Rivaroxaban FDA review of ROCKET AF data
Editor-In-Chief: C. Michael Gibson, M.S., M.D. [1]. Dr. Gibson has received research grant support from all major manufacturers of antiplatelets and antithrombins as shown here
Overview
The FDA conducted an advisory committee review regarding the approvability of rivaroxaban for reduction of stroke and systemic embolization in patients with non-valvular atrial fibrillation on September 8th, 2011.
Non-Inferiority Margin
An upper bound of the 95% CI of the hazard ratio was set at 1.38 by the FDA to establish non-inferiority. In no subgroup did the comparison between rivaroxaban and warfarin come near the non-inferiority margin. Non-inferiority was clearly observed in all treatment subgroups and in all forms of analysis.
Primary Efficacy Endpoint
The primary efficacy endpoint was stroke and systemic embolization. The efficacy in the trial was driven by a reduction in hemorrhagic stroke and systemic embolization.
Secondary Efficacy Endpoints
There were fewer non-fatal myocardial infarctions in the Rivaroxaban group.
Safety Endpoint(s)
There were more major bleeding events in the rivaroxaban group, however, the risk of fatal bleeds (H.R.=0.61) and hemorrhagic strokes was reduced in the rivaroxaban group.
Cross Trial Comparisons
Dr. Califf emphasized that cross trial comparisons are inappropriate given the variation in design (open label vs double blind), variation in patient risk (87% of patients were CHADS2 and above in the ROCKET trial vs a third of patients in the RELY and ARISTOTLE trials) and the variability in the distribution of countries with the attendant variation in TTR by country.
Resumption of Events Following Discontinuation of Study Drug
In its briefing document, the FDA expressed concern regarding the excess number of primary endpoint events in the rivaroxaban arm when study drug was discontinued at the end of the trial. There was no excess unmber of events if the drug was disontinued during the course of the study. Dr. Califf argued that the excess number of events among rivaroxaban patients is what would be expected given their high CHADS2 score (87% had a CHADS2 score > 3) while patients were awaiting therapeutic anticoagulation with coumdadin. In other words, these events represented a resumption of events in high risk atrial fibrillation patients awaiting therapeutic levels of anticoagulation versus rebound (which would represent a higher rate of events than if they had never been treated with Rivaroxaban).
Method of Transition at the End of the Trial
- In the rivaroxaban arm, the warfarin was started at a maintenance dose rather than using a loading dose
- There was no overlap with blinded study drug and warfarin
- No INR was checked for three days following the initiation of warfarin
- Heparin bridging was rarely used. This is important as Protein C levels may drop quicker with warfarin treatment and the patient may become pro-thrombotic. Often warfarin is started in the in-hospital setting with overlap with heparin which may minimize this pro-thrombotic risk. In the trial, warfarin was initiated on an outpatient basis.
The independent data monitoring committee (IDMC) overseeing ROCKET indicated that there was no problem with the transition method. At the time the IDMC evaluated the transition period data, there were actually more events in the warfarin arm, and there was no concern about the potential for excess events among patients treated with rivaroxaban. Dr. Califf stated that it is the opinion of the ROCKET investigators that the excess number of events in the rivaroxaban arm during the transition was due to the INR being under 2 for an average of 13 days in the rivaroxaban patients until they achieved therapeutic anticoagulation which was longer than the 3 days in the patients who had been treate . Following rug discontinuation there was a: 6.4% event rate per year in the rivaroxaban arm which is consistent with the rate in patients who are not anticoagulated 1.7 event rate per year in the warfarin arm consistent with effective warfarin therapy The efficacy of heparin bridging has not been established In the RELY trial, patients were rolled over into the RELIABLE study, and there was often not the opportunity to evaluate the patients for rebound. With drug discontinuation during the trial rather than at the end of the trial, there was no excess events.
Time in Therapeutic Range
Dr. Califf argued that:
- TTR is not a surrogate for anticoagulation risk benefit balance.
- TTR is associated with regional variations. The TTR in the North America was 64%, and the treatment benefit in North America was non-inferior.
- TTR has no effect on the benefit of a novel anticoagulant over warfarin. RELY, ARISTOTLE and ROCKET AF all showed no interaction between center TTR and the benefit of a novel anticoagulant over warfarin.
- The events rates in ROCKET were similar to those in RELY and ARISTOTLE when stratified by CHADS 2 score indicating the "skillful use" of warfarin.
How was TTR Calculated?
In RELY and ARISTOTLE, any interruption in drug resulted in exclusion of the TTR during that period. In contrast, ROCKET included TTR data in patients who discontinued drug for < 7 days. This lowers the TTR by 1-2% compared with other trials.
ROCKET included the total time on and off drug in all patients. Patients who were on drug longer contributed more to the TTR at that center. In the RELY and ARISTOTLE trials, which used the Connolly method, a patient on drug for 1 week contributed equally to the average as a patient who was on warfarin for 5 years.
Is TTR a Surrogate Endpoint
Califf argued that TTR does not meet the criteria to qualify as a surrogate endpoint based upon the Flemming article from the Annals of Internal Medicine in 1997. Therefore, the interpretation of TTR should be more circumscribed.
Predictors of Lower TTR in ROCKET
- The country the patient was monitored in is 3 to 4 times more important than all the other factors combined. Imbalances in the numbers of countries participating in the trial will lead to imbalances in the TTR across trials.
- No prior VKA
- Prior stroke
- Congestive heart failure (CHF)
- CHADS2 score > 3
- Female gender
- Greater numbers of concomitant medications
- Poor socioeconomic status
TTR in the North America by Trial
The median TTR in North America in the ROCKET trial was similar to that observed in comparable trials:
- ROCKET: 64% (63% US, 66% Canada)
- RELY: 67%
- ACTIVE: 66%
- Sportif V: 68%
TTR in ROCKET vs RELY Adjusting for Imbalances in Countries
Adjusting for imbalances in the distribution of countries, the TTR in ROCKET was 61.2% and was 64.&% in RELY.
Conclusions by Dr. Califf
- TTR varies substantially by region
- Some regions were slower to adjust the warfarin dose in response to an INR value. This reflects real world practice.
- Center TTR was not associated with the relative magnitude of treatment effect of a novel anticoagulant over a vitamin K antagonist in any recent trial (RELY, ROCKET, ARISTOTLE).
- TTR is not a treatment effect modifier, and in ROCKET there was a constancy in the benefit across all TTRs in a large number of analyses.
- The North American TTR was 64% and is similar to that in other comparable trials and better than the 55% observed in the North American / US registry experience reported by Baker et al.
- The risk adjusted (CHADS2 2 and baseline characteristic) event rates for warfarin patients in ROCKET were similar to those observed in other recent comparable trials indicating that warfarin was in fact skillfully managed in ROCKET.
Comments by Dr. Califf on the FDA Briefing Book
In comparing trials, the FDA briefing book did not point out that RELY was an unblinded trial whereas ROCKET was blinded and double dummy. In RELY, the patients and Drs. knew the INRs, and that could have affected INR management. Dr. Califf stated that drugs should not be "ranked" as a second or third line drug by the FDA based upon inappropriate cross-trial comparisons.
Why was Hemorrhagic Stroke Included in the Efficacy Endpoint?
Dr. Sanjay Kaul asked the above question. Dr. Califf indicated that when a patient dies of a stroke, caregivers often do not know whether it was a hemorrhagic or ischemic stroke, particularly if the patient died soon after hospitalization before imaging could be obtained. Therefore the two were lumped together in the primary endpoint. There is a grey zone in classifying patient's with ischemic stroke who undergo hemorrhagic conversion. 18 patients could not be classified as hemorrhagic or ischemic strokes.
Were the Sites Educated on the Importance of INR?
Ken Mahaffey, the study Principal Investigator indicated that yes, there was a significant amount of site education. There was, however, no algorithm as to how to manage drug dosing to achieve the target INR. The importance of INR management was emphasized to the sites during meetings and in newsletters. Dr. Steve Nissen of Cleveland Clinic asked why "third world" sites were not given more guidance in adjusting the INR? "You left them on their own?" Nissen was criticized for his somewhat demeaning use of the term "third world" by Califf. Nissen: "You could have instructed patients to come back quicker if the INR is < 1.5". Califf emphasized that this was a "real world" trial. Califf emphasized that the TTR in each country was similar to that in other trials. The low TTR in ROCKET reflected the participation of a larger number of sites from countries where TTRs are traditionally lower.
How Did TTR Affect Ischemic and Hemorrhagic Stroke?
Dr. Temple of the FDA pointed out that the benefit was largely driven by a reduction in hemorrhagic stroke and asked how over and under anticoagulation affected the benefit of rivaroxaban. Califf showed that those patients who were somewhat under-anticoagulation still had a reduction in ICH. This a powerful argument in favor of the safety of rivaroxaban.
Twice a Day Versus Once a Day Dosing of Rivaroxaban
A member of the FDA team argued that rivaroxaban should have been administered as a twice a day drug rather than a once a day drug based upon its half-life of about 9 hours. Dr. Califf argued that while the safety and efficacy of twice a day dosing of rivaroxaban might be quite good, once a day dosing might achieve greater compliance than twice a day dosing in real world clinical practice.
Commentary on TTR by Dr. Rose of the FDA
Dr. Rose indicated that the opinions he expressed were his own and not those of the FDA as a whole.
Dr. Rose referred to Dr. Califf as Dr. Cahill, which caused some confusion.
Dr. Rose highlighted that there was no algorithm for adjusting the VKA dose in ROCKET to achieve a target INR. In RELY, he indicated there was an algorithm. This indicates that ROCKET was more of a real world study.
The TTR was 55%. 30% of the time time warfarin patients were under-anticoagulated, and 15% of the time warfarin patients were over-anticoagulated. This could be interpreted as indicating that rivaroxaban is safer than a population of patients who were, if anything, slightly under-anticoagulated with warfarin.
The ROCKET investigators used an analysis that calculated TTR based upon all days all patients TTR was in the therapeutic range / divided by all days all the patient were in the trial. This method weights the contribution of patients based upon how long they were in the trial. The method of Connolly would average each patients TTR at each site, and patients who were in the trial a short period of time contributed equally to those patients who were in the trial a long time.
Dr. Rose argued that the goal should be to achieve a TTR of > 70% in clinical practice. This is in the face of registry data that shows a TTR of 50% in the US.
Dr. Rose acknowledged that the relatively low TTR in ROCKET was largely due to the fact that those parts of the world with lower TTRs enrolled the most patients. The FDA and the sponsor agree that this difference in site distribution drove the lower TTR observed in ROCKET.
Dr. Rose stated "The ROCKET study enrolled patients from parts of the world that did not use warfarin well, and that is our concern."
Again, it should be noted that the TTR in North America was 64% (which is similar to other contemporary trials) and that non-inferiority was demonstrated in the US population.
Commentary on Transition to Open Label VKA at the end of the study by Dr. Rose
Dr. Rose argued that the end of trial data are not inconsistent with a rebound phenomenon. It is Dr. Roses's own personal view that the sponsor should perhaps perform not a clinical outcomes but a pharmacodynamic study that excludes excess thrombin generation prior to approval.
Should Novel Anticoagulants be Compared to Each Other in the Approval Process?
Dr. Rose argued that low dose of dabigatran (110 mg) was not approved because it was inferior to high dose dabigatran (150 mg) even though the 110 mg dose of dabigatran was non-inferior to warfarin in the RELY trial.
Questions from the FDA Panel to the FDA
- Dr. Kaul asked the FDA what they meant by "as effective as"? Does this mean non-ineriority (you rule out harm) or equivalence (the two drugs are indistinguishable)? The FDA indicated that they are interested in seeing robust non-inferiority. The goal in evaluating the ROCKET data is to protect the population from stroke and an inferior outcome.
- The FDA argued that the warfarin arm had event rates that were consistent with historical controls so the constancy assumption was met.
- One panel member indicated that since this trial had such a different cohort of patients, is it fair to hold it to the usual constancy assumptions? Dr. Rose of the FDA responded that the 1.38 HR was derived from the square root of the effect size, preserving half of the effect size in prior placebo controlled trials with a broad range of INR targets (1.5 to 4.5). ROCKET falls within the range of INRs and the TTR of prior trials. Dr. Rose indicated that ROCKET met the constancy assumptions. Dr. Temple indicated that the results appear robustly non-inferior. Most of the advantage of rivaroxaban is on hemorrhagic stroke and a low TTR should if anything make warfarin perform better in terms of safety. Despite this conservative anticoagulation with warfarin, rivaroxaban was safer.
- One panel member asked if patients who have TTRs in the 60-70 range achieve non-inferiority. Rose responded that there are likely too few patients to answer this question.
- Nissen asked if novel antiplatelets had to meet the same non-inferiority margin. Temple indicated that the margin would be derived in the same way: to preserve half the benefit. The 95% CI margin would have been tighter for antiplatelet agents since the relative benefit of clopidogrel over placebo was smaller.
Panel Discussion
There were 9 questions posed to the advisory panel.
Question 1
Comment on the adequacy of design of ROCKET-AF
Was the planned warfarin management strategy reasonable?
Dr. Steve Nissen stated: "I think it was not". "Compared to contemporary trials the TTR was low". "There was not an algorithm to adjust the warfarin dose". "Measures were not taken to meet the highest contemporary standards."
Dr. Kaul: "Some may view lack of an algorithm as a strength. This may reflect more real world practice."
Dr. Lincoff: "The question was not whether the strategy was good or excellent, but rather if it was reasonable". Lincoff felt that stratgey was "reasonable". The lack of an algorithm was a an intentional real world strategy, not a lack of oversight.
Dr. Fox: Agreed with Dr. Lincoff.
Was it reasonable to assess a single regimen of rivaroxaban?
Dr. Sagan: Felt it was reasonable.
Dr. Steve Nissen: "The dose was selected for marketing advantages rather than for scientific reasons". "We don't know when the events happened and if thrombotic events were at a trough of drug levels." "The half life of 7-9 hours is too short for rivaroxaban to be once a day drug."
Dr. Fox: The dosing appeared reasonable. At the time the trial was designed, the regimen was rationale. No ATLAS data was in hand to adjust the dose to BID.
Other panel members thought a bid regimen would make better sense.
However, again, the question was if BID dosing was reasonable, not if it was better than once a day dosing.
For How Many Days Should One Count Events After investigational Drug is Discontinued?
It was 2 days in ROCKET.
Flemming: Dodged question on how many days after drug Dc'd. 30 days after drug is discontinued is a usual way of doing things.
Dr. Steven Nissen: "2 days is just too short." Can't tell you how many days to go.
Dr. McGuire: Questions the need to follow patients beyond thirty days after drug discontinued.
Dr. Kaul: The period of follow-up should be dictated by the pharmacodynamic effect of the drug, which in this case is 2 to 3 days. Kaul was concerned about rebound, and events should be evaluated through 2 weeks.
Dr. Seagan: 14 days
Dr. Temple of the FDA: For safety, if you go a long time after drug is discontinued, this drives things to a null result. The period of safety assessment should be defined prospectively. If you take a beneficial drug away, the two groups will converge, and become equal. The events here happened during the transition. The thromobtic events increased, what to do about it is unclear.
Dr. Emmerson: No problem if you stop the study on a given day with censoring at that time.
Dr. Lincoff: Informative censoring only applies to early discontinuation, not to planned discontinuation at the end of the trial. 2 days off drug is an adequate period.
Dr. Kaul: Since bleeding was the efficacy in this trial, you must follow bleeding closely.
Are There Other aspects of Study Design That Importantly Affect Interpretation of the Study?
Flemming: The transition could have been better. A 9% withdrawal of consent is a problem. Choice of warfarin instead of Dabigatran as a comparator is a limitation.
Other panel members: Waiting for 4 weeks in some countries to bring a patient back for an INR re check may have been too long.
Dr. Kaul: Double counting of ICH events toward both efficacy and safety is a limitation. If you count an ICH as an efficacy event, you should not be able to count this toward the safety endpoint as well. Historical results for the margin (M2) were based upon ischemic stroke reduction, not hemorrhagic stroke.
Dr. Lincoff: Old trials may not have included imaging, and you cannot exclude that more hemorrhagic strokes occurred in older trials.
Question 2
Was anticoagulation of warfarin good enough to show non-inferiority and superiority in the comparator (i.e was warfarin skillfully enough used as a comparator?)?
Dr. Steve Nissen: The TTRs are in mid to high 60s in previous studies. Warfarin wasn't being used as well as it was being used as well as in other contemporary trials. "It falls of the wagon and is a problem." Everyone is trying to make efforts to look in higher TTRs, we want to look at subgroups where warfarin was used well. "This is torturing the data." "I dont know what would have happened at high TTRs."
Dr. Emmerson: People seem to think this is inadequate treatment rather than different patient risk driving the difference.
Dr. Kaul: Even with the poor TTR, rivaroxaban is likely superior to placebo. Clearly not superior to warfarin.
Dr. Seagan: Confusing since rivaroxaban was safer than warfarin in a group of patients that were actuallys lightly under anticoagulated.
McGuire: "TTR was not a treatment modifier. I have no confidence that TTR reflects what we think it does." 55% reflects our 54% TTR at our hospital. This 64% TTR in North America is similar to other trials.
Lincoff: Agrees with McGuire.
Flemming: TTR does not invalidate but meaningfully compromises non-inferiority interpretation.
Nissen: We are better at the Cleveland Clinic. "It is not uncommon to see TTRs in the 70s in some clinics". "Where are we going to draw the line in the sand on TTR and non-feriority margin."
Lincoff: We should not be looking at confidence intervals in subgroups.
Disposition of patients: How does the disposition of patients affect your ability to address non-inferiority or superiority?
Flemming: Superiority not met. Again, rates of withdrawal may affect ability to interpret non-inferiority.
Another panel member: High rate of drug discontinuation may have affected effectiveness of rivaroxaban.
Was the follow-up of endpoints adequate in both treatment groups?
Are there other aspects of study conduct that importantly affect interpretation of the study?
Question 3
Please comment on the effectiveness of Rivaroxaban. How does rivaroxaban compare with warfarin in Rocket, the US, and when warfarin was well managed?
Kaul: Falls short on safety endpoint in non-inferiority.
McGuire: Confidence interval shows non-inferiority in US.
Nissen: It is common to interupt therapy, patients interupt therapy on their own. One of things about warfarin is that it is slow on and slow off. It is forgiving. Pretty bad things happen when you stop Riva due to the short half life. People in trials may be more compliant. "For the 80 year old who cant get a ride to the pharmacy, we have the potential for high event rates with rivaroxaban, and we have not studied how to ameliorate these events".
Kaul: Agrees that TTR may not be a treatment modifier. It is a marker though of quality care. Not sure what TTR reflects quality care. 60%? 80%?
McGuire: I think it is wrong to say that a low TTR is a marker of unskilled care. There are so many things that affect it.
Question 4
The agency does not require new therapies to be more effective than or be compared to other therapies particularly if there is a new patient population where the drug is being studied. Is the population in ROCKET sufficiently different than RELY to obviate a cross trial comparison?
Dr. Fox: Dabigatran was not approved during the course of the trial.
Dr. Kaul: There is no concern in comparing rivaroxaban to older agents, the only concern I have is in comparing the agent to dabigatran as it was studied in RELY.
Most panel members: The patients in ROCKET had higher CHADS2 score and were older but there was overlap and the condition (atrial fibrillaiton) is the same.
Dr. Temple of the FDA: Cross study comparison is not plausible here given the difference in populations between RELY and ROCKET. "Dabigatran makes me want to be sure the comparison to warfarin is good". "Dabigatran should not be the comparator."
McGuire: Do we compare to one drug or multiple ones?
Question 5
What does rivaroxaban need to be as effective as? Is the comparator warfarin? Is the comparator now dabigatran? Is it as effective as warfarin? What does "effective" mean operationally?
Flemming: "Effective as" means "you rule out equality". "Equality" is more stringent than non-nferiority. If no other agent is available, then OK to be non-inferior to warfarin. If another agent is superior to warfarin, why is it ok to be non-inferior to warfarin? Is dabigatran superior? The problem is RELY was unblinded and there were higher MI rates for dabigatran. Flemming to FDA: Was dabi superior?
Dr. Temple of FDA: Label does not explicitly state dabigatran is superior. The data in the tables in the label may provide p values that are consistent with superiority.
Dr. Lincoff: RELY had problems (open label, excess MI) and was performed in a different population (low risk). Therefore it is not an appropriate comparator.
Dr. Nissen: Would you have expected more benefit in a higher risk population like ROCKET?
Dr. McGuire: There are 4 agents competing in this space. It is not realistic to start and stop trials and change their designs as results are released. There have been unforeseen problems with dabigatran such as excess bleeding in certain populations.
Dr. Kaul: ROCKET meets the operational standards of non-inferiority.
Dr. Temple: Bioequivalence is 80-125% of the competing drug. This is not the same aas clinical trial equivalence. Non-nferiority does not mean equivalence.
Question 6
Are there adequate instructions for
- Use of the drug?
- For patients at high exposure risk?
- For transitioning?
- For actions to take if there is bleeding?
Dr. Nissen: "I don't know how to transition this drug." "We are not informed about that." In the absence of data, doctors may do the right or wrong thing (overlap rivaroxaban and warfarin, use UFH, or enoxaparin). "We don't have problems with events when we stop these drugs (like UFH, enox and coumadin) so transitioning is not as much as a problem".
Dr. Lincoff: We don't know what to do regarding the transition. We may need more data. There is no good guidance for any of these drugs regarding the treatment of bleeding.
Dr. McGuire: We don't have evidence based data for how to transition UFH or warfarin or Dabigatran.
Question 7
Should rivaroxaban be approved for reducing the risk of stroke and systemic embolizaiton in patients with non-valvular atrial fibrillation?
Yes: 9 people voted yes (Coukell, Krantz, McGuire, Kaul, Lincoff, Papademtriou, Kinzelski, McCall, Sager)
No: 2 people voted no (Emerson and Nissen)
Abstain: 1 voted to abstain (Flemming)
Dr. Flemming: abstained. Many strengths in this blinded trial. He indicated that the discussion points should provide a good perspective for the FDA in their deliberations, and that both the FDA and Dr. Califf were thoughtful.
Dr. Nissen: If you look at events after the drug is discontinued, the hazard ratio approaches 1. "The low TTR raises questions." "Not having an algorithm for adjusting the warfarin dose was a fatal flaw in study design." "Other trials used an algorithm." "I am concerned about non-inferiority creep." "What happens when the comparator warfarin is used with a TTR of 50 in the future? Should the drug be approved." "Is this a step backward or forward". "The drug should not be approved without transition data."
Dr. Lincoff: Rivaroxaban was clearly non-inferior. Dabigatran does not meet the standard to be the comparator (one open label trial with an increased risk of MI). Lincoff felt data was needed on the transition from rivaroxaban to warfarin before approval.
Dr. Kaul: I voted to approve. I was heartened to see a potential claim for patients failing other therapies. I hope this approval does not lower the standards for proof of non-inferiority. Rebound has not been ruled out. Not clear if transition should be addressed pre or post-approval.
Dr. Emerson: I voted no. This was not a firm no. i am concerned about bridging. I am concerned about non-inferiority creep.
Question 8
Should rivaroxaban be given a label as:
a) A superiorirty claim to warfarin
b) A claim as an effective alternative to warfarin
c) A claim as effective
d) A claim for patients failing other therapy? If so, what constitutes failure?
Dr. Kaul: No except on d.
Dr. Nissen: Ok with d.
Dr. McGuire: Ok for b, c and d
Dr. Lincoff: Agrees with McGuire
Flemming: No for a.
Other committee members: No for a. Mixed on b. Most OK for c and d. People debated if c meant to placebo or warfarin. Kaul said if c was relative to placebo, he would agree.
Quesiton 9
Any constraints on the population the drug should be used in? An additional data you would like to see?
Dr. Nissen: "I would like to see this drug tested as a bid drug." "It might look more like dabigatran if you gave the drug in a more optimal way." "If you do a study, get the TTR in the mid 60s."
Dr. Lincoff asked "Does anyone think the transition plan needs to be studied before approval?" It appeared that Dr. Nissen and 3 others said yes in this straw poll.