Risk of Rebleeding, Vascular Events and Death After Gastrointestinal Bleeding in Anticoagulant and/or Antiplatelet Users

Carlos Sostres; Beatriz Marcén; Viviana Laredo; Enrique Alfaro; Lara Ruiz; Patricia Camo; Patricia Carrera-Lasfuentes; Ángel Lanas


Aliment Pharmacol Ther. 2019;50(8):919-929. 

In This Article


The management of patients who develop a GI bleeding event during AP or AC therapy represents one of the most frequent problems in clinical practice. Today, hospitalisation due to either upper or lower GI bleeding in patients taking these compounds is very common due to increased indications and a growing elderly population worldwide.[21,22] The dilemma in the medical management of these patients is whether to interrupt AP or AC treatment to better control the bleeding event or maintain treatment to reduce the risk of thrombotic complications. Furthermore, a bleeding event increases the risk of thrombotic events due to compensatory mechanisms.[21,23] Based on a risk-benefit evaluation with data from the UK databases, interruption of aspirin for any reason is associated with a negative outcome since the number of GI bleeding events avoided was far less than the number of thromboembolic events induced.[24] In addition, preventing thromboembolic events in CV and cerebrovascular diseases are considered clinically more important than the increased risk of GI bleeding, since GI bleeding is associated with less mortality, morbidity or physical impairment and can be controlled endoscopically or radiologically in most cases. On the other hand, AC therapy is indicated in a growing population with atrial fibrillation and/or valvular heart diseases. Clinical management of these patients should depend on the balance between their thrombotic and bleeding risks. During acute bleeding, direct oral ACs or warfarin are withheld to facilitate haemostasis, but this is associated with an increased risk of thromboembolic events within the first 90 days.[14–17]

Current guidelines and expert consensus reports agree that the evidence guiding the management of patients on AC or AP agents is limited.[19,20] Despite this, guidelines recommend restarting AP or AC therapy sooner than later, but the exact or optimal time to resume therapy has not yet been defined.[25–29] The few studies available evaluated these effects separately for AP or AC therapy and mostly focused on upper GI bleeding events.[10,14–17,30] However, today there are more patients being admitted to hospitals due to lower vs upper GI bleeding events.[31] In the current study, we report data from one of the largest cohorts of patients and longest follow-ups published so far. The study includes patients who develop either an upper or lower GI bleeding while taking AP, AC or both therapies. Patients in our study were older than in other studies and had a high comorbidity index, which we believe represents the current typical clinical scenario worldwide. The influence of the location of the index bleeding in restarting therapy was an unexpected finding. Despite that both are life-threatening conditions, patients with lower GI bleeding were more likely to resume therapy than those with upper GI bleeding, even though there are more effective measures to reduce the risk of lesions in the upper GI tract (eg PPI therapy).

Overall, we have found that patients on AP or AC therapy who develop either an upper or lower GI bleeding event and do not resume therapy have both short- (90 days after the event) and long-term increased risk of death and thromboembolic events compared with those who resumed therapy after the event. These data agree with a small but seminal randomised clinical trial of Chinese patients[10] focused on aspirin therapy that provided data on short-term (30-day) outcomes after the event. In our study, the outcomes (death, rebleeding and thromboembolic events) were observed for patients on either AP and AC therapy; however, the risk of rebleeding was only significant for patients taking ACs. The differences in these three outcomes were observed for both upper and lower GI bleeding. The few observational studies published so far focused mostly on the short-term prognosis after an upper GI bleeding, separately for patients taking ACs or APs. All of them pointed out a benefit in the CV prognosis of these patients when restarting treatment, but their results disagree in terms of the risk of rebleeding and death.[10–14,16,17,30,32] Here, we demonstrated that both therapies offer similar short- and long-term beneficial outcomes when resumed, but the risk of rebleeding seems higher for patients on ACs compared with AP therapies. Most of these studies were focused on nonvariceal upper GI bleeding and only a few on lower GI bleeding (one with aspirin in a Chinese population[11] and another with AP or AC therapy in the United Kingdom[32]). In our study, we report that these events in patients who do not resume therapy are similar for both types of bleeding, irrespective of bleeding location. The reduction of the risk of ischaemic events and death among AC and AP users points out the importance of restarting treatment in patients with high thrombotic risk.

One interesting aspect reported here is the location of the rebleeding event compared with the initial event and the impact of PPI therapy on this outcome. There were more lower GI bleeding initial events than upper GI (46.7% vs 38.7%), but the differences increased when considering rebleeding events (52.50% vs 32.7%). The significant drop in peptic ulcers as the cause of rebleeding can be attributed to PPI therapy, since most patients were on PPI therapy after hospital discharge and this was maintained during the follow-up. These data agree with our previous report that indicated, after hospital discharge, most patients on dual AP therapy were on PPIs and most GI bleeding events were located in the lower GI tract.[33]

Diagnosis of GI bleeding arising from the small bowel is increasing in the last years. No data regarding rebleeding, ischaemic events or death after an episode of middle GI bleeding are reported in the literature. In our cohort, only 35 patients were diagnosed of small bowel bleeding; however, it is highly probable that the majority of patients with obscure GI bleeding were also originated from the small bowel. The risk estimates provided here for this specific group are consistent with those reported for the overall group of patients with lower GI bleeding which includes those with small bowel and colonic bleeding, which reinforce the suspicious that most cases with obscure bleeding are actually lower GI bleeds and most probably from the small bowel.

The best time to resume therapy, once it has been interrupted, remains unclear. One study from Hong Kong reported that immediate resumption of aspirin for high-risk cardiac patients did not increase the risk of fatal haemorrhage but significantly improved the 30-day survival.[10] These data agree with our results since early resumption was associated with a 90-day and overall (up to 6 years) decreased risk of thromboembolic events without a clear increase in the risk of rebleeding or death overall. When these data were broken down by type of therapy, differences did not reach statistical significance but the frequency of events suggested patients taking ACs had more differences among those who resumed therapy early vs late for the three outcomes of interest. However, in patients treated with AP agents that tendency was only observed for ischaemic events. These findings should be further investigated in future studies. Patients who interrupted AP or AC therapy and resumed it had no differences in outcomes when the cut-off for resuming treatment was set at < 7 days. This indicates that 1 week may be a good window to stop and start therapy before hospital discharge without affecting outcomes in the bleeding patient taking an AP or AC. When considering the location of the initial bleeding event, differences in frequency were observed for upper but not lower GI bleeding events, and this should also be investigated in future studies.

No data are available for the optimal timing for the resumption of non-aspirin AP agents or double AP therapy. Our study included all type of AP agents; although, aspirin was the most frequent and the number of non-aspirin AP agents did not allow a meaningful analysis. However, the frequencies of outcomes with aspirin and non-aspirin AP agents were similar. In the same way, our population of patients treated with double AP therapy or those taking an AP + AC was small, close to 10% of the whole cohort of patients. This percentage of patients do not permit a meaninful analysis, but showed that patients on dual AP therapy who resume it early after a GI bleeding event had a lower mortality compared with the rest of the patients.

Our study has strengths and weakness. Strengths include reporting data on consecutive patients with GI bleeding taking any AC or AP, as this provides overall and comparative outcomes for AP or AC therapy in an approach similar to what happens in clinical practice. Furthermore, we provided data on both upper and lower GI bleeding, even though information on lower GI bleeding has been limited, and describe how the location of the rebleeding event may change after resuming therapy. We also added information concerning the optimal time to resume therapy, which can facilitate individualised clinical decisions. In addition, we provide data on both short- and long-term outcomes that can serve as a basis to establish official recommendations regarding the time of interruption of these drugs after a GI bleed. Weaknesses include the retrospective nature of the study. This implies that we can only provide data that have been recorded or investigated during the clinical episodes. Despite collecting information on the initial event prospectively, follow-up data from databases of computerised medical records were reliable means of collection that allowed complete monitoring of each patient with low data loss rates.

However, there is a number of facts that need to be considered. Some patients with previously excluded upper and lower GI bleeding could not receive an exhaustive search for the source of bleeding, which may imply that some patients could be misclassified. Also, definitions of types of GI bleeding could have been different. In particular, lower GI bleeding, which here includes bleeding from the small (actual or presumed) and/or large bowel, could have been split as bleeding from the colon and bleeding from the small bowel. As explained above, obscure GI bleeding has been considered of presumed small bowel origin. We have tried to overcome this problem by reporting risk estimates of events by different combinations of types of bleeding in Table 5. As in other studies of our group, we have not included nonvariceal oesophageal bleeding, which in our experience is infrequent, mild and often nonhospitalised. The cause of death could not be identified in 13.5% of cases, since they occurred outside the hospital setting. Also, although the number of patients included in the cohort was one of the largest available, when analyses were broken down by type of drug and type of bleeding, the power of the analysis decreased and could not provide reliable data, which are still needed in this field. Data on AC therapy refer almost exclusively to vitamin K antagonists, since the number of patients on direct AC agents was still very small. Also, data on AP agents mostly concerned aspirin. Finally, the range of days for resuming therapy may seem wide, but 98.5% of patients resumed therapy within the first 30 days (median of 6 and mean of 7.6 ± 6.4 days). Also the event CV rates reported are high, but we need to point out that population of our study is of advanced age, with a high number of comorbidities and that the window of years observation is the largest reported for this type of studies to date. When the number of events is limited to 90-days after the initial bleeding episode, the event rates are similar to other previously reported (14, 17).

In conclusion, our study highlights the importance of weighing the CV and survival benefits of restarting treatment with ACs or APs with the GI risk in patients who develop a GI bleeding event. This needs to be established in patients who develop either an upper or lower GI bleeding event, as rebleeding is more common in the lower GI tract, even if the initial event was located in the upper GI. Our study provides information to balance the risks and benefits of the global management of these two pharmacological groups of drugs widely associated with increased GI bleeding risk. We demonstrated that the risk of rebleeding is especially present in patients taking AC agents. Further research is needed to define more precisely the best management strategy for stopping and resuming therapy according to the drug type and type of lesion and based on the individual's GI and CV risks, but here we provide data suggesting that restarting therapy within the first week after the bleeding event is associated with overall benefits for the patient. Possibly, risk scores and tools will be needed to help clinicians make the most appropriate recommendation due to the multiple variables involved.