Materials and methods
Search methods
We searched PubMed, Ovid MEDLINE, and Embase. To achieve maximum sensitivity, a combination of the MeSH terms “gastrointestinal bleeding”; “gastrointestinal hemorrhage”; “embolization”; “embolization, therapeutic”; and “surgery” were used ((“gastrointestinal bleeding” or “gastrointestinal hemorrhage”) and (“embolization” or “embolization, therapeutic”) and “surgery”). The search was performed in June 2018. Studies were retrieved and relevant studies were identified after reading the study title and abstract. Furthermore, the bibliographies of the selected studies were examined to identify any additional relevant studies.
Selection criteria
We used the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [
11] to formulate the basis of eligibility criteria using the PICO (P - Populations/People/Patient/Problem, I - Intervention(s), C - Comparison, O - Outcome) worksheet and search strategy (Table
1).
Table 1
PICO (Patients, Intervention, Comparison, Outcome) worksheet
Population | Adult patients with refractory NVUGIB (defined as failure of endoscopic hemostasis or rebleeding after successful endoscopic hemostasis) |
Intervention | Transcatheter angioembolization (TAE) |
Comparison | Direct comparison of TAE and surgery; report of at list one of the considered outcomes. If multiple trials or studies were published by the same center, only the most complete one was included. Studies regarding transpapillary bleeding were excluded |
Outcome | All-cause mortality with no time limit; rebleeding or continued bleeding; complications, both procedure-related and not procedure-related; need for further intervention for any reason |
RCTs, case-control, and cohort studies meeting the following criteria were included: (1) English language, (2) considering adult patients’ population with refractory NVUGIB (defined as failure of endoscopic hemostasis or rebleeding after successful endoscopic hemostasis), (3) direct comparison of TAE and surgery, and (4) report of at least one of the considered outcomes (mortality, rebleeding, complications, need for further intervention). If multiple trials or studies were published by the same center, only the most complete one was included. Studies regarding transpapillary bleeding were excluded. Case reports, editorials, letters, and studies containing duplicate data or data already published were excluded.
Data collection, assessment of study quality, and risk of bias
Studies were selected by two authors (A.T., Fa.C), and disagreements were resolved by collegial discussion. The full text of the included studies was obtained and reviewed, to determine the relevance and the quality of the paper. Risk of bias assessment was performed according to the Newcastle–Ottawa Quality Assessment Scale criteria.
Statistical analysis
Statistical analysis was performed using RevMan software (Review Manager version 5.3, Nordic Cochrane Centre, Copenhagen, Denmark). Odds ratio with a 95% confidence interval was used to compare outcomes. A fixed effect model was used in case of low heterogeneity, while a random effects model was used when significant heterogeneity was noted. Heterogeneity was assessed using the Cochrane Q square test (P < 0.1 was considered an indicator of significant heterogeneity) and the I2 estimates (< 25% moderate, 25–50% moderate, > 50% high heterogeneity).
Outcomes
The outcomes considered for analysis in the present study were all-cause mortality with no time limit, rebleeding rate, complication rate, both procedure-related and not procedure-related, and need for further intervention. Rebleeding was defined as bleeding from the same site after a successful index procedure or as a continuing bleeding, not controlled by index procedure. The need for further intervention comprehends all procedures performed after the index operation, either for a rebleeding or for a complication deriving from the index operation.
Discussion
In the case of non-variceal upper-GI bleeding, when medical and endoscopic treatment fails, surgery or transcatheter embolization is the available treatment option. Over the past few decades, the number of patients requiring surgical intervention has decreased enormously. In the 1990s, up to 13% of patients required surgery to control bleeding from peptic ulcer disease [
27], but with improved endoscopic hemostatic techniques and intravenous proton pump inhibitor infusions, the rate of surgical procedures has dropped to less than 2% in the present day [
28,
29]. In fact, endoscopic treatment is extremely effective in controlling NVUGIB, but despite adequate initial endoscopic therapy, refractory NVUGIB can occur in up to 24% of high-risk patients [
30] and mortality after a surgical salvage in the recent UK National Audit was still as high as 29% [
22]. The technological advances in interventional radiology are improving rapidly, whilst the experience of surgeons in the management of upper GI hemorrhage is declining. This trend is likely to continue in the future, so it is necessary to precisely determine the criteria that drive the choice between surgical and radiological treatment for NVUGIB.
In 1999, a prospective randomized study from Lau et al. [
31] compared endoscopic retreatment with surgery for rebleeding after initial endoscopy and found that in patients with peptic ulcers and recurrent bleeding, endoscopic retreatment reduces the need for surgery without increasing the risk of death and is associated with fewer complications than surgery. Furthermore, identification of the source of recurrent bleeding at a second endoscopy can lead to the placement of endoscopic clips at the ulcer rim; should endoscopic treatment fail to control the hemorrhage, the metal clips could provide the interventional radiologist with useful information to identify and embolize the culprit vessel in the absence of angiographic stigmata of bleeding [
32]. This implies that in the case of a bleeding peptic ulcer, surgical hemostasis or angiographic embolization should be performed after failure of repeated endoscopy. The same study from Lau [
31] found that hypotension at randomization (
P = 0.01) and an ulcer size of at least 2 cm (
P = 0.03) were independent factors predictive of the failure of endoscopic retreatment, suggesting that patients with hypotension and/or ulcer larger than 2 cm may undergo surgery/angiography without repeated endoscopy. This finding was confirmed by Wong et al. in 2011 in a retrospective review of a cohort of 3271 patients with peptic ulcer bleeding [
20].
Despite the limitations of the present study, this is the most updated and comprehensive meta-analysis, to our knowledge, that compares surgery to angioembolization for refractory NVUGIB and the results that arise from it are extremely interesting.
We found no difference in mortality rates between the two procedures, with a slight drift to a lower mortality in the TAE group (odds ratio [OD] = 0.77; 95% confidence interval [CI] 0.50, 1.18;
P = 0.05;
I2 = 43 [random effects]) (Fig.
4), despite the higher incidence of rebleeding after TAE (OD = 2.44; 95% CI 1.77, 3.36;
P = 0.41;
I2 = 4% [fixed effects]) (Fig.
5a, b). A statistically significant heterogeneity was found among the studies in mortality rates, and this could be due to the different time-frames for mortality assessment used between the different studies (30-day mortality was used in seven studies [
15,
18,
20,
21,
23,
24,
26], in-hospital mortality was used in four studies [
14,
17,
22,
25], and two studies did not report any time frame for mortality [
16,
19]). These results are even more important given the baseline differences in the two groups: frailer, older patients with more comorbidities tended to be treated with TAE. This undeniable selection bias is related to the methods for allocation of patients in the two arms of the meta-analysis: none of the included studies provides a clear list of the indication for TAE and the decision was mostly made by the attending surgeon on an individual basis; on the basis of personal experience, availability of an interventional radiologist, and operating room; and on the basis of patient clinical conditions and comorbidities. Defreyene and colleagues [
17] found that decision-making after endoscopic failure was significantly affected by the presence of a peptic ulcer: regardless of the bleeding severity or their clinical condition, patients with a peptic ulcer at endoscopy were 5.2 times more likely to be referred to surgery, and neither indirect parameters of hemodynamic instability nor coagulopathy independently influenced the choice of rescue. This bias was also reported by Beggs et al. [
33] that found a statistically significant higher incidence of coagulopathy and ischemic heart disease in patients undergoing TAE. On the other hand, a recent systematic review on embolization for NVUGIB found a mean technical success rate of 84% and a mean clinical success rate of 67%; while mean rebleeding rate, mean complication rate, and mean 30-day all-cause mortality were 27, 6, and 8%, respectively [
34]. The same study analyzed the factors related to angioembolization failure and found that the presence of coagulopathy and/or multi-organ failure has the worst impact on the outcome of embolization. In light of the above statement, the differences in patients’ comorbidities and coagulopathy at time of intervention between the two cohorts analyzed in the present study could then explain the higher rate of rebleeding in the TAE group. Furthermore, the study from Defreyne et al [
17] showed that rebleeding after TAE was observed only in the early post-procedural period (within 3 days), while after that time period, rebleeding episodes were observed only in the surgery group. This result further strengthens the hypothesis that rebleeding rate in the TAE group could be related to temporary physiological derangements. Another aspect that can explain this higher rate of rebleeding is the intermittent nature of this kind of hemorrhage, that makes it harder to identify the bleeding site during the angiographic study and could lead to notable variations in vessel diameters, thus making the embolic agent too small to completely occlude the vessel.
The complication rate is unarguably in favor of TAE (OD = 0.45; 95% CI 0.30, 0.47;
P = 0.24;
I2 = 26% [fixed effects]) (Fig.
6a, b), even if only six of the analyzed studies reported complications. Although the upper GI tract usually has a rich collateral blood supply, previous studies have shown ischemic complications to occur in 7 to 16% of cases [
35,
36] and they can either present acutely, with GI necrosis, or later, with ischemic duodenal stenosis. It is important to remember that multiple factors, such as previous surgery, pancreatitis, and radiation therapy, can interfere with collateral circulation and cause ischemia. Poultsides et al. [
35] reported four (7%) cases of bowel ischemia following embolization, all of whom had surgically altered anatomy. Other known causes of ischemia include the use of embolic agents, such as liquid agents (e.g., cyanoacrylate glue) or very small particles (e.g., gelatin sponge powder) that occlude more distally in the vascular bed. Only three of the included studies [
15,
20,
26] selectively analyzed TAE-related (i.e., pancreatitis, contrast-induced nephropathy, duodenal ischemia, coil misplacement, embolization of non-target vessels, access site arterial trauma, intimal dissection, or pseudo-aneurysm formation) and surgery-related complications (i.e., post-operative abscess, duodenal stump of anastomotic leakage, paralytic ileus, dehiscence of the fascia). The most accurate report is from Nykänen et al. [
26] that described, out of 53 embolizations, one (1.8%) iatrogenic dissection of superior mesenteric artery, four (7.5%) acute kidney injuries with three of them requiring dialysis, four (7.5%) gastroduodenal ischemic findings at follow-up endoscopy, and one (1.8%) migrating coil protruding through the duodenal ulcer into the duodenal lumen. These numbers are probably the closest to the real incidence of TAE-related complications; the tendency toward complications under-report in the other studies could be related to lack of data or difficulties in data collection, given the retrospective nature of the studies and the wide time spread of the analyzed cases (from 1986 to 2015). The increased complication rate after surgical treatment could also explain the absence of differences in mortality rate despite the increased risk of rebleeding after TAE: the choice between TAE and surgery seems to inevitably pose the dilemma of having an increased risk of rebleeding or an increased risk of complications. Consideration of these two possibilities must be weighted for individual patients and could be the element that drives treatment strategy, given the absence of strong, evidence-based recommendations.
In 2001, Defreyne et al. found that clinical parameters such as surgery after endovascular embolization failure negatively impacted survival [
37]. This result is extremely interesting and necessitates further investigation of this specific subset of patients, because it is plausible that the relative ischemia induced by a previous embolization could negatively affect the results of surgery. If this were the case, a meticulous evaluation of the rebleeding risk could be necessary and surgery as a first option could be taken into account for patients with a high rebleeding risk.
A recent retrospective study from the Karolinska Institute [
12], not included in this study because it did not provide raw data for the meta-analysis, included 282 patients with a 3-year follow-up, found a significant reduction in mortality and hospital length of stay after TAE, that could be explained by the comparability in terms of age and comorbidities of patients in the surgery and TAE groups. This is a further stimulus to implement TAE into everyday management of refractory NVUGIB.
The discussion of the results for the last analyzed outcome, i.e., the need for further intervention after the index operation, is limited by the nature of the included studies. In fact, only nine of them included this outcome in their analysis but a great degree of heterogeneity was found and this could be related to a selection bias, because some of the studies did not report the need for additional intervention in the case of procedure-related complications, but only in the case of rebleeding. Nevertheless, our meta-analysis revealed a significant reduction of further intervention in the surgery group (OD = 2.13; 95% CI 1.21, 3.77;
P = 0.02;
I2 = 56% [random effects]) (Fig.
6), but this result is likely to be related to the increased rate of rebleeding found in the TAE group.
Some confusion still remains on the applications of TAE in hemodynamically unstable patients, and none of the included studies analyzed this specific subset of patients. From the existing data, it seems that physician expertise and availability of equipment are the driving factors in the decision to perform TAE on hemodynamically unstable patients. More specifically, it is likely possible to perform TAE in these patients in facilities with a 24-h availability of an expert interventional radiologist and a hybrid OR, while it is not recommended in smaller facilities with limited resources.
It is also necessary to take into account the previous surgical history of the patient, because it could significantly alter the vascular anatomy, and the presence of post-surgical adhesions could make access to the gastroduodenal area arduous and time-consuming.
Limitations
The study has some limitations. First of all, the retrospective nature of the majority of included studies leads to an inevitable selection bias. Furthermore, the decision between TAE and surgery was made on an individual case-by-case basis by the attending surgeon, making group allocation and randomization difficult to achieve. This may lead to low external validity. Furthermore, although the most common cause of refractory bleeding is a peptic ulcer, there remain a variety of etiologies. TAE and surgical techniques also vary among the included studies. Regarding complications, only a few studies reported a complication rate for the analyzed procedures and some studies reported reintervention rates only in case of rebleeding after the index procedure; thus, the analysis of the reintervention and complication rates may not represent everyday reality. Lastly, the modality of mortality detection differs between the studies (in-hospital mortality, 30-day mortality, overall mortality, etc.). These limitations do not impair the power of the present study that represents the largest and most recent meta-analysis available.
Conclusions
The management of NVUGIB refractory to endoscopic treatment remains a clinical dilemma and the decision between TAE and surgery is usually made on an individual basis by the attending surgeon. The aim of this meta-analysis is to gather the currently available evidence, thus providing a guide for every clinician facing this emergency. Unfortunately, the quality of the studies currently published is low and their retrospective nature increases the likelihood of selection bias.
Nevertheless, the results of the present study show that TAE is a safe and an effective procedure and, when compared to surgery, TAE has a higher rebleeding rate, but this tendency does not affect the clinical outcome. In fact, the comparison of mortality rates for the two procedures highlights a slight drift toward a lower mortality for patients undergoing TAE, despite the fact that the TAE patient population usually includes those with greater comorbidities unfit for surgery.
The present study suggests that TAE could be a viable option as a first-line therapy for refractory NVUGIB and, in the absence of evidence of superiority of one specific approach, local factors, such as organization of surgical and radiological services, availability of specific radiological skills, services availability during night shift and weekends, etc., will continue to determine the therapeutic pathway. This study also aims to set the basis for the design of future randomized clinical trials. Another issue to be addressed in the future is the best treatment option for refractory NVUGIB in hemodynamically unstable patients.