Missing Data Lead FDA Panel to Vote Against Rivaroxaban for ACS

Reed Miller

May 23, 2012

May 23, 2012 (Updated May 24, 2012) (Silver Spring, Maryland) The missing data issues plaguing the ATLAS ACS 2 TIMI 51 trial of the factor Xa inhibitor rivaroxaban (Xarelto, Bayer Healthcare/Janssen Pharmaceuticals) have prevented the drug from earning the endorsement of the FDA Cardiovascular and Renal Drugs Advisory Committee.

At its May 23 meeting, the panel voted six to four (with one abstention) against recommending that the FDA approve rivaroxaban for reducing the risk of thrombotic cardiovascular events in patients with acute coronary syndrome or unstable angina in combination with aspirin, aspirin plus clopidogrel, or ticlopidine.

Janssen's application is based on the results of the ATLAS ACS 2 phase 3 and the ATLAS ACS TIMI 46 phase 2 trial. The placebo-controlled ATLAS ACS 2 showed rivaroxaban reduced the risk of both all-cause and cardiovascular mortality while increasing the risk of bleeding and intracranial hemorrhage, but the studies were hindered by early patient withdrawals and missing data.

We Don't Know What We're Missing

Based on the ATLAS ACS 2 results, FDA reviewer Dr Karen Hicks recommended approval of rivaroxaban for the requested indications except all-cause mortality. However, another FDA reviewer, Dr Thomas Marciniak, was adamant that the trial results are not interpretable because about 12% of the patients had incomplete follow-up, far higher than the 1% to 1.5% differences in the end-point rates between rivaroxaban and placebo. A total of 1294 subjects discontinued the trial prematurely, and the company was only able to contact 183, of which 177 were confirmed to be alive.

Because of the patient dropouts, the company adopted a "modified intention-to-treat analysis," whereby patients were observed for 30 days after randomization or the global end date for the trial, instead of observing all the patients until the end of the trial as the FDA originally suggested. Marciniak criticized the sponsor's efforts to follow the patients and said that three patient deaths not counted in the modified intention-to-treat analysis may just be the "tip of the iceberg." Because the percentage of patients whose ultimate vital status remains unknown is much greater than the reported differences in mortality rates, the claimed mortality benefits are not reliable.

The majority of the panel sided with Marciniak. For example, Dr Sanjay Kaul (University of California, Los Angeles) voted "no" because "there was enough uncertainty in the quality and robustness of the data that dissuaded me from voting yes. . . . The 'missingness' of the data doesn't invalidate it, but it certainly makes it hard to infer [the conclusion]."

Dr Steven Nissen (Cleveland Clinic, OH) said that the decision to use the modified intention-to-treat analysis had a "profound impact" on the interpretability of the data. "It's saying we don't care what happens after 30 days, [and] that colored the trial in ways we couldn't recover from." Given the risk of major bleeding, "I want to see better evidence that this strategy of adding an Xa inhibitor or a direct thrombin inhibitor or something else to a good antiplatelet agent is robustly better for the patient," Nissen said. He recommends that the companies run a new trial of the 2.5 twice-daily dose of rivaroxaban using a strict intention-to-treat approach, but, he said, "I don't expect the death benefit to be too robust."

Several panelists said they were concerned that the patients who dropped out of the trial were disproportionately likely to have a bleeding event, which led them to quit the trial, or a "protopathic" event, as statistician Dr Scott Emerson (University of Washington, Seattle) put it. "We're worried that an impending event is what is changing their behavior. We see that all the time in clinical trials--that regularly measured end points do not pick up [all of] the events," he said. He said that since the company was only able to contact 183 of the over 1200 patients who dropped out, it is possible that the dropouts skew the outcomes comparison of the trial.

"Differential event rates after dropout are the number-one thing we're afraid of, so you have to explore it" in a statistical sensitivity analysis of the potential impact of these unknown outcomes. "It would not surprise me if, at the end of the day, these data did not hold up under a proper sensitivity analysis," he said. "What I want to know is, among the people who had events, how differential was the follow-up, but I can tell you by just looking at it, there was a very slightly different amount of follow-up of the people in the treatment arm. But I don't know whether everyone in the treatment arm was cured and they were trekking in the Himalayas and everyone in the placebo arm went home to die. I don't know that that's not the case."

Dr Maury Krantz (University of Colorado, Denver) voted in favor of approval but said he does not know how rivaroxaban would perform in general clinical practice, especially when used with aspirin and clopidogrel. "I felt very much torn by this. This isn't a simple paradigm shift. It means going to triple therapy, which is really a three-headed monster in many ways. I think that what you're going to see in practice, if this is not done carefully with the proper labeling and secondary studies, is really dramatic magnification of bleeding and perhaps minimization of the efficacy benefit."