Article Text

Download PDFPDF
Endoscopy findings in patients on dual antiplatelet therapy following percutaneous coronary intervention
  1. Victor Galusko1,
  2. Majd Protty1,2,
  3. Hasan N Haboubi3,
  4. Sarah Verhemel4,
  5. Shantu Bundhoo1,
  6. Andrew D Yeoman5
  1. 1 Department of Cardiology, Royal Gwent Hospital, Newport, UK
  2. 2 Systems Immunity University Research Institute, Cardiff University, Cardiff, UK
  3. 3 Department of Gastroenterology, Cardiff and Vale University Health Board, Cardiff, UK
  4. 4 Department of Cardiology, Cardiff and Vale University Health Board, Cardiff, UK
  5. 5 Department of Gastroenterology, Royal Gwent Hospital, Newport, UK
  1. Correspondence to Dr Victor Galusko, Royal Gwent Hospital, Newport NP20 2UB, UK; vgalusko91{at}


Purpose of study This study examines the associations between dual antiplatelet therapy (DAPT) after percutaneous coronary intervention (PCI) and gastrointestinal bleeding (GIB), to explore possible predictors of outcomes.

Study design Retrospective analysis of 3342 patients who underwent PCI between 1 August 2011 and 31 December 2018 in a single centre was carried out. Oesophagogastroduodenoscopies (OGDs) for patients 12 months post-PCI were analysed.

Results Blood loss occurred in 2% of all (3342) patients post-PCI within 12 months. 128 patients (63% male, mean age (SD) of 69.8 (10) years) who had PCI subsequently underwent an OGD within 12 months of the index PCI procedure. GIB occurred within the first 30 days of DAPT in 36% (n=13/36) of cases. There were no thrombotic events associated with cessation of one antiplatelet agent. Increased age, haemoglobin (Hb) ≤109 g/L and Glasgow-Blatchford score ≥8 were associated with increased 12-month mortality. An Hb drop of ≥30 g/L was a sensitive and specific marker for significant pathology and evidence of bleeding on OGD (sensitivity=0.83, specificity=0.81).

Conclusions GIB bleeding occurred infrequently in the patients post-PCI on DAPT. Risk assessment scores (such as Glasgow-Blatchford and Rockall scores) are useful tools to assess the urgency of OGD and need for endoscopic therapy.

  • endoscopy
  • bleeding disorders & coagulopathies
  • ischaemic heart disease
  • anticoagulation

Data availability statement

Data are available upon reasonable request. Summary of the data available in the tables. Original data available on request.

Statistics from

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.


Patients undergoing percutaneous coronary intervention (PCI) require dual antiplatelet therapy (DAPT) in the form of aspirin and a P2Y12 inhibitor (clopidogrel, prasugrel or ticagrelor) for a period of up to 12 months.1 While this reduces the risk of adverse cardiac events, it increases the risk of bleeding2 with few predictive tools available.

Although aspirin and P2Y12 inhibitors both block platelet aggregation, they act on different platelet receptors. Aspirin irreversibly inhibits the cyclo-oxygenase (COX) enzyme which is responsible for generating thromboxane A2, a potent activator of platelets. However, blocking COX adversely reduces the production of prostaglandins in the stomach which impairs mucosal protective mechanisms and can lead to complications such as mucosal erosion or ulceration.3 In contrast, P2Y12 inhibitors target the platelet ADP receptors and have no direct action on the gastrointestinal tract. Consequently, P2Y12 inhibitors have been described to have a safer profile as demonstrated by reduced gastrointestinal bleeding (GIB) with clopidogrel compared with aspirin.4

Cardiac centres have varying preferences for combinations and duration of DAPT driven by a personalised approach to patient care in order to balance the risk of bleeding versus ischaemic events. When bleeding occurs on DAPT, it often necessitates its interruption. DAPT interruption may also be needed before performing urgent elective oesophagogastroduodenoscopy (OGD) when an invasive procedure needs to be performed (depending on the patient’s risk). Despite this, the evidence base guiding the cessation of DAPT in these circumstances remains weak.5

Previous studies investigating DAPT interruption for OGD after PCI demonstrated a greater than threefold increase in adverse cardiac events, but suggested that interruption of one agent can be done safely.6 However, very few patients included in these studies were on the newer P2Y12 inhibitors (prasugrel and ticagrelor), which have been shown to decrease ischaemic events further but come at a cost of increased bleeding.7 8 In addition, there are no studies reporting on OGD findings in patients on triple therapy (DAPT+oral anticoagulant), a combination that is also recognised for higher bleeding risk.9

The aim of the study is to identify indications for OGD following PCI, to describe rates, predictors and outcomes of bleeding on DAPT/triple therapy and the implications of DAPT interruption.


A retrospective analysis of 3342 patients who underwent PCI between 1 August 2011 and 31 December 2018 in a single centre was carried out. This was linked to the endoscopy database to identify patients who underwent an OGD within 12 months following PCI. Patient demographics, comorbidities, coronary disease patterns, PCI, DAPT/triple therapy agents, CRUSADE and HAS-BLED scores were extracted. For patients undergoing OGD, we obtained the dates of procedure, indications, serum haemoglobin (Hb) before and after OGD, Glasgow-Blatchford scoring, Rockall scoring (post-OGD), modifications in DAPT treatment, adverse events following the modifications in DAPT treatment and bleeding in the subsequent 12 months following OGD. Bleeding/rebleeding was defined as significant Hb drop or overt signs of bleeding requiring hospitalisation as per the Bleeding Academic Research Consortium type 2 or above. Hb differences were calculated from the blood on admission and compared with the previously stable Hb level measured in g/L on discharge post-PCI. OGD findings that were classed as ‘significant’ included: gastric/duodenal ulcers (non-healed), malignancy and varices. All causes of mortality were taken into analysis obtained through electronic patient records.

Patients were divided into two groups for subsequent subanalysis, in order to compare the characteristics of the group of patients who had clinical evidence of bleeding (as defined above) with patients who did not. Patients with large portions of data missing were excluded from the study.

Analysis was carried out in SPSS Statistics 26 (IBM) using t-test for parametric data and the Mann-Whitney U test for non-parametric data. Χ2 and Fisher’s exact tests were used for the analysis of the binary data. Multivariate logistic regression was carried out with RStudio (Open Source, V.1.2.5033) to investigate the association of outcomes with multiple covariates. Risk ratios were calculated for the different DAPT and triple therapy agents. Receiver operating characteristic curves were constructed and optimal cut-off values selected using Youden’s index (J).

The study was conducted as part of ongoing service evaluation of DAPT, PCI and OGD, aiming to improve the quality patient care. Local governance regulations were satisfied since the use of this data was anonymised and used to evaluate services.


Retrospective analysis of our database identified 128 (3.8%) patients out of a total of 3342 PCI patients who had an OGD within the subsequent 12 months. A total of 89.9% (n=115/128) of the PCI patients had a drug-eluting stent. A flow chart for patient selection is shown in figure 1. PCI was performed for acute coronary syndrome (ACS) in 59% of patients (n=75) and for the remainder 41% (n=53) the indication was stable angina. The majority of patients were on DAPT (88%) and 12% of patients were on triple therapy. Patients were all on DAPT for 12 months as standard, with a few exceptions where it was a minimum of 3 months (n=2) and 6 months (n=2).

Figure 1

Flow chart for processing and identifying relevant gastroscopies for analyses. Of the identified 3342 patients who underwent PCI in this period of time, 278 endoscopies were performed within 12 months of PCI. Of these, 163 were OGDs, and 128 initial OGDs were selected for final analysis. OGDs, oesophagogastroduodenoscopies; PCI, percutaneous coronary intervention.

Overall, 24% (n=31) of OGDs were performed for anaemia, 19% (n=24) melaena, 15% (n=19) dysphagia, 18% (n=23) dyspepsia and 5% (n=6) haematemesis. Other indications included abdominal pain and nausea or vomiting. Median PCI to OGD time was 144 days, ranging from 0 to 365 days. Median time to bleeding was 61 days.

The bleeding resulted in at least temporary interruption of DAPT. Aspirin was stopped permanently in 14 patients, while the P2Y12 inhibitors were stopped permanently in six patients (clopidogrel (n=4) and ticagrelor (n=2)). When aspirin was stopped, patients were continued on clopidogrel (n=9) or ticagrelor (n=5) single therapy. No adverse thrombotic events were observed.

PCI data

Patients were subdivided into two groups: (1) those with clinical evidence of bleeding and (2) those without. Baseline characteristics of the populations are displayed in table 1.

Table 1

Baseline characteristics of the study population divided into a group of patients presenting with clinical evidence of bleeding and without

Patterns of coronary disease and rates of stent implantation were similar in both groups. While CRUSADE scores in both groups were similar, the HAS-BLED scores were higher for the patients who then had clinical evidence of bleeding. CRUSADE score of ≥33 had the optimal predictive value for rebleeding (sensitivity=0.75, specificity=0.56, positive predictive value (PPV)=0.15, negative predictive value (NPV)=0.96) and 12-month mortality (sensitivity=0.75, specificity=0.56, PPV=0.15, NPV=0.96).

Clopidogrel was more prevalent in the group without clinical evidence of bleeding (73% vs 53%, p=0.008), while ticagrelor predominated in the group with evidence of bleeding (41.2% vs 17%, p=0.002). Figure 2 demonstrates the increasing rates of bleeding observed with the respective P2Y12 inhibitors when used in combination with aspirin. There was no statistical difference for bleeding observed between aspirin monotherapy and in combination with clopidogrel (risk ratio (RR)=1.45, CI 0.22 to 9.3, p=0.70). A significant increase in bleeding was observed with the aspirin and ticagrelor combination as compared with aspirin and clopidogrel resulting in an increase relative risk (RR=2.26, CI 1.29 to 3.96, p=0.004). Prasugrel in combination with aspirin increased the relative risk of bleeding as compared with aspirin and clopidogrel combination (RR=3.22, CI 1.31 to 7.92, p=0.011).

Figure 2

Rates of bleeding observed on the different DAPT therapies. The rate of bleeding observed is increased with ticagrelor and prasugrel as compared with clopidogrel. DAPT, dual antiplatelet therapy.

An upward trend in the relative risk for bleeding (RR=1.55, CI 0.77 to 3.09, p=0.22) and 12-month mortality (RR=2.51, CI 0.76 to 8.26, p=0.13) was observed with triple therapy as compared with standard DAPT (figure 3), but there was no statistical difference detected.

Figure 3

Rate of bleeding and 12-month mortality events observed in patients on DAPT and triple therapy. P values for both bleeding (p=0.25) and 12-month mortality (p=0.14) were not statistically significant. DAPT, dual antiplatelet therapy.

Characteristics of bleeding events

Patients who had clinical evidence of bleeding underwent OGD more urgently and as inpatients (68%). While patients without clinical evidence of bleeding mostly (86%) had outpatient investigations. Median time to bleeding was 61 days with 36% (n=13/36) of patients who had a bleed, experiencing it in the first 30 days. PCI to OGD time was quicker in the patients with evidence of bleeding (median 119 days). A haemostatic procedure was performed during OGD in 20.6% of patients with clinical evidence of bleeding, while none of the patients without clinical signs of bleeding required this (p<0.001).

The baseline Hb for the two patient groups was also similar. However, an Hb of ≤109 g/L had a sensitivity=0.73, specificity=0.85, PPV=0.31, NPV=0.97 for predicting 12-month all-cause mortality (online supplemental figure 1A). Following multivariate analysis, this parameter was independently associated with 12-month mortality along with left ventricular systolic dysfunction, smoking and cerebrovascular accident/transient ischaemic attack (figure 4).

Figure 4

Multivariate logistic regression analysis for the risk factors associated with 12-month mortality. Significantly different ORs are displayed in red, and include LV systolic dysfunction (LV SD), history of CVA/TIA, smoking, Hb ≤109 g/L (at baseline) and Glasgow-Blatchford (GB) score ≥8. AF, atrial fibrillation; CKD, chronic kidney disease; CVA, cerebrovascular accident; Hb, haemoglobin; HTN, hypertension; IHD, ischaemic heart disease; LV, left ventricular; TIA, transient ischaemic attack; T2DM, type 2 diabetes mellitus.

Mean Hb drop (∆Hb) was higher among the bleeding patients. A drop of ≥15 g/L Hb had a sensitivity=0.78, specificity=0.84, PPV=0.65, NPV=0.91 for predicting clinical bleeding, however was not as accurate at predicting the 12-month mortality. Following multivariate analysis, this parameter was independently associated with bleeding on OGD (figure 5).

Figure 5

Multivariate logistic regression analysis for the patient cohort with clinical evidence of bleeding. Significantly different ORs are displayed in red and include HTN, ∆Hb ≥15 g/L and Glasgow-Blatchford (GB) score of ≥8. AF, atrial fibrillation; CABG, coronary artery bypass grafting; CCF, congestive cardiac failure; CKD, chronic kidney disease; CVA, cerebrovascular accident; GI, gastrointestinal; Hb, haemoglobin; HTN, hypertension; IHD, ischaemic heart disease; LV, left ventricular; SD; systolic dysfunction; TIA, transient ischaemic attack; T2DM, type 2 diabetes mellitus.

Patients with clinical evidence of bleeding had more significant OGD findings (table 2). A ∆Hb of ≥30 had sensitivity=0.71, specificity=0.85, PPV=0.41, NPV=0.95% and 83% accurate for detecting significant findings on OGD (online supplemental figure 1B). An Hb drop of ≥30 g/L was also the optimal cut-off for detecting active bleeding on OGD (table 3) with an accuracy=81%, sensitivity=0.83, specificity=0.81, PPV=0.17, NPV=0.99 (online supplemental figure 2).

Table 2

Indications for OGD, Hb characteristics, Glasgow-Blatchford and Rockall scoring and findings in patients presenting with clinical evidence of bleeding and without

Table 3

Changes to antiplatelet therapy, treatment and adverse events as a consequence for patients with clinical evidence of bleeding on assessment and without

Glasgow-Blatchford and Rockall scores as well as the units of packed red cells transfused were significantly higher in the group which had clinical evidence of bleeding. Mortality at 12 months for the patients with evidence of bleeding was 8.3% (n=3/36).

Glasgow-Blatchford score of ≥8 predicted 12-month mortality (accuracy=79%, sensitivity=0.83, specificity=0.79, PPV=0.29, NPV=0.98) and was closely associated with rates of rebleeding observed (online supplemental figure 3A,B) (accuracy=42%, sensitivity=0.75, specificity=0.52, PPV=0.11, NPV=0.95). Rockall score ≥4 had sensitivity=0.58, specificity=0.64, PPV=0.14, NPV=0.94% and 63% accuracy for 12-month mortality; and sensitivity=0.83, specificity=0.66, PPV=0.20, NPV=0.98% and 68% accuracy for rebleeding.


This is a retrospective, single-centre study describing the characteristics of patients requiring OGD within 12 months of PCI. We found that ∆Hb is a useful guide to the likelihood of detecting significant pathology on OGD as well as active endoscopic bleeding requiring haemostatic intervention following PCI. Furthermore, Glasgow-Blatchford and Rockall scores are useful in predicting 12-month all-cause mortality and rebleeding in patients on DAPT. Given the high NPVs of both Glasgow-Blatchford ≥8 and Rockall ≥4, a score below this carries reassuring prognosis. There was more bleeding observed on the more potent P2Y12 inhibitors emphasising the importance of bleeding risk assessment before choosing DAPT post-PCI.

Bleeding is a poor prognostic sign

Major bleeding post-PCI has been identified as an independent risk factor for mortality10 and was found to be associated with increased rate of major adverse cardiovascular events (MACEs).11 Risk factors associated with higher risk of bleeding include increased age, renal failure, low baseline Hb, heart failure and haemodynamic instability among others, while prior proton-pump inhibitor (PPI) use was protective.12 Historically, it was not well delineated whether bleeding is an independent risk factor or whether these patients are inherently more frail and therefore at higher risk of death and MACE. A review analysing 42 studies identified that in fact many studies did not adjust for potential confounders but when reanalysed, taking these into account, still demonstrated a threefold increase risk in death.10 These and other studies have looked at all causes of bleeding and it remains unclear whether the association identified between upper GIB and mortality is an independent one.

Traditionally, a number of scoring systems have been developed to assess the risk of bleeding during the peri-procedural period of PCI including CRUSADE scoring. More recently, newer tools have emerged allowing binary classification of patients into being either low-risk or high-risk groups.13 Such risk assessments can be helpful allowing appropriate selection of DAPT/triple therapy post-PCI.

Prevention is better than cure

As demonstrated previously, more potent DAPT reduces the risk of ischaemic events but increases the risk of GIB.14 Both GIB and ACS carry significant morbidity and mortality and therefore it is important to find the appropriate balance between the two to reduce the overall risk to the patient.15 Recent evidence has emerged that reduction of DAPT duration in low-risk and high-risk patients is associated with a significant reduction of bleeding without resulting in increased ischaemic events.16 The TWILIGHT trial showed that after 3 months of DAPT, a P2Y12 alone (without aspirin) remained effective and resulted in reduction of bleeding in high-risk patients.17

Tailoring DAPT to the risk of individual patient’s bleeding risk may be a useful strategy to reduce the bleeding observed subsequently and may minimise the need to do OGD. CRUSADE scores and HAS-BLED scores may be useful to risk-stratify patients who will be prone to bleeding on DAPT and triple therapy. Although CRUSADE scoring is verified post-ACS and most predictive of major bleeding,18 we did not find a significant difference between the scores in the patients who experienced a bleed and those who did not. Despite this, higher scores ≥33 were predictive of mortality and rebleeding, which is consistent with previous studies.19 The HAS-BLED scores of patients who experienced bleeding were significantly increased in our study. This is in line with previous data that found them to be predictive of death and major bleeding in the context of PCI.20

We demonstrate here that Hb levels and ∆Hb are reliable predictors of 12-month mortality and bleeding. Previous studies show that 6% of patients had haemostatic intervention within 30 days of PCI and the median time to bleeding was 145 days.6 In our study, we observed 36% of bleeding occurring within 30 days of PCI with a median of 61 days. More intense monitoring around the critical period of time post-PCI may highlight the patients at risk of bleeding. A baseline Hb can be used as a tool to guide the intensity of monitoring required, given the high NPV of Hb >109 g/L for predicting 12-month mortality. If significant Hb drops are observed, it could be an early indication for urgent elective OGD to be organised allowing for early intervention. Early cessation of one antiplatelet agent may be clinically indicated to reduce the likelihood of significant bleeding and appears to have acceptable consequences in our study and previous literature.6 Despite this, clinical trials of a sufficient statistical power are still required to confirm these findings.

Does it matter which antiplatelet agent to stop?

We observed that it was common practice to temporarily stop or withdraw antiplatelet agents, and administer blood transfusion when significant blood loss occurred. We found that aspirin was the most often stopped (70.6%) antiplatelet agent during GIB. Aspirin is typically associated with GIB and side effects.21 This bleeding risk carries a substantial risk of death.22 23

A recent meta-analysis of antiplatelet monotherapy for stable coronary artery disease showed that there was no difference between MACE and bleeding for aspirin versus clopidogrel single therapy for the treatment of stable coronary disease.24 This appears to differ with patients who have undergone PCI. With these patients, we observe a decrease in ischaemic events, but a compensatory increase in bleeding with increased potency P2Y12 inhibitors. Recent reviews advocate for continuation of aspirin therapy and withdrawal of P2Y12 inhibitors as the adverse effects of gastric irritation can be protected by the use of PPIs.25

We did not observe any adverse events from stopping a single antiplatelet agent however due to the relatively low event rates, conclusions cannot be made on which antiplatelet to withdraw. Our study is consistent with previous literature where one antiplatelet agent was safest to withdraw from a thrombotic risk perspective, as withdrawing both increases the thrombotic risk.6 This needs to be balanced clinically against the severity of bleeding. Further studies are needed to evaluate the safety of withdrawing the different antiplatelet therapy agents.

Who requires OGD on DAPT?

Patients on DAPT who are admitted with a low Hb need a thorough risk assessment. Higher Glasgow-Blatchford and Rockall scores are well-known prognostic markers of poor outcomes and are associated with mortality and rebleeding.26 Glasgow-Blatchford score has been found to be at the very least non-inferior for predicting mortality and need for blood transfusion as compared with the Rockall score.27 28 In the latter study,28 a Glasgow-Blatchford cut-off score of ≥7 was predictive for the need for urgent OGD, endoscopic treatment and requirement for blood transfusion. Here we calculated that a score ≥8 was predictive of 12-month all-cause mortality and rebleeding, and was a good prognostic sign if <8.

Furthermore, we found that Hb drop of ≥15 g/L had an 82.3% accuracy for predicting bleeding. While a drop of ≥30 g/L was 83% accurate for predicting significant endoscopic findings and 81% accurate for predicting endoscopic evidence of bleeding. While a Rockall score of ≥4 was 68% accurate for predicting rebleeding as compared with the 52% accuracy of the Glasgow-Blatchford score ≥8.


This is a single-centre, small, retrospective observational study exploring the parameters of patients who had bleeding within 12 months of PCI. Moreover, due to the relatively low event rate, there were large CIs reported on the multivariate analyses. Nevertheless, several risk factors continued to be significantly associated with outcomes. Furthermore, some records were incomplete and were therefore excluded from the analysis, which may impact on the overall interpretation of the data. Further bigger studies are needed to evaluate the effects of single antiplatelet, DAPT and triple therapy post-PCI and effects of cessation of therapy in order to minimise the bleeding risk. Larger prospective studies will be able to establish more accurate cut-off values for the risk assessment of bleeding and to triage the urgency of OGD on DAPT.


In our study, we found that up to 4% of the cohort of patients undergoing PCI required OGD within 12 months, with 2% because of anaemia or bleeding. Patients on more potent P2Y12 (prasugrel or ticagrelor) inhibitors and triple therapy are more likely to present with GIB. Bleeding was an indication for OGD in 28% of cases requiring permanent interruption of DAPT 44% of the time at the time of bleeding (median 61 days). A Glasgow-Blatchford score ≥8 is a reliable predictor of 12-month mortality and rebleeding, with a high NPV. An Hb of ≤109 g/L was a prognostic marker associated with increased 12-month mortality, whereas a ∆Hb ≥30 g/L was strongly predictive of detecting pathology on OGD as well as active bleeding. Single antiplatelet withdrawal after GIB did not result in any major adverse cardiac events in this retrospective cohort.

Main messages

  • Risk stratification scores for bleeding should be used in patients started on dual antiplatelet therapy (DAPT) in order to choose appropriate therapy and its duration.

  • Only 2% of all patients started on DAPT presented with evidence of gastrointestinal (GI) bleeding within 12 months.

  • A haemoglobin drop of ≥30 g/L was a sensitive and specific marker independently associated with significant pathology and evidence of bleeding on oesophagogastroduodenoscopy (OGD).

  • Glasgow-Blatchford and Rockall scores are useful tools to risk-stratify patients on admission.

What is already known on the subject

  • DAPT initiation post-percutaneous coronary intervention can be associated with an increased rate of bleeding.

  • More potent P2Y12 inhibitors have been found to be associated with higher rates of bleeding.

  • When bleeding occurs from the GI tract, patients will undergo endoscopic investigations.

  • Triaging the urgency of OGDs can be challenging.

Data availability statement

Data are available upon reasonable request. Summary of the data available in the tables. Original data available on request.

Ethics statements

Patient consent for publication



  • Twitter @VGalusko

  • Contributors VG, MP, SB and HNH were involved in data collection. VG and MP were involved in data processing. All authors were involved in the writing of the manuscript and its final approval.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

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

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.