The Role of and Selection Process 
for the New Anticoagulants in 
Nonvalvular Atrial Fibrillation

Please subscribe to use our print features or to download PDF files.

In regard to the worst complication of anticoagulant therapy in patients with atrial fibrillation, which is a hemorrhagic stroke, rates of intracerebral hemorrhage were amongst the lowest ever reported in clinical trials. Warfarin actually performed very well compared to historical rates at 0.4% per year in the RE-LY trial. The rates on the two doses of dabigatran were approximately the same at 0.1% per year, lower than on warfarin with respect to this terrible complication of anticoagulation.

When you look at overall bleeding, which as I said, with the higher dose of dabigatran was statistically comparable to that with warfarin, the excess of bleeding on dabigatran was predominantly in the gastrointestinal (GI) tract. There was more GI bleeding with dabigatran 150 mg twice daily than warfarin, substantially less intracerebral hemorrhage, and was comparable between the two treatments for most other types of bleeding. There was a so called “signal,” that did not reach statistical significance for higher rates of myocardial infarction (MI) in patients given dabigatran compared to warfarin. When all MI events, including clinically silent events diagnosed based on new Q waves on the electrocardiogram, were included in the analysis, the differences were not statistically significant. It remains controversial whether or not there is an incremental risk of MI in patients with atrial fibrillation treated with dabigatran compared to warfarin. It’s worth emphasizing that regardless of how you interpret the statistical differences, the overall rates of MI in patients with atrial fibrillation are relatively low compared to the rates of stroke.

RE-LY was a game changer, because these trials were all designed with the idea of noninferiority, yet for the first time—and we thought this could not be done—we have an oral anticoagulant that achieved statistical superiority in the form of high-dose dabigatran, the major methodological limitation being the open-label nature of the trial. The FDA chose to approve only the higher dose and it remains debatable whether that dose is in fact too high a dose for some patients, particularly elderly patients, with respect to major extracranial bleeding.

Dr. Reiffel: Thanks Jon. Jerry, how would you comment on ROCKET with that background from RE-LY?

Dr. Naccarelli: ROCKET-AF was a prospective randomized trial.2 It had similar endpoints to the RE-LY trial using rivaroxaban. It was done a little bit differently in that where the RE-LY was in a prosepective, randomized, open-label, blinded-endpoint (PROBE) design as Jon just discussed, ROCKET-AF was a randomized, double-blind, double-dummy study. So patients taking either rivaroxaban or warfarin were not aware of what they were taking. The entry criteria were similar except for on purpose, the CHADS risk factors were higher, so the mean CHADS score in ROCKET-AF was 3.5, which was different than the 2.1 from the RE-LY trial.

The other subtle difference, or not so subtle difference, was that patients were randomized to rivaroxaban 20 mg a day if their creatinine clearance was 50 or higher. But if their creatinine clearance was 30 to 49, they were prospectively randomized to a lower dose adjustment of 15 mg per day, and the medicine was given with the evening meal. Now, if we look at the efficacy results related to the primary endpoint of stroke in non–central nervous system embolism, the intention to treat results were noninferior compared to warfarin, but they weren’t superior. There was a hazard ratio of 0.88. Interestingly, the hazard ratio was 21%, which was statistically superior, if one just used the on-treatment data.
If you look at bleeding, the results were very similar to what Jon just discussed in the RE-LY trial. Overall, there was no major difference; with major bleeding between rivaroxaban and warfarin, they were very similar. However, GI bleeds were higher, and when you look at fatal bleeding, it was lower. Intracranial hemorrhage was statistically lower. So these are very similar to the results from rivaroxaban; that is, there is no difference in the overall bleeding rate and increased risk of GI bleeding counterbalanced by a reduction in intracranial bleeding, and thus, fatal bleeding.

The other subtle differences in this trial, compared to what Jonathan just told you, is that the mean time in the therapeutic range for the warfarin arm was lower. It was about 55%. Some of this would be expected because of the higher CHADS score. So people with the higher CHADS score, if you look at studies in the literature, have a lower chance of keeping in the therapeutic range. The other subtle difference is that although hemorrhagic stroke was statistically reduced and similar to dabigatran, ischemic stroke was not any different between the rivaroxaban and warfarin arms.

The other—probably—piece of information that I think is important is when the RE-LY trial was done with dabigatran, patients were continued on dabigatran for a period of time and some were enrolled into a longer trial called Long Term Multi-center Extension of Dabigatran Treatment in Patients with Atrial Fibrillation Who Completed RE-LY Trial (RELY-ABLE).8 There was a little bit of a lesson maybe in trial design here, that at the end of the ROCKET trial, the centers were notified and said, “The study is over; thanks for participating. Since rivaroxaban is not commercially available, you need to discontinue the rivaroxaban and switch patients to whatever the therapy you or the patient decide on.” So this actually left patients after the trial, whom had been on rivaroxaban, exposed for a period of actually close to two weeks with a subtherapeutic INR. Some of the warfarin patients also had a short period of subtherapeutic INRs, because they went from double-blind, double-dummy warfarin to open-label warfarin.

There were 22 strokes in the rivaroxaban arm after the discontinuation of the trial compared to only six in the warfarin arm and some people had raised the issue of a rebound phenomenon.2 Although rebound phenomenon can’t be completely ruled out, what this probably told us is that people with a higher CHADS score, like were in ROCKET-AF, who discontinue their therapeutic anticoagulation, do expose themselves to some embolic events. Some people have actually done the math and come up with almost the same exact numbers of events if you had patients for weeks on end without therapeutic anticoagulation.

In summary, I think the differences were that this drug was not inferior. The bleeding rates were similar. This was done in a double-blind, double-dummy fashion. This raises the issue for all these drugs that you have to be careful that if they cause bleeding, not to stop them, because if you stop them, you expose the patients to embolic risk. So, I think that summarizes the bulk of the highlights from that trial.

Dr. Reiffel: Thanks, Jerry. Peter, would you give a similar brief review of ARISTOTLE?


Dr. Kowey: ARISTOTLE was probably somewhere in between these two in terms of its design.3 That is, it was a double-blind study and it was, obviously, a warfarin comparator study, but the patients who were enrolled in this study were not as vascularly sick as the patients in ROCKET. They were a bit more similar to the patients in RE-LY. That is, it was in a broader spectrum of vascular risk.

Like the ROCKET study, however, there was an opportunity to dose adjust based on clinical characteristics including renal function. This variability in dosing limited the exposure of patients who, otherwise, may have been overexposed to the drug. I think it really sets something of a new standard with regard to the results. What we saw in ARISTOTLE was not only superiority on the combined endpoint of ischemic and hemorrhagic stroke and systemic embolism, but what also looked to be an advantage with regard to bleeding.
One other interesting aspect of ARISTOTLE was that it had a companion study called A Phase III Study of Apixaban in Patients with Atrial Fibrillation (AVERROES),9 which was an aspirin-controlled study in patients who were deemed not to be suitable for warfarin. Most of us regard aspirin to be a fairly ineffective drug for this indication. So in some respects, it actually constitutes a placebo control. The reason why AVERROES is such an interesting study is because the bleeding was actually comparable to or even less for apixaban than it was for aspirin. This is remarkable. That result also held up in ARISTOTLE where the risk of bleeding was lower for apixaban compared to warfarin.
The final part of the study that I think is of interest is that there was also a strong trend—actually, a nominal P value for significance—for total mortality. So ARISTOTLE was successful in many respects with regard to what it proved for some meaningful endpoints. It did not prove a benefit in terms of ischemic strokes as did the RE-LY experience that you heard about, but it was similar to rivaroxaban in showing a reduction in hemorrhagic strokes. That really was the driver for the endpoint reduction that was seen in this study.

The final comment about ARISTOTLE is that the data haven’t yet been completely reviewed by regulatory agencies. We haven’t seen a public display of the data at an advisory committee as we have for rivaroxaban and dabigatran. After having been involved in regulatory medicine for a long time, I always like to see these things aired in public and at an advisory committee before I will accept all of the results as being as good as they look in a manuscript. That having been said, if the data do hold up and this drug has done what it’s been reputed to do in the manuscripts, it will be a big help for our patients, I believe.

Dr. Reiffel: Peter, in light of the last comments you made, it may be of interest—and very timely—that in this week’s issue of Stroke,10 new guidelines were published. And those guidelines say that warfarin is a Class I, level of evidence A; dabigatran, Class I, level of evidence B; apixaban, Class I, level of evidence B; and rivaroxaban, Class II-A, level of evidence B. So it says they are all indicated for the prevention of first and recurrent stroke in patients with nonvalvular atrial fibrillation. The selection of the antithrombotic agent should be individualized from the basis of risk factors, cost, tolerability, patient preference, potential for drug interactions, and other important clinical variables, including the time in an INR therapeutic range if the patient has already been on warfarin. It seems rather unusual, however, and perhaps inappropriate, that these new guidelines already include apixaban, given the fact that it has not yet been approved by the FDA, which has requested additional information from the manufacturer.

We’ve covered the background behind these new recommendations and it’s a very timely topic. Because the issues always come up when we talk about these three trials, let me now make a couple of comments as to whether one can compare one to the next. There are important similarities and the guidelines extrapolate from the trials and make them now all equivalent to warfarin, but there are also significant differences. And when we compare across trials, we have to be a little cautious to be sure that the comparisons are fair. So for example, if one has placebo control trials, which is not true in this case, and one wants to compare them, if the placebo rates are different you can be sure the populations are different. Then it becomes hazardous to compare the actual drug efficacy rates.

In these trials, the comparison was not placebo, but it was warfarin given essentially the same way. One would like to look at the warfarin event rates in the three trials if one wants to extrapolate the drug effect rates. You’ll see if you do that, the data from ARISTOTLE and RE-LY are very similar. The warfarin rates were similar. However the warfarin rates were higher in ROCKET-AF, and that tells us the population is different. And, in fact, ROCKET-AF had a higher mean CHADS score of 3.5 compared to about 2.1 for RE-LY and for ARISTOTLE. Thus, it’s a little hazardous to take the absolute event rates in each of these trials and make direct comparisons.

That having been said, there are clearly other issues—and I think we’ve covered them—the issues of drug interaction, issues of renal function—and they all have to come into play when we go ahead and make recommendations for particular patients. Peter, you mentioned that the FDA did not approve of the 110 twice daily dose for dabigatran and that perhaps was a shortcoming for clinicians in this country. It certainly is available in other countries, including Canada. Can you speculate for us why the FDA did that?

Dr. Kowey: Well, Jerry and I wrote an editorial in the American Journal of Medicine asking the same question and I’ve had an opportunity to discuss this at length with the regulatory people.11 The reason boils down to the fact that if you look at the hard numbers—if you just take a statistical approach—it’s very difficult to identify a subpopulation of patients for whom the 110 mg dose represents a therapeutic advantage over risk. Every way you look at this—remembering that the 150 mg dose is superior for the indication—it is very hard to find that there is so much more bleeding at the 150 mg dose that it outstrips the benefit of stroke reduction. I think there is also a belief that it’s much better to prevent a stroke than to prevent a bleed and, therefore, it’s better to err on the side of the therapeutic advantage. That’s not an arguable point. It’s completely true.

What Jerry and I said—and I know you and Jonathan have said the same thing—is that this is probably a decision that’s better left to the judgment of clinicians rather than to regulators. There clearly are some patients who are at such a high bleed risk, that we would probably want to spare them that bleeding risk, but still provide them with the drug that is at least noninferior to warfarin for the indication. That’s what rivaroxaban is—it’s noninferior for the indication. In most other countries where the drug has been approved, the 110 mg dose is available and I think American physicians need to have that kind of latitude. But again, as I said, if you take a very hardline mathematical approach to the question, the 110 mg dose really doesn’t pass muster.

Dr. Naccarelli: Jim, if you look at the Canadian use of the 110 mg dose, it’s as high as in one in four patients. Now if all those are appropriate, it shows that clinicians, based on the data that they know from these trials we just discussed, think it’s the right dose. Not having the 110 mg dose in the United States, works against dabigatran. What will happen is if you have a patient, say with borderline renal clearance, with a drug interaction like dronedarone, instead of going to the lowest dose of 75 mg twice a day, which was suggested in the package insert based on no data, no trials, patients will use a different drug. You could use rivaroxaban with a creatinine clearance of 38 saying, I know I’ve already dose adjusted in the trial and it’s a safe thing and I know what my outcome is. So I think it’s worked against dabigatran. I still think it was mistake. Peter and I have been pretty strong about that, but whether that will change in the United States, who knows in the future?