Here’s my live-blogg of the FDA’s Cardiovascular and Renal Drugs Advisory Committee meeting to consider the supplemental new drug application (sNDA) for rivaroxaban (Xarelto, Johnson & Johnson) for use in patients with acute coronary syndrome (ACS) already taking dual antiplatelet therapy. Here is a link to the FDA briefing documents.
4:48: Meeting adjourned! J&J has just issued a press release saying they will “ensure the questions raised today are addressed with the FDA.”
4:45: Temple is raising the possibility that the FDA might ultimately approve rivaroxaban for ACS. I would not be shocked if this were to happen.
4:38: Now discussing whether the FDA should say about the use of rivaroxaban with thienopyridine. All agree that people should be on a thienopyridine. What about prasugrel and ticagrelor? Nissen says the reality is that rivaroxaban will be given with these drugs, even though there’s no evidence. This was part of his reasoning in voting no. Sager said he would be more concerned about prasugrel, because of the excess risk of bleeding associated with prasugrel. But he notes there is no evidence to support use with either drug.
4:33: Moving on to questions about what to do if rivaroxaban is approved. Everyone agrees that the 2.5 dose is the dose that would get approved.
4:30: Kaul said he wished he had had the courage to abstain (to general laughter) and said he agrees with Fleming that he’d rather advise than vote, and that there was a very fine line between yes and no today.
4:24: Nissen thinks the decision to use mITT– which is counting events at 30 days after stopping treatment– is what doomed ATLAS. With mITT, he says, you’re telling the investigators and the patients that you don’t care so much about what happens to them long term. Nissen thinks the drug could be approvable if there were a second trial, but he also says its possible that the mortality benefit would not be replicated. He’s worried about exposing hundreds of thousands of patients to a three-drug regimen and causing lots of bleeding complications.
4:18: Results: Yes 4, No 6, 1 Abstain!
- Yes: Follman, Krantz, Mascette, Sager
- No: Coukell, Emerson, Kaul, Madden, Nissen, Wolfe
- Abstain: Fleming
4:16: Moving on now to first and most important voting question: should rivaroxaban be approved for ACS?
4:07: Now discussing bleeding rates. This is the last discussion question before voting.
4:06: Brief discussion now about all-cause mortality as a claim on the label. Although it was statistically significant, it was not pre specified, and most of the mortality was CV mortality anyway, say Kaul and Emerson.
4:01: Chairman Sager is reassured by the fact that the reduction in CV death is quite robust. Nissen says he would be more reassured if the same effect were seen in the 5 mg group, which undermines his confidence in the finding.
3:56: Nissen says that an alternative to improving ACS therapy with an anticoagulant is to use a better antiplatelet (ticagrelor, prasugrel).
3:53: Dean Follman appears to be more positive to the drug. Notes that its approval for other indications is reassuring.
3:49: Fleming would rather “advise” the FDA and not vote, particularly in this case. He praises aspects of the trial but has concerns about safety. Fleming notes that the most important of the composite endpoints, CV death, goes in the right way. Will he actually vote?
3:45: Kaul thinks the trial does not show a robust mortality benefit. And he also has concerns about the trial conduct raised by Marciniak. He also cites multiple clinical studies which failed to find a benefit for anticoagulant therapy in ACS. There are also safety concerns. So the results are not statistically convincing. Looks like another negative vote.
3:43: Now they’re talking about whether ATLAS is persuasive in demonstrating the beneficial effects of rivaroxaban. Nissen says it’s hard to make a compelling case for net clinical benefit. He really appears to be leaning toward a “no” vote.
3:40: Chairman Sager summarizes the problem of missing data. Says its a big overall problem. Although sensitivity analyses in ATLAS are assuring, there are still remaining questions and better sensitivity analyses would be reassuring.
3:33: Scott Emerson: doesn’t believe that the people with missing data we’re really 20 times more likely to die than the people for whom we have data.
3:28: I learned a new word today: “missingness.” Amazing that I’ve gone so many years without it, and yet today I’ve heard it a hundred times.
3:23: Fleming thinks Nissen’s concerns may be valid, in that the study design subtly encouraged investigators to allow patients to withdraw. Fleming makes a strong statement that this issue needs to be proactively addressed and gives credit to Marciniak for raising this issue.
3:15: Now the committee will formally discuss the conduct of the trial and the issue of missing data. Kaul notes that there’s no rule that stipulates an acceptable level of missing data. “The missingness in this trial is vexing and concerning,” says Kaul. “But I would not go to the extent of saying that this problem invalidates the results.” Kaul says that it “challenges” the data but does not invalidate it. He calls the data “fragile.”
3:12: Nissen says that the mITT “poisoned” the trial by encouraging investigators to give permission to study subjects to withdraw.
3:00: Now they’re talking about the primary analysis, the intention-to-treat analysis, and the modified intention-to-treat (mITT) analysis. FDA’s Bob Temple explains that the FDA accepts a modified ITT in order to gain a better sense of the true effect of the drug. Kaul notes that there’s not a lot of difference here between the ITT and mITT issue.
2:56: Now they’re discussing the endpoints in the trial, and the use of secondary endpoints which are variants of the primary endpoint. Again, this is important for the FDA in producing guidance for the future but will not play a role in the committee’s decision today.
5 Minute Break
2:41: Long discussion about statistics and different dose regimens, and how to design and interpret trials. Interesting stuff but will make absolutely no difference today.
2:28: Now they’re talking about testing the individual doses in the trial and whether the analysis of each dose separately is acceptable. Again, I doubt this will be an issue at this point.
2:23: Now they’re talking about whether the FDA should look at stratum 2 (DAPT) alone. It doesn’t look like there will be serious objections to this.
2:20: Moving on to FDA questions. Here’s the first part of the first question:
1.Please comment on the design of ATLAS ACS.
- a. ATLAS ACS enrolled subjects on aspirin alone (stratum 1) and aspirin plus thienopyridine (stratum 2). For the US, FDA asserted that results in stratum 2 were primary, because the overwhelming majority of US patients would be on a thienopyridine. In the rest of the world, it was both strata together.
1. FDA considers having region-specific differences in an analysis plan not to be a multiplicity issue. Do you agree?
The committee had no problem with this issue.
2:09: Kaul says he’s not satisfied by the explanations for the differences in outcome between the 2.5 mg and 5 mg rivaroxaban doses. There’s no “coherence” in the data, he says. The company concedes that it has no clear explanation. Excess bleeding– which you would expect for a higher dose of an anticoagulant– only explains about one-third of the difference between the doses, says Kaul.
2:00 PM: They’re getting closer to responding to the FDA questions. You can read the questions here.
1:57: Committee questions to the FDA are starting to wind down. There will never be a definitive answer about the missing data and the rest of the questions are not nearly as contentious or interesting.
1:50: Nissen raises concerns that people who were withdrawn from the trial did not have physical visits but may have been contacted by phone, thereby encouraging them not to stay involved in the trial. Marciniak says that he can’t tell how the patients were contacted. The company is immediately unable to clarify this. This may have been “a flaw in the design,” says Nissen.
1:43: Braunwald argues that the patients who withdrew were equally distributed among the 3 arms of the trial. He shows a Kaplan-Meier curve showing no imbalances in withdrawals between the 3 arms. He also shows that there were no more bleeding events before withdrawing. But the remaining concern, committee member Scott Emerson says, is about the data we don’t have, rather than the data that we do have.
1:33: Kaul asks Hicks to reconcile her recommendation for approval with her decision not to recommend a mortality benefit, since most of the benefit occurred in CV mortality and CV mortality was by far the largest component of total mortality. Hicks acknowledges the contradiction and refers to the prespecified agreement regarding the composite endpoint.
1:18: Lots of talk about the modified intention-to-treat (mITT) analysis. Is anyone else suspicious that this is part of the covert Republican agenda?
1:11: Forgot to mention: the FDA announced after lunch that the terms for Kaul and Sid Wolfe were expiring. But I’m guessing this won’t be the last we’ll see of either of them.
1:06: Nissen seeks clarification on the rate of withdrawal in TRITON-TIMI prasugrel trial. It also had a high rate, but lower than ATLAS, in part because it was a shorter trial.
1:04: Meeting Resumes. No open public hearing! Hooray! Committee discussion will now resume.
1:02 PM: One important point of clarification before we resume. I received a comment from a reader who thought I was predicting that the committee was sure to vote against approving rivaroxaban for ACS. So perhaps I haven’t been too clear. I try to avoid predictions, but in this case I’ll make an exception. I think it’s highly likely that the committee will recommend approval. I don’t have any sense yet that the committee members agree with Marciniak that the ATLAS results are invalid. Many of them are surely troubled by the issues he’s raised, but in the end I think they will adopt the principle that large, pragmatic trials, though necessarily imperfect, should guide decision-making unless proven wrong. They are innocent unless proven guilty, and Marciniak has not demonstrated beyond a reasonable doubt that ATLAS is guilty.
Also, we should remember last year’s ROCKET meeting. All the FDA reviewers in that case recommended turning down the NDA for rivaroxaban for stroke prevention in AF. In the end, the committee approved the NDA.
LUNCH BREAK– BACK AT 1 PM
12:06: Marciniak clarifies that he think the trial results are invalid even though he doesn’t believe the blind was broken. His point is that people who left the trial were more likely to bleed and thus more likely to be on rivaroxaban. In addition, Marciniak says there should be 100% assessment of vital status, and that if countries don’t allow you to assess vital status they should not be included in clinical trials.
11:59: Kaul asks Marciniak if his views “challenge” or “invalidate” ATLAS. Marciniak: “invalidate.”
11:48: Committee now questions the FDA reviewers. Sidney Wolfe asks Marciniak for his view of the company’s response to his questions about the missing data. Marciniak says he is not satisfied and that as far as he is concerned this constitutes the major question about this trial.
11:45: Straying from Hicks’s talk, I just wanted to point out that the problem with withdrawals and missing data highlighted by Marciniak is a direct product of the “outsourcing” of clinical trials. This is not an isolated incident, as Marciniak said. These sort of problems will not be going away anytime soon, and we’re going to be dealing with these issues for many years.
11:39: No surprise, Hicks shows that bleeding risk with rivaroxaban is greatest in women, the elderly, and lower body weight. She notes that “there are a number of ways FDA communicates bleeding risk; one of these is a boxed warning.”
11:28: The FDA clinical reviewer, Karen Hicks, is now up. The statistician didn’t address any of Marciniak’s issues and it doesn’t look like Hicks will either. She’s basically addressing the question of who should be eligible for the drug if rivaroxaban is approved.
11:15: FDA statistical reviewer now up. The drama level just went WAY down.
11:13: Marciniak conclusion: “ATLAS failed, not shrugged. The missing data and quality problems preclude ATLAS from providing substantial evidence of effectiveness.”
11:09: Marciniak criticizes trial because the clinical events committee (CEC) was not allowed to question the sites about the data, but had to go through J&J to get information about cases.
11:03: Marciniak assaults integrity of trial because of the 3 missing deaths that were not included with the trial data from the sponsor. He says he’s been accused of being paranoid but says that it’s his job to be paranoid.
11:00: Marciniak: because more bleeding w/ rivaroxaban, it’s likely that missing data would show more events with rivaroxaban.
10:58: Marciniak notes that event rates are much higher in the groups with missing followup, especially since people with major bleeding complications are more likely to withdraw from the trial.
10:53: Marciniak points out that the missing followup is greatly in excess of the difference in the major events in the trial, and that the missing info on vital status is much greater than the difference in mortality in the trial.
10:52: Marciniak talks about the growing problem of missing vital status in clinical trials. “There is absolutely no relationship between what you read in the medical journals” and what the FDA finds via site audits.
10:48: Marciniak starts, says that he hates talking about “missing data and questionable data.” Wants to know if this is an issue where “Atlas shrugged” or if the weight of the world is too great today? I have a feeling he’s going with the weight of the world.
10:31: Break. At 10:45 the FDA will make its presentation. The FDA Medical Team Leader Thomas Marciniak raised a lot of troubling questions in the review documents. He’s first up.
10:24: Why do the J&J people pronounce Sanjay Kaul’s name as if he were Simon Cowell?
10:22: Sanjay Kaul indicates he shares Fleming’s concerns. I think we’ll be hearing a lot more about this issue in the afternoon.
10:18: Statistician Thomas Fleming launches the first major assault on J&J’s presentation, pointing out that the marginal statistical significant benefit of rivaroxaban could easily be lost with only a few additional events going the wrong way in the missing patients. This appears likely to be the real focus of the day’s discussion.
10:14: More questions about withdrawals from the study. I didn’t know until today that in many places it is not allowed to determine the vital status of patients who withdraw from a study.
10:10: Gibson notes that there were no excess fatal ICHs in ATLAS, unlike in the prasugrel or ticagrelor studies.
10:08: Braunwald notes that this wasn’t tested in the trial but should be tested in the future. He points out the economic issue, noting that clopidogrel is now generic. Using rivaroxaban with either prasugrel or ticagrelor would be an expensive proposition with 2 branded drugs.
10:03: Sidney Wolfe, following on Nissen’s question, asks whether it would have been better to study rivaroxaban with a more potent antiplatelet like prasugrel or ticagrelor (though he stumbles several times pronouncing ticagrelor). Asks: is it possible to gain more clinical benefit by using a more potent antiplatelet inhibitor?
9:56: Very complicated statistical discussion going on now. Here’s one stat Braunwald presented: For every 10,000 patients treated over 2 years with rivaroxaban, there will be 25 additional ICHs, 14 nonfatal ischemic strokes, 0 fatal ICHs, 23 fewer nonfatal MIs, and 171 fewer CV deaths.
9:42: Probably worth pointing out that Nissen’s first question is based on an assumption that rivaroxaban will be approved. Good omen for J&J. Braunwald responds that it would be worth testing rivaroxaban with other anti platelets.
9;40: J&J presentation complete. Committee now asking clarifying question. Guess who’s up first? Nissen. He asks whether rivaroxaban should be used with prasugrel or other anti platelets. Gibson responds that there is no data for prasugrel or ticagrelor.
9:38: Braunwald: “We want out patients to come out the front door but we don’t want to cause ICH (intracerebral hemorrhage).”
9:35: But Braunwald does make an important point, that the concept of anticoagulant therapy has been floating around for a long time and never fully died. He notes that the APPRAISE trial with apixaban failed. In retrospect, they used a dose that was too high and included stroke patients. From phase 2 trials with rivaroxaban they learned to avoid higher doses of rivaroxaban.
9:33: Braunwald’s talk seems more a nod to a towering figure than a vital part of the proceedings. He’s now talking about ACS therapy in the 1950s. Anyone out there remember when Topol and Braunwald presented TPA for Genentech? And lost. Now that was drama.
9:30: Eugene Braunwald now presenting. Click here for a classic photo of this historic event.
9:25: Burton shows that patients who withdrew from study were not very different from the subjects who remained in the study. Very unlikely that the results would have been different if they had remained in the study.
9:15: Burton addresses the interesting problem of 3 study patients who died after withdrawing consent. In some countries it is not allowed to count these deaths. Very strange.
9:08: Burton now addressing some of the issues raises by the FDA.
8:57: J&J’s Paul Burton now presenting the safety data.
8:53: Gibson dealing with the observation that most of the benefit with riva is due to reduction in CV death but not MI. He contends that the reduction in MIs shows up as a reduction in deaths (from MIs). Reasonable argument but not sure how it moves the dial.
8:43: Gibson makes point that the ATLAS results are consistent across different sorts of statistical analyses. Ultimately, this point may help the committee reach a positive decision.
8:34: Gibson now addressing issue of missing data, which was a red flag raised by the FDA review. He’s trying to make the point that although there’s missing data from 8.9% of patients in the study, most of these patients were followed for most of the study, and that the withdrawals in the study were equally distributed across the study arms.
8:30: C. Michael Gibson now presenting the results of ATLAS ACS-TIMI 51. 93% of ATLAS patients received DAPT (dual antiplatelet therapy), the current standard of therapy.
8:18: Jessica Mega now introducing rivaroxaban and the ATLAS ACS TIMI development program.
8:13: J&J now starting its presentation. Jessica Mega, Mike Gibson, and the godfather himself, Eugene Braunwald, will be presenting for J&J.
8:10: Meeting now underway. After introductions and formalities, Norman Stockbridge is now setting out the agenda. He asks if “one study without an extreme p value suffices in this setting.” (He’s talking about ATLAS ACS-TIMI 51, of course)