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

Materials and Methods

Patient Population

We conducted a retrospective long-term cohort study in two general hospitals in Spain that included patients from 18 to 89 years old who used AC and/or AP agents and required hospitalisation due to GI bleeding (upper, lower or obscure) event, from January 2008 to December 2013. GI bleeding was defined as a decrease in haemoglobin of ≥ 2 g/dL with at least one of the following signs: haematemesis, melaena, haematochezia, bright red rectal bleeding, endoscopic active bleeding or bleeding stigmata (signs of recent bleeding present in the endoscopic procedure). Upper GI bleeding was defined as bleeding from a source proximal to the ligament of Treitz. We have included cases of nonvariceal GI bleeding from the stomach and duodenum. Patients with haematemesis were always classified with upper GI bleeding. Bleedings from the oesophagus were not considered in this study as they are usually infrequent, mild and due to oesophagitis. Lower GI bleeding was defined as bleeding from a source distal to the ligament of Treitz. Those patients with a macroscopically evident GI bleeding event without an identifiable source (eg, no source identified after complete endoscopic examination or patients who did not undergo a complete endoscopic examination due to age, comorbidities or patient's choice) were classified as obscure GI bleedings.

Exclusion criteria were variceal or tumoural bleeding, recurrent bleeding within the last year of the current bleeding event, bleeding from Mallory-Weiss syndrome or oesophagitis, coagulopathy (excluding drug-induced), intrahospital/in-patient bleeding, cancer diagnosed within the last 2 years and cancer treated within the last 6 months. Patients who died within the first 48 hours of hospitalisation were also excluded.

Data Collection

Data concerning the index cases of GI bleeding were obtained prospectively using a previous validated interview.[5] Information was collected during hospital admission and included the date of the event, location of the bleeding (upper/lower/obscure), diagnostic procedures and final diagnosis. In addition, information concerning, demographics, comorbidities, clinical presentation and laboratory results was prospectively collected for all bleeders. Drug use information, including AP and/or AC agents, nonsteroidal anti-inflammatory drugs, proton pump inhibitors (PPIs) and other concomitant drugs, was also prospectively collected during the bleeding event. Drug management during hospitalisation was also recorded in terms of interruption, resumption, time of interruption and drug use at discharge. Follow-up data were obtained from electronic databases from two different Spanish Health Care areas. Information was independently recorded by three trained researchers (BM, VL, CS), and reviewed by a single member of the research group (AL). All clinical information obtained was recorded in a database using Microsoft Office Access (Redmont, Washington, USA, 2010).

Outcomes of Interest

The main outcomes of interest were the first recurrent GI bleeding and then death from any cause and CV events. Recurrent GI bleeding was defined as a readmission at any hospital for any of the following situations: (a) macroscopically visible GI bleeding (upper/lower) with or without (obscure) an endoscopic diagnosis, or (b) suspected GI bleeding with and acute and serious (>2 g/dL haemoglobin drop) from the last known haemoglobin level but without evidence of macroscopically visible GI bleeding when assessed by medical or hospital staff, and no other non-GI potential sources of haemoglobin drop (eg haematological, renal, etc) that required hospitalisation. CV events were defined as any of the following: (a) acute coronary syndrome, (b) stable angina requiring hospitalisation, stroke or transient ischaemic attack, (c) venous thromboembolism, (d) deep vein thrombosis or (e) arterial thromboembolism. Deaths were classified as due to a CV event, GI event or other causes. Deceases in CV or GI rebleeding events were also considered CV events and GI rebleeding events respectively. If death occurred in the 30 days following a CV or GI event, it was considered to be related to the event and included in its category. Patients were censored at the time of having one of the outcomes of interest or at the end of the follow-up period (at least 1 year after the last patient was included in the study). Analysis for the outcomes was performed 90 days after the index GI bleeding and at the end of the follow-up, which concluded on 31 December 2014.

All patients were assigned to different groups according to drug discontinuation at the time of the initial GI bleeding event ('interruption' and 'no interruption'). Patients who interrupted therapy were assigned to one of the two groups defined by drug resumption ('resumption' and 'no resumption'). We categorised the 'resumption' group according to the timing of the interruption, generating two different groups: early (≤7 days) and late (>7 days) resumption. We also performed sensitivity analysis for the different periods of drug resumption. Patients who did not resume therapy at discharge but restarted it during the first 90 days of the follow-up period were included in the 'resumption' group and the date of restart was recorded.

Patients were also assigned to different groups depending on their drug use at the moment of the index GI bleeding. Patients treated with oral ACs (ie acenocoumarol, warfarin, new oral anticoagulants) were included in the AC therapy group. Patients in the AP group were treated with different AP agents, including low-dose aspirin, triflusal, thienopyridine class of APs and other agents. Patients using both AP and AC agents were included in the AP + AC group. Finally, patients using two different AP agents were included in the dual AP therapy group.

Statistical Analysis

An initial exploratory analysis of all variables included in the study was carried out. Qualitative variables were expressed as frequencies and percentages, whereas quantitative variables were expressed as means (SD) or medians (IQR). We determined the person-time incidence rate for each event, defined as the number of new events per population at risk related to the at-risk time of each patient. Comparisons between groups for qualitative variables were made using chi-squared or Fisher exact tests. Means between two independent groups were compared with the Mann-Whitney U test, and the Kruskal-Wallis test was used when the qualitative variable had more than two categories. The Kolmogorov-Smirnov test was performed to assess the normal distribution of the variables.

Survival among different groups was estimated using Kaplan-Meier survival curves and compared using the log-rank test. Cox proportional hazards models were used to evaluate the association of therapy resumption with time to rebleeding, vascular events and death. The hazard ratio (HR) with 95% CI was used to estimate the strength of the association. These analyses were adjusted for gender, age, Charlson comorbidity index and therapy. The statistical analyses were performed using SPSS software 22.0 for Windows (SPSS Ibérica, Madrid, Spain). For all tests, a two-sided P-value less than .05 was considered statistically significant.