Outcomes of Upper Gastrointestinal Bleeding Are Similar Between Direct Oral Anticoagulants and Vitamin K Antagonists

Claire Gouriou; Guillaume Bouguen; Pierre Lahmek; Agnes Pelaquier; Ramuntxo Arotcarena; Armand Garioud; Stephanie De Montigny-Lenhardt; Arnaud Pauwels; David Zanditenas; Claire Charpignon; Remi Combes; Stephane Nahon; Vincent Quentin


Aliment Pharmacol Ther. 2021;53(6):688-695. 

In This Article


This prospective multicentre study focusing on UGIB involved a large number of patients with almost 500 patients treated by an oral anticoagulant and showed no difference in the type of oral anticoagulant on patient outcomes whatever the judgement criteria (mortality, rebleeding or need for non-endoscopic treatment). In addition, endoscopic findings and management of UGIB were not different between VKA and DOACs.

To our knowledge, there is no large-scale prospective study published since the widespread use of new oral anticoagulants. The national audit in England in 2007[24] and a French study in general hospitals[25] only reported comedication with Warfarin ranging from 7% to 10.3%. A growing number of patients are treated in France with anticoagulants as observed in this study where roughly one out of five patients were treated with an anticoagulant at the time of bleeding a decade after the first cohort. The simultaneous increasing use of DOACs observed in our cohort (44%), consistent with the French tendency reported from healthcare data (38%), remains an issue for physicians caring for gastrointestinal bleeding.[26]

There are less data analysing endoscopic and peri-endoscopic management of patients treated by oral anticoagulant and experiencing digestive bleeding, hence current international guidelines are based on expert opinion.[27] In our cohort, there were no differences between groups in terms of initial gravity of the bleeding, endoscopic diagnosis and management and effectiveness of bleeding resolution.

In-hospital mortality rate in our study was 8.4%, slightly superior to the 7.5% rate of VKA-treated patients in the previous French study.[25] Nevertheless, in-patients, who have proved to be linked with mortality in this study, were not included. This remains in line with the overall mortality rate of UGIB reported from other former studies—UK audit in 2007 6.8%[24] and French audit in 2000 10.7%.[1] Mortality in the VKA group was higher than in the DOACs group (P = 0.0059), potentially related to the comorbidities that contraindicate the use of DOACs such as kidney failure and cirrhosis over-represented in the VKA group. However, after adjusting to the different factors, only the Charlson score and in-hospital status remain associated with mortality contrary to the type of oral anticoagulant. In a prospective study conducted in UK in the early 2000s, the 30-day mortality rate after UGIB was established as directly linked to the digestive bleeding for only 29% cases, suggesting that life-threatening comorbidities contributed to the overall prognosis.[28] UGIB of in-patients is a rare event (0.2%-0.4%) but is drastically associated with the risk of death ranging up to 26% compared to 7% of out-patients.[24] Herzig et al[29] developed a scoring system to evaluate in-hospital UGIB risk which retained in the final model major comorbidities and coagulopathy.

Despite this particular subgroup analysis on patients treated with anticoagulant the earlier in-hospital rebleeding rate was lower than the one reported before (7.5% versus 11.8%).[25] Overall, the rebleeding rate at 6 weeks of 11.8% seemed broadly stable as compared to other cohorts such as that the UK audit in which 13% bleeding rate was observed.[24] The severity of initial bleeding was surprisingly not associated with the rebleeding risk but with concomitant antiplatelet agents and beta-blockers intakes. After crossing different data, we have no explanation of this result and with a small number of events (n = 56 rebleeding), it might seem difficult to conclude of a protective effect.

The need for non-endoscopic treatment in our cohort was similar to the UK audit that included patients with and without anticoagulants (surgery 1.9%, radiologic treatment 1.2%).[24] The tumour origin of the bleeding was the only factor associated with this outcome as expected, related to the limited efficacy of endoscopic treatment in such cases.

The prospective enrolment over a 1-year period in French general hospitals spread over the country remains the main strength of this study. The fulfilment of e-CRF by physicians was almost perfect with only 1.5% of patients excluded from this cohort due to missing data. This large «real-life» setting cohort of patients brings useful data for daily practice, avoiding inclusion bias of selected patients in phase III trials. This was conducted in 2017–2018, several years after market authorisation of DOACs, giving a necessary hindsight of safety data. We also retrieved precise endoscopic findings often lacking in published studies with a sufficient follow-up of 6 weeks for a comprehensive overview.

Some limitations need however to be addressed. The date of discontinuation and recovery of oral anticoagulants and antiplatelet agents were not specifically collected in the electronical form that precluded drawing precise conclusion on their management during UGIB. Resumption of treatment was only recorded on the basis of a 5-day cut-off. Unfortunately, data about thrombotic events occurring after the gastrointestinal bleeding and death induced by a thrombotic episode were not recorded. This would have been useful for the day-to-day practice to help physician decision-making particularly by assessing the benefit-risk balance of the oral anticoagulation discontinuation/resumption. Apixaban appears to have the safest profile in meta-analysis[30] but heterogeneity might skew the results there is still no head-to-head comparison of the three DOACs available in France. Unfortunately, the small number of patients precludes a subgroup analysis for each oral anticoagulant.