Predictors and management of post‐banding ulcer bleeding in cirrhosis: A systematic review and meta‐analysis

Post‐banding ulcer bleeding (PBUB) is an understudied complication of oesophageal varices endoscopic band ligation (EBL). This systematic review with meta‐analysis aimed at: (a) evaluating the incidence of PBUB in patients with cirrhosis treated with EBL in primary or secondary prophylaxis or urgent treatment for acute variceal bleeding and (b) identifying predictors of PBUB.


| INTRODUC TI ON
Cirrhosis is the fifth leading cause of adult deaths worldwide.
Portal hypertension (increased pressure gradient between the portal vein and the inferior vena cava) is the main complication of cirrhosis and leads to the formation of porto-systemic collaterals, including gastroesophageal varices (GEV) that are present in over 50% of patients with cirrhosis at diagnosis. 1 GEV increase in size over time, and can rupture, leading to gastrointestinal haemorrhage. Variceal haemorrhage (VH) constitutes a major decompensating event in cirrhosis, and despite the advances in its treatment, it is still associated with significant mortality (15%-20% at 6 weeks). 1-3 Oesophageal varices endoscopic band ligation (EBL) is an endoscopic procedure aimed at eradicating varices by ligating them with rubber rings (bands). According to current international guidelines, EBL of oesophageal varices plays an important therapeutic role in three settings: (a) the prevention of a first VH as an alternative to non-selective beta-blockers (NSBB) in patients with contraindications or who cannot tolerate these drugs 1,4 ; (b) to achieve haemostasis in combination with vasoactive drugs (somatostatin, octreotide or terlipressin) in patients with acute VH; and (c) to prevent recurrent bleeding. In the latter case patients are treated both with NSBB and EBL. EBL is considered safe, but is not free of complications like chest pain, dysphagia, fever, and development of post-banding ulcers (PBU) that may lead to bleeding (PBUB). 5 After banding, when the ligation bands drop off, a superficial ulcer is formed that may bleed, which can be difficult to distinguish from VH due to portal hypertension. In these patients, bleeding is likely triggered by the EBL-induced ulcer that may damage the underlying mucosal vessels or varices. 5,6 PBUB induce significant mortality and morbidity in cirrhosis. 5,7,8 The aims of this systematic review were to summarize the existing data regarding the incidence of PBUB, and to identify predictors of PBUB in patients with cirrhosis. Additionally, we described the current strategies used to manage bleeding after PBUB.

| Protocol and registration
The review follows the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) recommendations. 9 Ethical approval was not sought because of the study design. The PROSPERO registered protocol number is CRD42022353449.

| Data extraction and quality assessment
From the eligible articles the reviewers extracted the following information: study characteristics (author, year, recruitment period, sample size); demographic characteristics of the population (age, sex); clinical characteristics of the population (cirrhosis aetiology, Child-Pugh score and MELD score); EBL and PBUB characteristics (setting of endoscopy, incidence of PBUB, interval ligation and risks factors for PBUB, management, follow-up duration and mortality associated with PBUB).

Key points
• MELD score and emergency EBL are the strongest predictors of post-banding ulcer bleeding (PBUB) in cirrhosis.
• The pooled post-PBUB mortality in the studies included was 22.3%.
• The strategies to prevent PBUB and the best treatment approach need to be prospectively investigated.

| Statistical analysis
Meta-analysis was performed to determine the incidence of PBUB in cirrhotic patients, overall and comparing urgent EBL treatment setting versus primary and secondary prophylaxis (elective EBL). The software Comprehensive meta-analysis (CMA), version 3 was used.
The odds ratio (OR) with 95% confidence interval (CI) is the measure of association used in this meta-analysis for the risk's factors of PBUB. Rates with 95% confidence interval were used for the incidence and mortality associated with PBUB.
We selected a random-effects model because of the differences between the studies (different population and setting of endoscopies). For the assessment of heterogeneity, clinical heterogeneity such as characteristics of population, losses to follow-up, outcomes and different management/treatments were considered.
Statistical heterogeneity was tested using the I 2 test. A I 2 ≥ 50% or p < .1 was suggestive of considerable heterogeneity among the studies. Publication bias was measured using Egger's regression.
3.2.1 | Incidence of post-banding ulcer bleeding with endoscopic band ligation in urgent treatment setting versus endoscopic band ligation in primary and secondary prophylaxis In a subgroup analysis of seven studies ( Figure 3), a total of 7527 patients with 515 bleeders from PBUBs, PBUB was more frequent in those having urgent EBL (OR 2.360, 95% CI 1.130-5.010, p < .001) compared with the primary and secondary prophylactic group ligation, with significant heterogeneity (I 2 87.6%, p < .01). The pooled incidence rate of PBUB among patients treated with urgent EBL was 8.3% (95% CI 6.4-10.7), with considerable heterogeneity (I 2 56%, p < .001). The pooled incidence rate of PBUB in the group with prophylactic EBL was 3.9% (95% CI 1. 8-8.4), with important heterogeneity (I 2 92%, p < .001). There is a significant difference of 4.5% between the two groups (p < .0001). One study 13 was the exception presenting a higher incidence of PBUB after prophylactic EBL compared with urgent EBL. This difference might also be related to the differences in sample size and in the characteristics of the population, since in this study patients treated with prophylactic EBL had a higher MELD score (mean MELD score 18.6 ± 5.1) compared with the other studies only with prophylactic EBL treatment, that present a mean MELD score below 14. 4,13,21,23 One study 5 was not included F I G U R E 2 Meta-analysis of the incidence of post-banding ulcer bleeding in patients treated with EBL for primary, secondary prophylaxis and urgent treatment for acute variceal bleeding. Retrospective studies are represented in black, prospective studies in grey. Rates are shown with 95% of confidence interval.
due to the lack of data about the number of patients undergoing to EBL in the elective versus urgent situations.

| Mean interval of post-banding ulcer bleeding occurrence
The pooled random effect describes a value for the time PBUB's occurrence in six studies of 11 days (95% CI 9.9-12) without heterogeneity (I 2 .000, p < .001) ( Figure S3). The rest of studies do not give mean and standard deviation (SD) values and thus could not be incorporated into the meta-analysis.
Articles showing a high Child-Pugh score as predictor of PBUB did not provide enough data to perform a meta-analysis (Table S6). 3.4.1 | High model for end-stage liver disease score A higher MELD score was a predictor of PBUB in six studies ( Figure S4), with a pooled random effect odds ratio (OR) of 1.162 (95% CI 1.047-1.291, p < .001). There was significant heterogeneity (I 2 67.752; with p = .015). The mean MELD score of patients with PBUB was 18 or more in three studies. 5,8,14 Two studies had an inferior MELD score of 13 12 and 10 15 in the group of patients with PBUB, but higher than the non-bleeding group in these studies. One study 16 was not included in this meta-analysis due to lack of sufficient data.

F I G U R E 3
Meta-analysis comparing the incidence of post-banding ulcer bleeding between patients with emergency EBL versus elective EBL. Odds ratios are shown with 95% of confidence interval.

| Ulcers with high-risk stigmata on endoscopy
Three studies 13,14,18 identified the endoscopic aspect of ulcers with high-risk stigmata (red wale sign, cherry spot, white nipple sign, platelet plug) as a predictor of PBUB. However, the pooled OR describes no significant effect of this factor on PBUB (OR 3.872, 95%, CI .794-18.881, p < .001). The studies showed a considerable heterogeneity (I 2 80.466; p = .06) ( Figure S6).

| Management of post-banding ulcer bleeding
Due to the considerable heterogeneity of included studies, a statistical analysis of the treatment of PBUB was not possible. Data were extracted and are described into the Table S7. PPI, sucralfate and antibiotics were used as prophylactic treatment in most studies.
Considering the treatment in case of acute PBUB, most studies used the vasoactive drug therapy used for variceal bleeding (somatostatin, octreotide or terlipressin) as well as intravenous PPI. More variability exists across the studies regarding the endoscopic type of treatment applied, with Re-EBL being used to treat PBUB in eight studies. 5,12,[14][15][16]19,23,24 Other types of endoscopic therapies were used as well (argon plasma coagulation, oesophageal variceal obliteration, like sclerotherapy and cyanoacrylate injection, epinephrine injection, hemoclip). In case of refractory bleeding, oesophageal selfexpandable metallic stent (SEMS) was employed in two studies, 14,22 and oesophageal balloon tamponade in six studies. 5,7,14,15,21,22 TIPS was used in severe bleeding in five studies 5,7,14,16,24 but the number of patients treated were reported only in three studies: one study 7 reported two patients receiving an emergency TIPS; in the second study, 16 TIPS was used successfully as initial treatment in two patients, and as a rescue treatment in four patients, and was successful in 3; in the third study 14 TIPS was performed in 19 patients, of whom 9 survived and 8 died despite TIPS. In one study, 14 two patients were transplanted.

| Mortality rate associated with post-banding ulcer bleeding
Fourteen studies reported the mortality associated with PBUB. The Although it is conceivable that not only the incidence of PBUB, but also the mortality was higher in emergency setting, this could not be inferred from this systematic review, as most studies did not provide separate mortality data in emergency versus elective EBL complicated by PBUB. In the study by Sinclair et al. (2015), 8 which reported the mortality in elective versus emergency procedures, mortality due to PBUB was higher in the emergency setting. Five deaths following PBUB were reported: one after primary prophylaxis, one post-secondary prophylaxis and three after band ligation in the setting of acute variceal bleeding.
Regardless of the high mortality, there is no standard-of-care treatment for PBUB. The current treatment remains empirical and based on the expertise of the individual centres. Despite there was no signal pointing at any influence of reflux esophagitis on PBUB, a prophylactic treatment with PPI was administered in the majority of the included studies. The literature describe that PPI may diminish the size of PBUs, but in most studies PPI had no effect on bleeding complications or in mortality. 5,18,[25][26][27][28] Only one study 29 reported a better outcome of EBL with long-term PPIs administration. On the other hand, a judicious use of these drugs is needed due to their known adverse effects in patients with cirrhosis (pneumonia, hepatic encephalopathy, spontaneous bacterial peritonitis, osteoporosis). 26,30,31 The recent BAVENO VII consensus, recommends that whenever PPIs are started before endoscopy, these should be discontinued if endoscopy does not show a clear indication. 32 Similarly, other study 7 used sucralfate as prophylaxis for PBUB, without a reduction in the risk of bleeding.
Whether antibiotic prophylaxis was used in the context of PBUB was not explicit in all included studies. Bacterial infections are an important complication in acute variceal bleeding. It remains a risk factor for rebleeding and bleeding control, especially in patients with cirrhosis Child-Pugh B and C. 33,34 When prophylaxis was introduced spontaneous bacterial peritonitis and bacteremia by enteric Gram-negative bacteria were frequent infections. 35 Nowadays, respiratory infections accounts for 50% of all cases of infection, in patients with acute variceal bleeding, under antibiotic prophylaxis. 34,36 International guidelines recommend antibiotic prophylaxis on admission to hospital, in cirrhotic patients with acute variceal bleeding because it reduces the risk of infection, improves control of bleeding, reduces rebleeding and mortality. 32,37 Whether the same should be applied to patients with PBUB, remains to be confirmed in future studies.
Medical therapy and different endoscopic procedures similar to those used for acute variceal bleeding were used to treat PBUB. TIPS was used in some studies in patients with severe bleeding. 7,14,16 Tierney et al. suggest considering an early TIPS in patients with massive haemorrhage secondary to PBUB. Nonetheless, in this study 16 patients treated with TIPS presented the highest rate of mortality, which is likely due to a more severe liver disease (higher proportion of patients with Child-Pugh C) or to a delayed decision for TIPS. It is likely that as it happens in spontaneous variceal bleeding in high-risk patients, 38 also in high-risk patients with PBUB the early use of TIPS could be associated with lower failure and mortality rates, but this should be investigated in adequate studies. Moreover, our data cannot support an evidence-based recommendation on which specific intervention should be chosen to manage PBUB after the acute haemorrhage is controlled (TIPS or continuing with NSBB and EBL), and the most appropriate therapy still needs to be assessed by specifically designed studies.
As rescue therapy oesophageal SEMS 14,22 or oesophageal balloon tamponade 5,7,14,15,21,22 were employed in the included studies. Balloon tamponade or SEMS are recommended as a bridge therapy to a more definite treatment in case of refractory variceal bleeding. SEMS are as efficacious and much safer than balloon tamponade. 32 The stent can be maintained for up to 1 week (and in some cases has been used for 2 weeks) until a definitive treatment. 39 In a retrospective analysis 22 included in this systematic review its efficacy in controlling the bleeding in patients with PBUB was of 82%.
The above-mentioned strategies are much the same used for variceal bleeding, implying that the authors have assumed that the mechanism of PBUB is similar to that of VH. The only specific treatment addressing a different possibility has been the use of PPI, but these have not been shown to be universally effective. 25,27,31 A possibility that deserves further study is whether defects in haemostasis related to the liver disease (low platelet counts, unbalanced coagulation status), 4,40,41  A limitation of this systematic review is that most data come from retrospective observational studies, which carry a high risk of bias. This is illustrated by the fact that the incidence of PBUB was higher in retrospective that in prospective studies, but the difference was mild (5.8% in retrospective series vs. 4.7% in prospective series).
Many other studies on rebleeding post-EBL could not be included since the incidence of PBUB and different causes of rebleeding were not specified. Clearly, a better knowledge of predicting factors, and specifically the issue of whether haemostatic defects contribute to PBUB, require further study.
In addition, the best treatment approach needs to be investigated in multicentric prospective studies. These should probably stratify patients presenting with PBUB according to the severity of the bleeding and severity of liver failure to define the best treatment approach for each subgroup.
In conclusion, this systematic review analysed data from 18 studies and 9034 patients with cirrhosis and portal hypertension treated with EBL in elective or in an emergency setting. We found that patients with a higher MELD score, or treated with emergency EBL for acute variceal bleeding are more prone to suffer from PBUB. PBUB carries a high mortality despite the variety of treatments used, which calls for prospective, specifically designed studies to improve outcomes of this severe iatrogenic complication.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors do not have any disclosures to report.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

E TH I C S S TATEM ENT
Ethical approval was not necessary because of the study design.

PROS PE RO R EG I S TR ATI O N N U M B E R
The PROSPERO registered protocol number is CRD42022353449.