Effectiveness and Safety of NOAC Versus Warfarin in Patients With Atrial Fibrillation and Aortic Stenosis.

Background Guideline recommendations on the use of non-vitamin K antagonist oral anticoagulants (NOACs) in atrial fibrillation (AF) patients with aortic stenosis are based on studies including a low number of patients with aortic stenosis. The aim of this study was to estimate the effects of NOAC versus warfarin on thromboembolism and major bleeding among AF patients with aortic stenosis. Methods and Results We emulated a target trial using observational data from Danish nationwide registries between 2013 and 2018. Thromboembolism was defined as a hospital diagnosis of ischemic stroke and/or systemic embolism, and major bleeding was defined as a hospital diagnosis of intracranial bleeding, gastrointestinal bleeding, or major or clinically relevant bleeding in other anatomic sites. Treatment effect estimates were based on an intention-to-treat and per-protocol approach. A total of 3726 patients with AF and aortic stenosis claimed a prescription for either a NOAC (2357 patients) or warfarin (1369 patients) and met the eligibility criteria for the trial. During 3 years of follow-up, the adjusted hazard ratios for thromboembolism and major bleeding were 1.62 (95% CI, 1.08-2.45) and 0.73 (0.59-0.91) for NOAC compared with warfarin in the intention-to-treat analyses. Similar results were observed in the per-protocol analyses. Conclusions In this observational study, we observed a higher risk of thromboembolism but a lower risk of major bleeding for treatment with NOACs compared with warfarin in patients with AF and aortic stenosis. This observation needs confirmation in large randomized trials in these commonly encountered patients.

intention‐to‐treat

Non–vitamin K antagonist oral anticoagulant

Per protocol

valvular heart disease

Clinical Perspective

What Is New?

In patients with atrial fibrillation and aortic stenosis, non–vitamin K antagonist oral anticoagulants may be less effective than warfarin for preventing ischemic events but safer with respect to bleeding.

What Are the Clinical Implications?

Based on the findings of our study and inconsistent data in the literature on the effectiveness and safety of non–vitamin K antagonist oral anticoagulants versus warfarin in patients with atrial fibrillation and aortic stenosis, the optimal oral anticoagulant strategy remains unclear.

The observed increased risk of thromboembolism in the non–vitamin K antagonist oral anticoagulants group in our study requires further investigation because it was not observed in the post hoc analyses of the ROCKET AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) and ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) trials.

Atrial fibrillation (AF) and valvular heart disease (VHD) often coexist, and aortic stenosis is one of the most prevalent VHD subtypes in developed countries, , , affecting a large proportion of the elderly population. The prevalence of both aortic stenosis and AF increases with age, and the number of patients diagnosed with aortic stenosis and AF will increase considerably in line with the rapidly increasing elderly population. , , Approximately 16% to 36% of all patients with aortic stenosis have concomitant AF, and most of these patients are in lifelong thromboprophylaxis with oral anticoagulant therapy. , , , Importantly, patients with aortic stenosis have been identified as a high‐risk subgroup in terms of the risk of thromboembolism and bleeding in anticoagulated patients with AF, which complicates the risk‐benefit ratio of oral anticoagulation. , , ,

Randomized, controlled trials have evaluated different oral anticoagulants, such as warfarin and non‐vitamin K antagonist oral anticoagulants (NOACs), for the prevention of thromboembolism in patients with AF; however, these trials excluded patients with significant VHD. , , , Patients with mechanical heart valves or moderate to severe mitral stenosis were excluded in all trials, and patients with any other VHD, such as aortic stenosis, were minimally represented. Only a few post hoc sub‐analyses of the existing randomized, controlled trials have examined patients with both AF and VHD, and the proportion of patients with aortic stenosis was underrepresented (6%–12%), , , , despite aortic stenosis being one of the most prevalent VHDs in recent patients with AF and VHD (17%–62% with aortic stenosis). , , , Consequently, the effectiveness and safety of NOAC versus warfarin in AF patients with aortic stenosis has not been specifically investigated, although guidelines currently allow for use of NOACs in AF patients with aortic stenosis (and without a mechanical heart valve or concomitant moderate/severe mitral stenosis). ,

The aim of the present study was to emulate a target trial using observational data from Danish nationwide registries to estimate the effects of NOAC versus warfarin on thromboembolism and major bleeding among AF patients with aortic stenosis.

Methods

Study Design and Data Sources

This study was conducted using the “target trial” principles. , Briefly, we specified the protocol of a target trial (a hypothetical randomized experiment) to estimate the effectiveness and safety of NOAC versus warfarin in AF patients with aortic stenosis and then attempted to emulate this trial using observational data from the Danish nationwide registries. The specifications of each component in the target trial and the emulated trial are provided in Table S1.

Four Danish nationwide registries were used: The Danish Civil Registration System, the National Prescription Registry, the Danish National Patient Registry, and the Danish National Laboratory Register. Data from these registries were linked via a unique personal identification number, which is used across all Danish nationwide registries. Data S1 provides a description of the registries.

Permissions to access data from the nationwide registries were obtained through the Danish Health Data Agency. Requests to access the data set may be sent to The Danish Health Data Agency at forskerservice@sundhedsdata.dk. Dr Melgaard, Mr Jensen, and Dr Nielsen had full access to all the data in the study and take responsibility for its integrity and the data analysis. The study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations.

Eligibility Criteria

We identified patients in the Danish nationwide registries who met the target trial eligibility criteria (Table S1). The study population included patients with a first‐time prescription for a NOAC or warfarin (baseline date) and a diagnosis of both AF and aortic stenosis at baseline or within 30 days after baseline. To ensure that patients were eligible for stroke prevention with oral anticoagulant therapy according to contemporary guidelines, a CHA2DS2‐VASc score level threshold of ≥1 for male and ≥2 for female patients was also an eligibility criterion. Patients with other indications for oral anticoagulant therapy or potential contraindications for NOAC or warfarin treatment were excluded. Patients with dispensation of both a NOAC and warfarin within the first 30 days after baseline were excluded. Lastly, only patients who were alive and event‐free after the first month were included due to the data setup and the statistical methods used (see Figure 1 for a flowchart of the study population). For details on the definition of AF, aortic stenosis, contraindications, comorbidities, and co‐medication, please see Table S2.

Figure 1

Flowchart of eligible patients to emulate the target trial.

CHA2DS2‐VASc indicates congestive heart failure, hypertension, age ≥75 years [doubled], diabetes, prior stroke/transient ischemic attack/systemic embolism [doubled], vascular disease [prior myocardial infarction, peripheral artery disease, or aortic plaque], age 65–74 years, sex category [female]; and NOAC, non–vitamin K antagonist oral anticoagulant.

Treatment Strategies

The Danish National Prescription Registry was used to identify patients who redeemed a first‐time prescription for a NOAC (apixaban, dabigatran, rivaroxaban, or edoxaban) or warfarin between January 2013 and October 2018 (oral anticoagulation naïve users only). Patients were considered exposed to treatment irrespective of any subsequent dosage changes.

Outcomes

The effectiveness outcome was a hospital diagnosis of ischemic stroke and/or systemic embolism defined as a composite endpoint of “thromboembolism.” The safety outcome was major bleeding leading to hospital admission (either intracranial bleeding, gastrointestinal bleeding, or major or clinically relevant bleeding in other anatomic sites). For details about the definition of the outcomes, please see Table S2.

Follow‐Up Period

Each patient was followed up in the registries for the outcomes of interest. Follow‐up started 30 days after treatment assignment (baseline) and ended at outcome diagnosis, death, administrative end of follow‐up (3 years or December 2018), or emigration (loss to follow‐up), whichever occurred first.

Causal Contrasts

To compare the 2 treatment strategies, we estimated the intention‐to‐treat (ITT) effect and per‐protocol (PP) effect.

Statistical Analysis

The baseline characteristics of the study population were described according to treatment exposure category (NOAC or warfarin) using means and standard deviation for continuous variables, and proportions for categorical variables. The exposure category (ie, NOAC or warfarin) of each patient was based on the prescription claim at the baseline date, and this category remained unchanged throughout the study duration.

Counterfactual outcomes were investigated at 3 years and data arranged in such a way that each patient‐month was represented by a single row (maximum of 36 rows per individual, corresponding to 3 years). To account for the non‐randomization of the treatment assignment, we derived stabilized inverse probability of treatment weights. To compute these weights, we estimated the propensity of being assigned each treatment by a logistic regression including the following baseline confounding factors: age (as a restricted cubic spline) and dichotomous covariates on sex, heart failure, hypertension, diabetes, myocardial infarction, ischemic heart disease, renal disease, prior bleeding, prior thromboembolic event, diagnosis of atrial fibrillation, aortic stenosis or valve surgery (including bioprosthetic valve implantation) within 60 days before or 30 days after the baseline date, and use of statin or antiplatelet therapy within the last year.

The assessment of outcomes was based on an ITT approach and a PP approach (please see details in Data S1). When estimating the ITT treatment effects, treatment status was assessed at the date of first prescription claim (NOAC or warfarin) and remained unchanged throughout follow‐up disregarding actual treatment. When estimating the PP treatment effects, treatment status was assessed continuously using a recommended daily dose and quantity of pills per pack in each prescription (a 60‐day grace period between each prescription claim was allowed). The variable dose regimen of warfarin was modeled by continuous adaption of an (individual) estimated daily dose. Patients were considered adherent to the initial treatment strategy (NOAC or warfarin) unless a clinical event that fully or partly contraindicated treatment or had a major clinical impact on the anticoagulant therapy strategy occurred. If such an event occurred, we stopped updating the censoring weight for that patient, but kept the patient in the analysis. For the ITT and PP analyses, pooled logistic regression models were used to estimate the average treatment effects by means of hazard ratios (HRs) for the outcomes. In detail, we derived odds ratios from pooled logistic regressions, which are approximations of HRs when the investigated outcome is rare in all time intervals. The calculated stabilized inverse probability of treatment weights were applied in pooled logistic regression models. For the PP analyses, we calculated stabilized inverse probability of censoring weights to account for the dependence between measured post‐baseline time‐varying prognostic factors (heart failure, hypertension, diabetes, ischemic heart disease, myocardial infarction, and use of statin or antiplatelet therapy [all included as dichotomous covariates]) and censoring, and these weights were multiplied by the stabilized inverse probability of treatment weights of baseline confounding factors and applied in the weighted pooled logistic regression models to estimate the PP treatment effects. In addition, standardized event‐free survival curves were constructed, which depict the estimated counterfactual event‐free survival had every individual receiving either treatment. All statistical analyses were performed using SAS 9.3 (SAS Institute) and Stata version 16 (StataCorp LP).

Sub‐Analysis and Sensitivity Analyses

Some patients had an aortic valve surgery/procedure before inclusion in the study, which may affect the treatment effects, especially if the surgery/procedure was performed close to the baseline date. Therefore, we performed a sub‐analysis in which we restricted the population to the following subpopulations and repeated the main analyses: (1) those who had an aortic valve surgery/procedure within 60 days before or 30 days after baseline, (2) those who had an aortic valve surgery/procedure at any time before or 30 days after baseline, and (3) those who never had an aortic valve surgery/procedure.

Two sensitivity analyses were performed to investigate the robustness of the analytical strategy in the main analyses. We performed a sensitivity analysis of the PP analysis in which we changed the assessment of continuous treatment status by allowing a grace period of 90 days as a treatment gap. Additionally, 2 “falsification outcomes” were examined, which were expected to have a null association with the exposure. In detail, we emulated an individual target trial with pneumonia as the outcome and an individual target trial with cancer as the outcome using the described features from the ITT analyses.

Ethical Considerations

The study was conducted in compliance with General Data Protection Regulation Article 30, recorded at Aalborg University Hospital and Aalborg University (project no. 2017‐40). No ethical approval or patient consent are required for studies based on data from administrative Danish registries according to Danish laws.

Results

Of 5303 patients with AF and aortic stenosis who claimed a prescription for either a NOAC or warfarin between January 2013 and October 2018, 3726 were eligible for the target trial emulation (Figure 1). The study group comprised 1369 patients who claimed a prescription for warfarin and 2357 patients claimed a prescription for a NOAC: apixaban, 1105; dabigatran, 323; edoxaban, 38; and rivaroxaban, 891. The baseline characteristics of the study population are summarized in Table 1.

Table 1

Baseline Characteristics of Eligible Patients to Emulate the Target Trial

	Warfarin	NOAC
N (%)	1369	2357
Women, n (%)	590 (43.1)	1170 (49.6)
Age in years, median (IQR)	79 (73–85)	82 (75–88)
Days since diagnosis of AF, median (IQR)	15 (6–235)	11 (4–170)
Days since diagnosis of aortic stenosis, median (IQR)	360 (23–1528)	515 (18‐1780)
Previous aortic valve intervention*, n (%)	432 (31.6)	497 (21.1)
Days since aortic valve intervention, median (IQR)	22 (13–100)	67 (17–1104)
Year of inclusion:
2013, n (%)	393 (28.7)	197 (8.4)
2014, n (%)	364 (26.6)	310 (13.2)
2015, n (%)	285 (20.8)	379 (16.1)
2016, n (%)	202 (14.8)	484 (20.5)
2017, n (%)	97 (7.1)	570 (24.2)
2018, n (%)	28 (2.0)	417 (17.7)
Comorbidities: n (%)
Heart failure	670 (48.9)	1008 (42.8)
Hypertension	957 (69.9)	1616 (68.6)
Diabetes	290 (21.2)	429 (18.2)
Prior thromboembolic event	276 (20.2)	571 (24.2)
Prior major bleeding event	272 (19.9)	500 (21.2)
Vascular disease	448 (32.7)	658 (27.9)
Prior myocardial infarction	270 (19.7)	380 (16.1)
Ischemic heart disease	599 (43.8)	881 (37.4)
Prior percutaneous coronary intervention	181 (13.2)	295 (12.5)
Prior coronary artery bypass grafting	206 (15.0)	231 (9.8)
Alcohol abuse	53 (3.9)	109 (4.6)
Chronic obstructive pulmonary disorder	231 (16.9)	430 (18.2)
Chronic kidney disease	159 (11.6)	152 (6.4)
CHA2DS2‐VASc score, median (IQR)	4.0 (3.0–5.0)	4.0 (3.0–5.0)
HAS‐BLED score, median (IQR)	3.0 (2.0–4.0)	3.0 (2.0–4.0)
Co‐medication: n (%) *
NOAC agent:
Apixaban	…	1105 (46.9)
Dabigatran	…	323 (13.7)
Edoxaban	…	38 (1.6)
Rivaroxaban	…	891 (37.8)
Aspirin	765 (55.9)	1185 (50.3)
Other antiplatelet therapy	224 (16.4)	472 (20.0)
Beta‐blockers	639 (46.7)	991 (42.0)
ARB/ACE‐inhibitors	755 (55.1)	1265 (53.7)
Calcium channel blockers	519 (37.9)	863 (36.6)
Amiodarone	48 (3.5)	44 (1.9)
Digoxin	93 (6.8)	135 (5.7)
Non‐loop diuretics	600 (43.8)	1012 (42.9)
Loop diuretics	553 (40.4)	838 (35.6)
NSAIDs	225 (16.4)	406 (17.2)
Statins	763 (55.7)	1184 (50.2)

ACE indicates angiotensin‐converting enzyme; AF, atrial fibrillation; ARB, angiotensin II receptor blocker; IQR, Interquartile range; and N, Number.

Primarily bioprosthetic aortic valve replacement.

Patients with a redeemed prescription within 180 days prior to or 30 days after the diagnosis of atrial fibrillation.

Effectiveness Outcome

During a median follow‐up of 14 months (interquartile range [IQR]: 6–23 months), 113 thromboembolic events were observed. In the ITT analysis, the adjusted HR for thromboembolism was 1.62 (95% CI, 1.08–2.45) for NOACs compared with warfarin (Table 2). In the ITT analysis, the estimated 3‐year thromboembolic‐free survival was 94.0% for NOACs and 96.0% for warfarin (Figure 2).

Table 2

Treatment Effects of NOAC Versus Warfarin on Thromboembolism and Bleeding After 3 Years of Follow‐Up

Analytical strategy	Intention‐to‐treat analysis		Per‐protocol analysis
	Warfarin	NOAC	Warfarin	NOAC
Thromboembolism
Event count	36	77	19	62
HR (95% CI)	Ref.	1.62 (1.08–2.45)	Ref.	1.92 (1.11–3.30)
Major bleeding†
Event count	171	184	119	163
HR (95% CI)	Ref.	0.73 (0.59–0.91)	Ref.	0.78 (0.60–0.99)

HR indicates hazard ratio; and NOAC, non–vitamin K antagonist oral anticoagulant.

Composite of intracranial bleeding, gastrointestinal bleeding, and major or clinically relevant bleeding in other anatomic sites.

Figure 2

Standardized survival curve free from thromboembolic events.

Thromboembolism‐free survival probability according to treatment strategy (NOAC or warfarin) for the intention‐to‐treat analysis and the per‐protocol analysis. NOAC indicates non–vitamin K antagonist oral anticoagulant.

In the PP analysis, the median follow‐up was 11 months (IQR: 5–20 months), and 81 events were observed. A total of 3079 patients (82.6%) had a censoring event. The adjusted HR for thromboembolism was 1.92 (95% CI, 1.11–3.30) for NOACs compared with warfarin in the PP analysis. In the PP analysis, the estimated 3‐year thromboembolic‐free survival was 93.9% for NOACs and 97.0% for warfarin (Figure 2).

Safety Outcome

During a median follow‐up of 13 months (IQR: 6–23 months), 355 major bleeding events were observed: 66 intracranial bleeds, 176 gastrointestinal bleeds, and 121 major or clinically relevant bleeds in other anatomic sites (some patients had more than one bleeding event on the same day). In the ITT analysis, the adjusted HR for major bleeding was 0.73 (95% CI, 0.59–0.91) for NOACs compared with warfarin (Table 2). In the ITT analysis, the estimated 3‐year major bleeding‐free survival was 87.6% for NOACs and 83.6% for warfarin (Figure 3).

Figure 3

Standardized survival curve free from major bleeding events.

Major bleeding‐free survival probability according to treatment strategy (NOAC or warfarin) for the intention‐to‐treat analysis and the per‐protocol analysis. NOAC indicates non–vitamin K antagonist oral anticoagulant.

In the PP analysis, the median follow‐up was 11 months (IQR: 5–20 months), and 282 major bleeding events were observed: 53 intracranial bleeds, 143 gastrointestinal bleeds, and 90 major or clinically relevant bleeds in other anatomic sites (some patients had more than one bleeding event on the same day). A total of 2931 patients (78.7%) had a censoring event. The adjusted HR for major bleeding was 0.78 (95% CI: 0.60–0.99) for NOACs compared to warfarin in the PP analysis. In the PP analysis, the estimated 3‐year major bleeding‐free survival was 87.4% for NOACs and 85.1% for warfarin (Figure 3).

In the sub‐analysis, we restricted the population to 512 patients who had an aortic valve surgery/procedure within 60 days before or 30 days after baseline, 888 patients who had an aortic valve surgery/procedure at any time before or 30 days after baseline, and 2838 patients who never had an aortic valve surgery/procedure. We observed too few events in the 2 subpopulations with a history of valve surgery/procedure to perform the pre‐planned analyses. However, we repeated the main analyses in the patients with no prior aortic valve surgery/procedure (ie, patients with native aortic stenosis) and found similar results as in the main analysis (Table S3).

In the sensitivity analysis, allowing a 90‐day treatment gap in the estimate of continuous treatment, we observed similar results as in the main analysis. The PP analysis yielded an adjusted HR for thromboembolism of 1.87 (95% CI, 1.12–3.10) and an adjusted HR for major bleeding of 0.71 (95% CI, 0.56–0.90). The estimated 3‐year survivals were materially unchanged, and reflecting results found in the main analysis (data not shown).

In the “falsification outcome” analyses using the ITT approach, the adjusted HR for the pneumonia outcome was 0.94 (95% CI, 0.80–1.11) for NOACs compared with warfarin and the adjusted HR for the cancer outcome was 1.15 (95% CI, 0.91–1.44) for NOACs compared with warfarin.

Discussion

In this study, we used observational data to emulate a target trial estimating the average treatment effects of NOAC versus warfarin on thromboembolism and major bleeding among patients with AF and aortic stenosis. We observed a significantly higher risk of thromboembolism in the NOAC group compared with the warfarin group in both the ITT and PP analyses. In addition, we observed a significantly lower risk of major bleeding in the NOAC group compared with the warfarin group.

Large randomized, controlled trials evaluating the efficacy and safety of NOACs versus warfarin for prevention of thromboembolic events in patients with AF demonstrated that the NOACs are associated with similar or lower rates of both ischemic stroke and major bleeding and less than half the risk of intracranial hemorrhage compared with adjusted dose warfarin. , , , Previous post hoc studies of these randomized, controlled trials examining NOACs versus warfarin in anticoagulated patients with AF with and without VHD generally observed comparable outcomes of stroke or systemic embolism and major bleeding in patients treated with regular doses of NOACs and patients treated with warfarin, with the exception of 20 mg rivaroxaban, which was associated with higher rates of major bleeding compared with warfarin. , , , Data on outcomes comparing NOACs to warfarin in AF patients with aortic stenosis have only been reported in 2 post hoc analyses of the ROCKET AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) and ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) trials. , In the post hoc analysis of the ROCKET AF trial, which included 214 AF patients with aortic stenosis, AF patients with aortic stenosis on 20 mg rivaroxaban daily had similar stroke or systemic embolism rates compared to patients on warfarin (HR not reported) but higher rates of major bleeding (HR, 1.73; 95% CI, 0.73–4.12) and major bleeding/non‐major clinically relevant bleeding (HR, 1.18; 95% CI, 0.70–1.97). In the post hoc analysis of the ARISTOTLE trial, which included 1150 AF patients with aortic valve disease, 384 of which had aortic stenosis, AF patients with aortic valve disease on 5 mg apixaban twice daily had lower risk of stroke or systemic embolism (HR, 0.55; 95% CI, 0.30–1.01) and major bleeding (HR, 0.72; 95% CI, 0.44–1.18) compared with patients on warfarin. In these 2 post hoc trials, , the finding of a lower or similar risk of thromboembolism in the NOAC group compared with the warfarin group is different from the findings in our study, where we observed a significantly higher risk of thromboembolism in the NOAC group compared with the warfarin group (HR, 1.62; 95% CI, 1.08–2.45 in ITT analysis; HR, 1.92; 95% CI, 1.011–3.30 in PP analysis). This observation may be explained by the non‐randomized setup in our study, though, we derived stabilized inverse probability of treatment weights to account for the non‐randomization of the treatment assignment, and the falsification outcome analyses revealed minimal risk of residual bias. Therefore, our observation could also reflect an actual increased risk of thromboembolism in the NOAC group.

The trial subgroups were small and the CIs wide, and recent studies examining the effectiveness and safety of NOAC versus warfarin in AF patients with aortic stenosis undergoing transcatheter aortic valve replacement also observed an increased risk of thromboembolism in the NOAC group and rivaroxaban‐related safety concerns. Furthermore, our study may better mirror the clinical reality, as we included a more diverse group of patients with AF and aortic stenosis than the ARISTOTLE and ROCKET AF trials.

The finding of major bleeding in our study is in line with the findings of the post hoc analysis in the ARISTOTLE trial, as we also observed a lower risk of major bleeding in the NOAC group compared with the warfarin group (HR, 0.73;95% CI, 0.59–0.91 in ITT analysis; HR, 0.78; 95% CI, 0.60–0.99 in PP analysis). In the ROCKET AF trial, the higher rates of major bleeding in patients treated with rivaroxaban were also observed in patients with other VHDs ; thus, the increased risk of major bleeding associated with rivaroxaban could be clinically important in patients with VHD, though we observed a considerably lower risk of major bleeding in the NOAC group in our study (38.0% of the patients in the NOAC group were in oral anticoagulant therapy with rivaroxaban).

Unfortunately, our data did not allow us to examine each NOAC agent or dose individually because of the limited sample size. In our study, the proportion of patients in the NOAC group using apixaban was 46.8%, so our findings are mainly driven by therapy with apixaban or rivaroxaban. The proportion of patients in oral anticoagulant therapy with either dabigatran or edoxaban was low (13.7% and 1.6%, respectively) and randomized and observational data on outcomes of dabigatran and edoxaban in patients with AF and aortic stenosis are lacking in general.

Clinical Implications and Future Directions

International guidelines for patients with AF recommend NOAC as an alternative to warfarin in patients with native VHD, including aortic stenosis. , However, these guideline recommendations on the use of NOAC in AF patients with aortic stenosis are based on post hoc trial analyses including a small and widely underrepresented number of patients with aortic stenosis. , , , Furthermore, the results of these post hoc analyses were inconsistent, as outlined above. , Generally, caution is warranted when interpreting post‐hoc trial analyses.

The observations in our study suggest that NOAC may be less effective than warfarin for preventing ischemic events, but safer with respect to bleeding, in the population with AF and aortic stenosis. Thus, individual assessment of the thromboembolic risk and bleeding risk in this population might be necessary before deciding which oral anticoagulant agent the patient should be prescribed. Generally, the clinician should be aware of the overall increased risk of thromboembolism and bleeding in this typically aged, multimorbid population.

Based on the findings of our study and the existing inconsistent data on the effectiveness and safety of NOAC versus warfarin in patients with AF and aortic stenosis, the optimal oral anticoagulant strategy is not clear at this point and more research is necessary. Similarly, the safety of every NOAC agent is questionable because rivaroxaban may be associated with an increased risk of major bleeding and major bleeding/non‐major clinically relevant bleeding in patients with AF and aortic stenosis, whereas apixaban seems to be associated with an appealing safety profile but data on dabigatran and edoxaban are lacking. Importantly, the observed increased risk of thromboembolism in the NOAC group in our study requires further investigation, preferably in a large randomized trial, as this finding was not observed in the post hoc analyses of the ROCKET AF and ARISTOTLE trials. ,

Strengths and Limitations

We examined the effect of NOAC versus warfarin on thromboembolism and bleeding using the “target trial” principle, which has the advantage of avoiding common pitfalls that occur when conducting comparative effectiveness analyses using observational data. , Due to the non‐randomized design, all confounding factors may not have been accounted for; for example, lifestyle factors were not available in the registries we utilized.

The diagnoses of AF and aortic valve disease have been validated with positive predictive values of 93% and 98%, respectively. , , Patients with aortic stenosis is a broad group of patients with varying severity of valve disease. We did not have access to echocardiographic data or individual blood pressure measurements; therefore, we did not have information about the severity of aortic stenosis or degree of hemodynamic influence.

The diagnoses of ischemic stroke and intracranial hemorrhage have been validated with positive predictive values of 80%–90% and 88%, respectively. , , No validation studies for the diagnoses of other major bleedings currently exist. However, we examined only diagnoses of major bleeding leading to a hospital admission to ensure that the bleeding was truly major and clinically relevant. By this approach, bleeds registered in outpatients were not examined, but some of these bleeds could have been clinically relevant.

We observed an increase in NOAC users and a decrease in warfarin users during the years of inclusion, which are in line with observations in other AF populations. , NOAC has gradually become the preferred oral anticoagulant drug in patients with AF, both among new‐users and prevalent users. , In our study, we only included patients with a first‐time prescription of any oral anticoagulant agent, and, therefore, the increase in NOAC users in our study reflects the increased use of NOAC as the first choice of oral anticoagulant agent in new‐users and not the switch to NOACs among prevalent users (which may have a different clinical profile and risks of adverse events than new‐users). However, the lack of randomization in our study is a major limitation and, therefore, our findings need confirmation in large prospective randomized trials.

Conclusions

In this observational study, we observed a higher risk of thromboembolism but a lower risk of major bleeding for treatment with NOACs compared with warfarin in patients with AF and aortic stenosis. This observation needs confirmation in large, randomized trials in these commonly encountered patients. Importantly, the clinician must be aware of the increased risk of thromboembolism and bleeding in this population in general.

Sources of Funding

This work was supported by “The BMS/Pfizer European Thrombosis Investigator Initiated Research Program (ERISTA) 2018” and the Obel Family Foundation. The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.

Disclosures

Professor Lip reports consulting and speaking for BMS/Pfizer, Boehringer Ingelheim, and Daiichi‐Sankyo. No fees are received personally. Professor Larsen is an investigator for Janssen Scientific Affairs, LLC, Bayer AG, and Boehringer Ingelheim, and has received speaker honorarium from Bayer, BMS, Pfizer, and Merck, Sharp & Dome (MSD). Dr Nielsen reports personal fees from Boehringer Ingelheim, grants and personal fees from Daiichi‐Sankoy, grants from BMS/Pfizer, and grants and personal fees from Bayer outside the submitted work. Mrs Melgaard reports grant support from BMS/Pfizer. Dr Overvad reports grant support from BMS/Pfizer. Dr Christensen is on the speaker bureaus for AstraZeneca, Boehringer Ingelheim, Pfizer, Roche Diagnostics, Takeda, Merck Sharp & Dohme (MSD), and Bristol‐Myers Squibb and an Advisory Board for Bayer and Merck Sharp & Dohme (MSD). The remaining authors have no disclosures to report.

Data S1

Table S1–S3

Click here for additional data file.