Transcatheter arterial embolization (TAE) is widely acknowledged for its effectiveness in treating brisk arterial bleeding20. It has largely replaced surgery as a first-line therapy for UGIB refractory to endoscopic treatment due to its high efficacy and favorable safety profile21. However, a subgroup of ulcers remains at high-risk of future rebleeding despite achieving hemostasis via endoscopy. In 2012, based on the available data at that time, the American College of Gastroenterology Guidelines only recommended medical treatment of the underlying etiology, for example, withholding aspirin, treating H.pylori, etc. Afterwards, studies of using PTAE to decrease this risk and mortality in selected patients have been published in the last five years8,22. These high-risk lesions are typically large in size (i.e. greater than 10 mm), belong to Forrest class I-IIb and/or were treated in hemodynamically unstable patients8,13,23. Whereas residual blood flow underneath a treated lesion may serve as a source of rebleeding, the intermittent nature of UGIB bleeding is dependent on a patient’s hemodynamic status. For example, a hypotensive patient may present with a lesion that appeared hemostatic at endoscopy however following blood transfusion the normalized blood pressure could initiate rebleeding24–26. As such, most authors considered PTAE as an option if the ulcer is considered to be high-risk based on based on endoscopic findings (i.e. large ulcers, Forrest class I-IIb) and/or clinical presentation (i.e. hypotension, decreased hemoglobin during endoscopy) 8, 11–13,18.
According to the present meta-analysis, PTAE was highly effective in reducing the risk of rebleeding compared to EO (2.34 [95%CI: 1.33–4.13], p = 0.003). Compared to patients who did not receive additional embolization after endoscopy, the PTAE group has a statistically significantly lower rebleeding rate of 6.8%. This value is also lower than the reported rates of emergent embolization for nonvariceal UGIB that range from 9 to 47%27–31. For nonvariceal UGIB recalcitrant to TAE, surgery is considered to be the final, definitive treatment option but confers a mortality risk as high as 43%34,35. Three out of the five included study cohorts required surgical intervention, with a higher cumulative rate in patients that did not receive PTAE after endoscopy (14.4% vs 3.0%; OR: 2.898 95%CI [1.374–6.112], p = 0.005). Further, the PTAE group also has a statistically significantly lower mortality rate than its EO counterpart (4.5% vs 8.8%, OR: 2.11, 95%CI [1.07–4.15], p = 0.032), though publication bias exists (Fig. 6C). A total of twelve PTAE patients from four studies died. According to Laursen et al, the majority of deaths were non-UGIB related (5/7) such as sepsis and malignancy18. In one case, an endovascular coil migrated to the hepatic artery, leading to acute hepatic failure and death. Of note, this patient had poor baseline liver function at admission due to advanced cirrhosis as isolated coil migration to the hepatic artery is not expected to routinely cause fulminant hepatic failure due to the dual blood supply of the liver. Coil migration also occurred in nine other patients, which were all asymptomatic without elevation of liver enzymes. Eight of these nine patients were from the same study, so this type of complication may have been operator dependent. Nevertheless, only two studies observed complications from PTAE with a major and minor complication rate of 0.75% and 3.4%, respectively13,18. In comparison to the differences of mortality and rebleed rates between PTAE and EO groups, the risk of PTAE-related major complication was considered rather minimal. Based on subgroup analysis (supplement table 4), patients with NSAID usage is more likely to benefit from PTAE, because of the higher rebleeding rates in this population32. Compared to patients with hemoglobin greater than 8.0 g/dl, patients with lower hemoglobin level are less likely to benefit from PTAE, which is consistent with the finding of previously published literature on TAE33. However, these results were based on study-level evidence. Stratified analysis using patient-level data are warranted in future studies.
There are several limitations of this study. First, only five studies were included, and three of them were retrospective cohorts, which are inferior to RCT and prospective studies. Due to limited evidence available, study removal during sensitivity analysis could render the overall results loss of statistical significance (supplement Fig. 1.1, 1.2). Second, the length of follow-up time varied among studies. Subgroup analysis on 30-day follow-up did not suggest statistically significant advantage of PTAE in terms of rebleeding and mortality reduction, despite the cumulative overall results did suggest its efficacy (supplement table 4). The rebleeding and mortality rates are expected to likely increase cumulatively over time. Only Lau et al implemented a Kaplan-Meier curve to characterize these evolutions8. Finally, limited evidence was available regarding the risk factors of rebleeding and deaths after PTAE. None of the authors provided the baseline characteristics separately between successful and failed PTAE subgroups.
Given these limitations, several important aspects should be addressed during future study design. In addition to increasing sample size, future RCTs with intention-to-treat protocols should aim to reduce the number of patients allocated to PTAE that fail to receive the procedure. For example, increased access to centers with expert resources in interventional radiology or modification of protocols to conduct angiography immediately following endoscopy should be considered. Further, standardization of technique is also important with respect to embolic(s) used and method of embolization. As suggested by Loffroy et al, variations may contribute to varied rebleeding risk, and few authors from prior literature have specified procedure details14,27. Finally, stratification based on lesion type and size should also be performed, as the post-hoc analysis by Lau et al suggested that PTAE reduced rebleeding risk in ulcers greater than 15mm8. Additional clinical trials with long-term clinical outcomes are needed to confirm these observations.