigshospitalet, Copenhagen, Denmark. Rigshospitalet is a referral center for Zealand, with a population of approximately 2.6 million inhabitants. Patients with peptic ulcer bleeding are typically treated endoscopically at the local and regional hospitals, and referred to Rigshospitalet for embolization. Data collection was performed between September 21 and December 28, 2022. We included all patients who had undergone attempted prophylactic transarterial embolization at Rigshospitalet between January 1, 2016, and December 31, 2021. Patients were followed up until December 28, 2022. For patients to be included into the study, they had to have undergone attempted transarterial embolization following a peptic ulcer bleeding, within the same hospital admission. Patients were excluded if they had incomplete medical records. Patients were identified through a local registry at the Department of Radiology, Rigshospitalet. Data were collected through electronic health records as well as radiological imaging software. Follow-up was also performed through the electronic health records.

The following variables were extracted from the patient history in the electronic health record: age, sex, comorbidities, smoking and alcohol history, plasma hemoglobin prior to upper endoscopy, weight, height, use of non-steroidal anti-inflammatory drugs and anti-coagulants, blood pressure and pulse prior to upper endoscopy (pulse < 100, systolic blood pressure > 100; pulse > 100, systolic blood pressure > 100; systolic blood pressure < 100), American Society of Anaesthesiologists score [12], time of last upper endoscopy prior to embolization, placement of ulcer, size of ulcer, endoscopic method of bleeding control, if bleed control was achieved (yes/no), Forrest classification [7], diagnosis as per the Rockall-score [6], Helicobacter pylori status, number of upper endoscopies prior to embolization, rebleed after embolization, date of rebleed, need for reintervention, type of reintervention, death at time of follow-up, date of death, cause of death, discharge date from hospital, readmission related to bleeding ulcer, and date of readmission. The following variables were extracted from the digital subtraction angiography (DSA) by a trained interventional radiologist: time of embolization, active bleeding on digital subtraction angiography, visible hemoclips during digital subtraction angiography, type of embolization material, standard embolization (yes/no), any anatomic variants on digital subtraction angiography, visualization of cystic artery (yes/no), embolization of cystic artery, non-target embolization (yes/no). We defined a standard embolization as coiling of the gastroduodenal artery and side branches, as well as any actively bleeding sites (Fig. 1), as this is the standard embolization technique employed at our institution. Consequently, non-standard embolization includes all embolizations not following the above, e.g., attempted embolizations with technical failure, anatomical variants making a standard approach impossible, different embolization materials. All data were entered into a Research Electronic Data Capture (REDCap) database.

Fig. 1
figure 1

Overview of arteries involved in prophylactic transarterial embolization with embolization of the gastroduodenal artery and side branches. AA abdominal aorta, CeT celiac trunk, LGA left gastric artery, SA splenic artery, CHA common hepatic artery, RGA right gastric artery, PHA proper hepatic artery, LHA left hepatic artery, RHA right hepatic artery, CA cystic artery, GDA gastroduodenal artery, SPDA superior pancreaticoduodenal artery, RGEA right gastroepiploic artery, SMA superior mesenteric artery, IPDA inferior pancreacitoduodenal artery

Full size image

Normally distributed data were reported as mean (SD) and non-normally distributed data were reported as median (range). The Rockall-score was grouped (≤ 3, 4–5, ≥ 6) based on the Danish Society of Gastroenterology and Hepatology guidelines [10]. Both primary outcomes (rebleeding and 30-day overall mortality) were dichotomous outcomes. To investigate the association between rebleeding and 30-day overall mortality, we performed a χ2-test. If there was an association, we then calculated the odds ratio to quantify the association. Finally, we performed a binomial logistical regression, evaluating the effect of several variables on the likelihood of rebleeding and 30-day mortality. The variables included in the model were: active bleeding during transarterial angiography, visible hemoclips following endoscopy during transarterial angiography, Rockall score (grouped), normal anatomy, standard embolization procedure, and number of endoscopies performed prior to embolization (grouped into either one endoscopy or more than one endoscopies). Results with p < 0.05 were considered statistically significant.

Results

We identified a total of 302 transarterial embolizations performed at Rigshospitalet from 2016 to 2021. Of these, 277 were unique patients. We excluded 101 patients for the following reasons: 12 were missing medical records, 85 had other diagnoses than peptic ulcer bleeding, and 4 had not had an upper endoscopy prior to transarterial embolization. In total, 176 patients were included in the study, see Fig. 2. All embolizations were performed by one of seven experienced interventional radiologists at Rigshospitalet. For descriptive and outcome data, see Table 2.

Fig. 2
figure 2

Flowchart of cohort selection. ERCP endoscopic retrograde cholangiopancreatography

Full size image
Table 2 Characteristics and outcomes of patients included
Full size table

Of the 26 patients who died within 30 days of embolization, 15 (58%) had experienced a rebleeding following embolization. The χ2-test of independence showed an increased likelihood of 30-day mortality in patients who had suffered a rebleeding following transarterial embolization (p < 0.001), and the odds ratio of 30-day mortality in patients who had suffered a rebleeding following transarterial embolization was 5.6897 (95% confidence interval 2.3665–13.6795).

None of the included patients had significant complications directly related to the transarterial embolization, i.e., none suffered from ischemia or perforation, and no complications arose in the patients who had non-target embolization or coiling of the cystic artery.

The rebleeding binomial logistic regression model was statistically significant, χ2 (6) = 19.207, p = 0.004. The model explained 15.3% (Nagelkerke R2) of the variance in rebleeding and correctly classified 76.7% of cases. Sensitivity was 18.2%, specificity was 96.2%, positive predictive value was 61.5% and negative predictive value was 77.9%. Of the six predictor variables (active bleeding, visible hemoclips, Rockall score, normal anatomy, standard embolization procedure, and number of endoscopies prior to embolization) two were statistically significant, standard embolization procedure and number of endoscopies prior to embolization, see Table 3. Patients who did not undergo a standard embolization procedure had 3.298 (1.484–7.329) higher odds (95% confidence interval) of rebleeding compared with patients who underwent a standard embolization procedure. Patients who had more than one endoscopy prior to embolization had 2.369 (1.088–5.158) higher odds (95% confidence interval) of rebleeding compared with patients who only had one endoscopy.

Table 3 Logistic regression predicting likelyhood of rebleeding based on active bleeding, visible haemoclips, Rockall score, normal anatomy and standard embolization procedure
Full size table

The 30-day overall mortality binomial logistic regression model was not statistically significant, χ2 (6) = 12.082, p = 0.06. However, the Hosmer and Lemeshow goodness of fit test was not significant (p = 0.379), indicating that the model is not a poor fit, for predicting categorical outcomes. The model explained 11.7% (Nagelkerke R2) of the variance in 30-day overall mortality and correctly classified 85.8% of cases. Sensitivity was 3.8%, specificity was 100%, positive predictive value was 100% and negative predictive value was 85.7%. Of the six predictor variables (active bleeding, visible hemoclips, Rockall score, normal anatomy, standard embolization procedure, and number of endoscopies prior to embolization) two were statistically significant, Rockall score and standard embolization procedure, see Table 4. Patients who underwent a non-standard embolization procedure had 3.340 (1.273–8.765) higher odds (95% confidence interval) of 30-day mortality compared with patients who underwent a standard embolization procedure. Further, a higher Rockall score was associated with an increased likelihood of 30-day mortality.

Table 4 Logistic regression predicting likelyhood of 30-day mortality based on active bleeding, visible haemoclips, Rockall score, normal anatomy and standard embolization procedure
Full size table

Finally, we performed a qualitative analysis of embolizations defined as non-standard. In Table 5, reasons for them being defined as non-standard are listed, and the procedures were divided according to if the patient experienced rebleeding following the non-standard embolization or not.

Table 5 Reasons for embolizations to be considered non-standard, divided into patients who experienced rebleeding following embolization
Full size table

Discussion

This retrospective cohort study included 178 patients who underwent prophylactic transarterial embolization following endoscopic treatment for peptic ulcer bleeding. Following transarterial embolization, 25% of patients suffered from rebleeding and overall 30-day mortality was 14%. The risk of 30-day mortality was significantly increased in patients who suffered from rebleeding. Not undergoing a standard treatment, embolization of the gastroduodenal artery and side branches as well as actively bleeding sites (if any), increased the odds of rebleeding by 3.298, and the odds of dying within 30 days of the procedure by 3.340. Having undergone more than one endoscopy prior to embolization increased the odds of rebleeding by 2.369, and a high Rockall-score at the time of the last endoscopy prior to transarterial embolization was associated with a high likelihood of overall 30-day mortality.

A strength of this study is the near complete follow-up of all patients due to electronic health records and the unique Danish patient identification number system. Furthermore, the authors had access to both surgical reports and digital subtraction angiographies, so procedures could be reevaluated. This study was limited by the nature of a retrospective design, by having no control group, and possible selection bias in the patients who were lost to follow-up due to incomplete medical records.

To our knowledge, there is no definitive consensus on which vessels to embolize, and we found no firm guidelines dictating if the gastroduodenal artery always should be embolized, even when a visible actively bleeding vessel is identified during the digital subtraction angiography [5, 13, 14].

Compared with previous studies, we had a high rate of rebleeding, similar mortality rate, and similar adverse event rate [5, 8, 9, 14, 15]. Most patients in our study had undergone more than one endoscopy prior to embolization, which was not the case in most comparable studies. The Rockall scores of our patients were comparable with another retrospective study [15], but our rate of rebleeding was higher. Compared with the randomized clinical trial, also performed in Denmark, most patients in the current study had undergone more than one endoscopy, their clinical condition was worse, and they had more comorbidities [8]. Overall, we attribute the higher rate of rebleeding in our cohort to the worse overall clinical condition of our patients, possibly due to delay of embolization as patients had undergone more endoscopies prior to embolization. As in other studies, we found no adverse events specifically related to the transarterial embolization procedure in our study.

In general, characeristics of patients, who undergo prophylactic transarterial embolization following peptic ulcer bleeding, might vary extensively between different countries. This to a high extent depends on the accessibility of endovascular treatment. Naturally, this limits the external validity of our study.

In conclusion, our study in patients with high-risk peptic ulcer showed a high risk of rebleeding (25%) and mortality (15%) following prophylactic transarterial embolization, with higher likelihood of mortality in patients with rebleeding. We found that patients who had undergone a non-standard prophylactic transarterial embolization had a higher risk of rebleeding and 30-day mortality, patients who had undergone more than one endoscopy prior to embolization had higher odds of rebleeding, and patients with a high Rockall-score had a higher risk of 30-day mortality. However, comparison between our study population and the populations of previous trials was difficult, warranting further research to identify patients who benefit from prophylactic transarterial embolization, and to identify patients at high risk of rebleeding and mortality.