The patients taking DOACs are at high risk for both ischemic stroke and delayed bleeding in upper GI ESD [8, 13, 14], making it essential to identify the optimal DOAC considering these adverse events. Evaluating the risk of ischemic stroke proved challenging due to the need for a very large number of cases. Thus, this study utilized the DPC database to address these issues.

This study includes two key clinical insights. First, dabigatran exhibited the lowest risk for delayed bleeding in gastroduodenal ESD among the four DOACs, with rivaroxaban following. The low risk associated with dabigatran in gastric ESD aligns with findings from a prior nationwide multicenter study [21]. However, that earlier study did not evaluate the risk of delayed bleeding for other DOACs because of a limited number of cases nor did it examine dosage variation for each DOAC, even though dosage had minimal impact on the differences in delayed bleeding rates among the four DOACs. The different targets, i.e., thrombin for dabigatran and factor Xa for the three other DOACs, may have affected the varying rates of delayed bleeding among these medications. The risk of delayed bleeding in esophageal ESD followed a trend similar to that in gastroduodenal ESD. However, the small sample size likely contributed to the lack of significant results. A larger sample would be required to draw firm conclusions regarding esophageal ESD. Nonetheless, the trend in delayed bleeding risk among the four DOACs may remain consistent in upper GI ESD, given the relatively similar ORs between gastroduodenal and esophageal ESD.

A recent study on colorectal ESD in patients taking DOACs showed different findings [38], indicating that dabigatran had the highest risk of delayed bleeding in this context. Dabigatran’s behavior as a prodrug and the differences in acidic conditions between patients with gastric and colorectal ESD may explain these conflicting results. Dabigatran etexilate is more effectively absorbed in an acidic milieu in the GI tract. While it is formulated together with tartaric acid to reduce the variability in absorption, acidic conditions in the stomach can still impact its absorption. Indeed, two studies have shown that such conditions can significantly reduce the trough and peak plasma levels of dabigatran [39, 40]. Many cases with early gastric cancer have a nonacidic condition in the stomach due to gastric atrophy, which could reduce the absorption of dabigatran etexilate, resulting in a lower delayed bleeding rate and higher ischemic stroke rate after gastric ESD. Although there is no data on the gastric acidity in patients with colorectal ESD, differences in the acidity may contribute to the contradictory findings between gastric and colorectal ESD.

Second, and more importantly, we first revealed a lower risk of ischemic stroke with rivaroxaban and apixaban compared with dabigatran in upper GI ESD, especially when the DOACs were limited to the standard dose. A previous multicenter study involving over 10,000 cases with gastric ESD revealed a thromboembolic event rate of 0.03% with all events classified as ischemic stroke [13]. Further, all events occurred in patients on anticoagulants, regardless of the anticoagulant status, leading to an incidence rate of 0.52%. Thus, our study results are crucial, considering the risk of thromboembolic events in upper GI ESD. Although no studies have directly compared the risk of ischemic stroke among DOACs in patients with upper GI ESD, several studies have compared its risk in those with atrial fibrillation [18,19,20]. However, the results in the risk of ischemic stroke for each DOAC were inconsistent across the previous studies. Our results were also inconsistent with these reports. However, the different targets in the coagulation pathway between dabigatran and three other drugs, i.e., thrombin and factor Xa, respectively, may potentially explain the higher risk of ischemic stroke in dabigatran in this study.

Regarding an optimal DOAC for upper GI ESD, dabigatran exhibited the lowest risk of delayed bleeding in gastroduodenal ESD and also showed the lowest rate of delayed bleeding in esophageal ESD, although this finding was not statistically significant. However, it presented the highest risk of ischemic stroke in upper GI ESD, especially at standard doses. The guidelines [22] indicate that the risk of thromboembolism draws more attention than the risk of bleeding due to its seriousness. Thus, the risk of ischemic stroke should be prioritized over that of delayed bleeding, making dabigatran less preferable for use in upper GI ESD. Rivaroxaban, conversely, demonstrated a low risk of ischemic stroke in upper GI ESD, a second-lowest risk of delayed bleeding in gastroduodenal ESD, and a lower trend in risk for delayed bleeding in esophageal ESD. Altogether, rivaroxaban might be optimal for upper GI ESD. Although this study does not explain the reasons for the favorable results of rivaroxaban compared with the three other DOACs, especially apixaban and edoxaban, these results may help clinicians select a DOAC for upper GI ESD. In clinical practice, gastroenterologists typically do not prescribe DOACs but can recommend switching to rivaroxaban during the perioperative period for prescribing physicians, making our results relevant to clinical practice.

The strength of this study lies in the large number of cases, which enabled us to clarify the risk of ischemic stroke in each DOAC for the first time for upper GI ESD. Furthermore, the use of IPW and a generalized boosted model resulted in the best balance of baseline characteristics, including the risk of ischemic stroke and bleeding, among the four DOAC groups, ensuring reliable results.

This study had several limitations. First, this is not a randomized controlled trial but is a retrospective study, with the potential for selection bias. Second, the perioperative management strategy of DOACs (i.e., continuation, interruption, and heparin bridging) varied among patients. While heparin use was included as a covariate for IPW, this analysis did not consider DOAC continuation or interruption. Furthermore, the timing of interruption and resumption is important. Although the guidelines recommended stopping DOACs on the morning of ESD and resuming them the following morning [41], it is unclear how many cases complied with this recommendation. Third, some lesion-based factors, such as tumor size and location, which were reported as risk factors for delayed bleeding in gastric ESD [14], were not considered as covariates. Fourth, this study faced potential inaccuracies in coding including ischemic stroke, although a previous validation study demonstrated that the reliability of this database was relatively high in general [29] and the rate of ischemic stroke in patients on DOACs was similar to those in previous reports [13, 33]. Furthermore, the severity of ischemic stroke could not be assessed in this study, despite that it varies among the cases. Fifth, the results in cases that did not use heparin may best reflect the actual situation in the current clinical practice because the use of heparin was less recommended in the guidelines published in 2017 (in Japanese) [41], although this analysis was set as one of the sensitivity analyses in the present study. Indeed, only 8.9% of the cases used heparin after 2018. Nevertheless, the results in cases limited to no use of heparin were relatively consistent with the main analysis. Lastly, the results in some sensitivity analyses were not consistent with the main results, although the advantage of rivaroxaban was consistently observed in both the main and sensitivity analyses.

In conclusion, this large-scale database study revealed that rivaroxaban had the second lowest risk of delayed bleeding after dabigatran in gastroduodenal ESD, with relatively similar results observed in esophageal ESD. Additionally, rivaroxaban presented a lower risk of ischemic stroke than dabigatran in upper GI ESD, especially when limited to the standard dose. Given the greater importance of the risk of ischemic stroke over that of delayed bleeding, rivaroxaban might be optimal among the four DOACs for upper GI ESD.