Introduction
Direct oral anticoagulants (DOACs) are recommended in management guidelines globally for atrial fibrillation (AF) as the preferred oral anticoagulant (OAC) to reduce the risk of stroke and thromboembolism.1–3 However, DOACs are associated with a risk of bleeding.4 After a bleeding event while on DOAC, clinicians face a dilemma between the fear of stroke while DOAC is interrupted and the risk of recurrent bleeding if DOAC is restarted.5
While it can be tempting to interrupt DOAC after a bleeding event, previous studies have shown both a substantial stroke risk and bleeding risk following both transient interruption6 and permanent discontinuation of OAC.7 The increased stroke risk after discontinuation might be explained by the short half-life of DOACs, meaning a rapid reversal of the anticoagulant effect shortly after treatment interruption.8 However, other explanations such as a thrombogenic effect or a rebound hypercoagulability effect of OAC-associated bleedings shortly after discontinuation have also been postulated.9
The problem is further enhanced by studies showing that a bleeding event itself can increase the risk of stroke but also recurrent bleedings, given shared risk factors for both.10 In fact, a bleeding event itself, for example, intracranial bleeding, entails increased risk of subsequent ischaemic stroke.11 12 In a prior report from the Danish registries, restart of Vitamin K Antagonists (VKA) after an intracranial bleeding reduced stroke risk.13 The same applies to restarting VKA after gastrointestinal bleeding, which showed a reduction in both mortality and thromboembolism without increasing recurrent gastrointestinal bleeding.14 The latter study by Qureshi et al investigated early (after 7 days) versus late (after 30 days) restart of VKA after gastrointestinal bleeding and found that early restart of VKA was associated with a decreased risk of death and thromboembolism and was not associated with an increased risk of gastrointestinal bleeding compared with late restart. Whether all these findings also apply to the restart of DOAC after a serious bleeding event has not been well investigated. There is limited evidence about which strategy is best when an AF patient experiences a bleeding event during DOAC treatment. It is thus unknown if it is best to continue DOAC, make a temporary interruption and for how long time, switch to another antithrombotic treatment or discontinue DOAC permanently.
In this study, we first investigated the risk of stroke and recurrent bleeding events in patients with AF who experience a serious bleeding event while on DOAC. Patients who had a short (maximum 60 days) treatment interruption period of DOAC after their first serious bleeding event (index bleeding event) were compared with patients who had a long treatment interruption period (more than 60 days). The 60-day period was chosen as most patients are expected to restart DOAC within 2 months after an initial bleeding event.13 Second, we investigated time-varying antithrombotic exposure periods with different antithrombotic regimens regarding risks of stroke and recurrent bleeding after index bleeding event while on DOAC in AF patients.
Methods
Data sources
The Danish National Patient Registry15 includes data on all hospital admissions and discharges since 1977 and outpatient contacts since 1995 in Denmark. This registry was thus used to obtain information on the incidence of AF and atrial flutter by using diagnostic codes from the International Classification of Diseases (ICD) system, 10th revision. The Danish National Prescription Registry16 possesses information on medications dispensed from all Danish pharmacies since 1995, including information on date of filling, strength and quantity filled, and was thus used to obtain information on prescriptions by using the Anatomical Therapeutical Chemical (ATC) classification system. The Danish Civil Registration System17 was used to obtain information on patients’ vital status, date of birth and date of death. The Clinical Laboratory Information System was used to obtain information on laboratory tests.18
In Denmark, every person is assigned a unique and permanent civil registration number on birth or immigration. An encrypted form of this civil registration number is used by the abovementioned registries, so we cross-linked information to obtain information on diseases, medications, laboratory data and vital status.
Study population
We used the Danish National Patient Registry to include all patients discharged alive with a first-time diagnosis of AF who initiated DOAC between 2012 and 2021. We then proceeded to only include patients who had experienced a serious bleeding event requiring hospital contact or admission while on DOAC and had a CHA2DS2-VASc score of ≥2 at baseline. The CHA2DS2-VASc score is a clinical tool used for estimating stroke risk in people with AF19 (see online supplemental material, table 1c). Higher scores are associated with higher stroke risks. The ATC and ICD codes used are listed in the online supplemental material, table S1.
Baseline comorbidities and medications
Comorbidities, including those related to the CHA2DS2-VASc score, were identified in the Danish National Patient Registry by using ICD codes from hospital contacts. Hypertension was defined as the use of at least two antihypertensive drugs defined by ATC codes.20 The time frame for baseline comorbidities used in the analysis of early versus late restart of DOAC was anytime a comorbidity was diagnosed before inclusion. The time frame for comorbidities used in the analysis of time-varying antithrombotic therapy was anytime during follow-up a comorbidity was diagnosed.
To find baseline medication, we used ATC codes to find medications in the Danish National Prescription Registry. For medications other than DOACs, prescriptions claimed during the 180 days prior to inclusion were included. Relevant ATC and ICD codes are shown in the online supplemental material, table S1.
Exposure
Early versus late restart of DOAC after index bleeding event
We investigated the effect of restarting DOAC within a maximum of 60 days after the index bleeding event requiring hospital contact or admission. Patients were thus divided into two groups according to whether they restarted DOAC no later than 60 days after index bleeding event (early restarters), and those who did not restart DOAC within 60 days from index bleeding event (late restarters). Figure 1 shows a flow chart for the inclusion of patients. To test the robustness of our study results, we performed a sensitivity analysis in which we changed the 60-day cut-off to a 30 s days cut-off.
Figure 1Flow chart for inclusion and exclusion of patients. CHA2DS2-VASc, Congestive heart failure, Hypertension, Age ≥75 (doubled), Diabetes, Stroke (doubled), Vascular disease, Age 65-74, and Sex category (female); DOAC, direct oral anticoagulant. * 11.565 patients were included in the analysis of risks of stroke and subsequent bleeding event associated with time-varying treatment periods of different antithrombotic treatment regimens after a first-time serious bleeding event ** 10.291 patients were included in the analysis of early versus late restart of DOAC therapy after a first-time serious bleeding event.
Time-varying periods of antithrombotic treatment
We investigated the risks of stroke and a recurrent bleeding event associated with a time-varying exposure, that is, time-varying antithrombotic treatment periods to which patients were exposed after their index bleeding event. Patients were thus allowed to switch antithrombotic treatment during follow-up. Exposure could thus vary over time, and follow-up time for each patient was subdivided into several exposure periods according to the actual antithrombotic treatment taken. Six exposure categories were adopted as explained in the section below. As we wanted to consider multiple exposures over time, time‐dependent changes of confounders under the follow-up period were taken into account.
Construction of treatment periods
Treatment periods were defined as periods with no antithrombotic therapy, periods on DOAC monotherapy, DOAC in combination with aspirin, DOAC in combination with a P2Y12 receptor antagonist, DOAC in combination with both aspirin and a P2Y12 receptor antagonist, and lastly, periods with aspirin and/or P2Y12 receptor antagonist.
First, follow-up time for each patient was split into periods according to confounders that could vary over time and which could affect the outcome. These time-varying confounders were chronic kidney disease (CKD), myocardial infarction, any malignant disease and becoming 65 years or 75 years old. Whenever any of these time-varying covariates occurred, the patient was classified into a new risk period. Then, these risk periods were subdivided into treatment periods depending on which antithrombotic treatment the patients were taking.
Date, strength and amounts of prescribed drugs are available in the Danish National Prescription Registry, but we had no information about the prescribed dose and duration. Thus, for each prescription, an exposure period was calculated based on information of date, strength and quantity purchased. The daily quantity used was estimated on the assumption that patients took one defined daily dose of the prescribed drug per day, which is the assumed maintenance dose per day for a drug used for AF as indication in adults. Overlapping refills were adjusted such that an overlapping refill date was shifted forward to start after the end of the previous refill’s supply was exhausted, based on the assumption that a new refill only started after all tablets from the previous refill were taken fully.
Study outcomes
The efficacy endpoint was the composite of ischaemic stroke or transient ischaemic attack (TIA), and the safety endpoint was a recurrent serious bleeding event requiring hospital visit or admission. Study outcomes could be fatal or non-fatal. A serious bleeding event (both index and recurrent bleeding event) was defined as the composite of gastrointestinal bleeding, urogenital bleeding, intracranial bleeding or epistaxis requiring hospital contact or admission.
In the analysis of early versus late restart of DOAC, patients were followed from day 60 after the index bleeding event until the occurrence of an outcome (ischaemic stroke/TIA or bleeding event), death, 1 year of follow-up or study-end (31 December 2021), whichever occurred first.
In the analysis of time-varying exposure, patients were followed from index bleeding event until the occurrence of an outcome (ischaemic stroke/TIA or bleeding event), death or study end (31 December 2021), whichever occurred first. Only events after day 30 (blanking period) from the index bleeding event were included, as events registered within the first 30 days from index bleeding event were potentially considered associated with the index bleeding event.
We also did a further analysis looking at a composite end point of ischaemic stroke/TIA and serious bleeding event requiring hospital contact or admission.
All outcomes were identified using ICD codes from hospital contacts. Thus, only stroke/TIA and serious bleeding events (ie, bleeding events that required hospital visit/admission) were included. The ICD codes used to define the study outcomes can be found in the online supplemental material, table S1.
Statistical analysis
Continuous variables are presented as means with SDs or medians with IQRs, as appropriate. The Kruskal-Wallis test was used for comparison of the different groups no matter distributional form. Categorical data are presented as numbers and percentages and were compared using the χ2 test. The ‘loglogplot’ function in the ‘survival’ package in R was used to assess the proportional hazards assumption.
In the analysis of early versus late restart of DOAC, follow-up time was calculated from 60 days after the index bleeding event to the date of an outcome of interest, death, 1 year of follow-up or study end, whichever occurred first. We used Cox proportional hazards regression models to calculate HRs with 95% CIs for the two groups, where patients who did not restart DOAC therapy within 60 days from the index bleeding event comprised the reference group.
In the analysis of time-varying antithrombotic treatment, follow-up time was calculated from the index bleeding event to the date of an outcome of interest, death or study end, whichever occurred first. The cumulative incidence of events was estimated using the Aalen-Johansen estimator, which takes into account the competing risk of death from other causes. We used Cox proportional hazards regression models with time‐varying covariates to adjust for time‐varying confounders, assuming that the time‐varying confounders were not affected by previous treatment periods (ie, they are not intermediates between the exposure and the outcome). We calculated HRs with 95% CIs for the six predefined treatment exposures, periods with no antithrombotic treatment being the reference group.
The Cox models were adjusted for sex, age, hypertension, CKD, chronic liver disease and alcohol abuse. For both analyses, we reported HRs, which were computed by using a cause-specific Cox regression model, considering death as a competing risk.
A two-sided p value threshold of p<0.05 for statistical significance was used. Data management and statistical analyses were performed using R (V.4.2.1 for Windows, R Foundation for Statistical Computing).
Results
Characteristics of the study population
There were 139 429 patients with first-time AF, who initiated DOAC therapy between 2012 and 2021. Of those, 12 282 patients experienced at least one serious bleeding event requiring hospital contact or admission while receiving DOAC. More detailed descriptions and more detailed results can be found here.21 Only patients with CHA2DS2-VASc score ≥2 at baseline were included, yielding a study population of 11 565 patients (figure 1).
Early versus late restart of DOAC after index bleeding event
Within the first 60 days after index bleeding event, 4955 bleedings and 389 strokes were registered (not included in the analysis, online supplemental material, table 2). In addition, 1274 patients died within the first 60 days after the index bleeding event (of whom 1014 patients had already restarted DOAC before death). Thus, 10 291 patients who survived the first 60 days after the index bleeding event were included for further analysis, with a median duration of follow-up of 1.0 year (IQR 0.6–1.0 years).
Of the 10 291 patients included in the study, 5970 patients restarted DOAC within 60 days (early starters), while 4321 did not (late starters) (figure 1). Table 1 summarises patient characteristics.
Table 1Baseline characteristics of the study population
As shown in table 1, early restarters had significantly lower prevalence of previous malignant disease compared with late restarters, 24.7% vs 26.7%, p<0.05. Otherwise, the two groups had comparable baseline characteristics regarding CHA2DS2-VASc score, HAS-BLED score, previous stroke/thromboembolism, CKD, previous peptic ulcer, liver disease, use of antiplatelet agents and other cardiovascular medications at baseline.
Mortality during follow-up stratified by group is depicted as Kaplan-Meier plots in online supplemental material, figure 1.
Stroke
Cumulative incidence rate of stroke is depicted as Aalen-Johansen plot stratified by group in figure 2. During the first year of follow-up, 242 patients experienced a stroke in the group of early restarters vs 194 patients in the group of late restarters, corresponding to an event rate of 4.7 and 5.2 per 100 patient-years, respectively (table 2). There was weak evidence of benefit in the form of reduced stroke risk in the group of early restarters compared with late restarters, adjusted HR 0.89 (95% CI 0.74 to 1.08). Changing the 60-day cut-off to 30 days resulted in an almost unaltered adjusted HR of 0.96 (95% CI 0.80 to 1.15) for stroke risk for early restarters compared with late restarts.
Figure 2Cumulative incidence rates of stroke, recurrent bleeding and composite event.
Table 2Event rates per 100 patient-years
Bleeding
Cumulative incidence rate of recurrent bleeding is depicted as Aalen-Johansen plot stratified by group in figure 2. During the first year of follow-up, 752 patients experienced a serious bleeding event in the group of early restarters vs 451 patients in the group of late restarters, corresponding to an event rate of 15.4 and 12.8 per 100 patient-years, respectively (table 2). Compared with the group of late restarters, early restarters had significantly increased risk of recurrent bleeding, adjusted HR 1.21 (95% CI 1.07 to 1.36). Changing the 60-day cut-off to 30 days resulted in an adjusted HR of 1.02 (95% CI 0.92 to 1.13) for risk of recurrent bleeding for early restarters compared with late restarts. As such, the risk of recurrent bleeding did no longer differ between the two groups.
Composite event
Cumulative incidence rate of composite event is depicted as Aalen-Johansen plot stratified by group in figure 2. Within 1 year after the index bleeding event, 954 patients in the group of early restarters and 609 patients in the group of late restarters experienced an event (stroke/TIA/bleeding), corresponding to an event rate of 20.0 and 17.6 per 100 patient-years, respectively. Compared with late restarters, early restarters had significantly increased risk of composite event, mainly driven by recurrent bleeding events, adjusted HR 1.13 (95% CI 1.02 to 1.26). Changing the 60-day cut-off to 30 days resulted in an adjusted HR of 1.02 (95% CI 0.93 to 1.12) for the risk of a composite event for early restarters compared with late restarts. Therefore, this change meant that the risk of the composite event was no longer significantly different between the two groups.
Time-varying exposure of antithrombotic therapy
Stroke
There was a total of 143 027 treatment periods before first stroke or end of follow-up; for detailed description of treatment periods including average treatment durations according to stroke as outcome of interest, see online supplemental material, table 4. Compared with periods with no treatment, only periods with DOAC monotherapy were associated with a significantly reduced stroke risk, HR 0.78 (95% CI 0.68 to 0.89) (figure 3).
Figure 3HRs for stroke, recurrent bleeding and composite event for different antithrombotic treatment regimens. DOAC, direct oral anticoagulant.
Bleeding
There was a total of 146 600 treatment periods before a recurrent bleeding event or end of follow-up; for detailed description of treatment periods according to recurrent bleeding as outcome of interest, see online supplemental material, table 5. All periods where patients received DOAC with or without aspirin and/or a P2Y12 receptor antagonist were associated with increased risk of bleeding compared with no treatment (figure 3). This also applies to periods with DOAC monotherapy, which were associated with a significantly increased risk of recurrent bleeding compared with periods with no treatment, HR 1.26 (95% CI 1.15 to 1.38).
Composite event
There was a total of 120 655 treatment periods before first event or end of follow-up. For detailed description of treatment periods, see online supplemental material, table 6. Most of the time, patients were on either DOAC monotherapy or received no antithrombotic treatment. All periods where patients received DOAC with or without antiplatelet therapy were associated with increased risk of composite event compared with no treatment periods (figure 3).
Discussion
To our knowledge, our analysis is the first to provide nationwide information on the trends of DOAC use in AF patients after a first-time serious bleeding event. In this large nationwide study, we investigated the effectiveness and safety of DOAC after a first-time serious bleeding event requiring hospital contact/admission while receiving DOAC in patients with AF. The main findings of our analysis are as follows: (1) There was weak evidence of benefit in the form of reduced stroke risk in the group who restarted DOAC within 60 days from index bleeding event compared with patients who did not restart DOAC within 60 days from index bleeding event. but they had a significantly increased risk of recurrent bleeding and (2) Compared with periods with no antithrombotic treatment, only periods with DOAC monotherapy were associated with a significantly reduced stroke risk. However, this was at the expense of the risk of recurrent bleeding, where periods with any (combination of) antithrombotic treatment were associated with increased risk of recurrent bleeding compared with periods with no antithrombotic treatment.
To our knowledge, our analysis is the largest to evaluate DOAC interruption among patients with AF after a first-time bleeding event.
Previous studies22–25 have investigated trends in temporary interruptions or permanent discontinuation of DOAC. In a population-based cohort study from UK primary care, they found that AF patients receiving DOAC had high discontinuation rates (defined as a gap in DOAC therapy of >30 days) within the first year, ranging between 26% and 40%, but that the vast majority resumed treatment after interruption and that less than 10% of patients stopped DOAC permanently.25 In contrast to our study, they included all interruptions/discontinuations in DOAC-treated AF patients and not only patients who interrupted/discontinued DOAC after a serious bleeding event. A bleeding event will usually influence patients’ DOAC therapy or prescribing patterns. A nationwide Danish registry-based study26 investigated reasons behind DOAC discontinuation and found that treatment changes during DOAC in AF are very common, and bleedings were among the leading causes for treatment change.26 In fact, it has been shown that up to 44% of discontinuations of DOAC treatment were due to bleeding events.27
The risks of stroke, major bleeding and recurrent gastrointestinal bleeding associated with antithrombotic treatment after a first-time gastrointestinal bleeding event in AF patients were investigated in a previous Danish nationwide study.28 The study found that the restart of OAC treatment reduced the risk of thromboembolism but increased the risk of major bleeding without increasing the risk of recurrent gastrointestinal bleeding significantly. In our study, we also found that periods with DOAC monotherapy after a serious bleeding event were associated with a lower stroke risk, but higher bleeding risk compared with periods with no treatment. This is despite a main difference between the two studies, where we only included patients receiving DOAC while the mentioned study included mainly VKA patients. Another difference is that the mentioned study used gastrointestinal bleedings as the index bleeding event, while we defined the index event as any event included in the definition of composite of bleeding events. Thus, the observed difference in bleeding risk could be due to the fact that we only included DOAC but not VKA treatment, we included all bleeding events as outcome of interest (not GI bleedings separately), and lastly, due to the definition of the index event, as we defined index bleeding as index event, not only GI bleedings.
The optimal treatment option for AF patients who survived an intracranial bleeding was also investigated by another Danish nationwide study.13 The study found that OAC restart was associated with a significantly lower risk of ischaemic stroke and all-cause mortality, and therefore, encourages restarting OAC after an intracranial bleeding as soon as possible. They also found that the risk of recurrent intracranial bleeding was not significantly different between non-OAC-treated and OAC-treated patients, which is in contrast to our study as we found a significantly higher risk of recurrent bleeding associated with DOAC restart. An explanation for this difference is that we looked at recurrent bleeding as a composite endpoint, not investigating every bleeding site separately. But again, the mentioned study included mainly patients receiving VKA, while we only included patients receiving DOAC treatment.
In a previous study with VKA,29 drug combinations involving antiplatelet agents and VKA were associated with a high risk of gastrointestinal bleeding beyond that associated with each drug alone. VKA in combination with an antiplatelet agent is associated with increased risk of both fatal and non-fatal bleeding compared with VKA monotherapy.30 This is in line with our analysis, where the risk of recurrent bleeding was highest if DOAC was combined with antiplatelet agents.
Usually, observational studies define exposure as a baseline covariate. However, exposure may vary over time, yielding several exposure periods. Another type of studies is studies with time-varying exposure. Both study types, that is, fixed-varying and time-varying study types, have benefits and disadvantages. In a time-varying analysis, it is critically important to define treatment episodes that reflect the actual exposure of included subjects to avoid misclassification as good as possible. A misclassification of exposure can give rise to bias towards neutral in relation to outcomes. In time‐varying studies, it might be very complicated to define exposure, which can lead to the construction of different treatment episodes and hence potentially different conclusions. We faced many challenges when constructing exposure episodes based on prescriptions. First, no information about dosing or treatment duration is included in our registry data, so we estimated the duration of treatment episodes based on information about purchasing date, strength and quantity purchased. Second, we faced challenges with temporal gaps and overlaps between prescriptions. Using different methods to handle temporal gaps and overlaps between prescriptions can lead to different estimates of treatment duration and exposure periods, which in turn can lead to different conclusions regarding the effectiveness and safety of a drug.31 On the other hand, studies with a fixed exposure use methods that do not take into account time‐varying confounding factors, attempting to estimate the association between the treatment and the outcome conditioning only on the baseline characteristics of included subjects. These studies assume that the randomisation balance of the treatment is preserved, considering that each subject receives the randomly assigned treatment during the entire follow‐up period. However, if we do not adjust for confounders that vary over time, we might not achieve a correct estimation of treatment effect, as the comparability over time between groups is not guaranteed. As the focus of our second analysis was to consider multiple exposures over time, time‐dependent changes of exposure and confounders were taken into account. As in all studies, some confounders can be measured and adjusted for in the analysis, whereas others are unmeasured, which might lead to residual confounding. Despite adjusting for risk factors, confounding by indication might still have affected our results, which might have led to an underestimation of stroke risk and an overestimation of risk of recurrent bleeding. Some prognostic factors for recurrent bleeding also influence the decision to restart DOAC. For instance, fragile elderly patients have on the one hand high bleeding risk and on the other hand their fragility might encourage clinicians to restart DOAC to avoid further deterioration of health by a stroke event in the future. Thus, the restart of DOAC might seem to have increased the risk of recurrent bleeding—thus causing an overestimation of bleeding risk. In our analysis, we could not assess the severities of risk factors, which may have influenced clinicians’ choices of DOAC restart. We also do not know how potentially correctable risk factors for bleeding were handled by physicians during treatment interruption periods. In addition, switching from one antithrombotic regimen to another cannot be considered as an unrelated random process because it may be affected by many variables under follow-up, which may affect the outcome of interest. To avoid bias in our investigation of drug effects, we should optimally have accounted for the mechanism behind the switching and not considered it as a random process, which was not possible by using only registry data.
Although our analysis advocates the importance of restarting DOAC monotherapy in AF patients after a serious bleeding event to avoid stroke, our analysis does not define the best time window to restart DOAC. Further studies are needed to investigate the best time window for the restart of DOAC therapy after a serious bleeding event.
A main strength of our analysis is that it is a large population-based study representative of the Danish population as a whole. There is no loss to follow-up, and Danish registries are known for high validity of the AF diagnosis.32 A main limitation of our analysis is confounding by indication, which might have affected our results. Bias related to confounding by indication is difficult to fully take into account; however, we believe that the included confounders addressed this issue to a sufficient extent. We also have limitations related to the use of registry data, for example, we included prescriptions dispensed from pharmacies, but some may not have been subsequently taken by the patient and were thus misclassified. This misclassification can potentially lead to an overestimation of events in the groups who restarted antithrombotic drugs.
In addition, we do not know the reasons for patients’ switch/discontinuation, which can range from side effects, patient compliance, unwillingness to continue, physician’s suggestion or other causes. Although we considered initial DOAC interruption related to the index bleeding event, we do not know the reasons for subsequent DOAC interruptions. In fact, a Danish nationwide registry-based study investigated clinical events preceding switches to and from DOAC as well as discontinuation of DOAC among patients with AF found that the majority of patients changed treatment for reasons not recorded in health registries.26 In addition, we only included major events, so minor events handled outside the hospital setting, for example, primary care, are not included as bleeding events. Another limitation in our analysis is that we did not differentiate between the four DOACs available in the market, which is due to our sample size not allowing for the analysis to be properly done.
We do also not know which investigations have been conducted during interruption periods or if bridging with heparins was used during these periods. Although we tried to adjust for confounders, there might still be residual confounding factors not accounted for in our analysis. As our study is an observational study, we cannot conclude for a causation effect but a potential association between restart of antithrombotic treatment and investigated outcomes. Due to these limitations, our results should be interpreted with caution. A randomised controlled trial on restarting DOAC in AF patients after serious bleeding events to assess the risk of stroke and recurrent bleeding is thus warranted, although such a trial can potentially face many challenges, both practical and ethical.
Conclusions
Restart of DOAC within 60 days from index bleeding event was associated with some evidence of harm with increased risk of recurrent bleeding, while there was some weak evidence of a benefit of reduced stroke risk, although not statistically significantly different compared with later restart of treatment. Compared with periods with no treatment, periods with DOAC monotherapy or in combination with antiplatelet agents were associated with increased bleeding risk, while DOAC monotherapy showed best protection against stroke.
