Introduction
Rivaroxaban, an oral direct factor Xa inhibitor, is approved for the prevention of stroke and systemic embolism (SE) in adults with non-valvular atrial fibrillation (NVAF) with one or more risk factors (e.g., prior stroke) [
1], based on the phase 3, randomized, controlled trial ROCKET AF (NCT00403767) [
2]. In ROCKET AF, rivaroxaban (20 mg once daily [OD], or 15 mg OD if creatinine clearance [CrCl] was 30–49 mL/min) was non-inferior to dose-adjusted warfarin for the prevention of stroke or SE, and similar with respect to the risk of major bleeding or a composite of major or non-major clinically relevant (NMCR) bleeding.
Advanced age and impaired renal function are associated with increased rivaroxaban exposure [
1] and are also independent risk factors for NVAF-related thromboembolism and for major bleeding events in anticoagulant-treated patients [
3‐
6]. It has been proposed that therapeutic drug monitoring (i.e., plasma concentration-based dose adjustment) may help guide anticoagulant dosing for individual patients. This post hoc exposure–response analysis aimed to explore this possibility and to quantify the associations between predicted rivaroxaban exposures, patient characteristics and clinical outcomes in patients with NVAF using data from ROCKET AF.
Discussion
This analysis evaluated rivaroxaban exposure–response relationships in over 7000 patients with NVAF to assess the potential of monitoring drug levels and evaluating patient characteristics in optimizing the benefit–risk profile of treatment.
Warfarin, which requires monitoring, has a clear delineation between international normalized ratio values that are associated with maximum efficacy and those that are associated with increased bleeding risk (i.e. a narrow therapeutic window) [
16].
In this analysis, rivaroxaban showed no clear lower limit of exposure that resulted in loss of efficacy, indicating a wide therapeutic window for efficacy in the NVAF indication. Several patient characteristics were significantly associated with the composite efficacy outcomes, but the CHADS2 score showed no significant association. A likely explanation is that history of stroke (which showed significant associations with both composite efficacy outcomes) was included as an independent risk factor in the model. Impaired renal function (CrCl < 50 mL/min) showed significant associations with both composite efficacy outcomes.
Increasing predicted rivaroxaban Ctrough from the median to the 95th percentile was associated with a significant increase in the risk of major or NMCR bleeding, with a HR of 1.26. The HR for major bleeding was similar (1.25) but the association between Ctrough and the risk of major bleeding was not statistically significant. This may reflect the smaller number of major bleeding events compared with the composite of major or NMCR bleeding events (395 vs. 1475). Thus, the significance of the association between rivaroxaban exposure and major bleeding and the extent to which this contributes to the association between exposure and the composite of major or NMCR bleeding remains uncertain. However, the present analysis does show that the exposure–response relationships for both major bleeding and the composite of major or NMCR bleeding were shallow, with a gradual increase in bleeding risk across a wide range of predicted exposures and no clear threshold of exposure above which the increase in bleeding risk accelerated. The expected increase in the HR of the composite of major or NMCR bleeding, and possibly major bleeding, is therefore small relative to the change in rivaroxaban plasma concentration, which means that any potential gain from measuring rivaroxaban levels and forcing a change in dose would be limited. The CIs around the 1-year estimates of bleeding event rates were wide for any given rivaroxaban concentration and overlapped within the 5th and 95th percentiles of exposure. Taken together, these results suggest that therapeutic drug monitoring would be of limited benefit in patients with NVAF receiving rivaroxaban under the prescribed regimen.
Our analysis identified age, NSAID or aspirin use, history of GI bleeding and low baseline hemoglobin as the components of the HAS-BLED and other bleeding scores [
4,
8,
11], which were statistically significant risk factors for major bleeding. These patient characteristics therefore appeared to be more important determinants of risk than rivaroxaban exposure. The increased risk of major bleeding in North American patients compared with those from Western Europe observed in this analysis may be due to ascertainment bias or other confounding factors, such as comorbidities [
12]. For major or NMCR bleeding, patient characteristics such as history of GI bleeding and age were statistically significant risk factors, with an impact similar to or greater than rivaroxaban exposure. For example, increasing rivaroxaban C
trough from the median to the 95th percentile (from 52.55 to 124.13 µg/L) increased the risk of the composite of major or NMCR bleeding by 26%, whereas having a history of GI bleeding increased this risk by 47%.
Similar findings regarding the effects of exposure and patient characteristics on bleeding risk have been reported for edoxaban, another direct factor Xa inhibitor. In separate analyses of phase 2 and phase 3 trial data, there were significant increases in bleeding risk with increasing edoxaban exposure in patients with NVAF [
17,
18]. In contrast to the present results for rivaroxaban, the relationship between edoxaban exposure and bleeding risk was steep over the exposure range [
18]. However, edoxaban dose reductions based on patient characteristics in the phase three trial were associated with preservation of efficacy and further reductions in the incidence of major bleeding compared with warfarin (dose reduction vs. no dose reduction; p interaction ≤ 0.023), leading the authors to conclude that the data validate the strategy of tailoring the dose based on clinical factors alone and that such a strategy obviates the need for drug monitoring [
19]. The significant variability in exposures in both the edoxaban and dabigatran trials and thus the potential difficulty in selecting threshold drug concentrations for guiding dose changes was also highlighted [
18,
19].
The dosage of rivaroxaban in ROCKET AF was tailored based on renal function (20 mg OD reduced to 15 mg OD in patients with a CrCl of 30–49 mL/min) and these dosages were subsequently approved for the NVAF indication [
2,
20]. Renal function is also a key consideration in decision-making regarding peri-procedural management of rivaroxaban therapy [
21]. While monitoring of coagulation and plasma drug concentrations has been proposed in some patients for guiding pre- and peri-procedural management of direct oral anticoagulants [
22], expert consensus from the American College of Cardiology (ACC) focuses on the importance of patient and procedural risk factors. The ACC recommends that patient risk factors for bleeding followed by bleeding risk of the procedure be considered for the decision on whether or not to interrupt therapy, and that the specific drug and level of renal function then be used to guide the timing and duration of interruption to therapy [
21]. Results from the present analysis support the central role of patient characteristics in decision-making processes regarding bleeding risk with rivaroxaban and the limited likely value of adding drug monitoring into management pathways.
Limitations of this analysis include the paucity of direct rivaroxaban plasma concentration measurements in ROCKET AF, although this was partially offset by the PT adjustment in some patients [
14,
15]. The predicted C
trough values showed moderate between-patient variability (coefficient of variation: 54%) and were consistent with the previously published ROCKET AF popPK model [
23]. In addition, because ROCKET AF was not designed to evaluate exposure–response relationships, the current analysis may have been underpowered to detect statistically significant differences for some outcomes. Finally, the exposure–response analysis included baseline use of antiplatelet agents and NSAIDs but did not evaluate the impact of their continued use during follow-up.
Acknowledgements
This work was conducted at Bayer AG, Bayer U.S., LLC, and Janssen Research & Development, LLC. E. Bolton of Oxford PharmaGenesis, Oxford, UK, provided medical writing support, which was funded by Bayer AG, Berlin, Germany. The authors would like to thank D. Garmann (Bayer AG, Wuppertal, Germany), T. Spiro (Bayer U.S., LLC, Whippany, NJ, USA), C. Nessel, A. Sharma and M. Samtani (Janssen Research & Development, LLC, Raritan, NJ, USA), J. Piccini (Duke Clinical Research Institute, Durham, NC, USA) and R. C. Becker (University of Cincinnati College of Medicine, Cincinnati, OH, USA) for their input and helpful comments during the preparation of the manuscript.
Compliance with ethical standards
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.