Effectiveness and safety of rivaroxaban therapy in daily-care patients with atrial fibrillation

Summary The effectiveness and safety of rivaroxaban for stroke prevention in atrial fibrillation (SPAF) demonstrated in ROCKET AF needs to be confirmed in daily care. To evaluate effectiveness and safety of rivaroxaban therapy in SPAF patients in daily care, we used data from an ongoing, prospective, non-interventional registry of more than 2700 patients on novel oral anticoagulants in daily care. Between October 1, 2011 and February 28, 2013, a total of 1204 SPAF patients receiving rivaroxaban were enrolled. During a mean follow-up of 796.2 ± 207.3 days, the combined endpoint of stroke/transient ischaemic attack/systemic embolism occurred at a rate of 2.03/100 patient-years in the intention-totreat analysis (95 % confidence interval [CI] 1.5–2.7) and at 1.7/100 patient-years in the on-treatment analysis (events within 3 days after last intake). On-treatment rates were higher in patients selected for 15 mg rivaroxaban (n=384) once daily [OD] compared with the 820 patients selected for 20 mg OD (2.7 [95 % CI 1.6–4.2] vs 1.25/100 patient-years [95 % CI 0.8–1.9]). On treatment, major bleeding occurred at a rate of 3.0/100 patient-years and significantly more often in patients receiving the 15 mg OD dose compared with the 20 mg OD dose (4.5 vs 2.4/100 patient-years). Rivaroxaban treatment discontinuation occurred in a total of 277 patients during follow-up (12.0/100 patientyears in Kaplan-Meier analysis). Our data contribute to the confirmation of effectiveness and relative safety of rivaroxaban in daily-care patients. Furthermore, rivaroxaban discontinuation rates were considerably lower than those reported for vitamin K antagonists. Supplementary Material to this article is available online at www.thrombosis-online.com.


Introduction
Direct-acting, non-vitamin K antagonist (VKA) oral anticoagulants (NOACs) are currently replacing VKAs, which for decades have been the standard long-term anticoagulation therapy for stroke prevention in atrial fibrillation (SPAF) patients (1). VKA are difficult to manage due to inter-individual variations in metabolism, common drug-drug interactions, and the interaction with the dietary intake of vitamin K (2). Therefore, VKA need to be individually dosed according to the measurements of prothrombin time or international normalised ratio (INR) but still, the 'time in therapeutic range' of VKA patients in daily care is as low as 50-60 % (3), which indicates frequent periods of over-and underdosing and the need for frequent dose adjustments (4).
By contrast, NOACs such as the Factor Xa inhibitor rivaroxaban demonstrate a much better dose-response relationship, which makes routine monitoring and frequent dose adjustments unnecessary (5). In the large phase III ROCKET AF trial for SPAF, both VKA and rivaroxaban demonstrated high efficacy and safety in a predefined high-risk population that was characterized by a high stroke risk (mean CHADS 2 score 3.5) (6) and a significant risk for major bleeding complications (median HAS-BLED score 3) (7).
However, the external validity of phase III trials needs to be confirmed in daily-care settings, in which patients may have significant co-morbidities and are treated without a strict protocol under less intense surveillance. Consequently, observational studies in unselected patients are urgently needed to evaluate the effectiveness and safety of rivaroxaban in daily-care settings.
Using data from a large multicentric registry, we prospectively evaluated the management and outcome of patients with SPAF treated with rivaroxaban.

Patients
The Dresden NOAC Registry (NCT01588119) is a large, prospective registry in the administrative district of Dresden (Saxony), Germany. In this ongoing project, a network of more than 230 physicians from private practices and hospitals are enrolling consecutive NOAC patients, who are prospectively followed up by the central registry office. The design and methodology of the Dresden NOAC Registry has been published previously (8)(9)(10)(11)(12). All patients who are treated with NOACs for at least three months and provide written informed consent are eligible for participation in the registry. No exclusion criteria apply. Patients are followed up by telephone interview 30 days after enrolment and quarterly thereafter to collect data on the efficacy, safety, and management of NOAC therapy in daily care. All suspected outcome events are reviewed by a central adjudication committee.

Outcome measures
To assess effectiveness of rivaroxaban therapy in SPAF, the annualised rate of the combined endpoint of stroke, transient ischaemic attack (TIA), or systemic embolism was evaluated.
The main safety outcome was the annualised rate of major bleeding according to the International Society on Thrombosis and Haemostasis (ISTH) definition (13). Further safety outcomes were rates of ISTH non-major clinically relevant (NMCR) bleeding, minor bleeding and all-cause mortality.

Treatment discontinuation
Treatment discontinuation was defined as a permanent discontinuation of rivaroxaban or an unscheduled interruption of rivaroxaban for longer than four weeks without the initial plan to restart rivaroxaban (14). Treatment persistence was defined as the continuation of rivaroxaban therapy over the entire follow-up period, allowing for temporary interruptions. At every visit, any change in anticoagulant therapy was assessed, and the reasons for this decision as well as the future treatment plan were obtained from patients or attending physicians.

Statistics
Two different analysis sets were defined and evaluated: • The overall rate of major cardiovascular events was evaluated in the intention-to-treat analysis, including all patients who were enrolled in the registry receiving rivaroxaban at baseline. All effectiveness outcome events were included that occurred throughout the follow-up period, including those occurring at any time after interruption or discontinuation of rivaroxaban.
• Rates of major cardiovascular events on treatment and rates of bleeding complications (all, major, and NMCR bleeding) were evaluated in the on-treatment analysis. All patients enrolled in the rivaroxaban group at baseline were included, but only outcome events that occurred during rivaroxaban treatment or within three days after interruption or discontinuation of treatment were evaluated. Patients with permanent discontinuation of rivaroxaban were censored at day 3 after last intake.
For comparison of means assuming parametric distribution, a t-test for independent samples was performed. For comparisons using non-parametric assumption or frequencies, the Kruskal-Wallis test or Chi-square test was used, as appropriate.
The 95 % confidence intervals (CIs) of proportions are calculated according to the Clopper-Pearson method.
Data are presented as absolute and relative frequencies, mean and standard deviation, or median with interquartile range as difference between 25th and 75th percentile, where appropriate. All p-values presented are exploratory in nature; thus, no adjustment of type I error for multiple testing is conducted. A p-value of ≤ 0.05 was considered to be statistically significant.
In both the intention-to-treat and the on-treatment analysis set, outcome event rates were calculated using Kaplan-Meier time-tofirst-event analysis, with data presented as events per 100 patientyears with their 95 % CIs, using the following formula: Event rate = number of events / total time under risk (defined as the sum of all days from inclusion in the registry until the day of the first event divided by 100 × 365 days and 100 patient-years as its unit). Corresponding CIs and P-values are calculated using the Poisson distribution.
All statistical analyses were carried out using the IBM ® SPSS ® Statistics version 19, software package SAS release 9.4 (SAS Institute), and R version 3.1.0 (Comprehensive R Archive Network) with RStudio version 0.98.953.

Ethics
The study protocol of the Dresden NOAC Registry was approved by the local ethics committee at the Technical University Dresden (AZ EK 349092011) and registered at ClinicalTrials.gov (NCT01588119). The study complies with the principles and requirements of the Declaration of Helsinki. All patients provided written informed consent, including a data protection waiver, before enrolment.

Cohort characteristics
Between October 1, 2011 and February 28, 2013, SPAF patients receiving rivaroxaban (n=1204) were enrolled into the registry. The baseline characteristics overall and by treatment group receiving rivaroxaban 20 mg or 15 mg once daily (OD) are shown in ▶Table 1.
Patients receiving the 15 mg OD dose were found to be significantly more often female, were older, and more often had chronic heart failure, diabetes, peripheral arterial occlusive disease/coronary artery disease, impaired renal function (defined as creatinine clearance ≤ 50 ml/minute [min]), CHADS 2 score ≥ 2, CHA 2 DS 2 -VASc score ≥ 2, and HAS-BLED score ≥ 2. The characteristics of patient subgroups of VKA pre-treatment or newly anticoagulated patients are presented in Suppl. ▶Table 1 (available online at www.thrombosis-online.com).

Effectiveness of rivaroxaban therapy
In the intention-to-treat analysis, the overall rate of the composite effectiveness endpoint (stroke/TIA/systemic embolism) was found to be 2.03/100 patient-years (95 % CI 1.5-2.7).
In the on-treatment analysis (all events during or within 3 days after last intake of rivaroxaban), stroke/TIA/systemic embolism occurred at a rate of 1.7/100 patient-years (95 % CI 1.2-2.3), which was lower for patients receiving the 20 mg compared with the 15 mg OD dose (1.3 vs 2.7/100 patient-years; p = 0.016). Rates and confidence intervals of all effectiveness endpoints are shown in ▶Table 2.  HAS-BLED score ≥ 2, n (%) BMI, body mass index; IQR, interquartile range; NSAID, non-steroidal anti-inflammatory drug. PAOD/CAD, peripheral arterial occlusive disease/coronary artery disease; SD, standard deviation. a Impaired renal function was defined as current or a history of creatinine clearance < 50 ml/min.

Safety of rivaroxaban therapy
In the on-treatment analysis, the annual rates of major bleeding according to the ISTH definition were 3.0/100 patient-years (95 % CI 2.3-3.8 %) and numerically higher in patients receiving rivaroxaban 15 mg (4.5/100 patient-years; 95 % CI 3.0-6.4) vs 20 mg OD (2.4/100 patient-years; 95 % CI 1.7-3.3; ▶Table 3). ▶Figure 2 presents the Kaplan-Meier curves for major bleeding for all patients and for subgroups of patients receiving rivaroxaban 20 mg or 15 mg OD.   Of the 1461 bleeding events that were observed during followup, only 78 (5.3 %) were ISTH major bleedings. Of those, the majority could be conservatively managed and only 34/78 major bleeding events (43.6 %) required surgical or interventional treatment. Prothrombin complex concentrate (PCC) was given in 11 cases (0.75 % of all bleeding and 14.1 % of all major bleeding events). PCC was regularly used to treat intracranial bleeding (6/10 ICB; 60 %) but rarely in patients with extracranial haemorrhage.
Details regarding the site, severity, and management of bleeding during rivaroxaban therapy are provided in ▶Table 4.
A total of 141 patients died during follow-up, of which 80 deaths occurred on treatment with rivaroxaban (during or within 3 days after last intake). In the on-treatment analysis, these 80 deaths translated into an all-cause mortality rate of 3.5/100 patient-years.
During follow-up, a total of 277 patients discontinued rivaroxaban, which translated to a discontinuation rate of 12.0/100 patient-years in Kaplan-Meier analysis (95 % CI 10.6-13.5). Discontinuation rates were higher at the beginning of the study period and declined rapidly over time with no statistically significant differences for newly anticoagulated patients or those receiving rivaroxaban after VKA pre-treatment (▶Table 5). Rates

Patient characteristics
Our findings, which are based on prospectively collected data derived at patient level, contribute to the recently published results regarding the daily-care effectiveness and safety of rivaroxaban treatment. Although our patient cohort was older and had slightly higher CHADS 2 -and HAS-BLED scores compared to some of the other recently published daily-care SPAF cohorts (15-23) these dif-ferences were marginal, which is important for the external validity of our findings.
In contrast, compared with the cohorts treated in the ROCKET AF trial (6), patients in our daily-care registry were older (73 vs 75 years) and more often female (40 % vs 48 %). Furthermore, patients in our cohort less often had a history of stroke/TIA/systemic embolism (55 % vs 15 %) or hypertension (93 % vs 83 %), which also resulted in a difference in mean CHADS 2 score (3.5 vs 2.4).    The most appropriate comparison may be performed with the recently published prospective XANTUS registry that evaluated the effectiveness and safety of rivaroxaban therapy in 6784 patients from 311 centers in Europe, Israel and Canada (19). Using a similar design and similar outcome definitions, event rates in this slightly healthier cohort (mean CHADS 2 score 2.0 compared to 2.4 in our cohort) were 1.8/ 100 patient-years for stroke/TIA/systemic embolism and myocardial infarction, and 2.1/ 100 patient-years for ISTH major bleeding. Of note, the mean follow-up duration in XANTUS was considerably shorter compared to our analysis (329 days vs 796 days).

Dosing pattern
Comparing characteristics of patients receiving either 20 mg or 15 mg rivaroxaban OD, our data suggest that physicians tend to use the lower dose of 15 mg rivaroxaban OD more often in elderly or female patients and in those with significant co-morbidities that increase the risk of complications. Such a selection is not surprising because, compared with patients who receive 20 mg OD, these

Thromboembolic and bleeding outcomes
In line with the differences in risk profile, the rates of centrally adjudicated stroke/TIA/systemic embolism (2.03/100 patient-years in the intention-to-treat analysis and 1.7/100 patient-years in the on-treatment analysis) were considerably lower than those in the ROCKET AF trial (6). Furthermore, in the on-treatment analysis we found event rates to be low for patients receiving 20 mg OD (1.25/100 patient-years), but considerably higher for those receiving 15 mg OD (2.7/100 patient-years). Direct comparisons between phase III trial and registry data and between observational studies need to be made with great caution, because registry data are subject to selection bias since they are not derived from randomised (and, therefore, directly comparable) cohorts.
However, even within these limitations, our effectiveness and safety data for rivaroxaban are in line with those recently published from other prospective cohort studies or retrospective database analyses to evaluate effectiveness and safety of NOACs in daily care in the US, Scandinavia, France or Germany (▶Table 6). patients also have a higher risk of renal impairment (29.7 % vs 4.5 %) and a significantly higher HAS-BLED score (73 % in the 15 mg cohort had a score ≥ 2 vs 57 % in the 20 mg cohort). Interestingly, also XANTUS reported a relevant proportion of patients (21 %) receiving reduced dosages of rivaroxaban, many of which did not have moderate to severe renal impairment (19). Because reduction of rivaroxaban dosage is only recommended for patients with moderate-to-severe renal impairment these data seem to indicate that physicians in daily care consider individual patient characteristics beyond renal function for their selection of NOAC dosage. A similar dosing pattern has been described in dabigatran studies. In the Dresden NOAC Registry, 341 dabigatran patients were prospectively followed, of whom 183 (53.7 %) received the reduced dose of 110mg twice daily (BID) (11). These patients were significantly older and had higher CHADS 2 and HAS-BLED scores than the 158 patients receiving the standard dose of 150 mg dabigatran BID. Similarly, in the Danish nationwide atrial fibrillation study, 59.1 % of patients receiving dabigatran were treated with the lower dose of 110 mg BID. These patients were also significantly older and had higher HAS-BLED and CHADS 2 scores than patients receiving 150 mg BID (16). While we lack detailed information on the reasons for NOAC dose reduction beyond renal impairment, it is likely that fear of bleeding is the dominant driver for this decision, since higher HAS-BLED scores at baseline and higher bleeding rates during follow-up have been consistently described for cohorts of patients treated with reduced NOAC dosages (11,16,19,23,24). In this context it is important to understand that, while major bleeding with NOAC seems rare in daily routine, non-major bleeding events are frequently observed and may have a relevant impact on patient`s health and adherence to medication. Therefore, the overall bleeding risk should also not be underestimated for NOAC treatment.
Our data indicate that bleeding complications are common in patients receiving chronic anticoagulant therapy. We found that major bleeding events during rivaroxaban therapy occurred at a rate of 3.0/100 patient-years, which is lower than the major bleeding rate in the ROCKET AF trial (3.6/100 patient-years) (6). Therefore, our data at least indicate that the daily-care safety of rivaroxaban is not worse than that reported for selected patients in a large randomised trial. This is in contrast to the increase in major bleeding seen with VKA in daily-care settings (> 5 %/year) (25)(26)(27)(28) compared with the much lower rates seen in clinical trials.
In addition to this, we recently published safety data for a cohort of carefully selected (low bleeding risk) and well-managed (time-in-therapeutic range 75 %) VKA patients from the Dresden NOAC Registry (12). In this cohort of stable VKA patients, the rate of major bleeding was found to be as high as 4.2/100 patientyears. Table 6: Overview on the design, mean follow-up and outcomes of the presented analysis in comparison to ROCKET-AF and published "realworld" studies on the effectiveness and safety of NOAC therapy in AF.

Potential of interventional bias
It could be argued that our quarterly phone calls to the patient were 'interventions' that influenced treatment persistence and contributed to the low event rates. If this really is the case, it would be a favourable effect in real life: if a simple, short contact with the patient once a quarter is sufficient to achieve high treatment persistence with NOACs this would be easy to perform, given that most patients will come to see their physician for a new prescription at least every three months. Consequently, we would Hecker et al. Stroke prevention with rivaroxaban in daily care

Treatment persistence with rivaroxaban therapy
In daily care, drug persistence to oral anticoagulation is an important topic, given that most atrial fibrillation patients are at life-long risk for major stroke. Our data indicate that, after 2.5 years, 78 % of patients were still taking rivaroxaban. Therefore, we could confirm previous findings that indicated that treatment persistence with rivaroxaban in daily care is better than that reported for dabigatran or VKA in our registry (11,12,14). Further details on rivaroxaban discontinuation are provided in a dedicated publication (14). Our findings are also in line with other real-world studies that demonstrated high treatment persistence with rivaroxaban (17,19,29,30).

Limitations
There are several limitations to our study.

Sample size
Although the reported cohort of SPAF patients treated with rivaroxaban consisted of 1204 patients, the sample size may still have been too small to detect statistically significant signals. However, even with this number we were able to detect significant differences between the 20 mg and 15 mg treatment groups. Although we agree that the small sample size limits the generalisability of our data as a stand-alone study, our data are in line with larger observational studies that used very different trial designs (15-17, 29, 30). Therefore, our study, performed at patient level, can be regarded as a confirmation of findings from large cohort studies that used database-derived data without central event adjudication.

Design and potential for bias
The design of our registry introduces the possibility of a selection bias, because local physicians within the network are not instructed on which of their patients should receive which type or dosage of oral anticoagulant therapy. While the potential for selection bias can never be avoided in observational cohort studies such as daily-care registries, we previously reported that this selection is, if anything, in disfavour of NOAC, since patients in the NOAC cohort were at higher risk of cardiovascular or bleeding complications (8-10) compared to those stable patients that are still prescribed VKA by the participating physicians (12). Despite this patient selection bias disfavoring NOACs, our results show that, for our specific cohort, the overall rates of outcome events are similar or even lower compared with the event rates found in the respective phase III trial (6) and lower than those observed in VKA patients treated in real-world settings (26).

Lack of randomised comparator
The lack of a direct randomised comparator group could be regarded as a limitation. However, several large VKA cohort studies in daily care (31)(32)(33) as well as large database studies comparing recommend using these visits opportunities for brief physician contact to remind patients of the risks associated with atrial fibrillation and the importance of regular drug intake, as well as to perform a short assessment of problems or potential side-effects.
Despite these potential limitations, the duration of follow-up, the prospective data collection at patient level in unselected dailycare patients, and the central adjudication of effectiveness and safety outcomes are significant advantages of our study compared with other observational studies on rivaroxaban in SPAF. Additionally, the use of a central adjudication process contributes to the strength and clinical impact of our data.

Conclusion
Our study contributes to the confirmation of effectiveness, safety of and persistence with rivaroxaban therapy in unselected SPAF patients from daily care.