Author + information
- Matthew R. Reynolds, MD, MSc⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Matthew R. Reynolds, Harvard Clinical Research Institute, VA Boston Healthcare System, 930-W Commonwealth Avenue, Boston, Massachusetts 02215
Late in 2011, rivaroxaban became the first factor Xa inhibitor to receive regulatory approval for the prevention of stroke in patients with atrial fibrillation (AF), based largely on the results of the phase III 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) trial (1), which enrolled 14,264 patients in a double-blind, double-dummy design. The ROCKET AF trial clearly established the noninferiority of rivaroxaban compared with warfarin for the primary endpoint of stroke or systemic embolism. The superiority of rivaroxaban was less clear: statistical superiority was shown (hazard ratio: 0.79, p = 0.02) in the as-treated safety population, which counted only on-treatment events (during treatment or 2 days or fewer after study drug discontinuation), but not in a full intention-to-treat analysis, which also considered events occurring outside the on-treatment window (hazard ratio: 0.88, p = 0.12).
In their initial trial publication (1), the ROCKET AF investigators reported that stroke or systemic embolism event rates were numerically higher, but not statistically different, between rivaroxaban and warfarin after early, permanent drug discontinuation, which occurred in nearly one third of the patients during the trial. In contrast, a significantly higher event rate was seen with rivaroxaban as compared with warfarin patients from 3 to 30 days after study drug discontinuation at the end of the trial, when most patients were transitioned to open-label vitamin K antagonist (VKA) therapy.
This latter finding raised particular concern during the Food and Drug Administration (FDA) review process and led some to wonder whether rivaroxaban discontinuation may be associated with a prothrombotic rebound effect (2). The notion of a rebound effect has been hypothesized for warfarin, but is of questionable clinical relevance (3). In the case of rivaroxaban, there is not a clear mechanistic hypothesis for a rebound effect. Nonetheless, the clinical events at the end of the ROCKET AF trial have raised questions about this, as well as about the appropriateness of once-daily dosing of this drug, which has a half-life of 5 to 9 h, in AF patients. The FDA's concern about these end-of-study events culminated in a black box warning of “an increased risk of thrombotic events” with rivaroxaban discontinuation.
In this issue of the Journal, Patel et al. (4) begin to bring more perspective and clarity to the previous observations by presenting more comprehensive and detailed data surrounding clinical events that occurred during all interruptions or discontinuations of study medication during the ROCKET AF trial. The authors report on the rates of stroke and systemic embolism as well as other secondary outcome measures under 3 sets of circumstances: 1) temporary interruptions of the blinded study drug, which were common (> 8,000 occurrences) and had a median duration of 6 days; 2) 3 to 30 days after early permanent discontinuation of the study drugs (n = 4,895); because of adverse events, withdrawals, and suspected or confirmed primary endpoint events; and 3) 3 to 30 days after permanent study drug discontinuations at the end of the study in surviving patients without suspected primary end point events or early permanent discontinuation (n = 9,239).
Temporary interruptions of the study drug were associated with event rates of roughly 5 to 6 per 100 patient-years and did not differ between rivaroxaban and warfarin patients. Early permanent discontinuations were associated with much higher event rates (23 to 25 per 100 patient-years), but again did not differ based on assigned therapy. It should be noted that event rates in this group were exaggerated by the inclusion of patients whose study drug discontinuation was prompted by a suspected primary endpoint event: approximately 8% of these patients went on to have a first or new primary endpoint event between 3 to 30 days after stopping their blinded therapy. After excluding this group, event rates were 8.1 and 9.4 per 100 patient-years in the rivaroxaban and warfarin arms, respectively.
As previously reported (1), and in contrast to the results during temporary or early permanent discontinuations of study drug described above, a relatively large difference in stroke or systemic embolism rates for rivaroxaban versus warfarin patients was seen 3 to 30 days after the end of the study (22 vs. 6 events and 6.4 vs. 1.7 per 100 patient-years, hazard ratio: 3.72, p < 0.01). Somewhat surprisingly, the new analysis also reveals an increased risk of major bleeding in rivaroxaban patients (7.3 vs. 2.0 events per 100 patient-years, hazard ratio: 3.62, p = 0.003) during this same timeframe.
What explains this observation of increased risk for rivaroxaban patients only at the end of the study? Although approximately 92% of patients in both study arms began open-label VKA therapy after the trial ended, as the authors show, there seems to have been an almost immediate imbalance in anticoagulation coverage between the groups during this transition. Roughly 60% of patients who had received warfarin during the trial had an international normalized ratio (INR) of 2.0 or more when next measured 3 days after trial end, and 81% had an INR ≥2.0 at least once in the 30-day post-trial period. In contrast, only approximately 25% of rivaroxaban patients had an INR ≥2.0 in the first 7 to 10 days after study end, and only 49% ever became therapeutic within 30 days. Independent analysis of this data by the FDA corroborates that the quality of VKA anticoagulation in the rivaroxaban patients was poorer during this timeframe than in the warfarin patients—and also shows that the last recorded INR preceding end-of-study stroke events was subtherapeutic roughly two thirds of the time (5).
The authors' implication from these data is that the most likely explanation for the observed risk in the post-study transition period is not that rivaroxaban has some property resulting in a rebound effect, but rather that the high-risk patients enrolled in the ROCKET AF trial (mean age: 73 years, mean CHADS2 score: 3.5) had a substantial difference in anticoagulation coverage during this period, and the event rates merely reflect the unmasking of their underlying risk. Indeed, the event rates reported in the current report seem to be consistent with this notion: during temporary interruptions of either study drug and during the post-study transition period for rivaroxaban patients, the event rates were approximately 5 to 6 per 100 patient-years, a rate quite close to what would be predicted based on their baseline CHADS2 scores (6). During the post-study transition period for warfarin patients, the event rate of 1.7 per 100 patient-years was close to that observed in the on-treatment analysis during the trial (1).
Although this explanation is fairly persuasive, the evidence that the post-study excess stroke risk in rivaroxaban patients was the result of inadequate VKA therapy remains somewhat circumstantial. The INRs were not collected as carefully during the post-trial period as during the in-trial period (5); thus, a statistical analysis adjusting for INR values or time in therapeutic range may not be possible. More importantly, the authors do not provide any information on the use of bridging therapies. Bridging with unfractionated or low–molecular-weight heparin was raised as a consideration, but was not mandated by the study protocol during either temporary interruptions or at the end of the study. The bleeding rates reported in the current study also are counterintuitive: if the excess strokes in rivaroxaban patients were the result of underanticoagulation in the post-trial period, then why did these patients also have a higher bleeding risk? The investigators will need to scrutinize this large trial database further to understand these issues more fully.
What is clearest from this important subanalysis of the ROCKET AF trial is that bad things will happen to high-risk AF patients if they are left untreated with effective anticoagulant therapy for sustained periods—and that in a population as large as this one, it does not take much time for those events to begin to accumulate. Although the black box warning may seem to single out rivaroxaban unfairly in delivering this message, it at least serves as a reminder that interruptions and transitions with short-acting anticoagulant drugs must be planned and managed carefully. Of course, defining what carefully means requires much more evidence than is available currently. In invasive electrophysiology, we have learned only recently that bridging strategies generally do more harm than good compared with doing procedures such as cardiac rhythm device implantation or AF ablation under uninterrupted warfarin therapy (7,8).
The current study also highlights the particularly high short-term risks present under the circumstances that resulted in early permanent discontinuation of the study drugs, particularly those associated with adverse events or suspected strokes or transient ischemic attacks. These challenging scenarios point to continued unmet needs in the area of stroke prevention in AF.
The author is a consultant to Boehringer Ingelheim.
↵⁎ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
- American College of Cardiology Foundation
- Patel M.R.,
- Hellkamp A.S.,
- Lokhnygina Y.,
- et al.
- United States Food and Drug Administration
- Santangeli P.,
- Di Biase L.,
- Horton R.,
- et al.