Introduction

Gastrointestinal bleeding (GIB) complications (such as the need for blood transfusion, hospitalisation, medium–short-term mortality) in atrial fibrillation (AF) patients treated with direct oral anticoagulants (DOACs) remain unclear [1, 2].

Landmark trials have shown that DOACs are not inferior in thromboembolic risk prevention compared to warfarin in AF [3,4,5]. However, the selected patients showed slightly higher GIB risk [6, 7], especially those aged > 75 years who were treated with concomitant antiplatelet therapy [6, 8]. A meta-analysis published in 2013 by Holster et al. that included previous randomized controlled trials (RCTs, total 75,081 patients) proved that DOAC users had a higher risk of GIB compared to patients treated with standard care, with an odds ratio (OR) of 1.45 [9].

Post-approval observational studies performed with administrative datasets have shown conflicting results about GIB risk. Some studies proved greater safety with warfarin [10, 11]; others described a non-significant difference between DOACs and warfarin [12, 13]. Hence, the original aim of these studies was not to assess the outcome of GIB. Their main focus was evaluating the efficacy of these drugs in reducing stroke incidence [12,13,14]. Therefore, these studies did not use a standard definition of GIB, which led to reduced data interpretation. Moreover, patients enrolled in clinical trials are frequently not representative of real-world practice. These groups often have strict inclusion and exclusion criteria, leading to complicated interpretation of the data from all of our patients [13, 14].

A recent observational real-world study found that warfarin users had fourfold risk of GIB and described lower intra-hospital mortality in DOAC-treated patients [14]. However, there is some evidence on the prognosis in DOAC-treated patients with GIB.

Here, we aimed to evaluate the difference in 30-day mortality risk between GIB patients treated with DOACs or warfarin.

Materials and methods

We considered all patients who were evaluated at the emergency departments (ED) of the Policlinico Universitario (Verona, Italy), Ospedale Civile Maggiore (Verona, Italy) and Ospedale Girolamo Fracastoro (San Bonifacio, Italy) for GIB. These EDs all have a 24-h endoscopy service. The analysis spanned from 1 January 2016 to 31 December 2017.

GIB was defined according to the International Society of Thrombosis and Haemostasis and subsequently divided into major bleeding and/or clinically relevant non-major bleeding [15, 16].

Clinically major GIB was defined as overt or occult GI blood loss resulting in hospitalisation and was associated with a decline in haemoglobin by ≥ 2 g/dl, hemodynamic instability (systolic blood pressure < 90 mmHg or heart rate > 100 beats per minute) within 24 h of presentation, and/or the need for endoscopic evaluation, angiography or surgery [15].

Clinically relevant non-major GIB was defined as any sign or symptom of haemorrhage that did not fit the criteria for the definition of major bleeding but that did meet at least one of the following criteria: required medical intervention by a healthcare professional and leading to hospitalisation, increased level of care or prompting a face-to-face (i.e., not just telephone or electronic communication) evaluation [16].

During the study period, all patients with ICD-9-CM (International Classification of Diseases, Ninth Revision, Clinical Modification) diagnosis codes related to GIB were extracted using First Aid software (Dedalus Healthcare System Group, Firenze, Italy). All identified cases were then evaluated via manual review of the patient’s medical records to ascertain presence of GIB and concomitant oral anticoagulant therapy. Patients with assessed GIB were divided into two groups according to the ongoing therapy: DOACs and warfarin. We excluded patients treated with anticoagulant therapy for other reasons than AF (i.e., mechanical heart valve, previous pulmonary thromboembolism).

Baseline demographic characteristics, comorbidities, pharmacologic therapy and laboratory data were recorded at the moment of manual revision of the ED folders.

The primary outcome was 30-day mortality from ED evaluation. The survival rate was obtained directly from the registry office.

The secondary outcomes were the need for blood transfusion in ED, urgent endoscopic evaluation and hospital admission after ED stabilisation. This retrospective observational study was performed according to the Declaration of Helsinki and under the terms of the relevant local legislation.

Statistical analysis

Continuous variables are reported as the median and interquartile ranges. Categorical variables are reported as the percentage and number of events.

The two treatment groups (DOACs vs. warfarin) were compared to evaluate any possible imbalance in the past medical history or clinical characteristics. Here, the use of propensity score matching had to be considered to equilibrate the two groups and to obtain a homogeneous cohort of patients for prognostic evaluation.

The 30-day mortality is reported as the percentage and number of total events. Comparison with the clinical and past medical history variables was performed using the Mann–Whitney U test and with Fisher’s exact test, as appropriate. Cox regression, adjusted for all variables that were significant to the previous univariate analysis, was performed to verify differences in mortality between the two treatment groups. Finally, the Kaplan–Meier method was used to compare 30-day survival between DOAC and warfarin users. Each analysis was considered significant at p < 0.05. The statistical analysis was performed using STATA 14.0 (StataCorp, College Station, Texas, USA).

Results

During the study period, 284 patients with ongoing oral anticoagulant therapy and concomitant GIB were evaluated. We excluded 34 patients with no confirmed bleeding at the moment of the visit, 7 patients because they lacked past medical history data and 5 patients without follow-up.

Among the 284 patients, 39.4% (112/284) were treated with DOACs, and 60.6% (172/284) were treated with warfarin. Among the DOAC-treated patients, 29.5% (33/112) were treated with dabigatran, 25.9% (29/112) with rivaroxaban, 37.5% (42/112) with apixaban and 7.1% (8/112) with edoxaban.

Within 30 days from ED evaluation, 8.1% of the patients (23/284) died.

Table 1 lists the demographic, past medical history and clinical characteristics of the two treatment groups (DOACs vs. warfarin) (Table 1).

Table 1 Clinical history, demographic and laboratory data recorded in 284 patients admitted to the ED for a GIB
Full size table

Propensity score matching was not needed, as the two groups did not show significant differences except for the concomitant platelet therapy (11.6% vs. 2.7%, p = 0.007) (Table 1). Table 2 shows the past medical history and clinical characteristics recorded at ED admission and 30-day mortality univariate analysis. The factors associated with 30-day mortality risk were age, history of chronic renal disease, active cancer, HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/alcohol concomitantly) value, and bleeding type (major GIB) and site (upper GIB). After the univariate analysis, neither of the two anticoagulant treatments resulted in higher 30-day mortality risk (warfarin 8.7% vs. DOACs 7.1%, p = 0.824) (Table 2).

Table 2 Univariate analysis for predicting 30- and 90-day mortality after ED admission for GIB
Full size table

Cox analysis adjusted for age, chronic renal disease, major GIB, upper GIB and baseline HAS-BLED showed no difference in mortality within 30 days of the GIB episode between the two groups [p = not significant (ns)] (Fig. 1). The Kaplan–Meier curves showed no difference in 30-day survival after the GIB episode between warfarin and DOAC users (p = ns) (Fig. 2).

Fig. 1
figure 1

Cox regression analysis of 30-day morality risk between DOAC-treated patients versus warfarin-treated patients

Full size image
Fig. 2
figure 2

Kaplan–Meier curves for differences in 30-day survival between DOACs and warfarin users

Full size image

No difference was recorded between warfarin and DOACs users for secondary outcomes (Table 1).

Discussion

This real-world retrospective study involved 284 episodes of acute GIB. All patients had concomitant anticoagulant therapy (DOACs, 112 vs. warfarin, 172) for preventing thromboembolism in non-valvular AF.

No difference between DOAC- and warfarin-treated patients was recorded in 30-day mortality after an acute GIB episode evaluated in ED. Even after adjustment for variables associated with the risk of 30-day mortality (age, chronic renal disease, major GIB, upper GIB, HAS-BLED), no differences were recorded within 30 days of the GIB episode between the DOAC and warfarin users.

GIB is a common problem that ranges from self-limited bleeding to haemorrhagic emergency [17]. In the latter, patients can present hemodynamic instability with altered mental status and rapidly progress to generalised shock [17, 18]. Even haemodynamically stable patients need accurate diagnostic evaluation and clinical follow-up [19]. The estimated global mortality after a GIB is > 10% [20, 21] and its severity is influenced by old age [22, 23] and comorbidities [22, 23]. Oral anticoagulant therapy (warfarin) represents a risk factor for the prognosis of patients with GIB, with a fatality rate of almost 6% [24, 25].

Interestingly, several studies have evaluated the GIB risk for DOACs compared with warfarin in patients with non-valvular AF, and reported contradictory findings [14, 26, 27]. Moreover, none of these studies explored prognostic differences after an acute episode of GIB in patients receiving oral anticoagulant, comparing DOACs and warfarin users [14, 26]. In the Randomised Evaluation of Long-term anticoagulation therapY, dabigatran (RE-LY) study, patients treated with dabigatran (150 mg) has higher GIB risk compared with warfarin-treated patients [hazard ratio (HR): 1.50, 95% confidence interval (95%CI): 1.19–1.89, p < 0.001] [3]. Similarly, the Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in AF, rivaroxaban (ROCKET AF trial) reported an increased rate of GIB in patients taking rivaroxaban compared to warfarin (HR 1.42; 95%CI 1.22–1.66) [5]. However, in the Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation (ARISTOTLE trial), the risk of major GIB with 20 mg apixaban was similar to that of warfarin, but patients taking apixaban rather than warfarin had a lower risk of non-major bleeding (HR 0.69; 95%CI 0.63–0.75) [4, 28]. Lastly, the ENGAGE study (edoxaban) showed a higher GIB risk for high edoxaban doses compared to warfarin (HR 1.23, 95%CI 1.02–1.50, p = 0.03) [29]. These landmark trials were not originally designed to assess the outcome of GIB, and no standard definition of GIB was used across these studies, restricting the data interpretation [14, 26]. Moreover, none of these landmark studies has specified prognosis or mortality after acute GIB during the follow-up of DOAC- or warfarin-treated patients [3,4,5, 28].

A real-life retrospective study in New Zealand reported that dabigatran-treated patients had slightly higher annual GIB rate compared to the warfarin-treated patients (3.70% vs. 3.35%), but the two treatments had similar relative risk for GIB [30]. The revision of other national registries demonstrated no differences between DOACs and warfarin, except DOACs in older patients and with higher dosages [31, 32].

In a retrospective study of 46,163 patients (dabigatran, 4907; rivaroxaban, 1649; warfarin, 39,607), Chang et al. demonstrated that the dabigatran group had higher GIB incidence (9.01 per 100 person-years), but the cumulative risk for GIB adjusted for confounding variables was not different for the DOAC and warfarin users [12]. Abraham et al. showed that, in a cohort of 758 patients well adjusted for the past medical history, the relative risk for GIB for dabigatran and rivaroxaban was 0.79 (95%CI: 0.61– 1.03) and 0.93 (95%CI: 0.69–1.25), respectively, compared with warfarin [13].

Recently, Cangemi et al. performed an observational study (163 GIB cases) and showed that GIB risk was > fourfold higher for patients using warfarin compared to those using DOACs (2.5% vs. 0.6%, OR   4.13; 95%CI 1.69–10.09) [14]. Furthermore, they show for the first time some prognostic indications after GIB in warfarin users with higher transfusion incidence (64.6% vs. 20%, p = 0.042), longer hospital stay (mean 7.7 vs. 3.8 days, p = 0.068) and higher 90-day mortality in hospitalised warfarin users (warfarin group, 7.6%) [14].

The present study has some interesting features. It is a real-life study performed in three different centres with an endoscopy unit working 24 h a day, analysed a high number of GIB patients, compared warfarin to all DOAC therapies available (dabigatran, apixaban, rivaroxaban, edoxaban) and explored short-term mortality. Unlike previous analysis based on anticoagulated patient registries, the present study directly considers individual episodes of acute GIB in the ED. Moreover, we considered only AF anticoagulation patients compared to previous studies that considered all therapeutic anticoagulation indications, thus including patients with higher warfarin dosage (i.e., mechanical heart valve) and patients with major bleeding risk (i.e., pulmonary thromboembolism).

The cohort of patients enrolled did not show any differences in baseline characteristics between DOACs and warfarin users. Therefore, there was no need for statistical matching (for example, propensity score matching).

To the best of our knowledge, our results are the first to explore the prognosis of DOAC-treated patients with GIB. The short-term mortality is aligned with the mortality data in other studies in the case of GIB in patients treated with oral anticoagulants (warfarin, anti-vitamin K [AVK]) [19,20,21,22,23].

Limitations

Our study has some limitations. First, the retrospective design exposes it to biases typical for these studies. Second, major GIB and clinically relevant non-major GIB were initially distinguished but later grouped under one single outcome. Recent studies have suggested that all non-major bleeding occurring during oral anticoagulant therapy are not trivial for either patients or health care systems [16, 19, 33,34,35]. Finally, we considered every therapeutic procedure that was carried out in the ED. Subsequent therapeutic procedures after patient stabilisation were not considered for statistical analysis.

Despite these limitations, the lack of prognostic information in the case of GIB in patients treated with oral anticoagulants is important data for ED evaluation of the anticoagulated patient. The rapid diffusion of these drugs should involve the gathering of more real-life evidence. However, further studies should take into account subsequent bleeding therapies in anticoagulated patients with GIB.

Conclusions

The present study shows no difference in short-term mortality between DOAC and warfarin users. DOACs represent a real innovation of the last decade. Despite their rapid diffusion, the available evidence is lacking, and the GIB prognosis of anticoagulated patients remains unclear. We believe that all information from actual clinical settings are important, given the global burn of thromboembolic pathology and the growing anticoagulant necessity in populations with comorbidities.