Article Text

Download PDFPDF

Stopping antithrombotic therapy after acute upper gastrointestinal bleeding is associated with reduced survival
  1. Keith Siau1,2,
  2. Jack L Hannah3,
  3. James Hodson4,
  4. Monika Widlak5,
  5. Neeraj Bhala6,
  6. Tariq H Iqbal6
  1. 1 Joint Advisory Group in Gastrointestinal Endoscopy, Royal College of Physicians, London
  2. 2 Department of Gastroenterology, Dudley Group Hospitals NHS Foundation Trust, Dudley
  3. 3 Department of Medicine, Countess of Chester Hospital NHS Foundation Trust, Chester, UK
  4. 4 Department of Medical Statistics, Institute of Translational Medicine, Queen Elizabeth Hospital, Birmingham, England
  5. 5 Department of Gastroenterology, University Hospitals Coventry and Warwickshire NHS Trust, Coventry, England
  6. 6 Department of Gastroenterology, University Hospitals Birmingham NHS Foundation Trust, Birmingham, England
  1. Correspondence to Dr Keith Siau, Department of Gastroenterology, Dudley Group NHS Foundation Trust, Dudley DY1 2HQ, UK; keithsiau{at}


Introduction Antithrombotic drugs are often stopped following acute upper gastrointestinal bleeding (AUGIB) and frequently not restarted. The practice of antithrombotic discontinuation on discharge and its impact on outcomes are unclear.

Objective To assess whether restarting antithrombotic therapy, prior to hospital discharge for AUGIB, affected clinical outcomes.

Design Retrospective cohort study.

Setting University hospital between May 2013 and November 2014, with median follow-up of 259 days.

Patients Patients who underwent gastroscopy for AUGIB while on antithrombotic therapy.

Interventions Continuation or cessation of antithrombotic(s) at discharge.

Main outcomes measures Cause-specific mortality, thrombotic events, rebleeding and serious adverse events (any of the above).

Results Of 118 patients analysed, antithrombotic treatment was stopped in 58 (49.2%). Older age, aspirin monotherapy and peptic ulcer disease were significant predictors of antithrombotic discontinuation, whereas dual antiplatelet use predicted antithrombotic maintenance. The 1-year postdischarge mortality rate was 11.3%, with deaths mainly due to thrombotic causes. Stopping antithrombotic therapy at the time of discharge was associated with increased mortality (HR 3.32; 95% CI 1.07 to 10.31, P=0.027), thrombotic events (HR 5.77; 95% CI 1.26 to 26.35, P=0.010) and overall adverse events (HR 2.98; 95% CI 1.32 to 6.74, P=0.006), with effects persisting after multivariable adjustment for age and peptic ulcer disease. On subgroup analysis, the thromboprotective benefit remained significant with continuation of non-aspirin regimens (P=0.016). There were no significant differences in postdischarge bleeding rates between groups (HR 3.43, 0.36 to 33.04, P=0.255).

Conclusion In this hospital-based study, discontinuation of antithrombotic therapy is associated with increased thrombotic events and reduced survival.

  • adult gastroenterology
  • thrombosis
  • upper gastrointestinal bleeding
  • mortality
  • outcomes
  • antithrombotics
  • gastrointestinal haemorrhage

Statistics from


Acute upper gastrointestinal bleeding (AUGIB) is the the most common gastroenterological emergency, with an estimated inpatient mortality rate of 10%.1 Antithrombotic therapy (AT), comprising antiplatelets and anticoagulants, is increasingly prescribed for treatment and secondary prevention of thromboembolic events, and was traditionally believed to be contraindicated following AUGIB. Recent data have challenged this perception.2 3 If there is compelling indication for AT therapy, early resumption of aspirin or anticoagulant therapy have been shown to be protective against cardiovascular events and all-cause mortality,2 3 provided that haemostasis has been achieved. The National Institute for Health and Clinical Excellence (NICE) have recommended continuation of aspirin therapy following haemostasis in AUGIB if indicated for secondary prevention,4 but their position on non-aspirin ATs (thienopyridines and anticoagulants) is less clear. Despite this, ATs are often stopped following AUGIB.3

We hypothesise that the  thrombotic risk in this subgroup of patients is particularly high due to various pathophysiological factors, including relative myocardial ischaemia, cardiovascular strain complicating those with pre-existing ischaemic heart disease and a prothrombotic state secondary to systemic stress, anaemia, blood transfusions and underlying comorbidities. Data exploring outcomes in this high-risk patient group is scarce, with no previously published UK-based data.


The aim of this study was to compare outcomes in a large cohort of patients admitted to a single university teaching hospital who were on established AT at the time of their presentation with AUGIB. We aimed to assess if restarting maintenance antithrombotic therapy on discharge was associated with different clinical outcomes, compared with patients in whom antithrombotics were not restarted prior to discharge. We also explored prevalence, predictors and reasons for AT discontinuation, and concordance with NICE recommendations.



We retrospectively identified a cohort of patients on established AT therapy who underwent gastroscopy for suspected AUGIB (variceal and non-variceal) at University Hospital Birmingham, UK, between May 2013 and November 2014 and performed follow-up until March 2015. Exclusion criteria included: age< 18, outpatient referrals, incomplete endoscopy, failure of haemostasis, alternate cause for bleeding found (eg, epistaxis or lower gastrointestinal (GI) bleeding), in-hospital mortality, established upper GI malignancy, patients with palliative intent and patients with a non-GI contraindication to AT therapy (eg, intracerebral bleed). Data were collected from hospital case notes, endoscopic records, discharge letters and general practitioner records.


Clinical outcomes were measured after discharge and included cause-specific mortality, thrombotic events, bleeding events and any serious adverse event (composite outcome of mortality, thrombotic event or bleeding event). Outcome data arising from repeat hospital admissions, both locally and regionally, were identified via the University Hospitals Birmingham informatics department via Hospital Episode Statistics using unique National Health Service (NHS) identifiers. Causes of mortality were verified with the coroner’s office.


Antithrombotic therapy was defined as any antiplatelet agent (including aspirin, thienopyridines—clopidogrel, prasugrel, ticagrelor) or anticoagulant (vitamin K antagonists, novel oral anticoagulants, therapeutic heparin dosage). Thrombotic events included acute coronary syndrome, ischaemic cerebrovascular event, venous thromboembolism, mesenteric ischaemia and acute arterial ischaemia. GI bleeding was defined as presentation with haematemesis and/or melaena. Rebleeding was defined as haematemesis and/or melaena with reduction in haemoglobin of ≥20 g/L, or the need for repeat endoscopy or an alternative modality (eg, CT angiography) after readmission with GI bleeding. Bleeding events included recurrent AUGIB, and bleeding from alternate sources, for example, lower GI bleed, haematoma, intracerebral bleed. AT maintenance was defined as continuation of AT agents, without being downgraded at discharge. For patients on dual therapy, reduction to single-agent therapy was considered as AT discontinuation. AT was deemed to be maintained if class switching had occurred, for example, conversion of clopidogrel to aspirin. Patients were stratified according to whether AT therapy was maintained or discontinued at the point of discharge.

Statistical methods

Continuous variables were compared between groups using independent samples t-tests, and reported as mean±SD where normally distributed, and Mann-Whitney tests with medians and IQRs otherwise. Dichotomous variables were compared between the groups using Fisher’s exact tests. Time-to-event outcomes were assessed using a Kaplan-Meier approach, with comparisons between groups made using log-rank tests. In addition, Cox regression models were produced in order to calculate HRs which were reported alongside 95% CIs. A subgroup analysis was also performed, considering different AT regimes. In order to account for confounding factors, multivariable Cox regression models were produced. Any factors found to differ significantly between groups were included in this analysis. All analyses were performed using SPSS (V.22, IBM Corp., New York, USA), with P<0.05 deemed to be statistically significant throughout.


A total of 118 patients with AT-associated AUGIB were included for analysis, after excluding 30 patients for inpatient mortality (14), no endoscopy (10), end-of-life care (4) and epistaxis (2). The median age of the cohort was 76 years, with men (n=77) constituting 65.3% of the study population. Patients were followed up for a median of 259 days from discharge (range 3–602). Findings on index endoscopic evaluation for AT-associated AUGIB comprised: peptic ulcer disease (PUD) [n=42; 35.6%], no cause found (n=33; 28.0%), gastritis/duodenitis (n=21; 17.8%), oesophagitis (n=7; 5.9%), Mallory-Weiss tear (n=6; 5.1%), blood only (n=4; 3.4%), varices (n=2; 1.7%), Dieulafoy’s lesion (n=2, 1.7%) and oesophageal carcinoma (n=1; 0.8%) (online supplementary figure 1). AT type consisted mainly of aspirin monotherapy (n=50, 42.4%), warfarin (n=26; 22.0%), dual antiplatelet therapy (n=18; 15.3%) and thienopyridine (n=11; 9.3%). Primary indications for AT included: ischaemic heart disease (n=48; 40.7%), atrial fibrillation (n=20; 16.9%), unclear indication (n=14, 11.9%), venous thromboembolism (n=11; 9.3%), peripheral vascular disease (n=8; 6.8%), cerebrovascular disease (n=7; 5.9%), secondary prevention (n=6; 5.1%) and prosthetic heart valve (n=4; 3.4%). The type of AT used and corresponding indications are tabulated in online  supplementary table 1.

Supplementary file 1

Supplementary file 2

AT discontinuation at the point of discharge occurred in 58/118 patients (49.2%). Baseline population characteristics, stratified according to AT maintenance or discontinuation, are summarised in table 1. On univariable analysis, older age (P=0.044), aspirin monotherapy (P=0.009) and peptic ulcer disease (P=0.007) were significant predictors of AT discontinuation, whereas dual antiplatelet use (P=0.011) predicted AT maintenance. Reasons for AT discontinuation were specified in 21 (36%) cases, and included: specialist advice (n=9, 43%), pending further review (n=8, 38%), patient choice (n=1, 5%) and unclear indication (n=3, 14%).

Table 1

Patient demographics

A total of 12 thrombotic events were observed at a median of 59 days postdischarge (range: 8–414). These consisted of myocardial ischaemia/infarction (n=7), cerebrovascular event (n=3), venous thromboembolism (n=1) and ischaemic foot (n=1). AT discontinuation was associated with a significant increase in thrombotic events (HR 5.77, 95% CI 1.26 to 26.3, P=0.010), with 10 events occurring in the AT discontinuation group, compared with two events in the AT maintenance group, giving a 1-year incidence of 17% versus 2% (figure 1A). Multivariable analysis (table 2) found that this effect remained significant after adjustment for both patient age (P=0.018) and PUD (P=0.025).

Figure 1

Kaplan-Meier curves showing freedom from thrombotic events (A), overall survival (B) and freedom from adverse events (C), according to whether antithrombotic(s) were restarted on discharge.

Table 2

Multivariable analysis adjusting for age and presence of peptic ulcer disease

There were 16 deaths during follow-up, giving a 30-day postdischarge mortality rate of 3.4%. Death was due to thrombotic causes in eight patients (50%), as a result of myocardial infarction (7) and cardiac thrombus (1). AT discontinuation was associated with significantly increased mortality (HR 3.32, 95% CI 1.07 to 10.3, P=0.027), with 12 deaths in the AT discontinuation group, compared with four in the AT maintenance group, giving 1-year mortality of 18% versus 5% (figure 1B). Multivariable analysis (table 2) found that this effect remained significant after adjustment for both patient age (P=0.040) and PUD (P=0.046).

Postdischarge bleeding occurred in four patients (3.4%), at a median of 50 days after discharge (range: 14–177). The sources of bleeding were tumour-related (n=1), oesophageal varices (1) and gastroduodenitis (1), with no cause identified in the fourth case. No significant difference in postdischarge bleeding rates was detected between the groups (HR for AT discontinuation: 3.43, 95% CI 0.36 to 33.04, P=0.255), with three events in the AT discontinuation group, compared with one event in the AT maintenance group, giving a 1-year incidence of 6% vs 2%.

A total of 29 patients had at least one serious adverse event (mortality, thrombotic or bleeding event). Serious adverse event rates were significantly higher in the AT discontinuation group (HR 2.98, 95% CI 1.32 to 6.47, P=0.006), with 21 events compared with 8 in the AT maintenance group, giving adverse event rates of 34% versus 10% at 1-year post discharge (figure 1C). This effect remained significant after adjustment for both patient age (P=0.008) and PUD (P=0.010) (table 2).

After stratifying groups by AT subtype (table 3), discontinuation of non-aspirin AT regimes was associated with a significant increase in overall adverse events (rate at 1-year post discharge: 43% vs 8%, P=0.009) and thrombotic events (23% vs 0%, P=0.016). Cessation of aspirin monotherapy also appeared to increase thrombotic events (32% vs 11%, HR 2.35, 95% CI 0.65 to 8.44), although this did not reach statistical significance (P=0.176).

Table 3

Patient outcomes stratified according to type of antithrombotic (AT)


AUGIB is a common complication of AT therapy. We demonstrate that stopping AT therapy at the point of hospital discharge is associated with poorer thrombotic outcomes and reduced survival, but with no significant reduction in rebleeding complications in this single-centre study. Our in-hospital mortality rate of 10.6% (14/132) appears comparable to the mortality figure from the 2007 UK AUGIB national audit of 10%.1 That audit also supported our findings that patients were more likely to die from thrombotic causes rather than haemorrhage. Additionally, we report the endoscopic findings of AT-associated AUGIB. Compared with data from the UK national audit,1 our AT cohort had higher rates of PUD (36% vs 26%), which correlate with the adverse effect profile of ATs. The lower rates of variceal bleeding (2% vs 8%) may reflect the lower prevalence of AT use in cirrhotic patients. Similar to other studies,5 6 there was a substantial proportion without causes of bleeding found (28%), which may be attributable to the higher incidence of lower GI and small bowel bleeding which may also present as melaena.

Our results echo the findings of several studies which have explored outcomes in this group of patients. In a landmark randomised trial conducted by Sung et al involving 156 patients admitted with AUGIB while on aspirin, discontinuation of aspirin was associated with increased all-cause mortality at 8 weeks compared with aspirin maintenance (10.3% vs 1.3%, P=0.001).2 Mortality mainly arose due to thrombotic events, without a significant difference in rebleeding rates. Despite these data and NICE guidelines, aspirin monotherapy was the most commonly withheld AT in our study, and was stopped in 64% (32/50). This may be due to concerns over its potential for inducing ulceration. Similarly, in a prospective cohort study by Sengupta et al, continuation of anticoagulation was associated with reduced 90-day risk of thrombotic episodes (HR 0.121, P=0.03), but without significant increase in risk of recurrent AUGIB.3 The effect of thienopyridine maintenance following AUGIB has not previously been evaluated.

The benefits of antiplatelet therapy for secondary prevention of cardiovascular events are well established. In a meta-analysis involving over 50 000 patients with ischaemic heart disease, discontinuation of aspirin when indicated for primary or secondary prevention was associated with increased thrombotic events (relative risk 3.14, 95% CI 1.75 to 5.61, P=0.0001).7 The average time between aspirin withdrawal and thrombotic event in this outpatient population was 10.7 days.7 The risk was significantly greater for patients who had previous intracoronary stenting, with premature interruptions to antiplatelet therapy being the most important predictor of stent thrombosis (HR 89.8; 95% CI 29.9 to 269.6; P<0.001).8

We acknowledge the limitations of our study. There was a small number of outcomes in this defined population, which limits the statistical power for the assessment for bleeding outcomes. The retrospective nature of this study, with the absence of matched controls, needs to be considered as a source of potential bias, but this was initially designed as an audit. We did not stratify cardiovascular risk in patients with IHD, as difficulties were encountered in attaining cardiovascular risk modelling, accurate smoking history and, in some cases, the antithrombotic indication, which may lead to selection biases. Among other confounders and biases, we did not have records for socioeconomic deprivation, comorbidities and transfusion requirements. Finally, we were not able to measure antithrombotic compliance or alterations to antithrombotic regimen in the primary care settings as access to general practitioner electronic health records was not available for all patients.

However, our study does have its strengths. Despite relatively small numbers, it did demonstrate a statistically significant difference in clinically important outcomes after discharge. Our patient selection, which only included patients with endoscopic evidence of haemostasis and excluded patients with palliative intent, limits confounding of mortality due to causes such as GI malignancy, and allows our results to be interpreted as ‘real world data’. The informatics approaches for collection of the outcomes are also robust due to the use of NHS identifiers for mortality and readmission, as the serious adverse events collected are clinically important. Multivariable analysis was also performed to adjust for other confounding factors, although these models could only accommodate one additional variable at a time on account of the small number of outcomes. We had previously presented data on postendoscopic AT maintenance, to include outcomes prior to discharge, which yielded similar results.9 In this study, we excluded inpatient mortality to reduce potential for confounding, such as inpatient comorbidity and deaths occurring shortly after endoscopy, which may have occurred regardless of postendoscopic antithrombotic resumption. Our results on thromboprotective outcomes are consistent with the very limited data in the literature. Hence, our data could provide a stimulus to further hypotheses to evaluate thrombotic outcomes following AUGIB in larger studies.

Our study did not answer the question of when to restart antithrombotics after endoscopic haemostasis. The European Society of Gastrointestinal Endoscopy (ESGE) have released guidelines for aspirin and warfarin resumption after AUGIB.10 For aspirin, immediate resumption was recommended for low-risk lesions (Forrest IIc and Forrest III peptic ulcers), and for reintroduction by day 3 for high-risk peptic ulcers (Forrest 1a to Forrest 2b).10 For warfarin, reintroduction should be commenced between 7 and 15 days following the bleeding event for low-risk indications, with earlier resumption within the first 7 days for those with high thrombotic risk.8 In patients with high-risk lesions, a relook endoscopy may be helpful to aid decision making with regards to antithrombotic resumption. Furthermore, all patients should also be coprescribed antisecretory therapy after AUGIB,8 although in our study, this was the case in only 76% of cases.

We hereby present novel data from the UK to support AT maintenance following haemostasis for AUGIB. Although we understand the hesitation with restarting AT in view of rebleeding concerns, the risk of death from stopping such therapy appears to be greater. Our rebleeding rate of 3.4% is comparable with recent data reporting a 3.1% readmission rate of rebleeding within 30 days of hospital discharge.11 We therefore strongly urge clinicians to review the indications for AT therapy and consider resumption at the point of discharge. From our study, only 17% of patients had a plan detailing AT resumption on the endoscopy report. We believe that there should also be more onus on the endoscopist and systems (eg, endoscopy reporting software) to have a more proactive role in outlining an AT plan after endoscopy for AUGIB.

Despite advances in endoscopic and medical therapies over the last decade, mortality from AUGIB has remained high. Proactive resumption of antithrombotic therapy may be the upcoming strategy for improving post AUGIB outcomes.


In this high-risk patient group, mortality from thrombotic causes following AUGIB is high. Discontinuation of antithrombotic therapy is associated with thrombotic events and reduced survival. Antithrombotic resumption is an independent factor which may improve survival after AUGIB.

Main messages

  • Antithrombotic drugs were stopped in 49.2% after an admission with acute upper gastrointestinal bleeding (AUGIB).

  • Resumption of antithrombotic therapy after AUGIB was associated with reduced mortality and thrombotic events, thereby outweighing the potential risks of haemorrhage.

  • As antithrombotic resumption improves outcomes, clinicians should address antithrombotic therapy as the upcoming strategy to improve post-AUGIB outcomes.

Current research questions

  • What proportion of patients have aspirin restarted after haemostasis for acute upper gastrointestinal bleeding (AUGIB), as per National Institute for Health and Clinical Excellence guidance?

  • Is stopping antithrombotic therapy after AUGIB beneficial?

  • What are the predictors of antithrombotic cessation?



  • Twitter Follow Keith Siau @drkeithsiau

  • Contributors THI and NB jointly supervised the design of this study, and are both guarantors of the article. KS, JLH and MW were involved in data collection and helped with the literature review. KS and JH analysed the data and wrote the majority of the paper. All authors approved the final version of the manuscript.

  • Competing interests None declared.

  • Ethics approval UHB Audit Department.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Presented at Similar work was presented at United European Gatroenterology Week 2015 which was awarded the Oral Free Paper Prize.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.