Validity of Pivotal ROCKET AF Rivaroxaban Trial Questioned

Deborah Brauser

February 05, 2016

LONDON, UK — The key ROCKET AF trial used a faulty point-of-care device to measure INR in its comparator arm of patients taking warfarin, which may have made warfarin results look worse than they otherwise would seem, alleges a new report released this week in the BMJ[1].

The findings from ROCKET AF, which were first published in the New England Journal of Medicine in August 2011, showed that the novel oral anticoagulant (NOAC) rivaroxaban (Xarelto, Bayer/Johnson & Johnson) was noninferior to dose-adjusted warfarin for the prevention of stroke or major embolism in more than 14,000 patients with atrial fibrillation (AF). Three months later, the FDA approved rivaroxaban for the treatment of AF based solely on these results.

But as described in the BMJ, the warfarin arm used an INR-measuring device that was included in a December 2014 recall because of the possibility of delivering "clinically significantly lower" results than those delivered from a plasma-based laboratory reading.

"A falsely low reading could mean that patients had their warfarin dose unnecessarily increased, leading to a greater risk of bleeding," writes BMJ associate editor Deborah Cohen. "It could [also] make rivaroxaban seem better than it was at reducing the risk of bleeding and throws doubt onto outcomes used to support the use" of rivaroxaban.

However, in a research letter in the NEJM[2], also published this week, ROCKET AF lead investigator Dr Manesh R Patel (Duke Clinical Research Institute [DCRI], Durham, NC) and other executive committee members report that, after conducting a series of post hoc analyses, new results "are consistent with the overall trial findings" and any possible malfunctions in the device "did not have any significant clinical effect on the primary efficacy and safety outcomes in the trial."

"We recognize that there are people who may have concerns with these findings. We hope that the diligent, thorough analysis we've done will answer a lot of those concerns," Patel told heartwire from Medscape.

In the BMJ report, Dr Harlan Krumholz (Yale University, New Haven, CT) noted that the NEJM should place "an immediate expression of concern" on the original ROCKET AF paper—and that the study "should be considered of uncertain validity until a more thorough review can be done."

Do the new data released by Patel et al alleviate his worries? Krumholz told heartwire "no." Although he noted that the DCRI provided a "nice analysis to prove their trial valid," he wished they had done further analyses; he stressed that all data should be made public. "Why not share the data and let others take a look?" he asked.

On the other hand, Dr Lars Wallentin (Uppsala Clinical Research Center, Sweden) told heartwire by email that although the time in therapeutic range (TTR) was lower than targeted in the warfarin arm of ROCKET AF, the standards of warfarin care still seemed "rather similar" to real-life situations.

"Based on the totality of available knowledge, there are no compelling reasons to change the current treatment recommendations concerning any of the new oral anticoagulants for stroke prevention in atrial fibrillation."

Class I Recall

The FDA issued a class I recall notice in December 2014 for the Alere INRatio Monitor System and announced that the manufacturer had received 924 reports of device malfunctions. But Patel said the ROCKET AF executive committee did not receive notice that the device used in their warfarin arm was part of this recall until the fall of 2015.

Cohen writes that the BMJ did not receive any response from the DCRI, but the drug manufacturers told the journal they were taking the concerns seriously. After she contacted the European Medicines Agency (EMA), the EMA responded in November 2015 that it was investigating and that "the INR device may have impacted the clotting results" in some of the warfarin-treated patients.

Today, the EMA issued a statement concluding that the issue with the INR monitoring device in ROCKET-AF "does not change its conclusions on the overall safety or benefit-risk balance" of rivaroxaban in the trial[3]. "This means that Xarelto can continue to be used as before, in line with the current prescribing information."

In addition, the FDA told the BMJ that it was "reviewing relevant data" in regard to the defective device's possible effect on ROCKET AF's results.

Patel said the investigators began their new analyses in October and finished in December. A short summary statement was released by Bayer/Johnson & Johnson soon after, reporting that the results confirmed the ROCKET AF findings and the positive benefit/risk profile of rivaroxaban in the treatment of nonvalvular AF.

At that time, there was a clamor for more details to be released immediately instead of waiting for publication. But Patel notes that there were concerns about making a statement without having all of the data ready.

"We spent a lot of time with two independent statisticians, a group of individuals reviewing the data in a blinded fashion, doing the reanalysis, and then we wanted it to go through peer review," he said. "We tried to act as fast as possible, with substantial analysis now in press 4 months after we first received notice."

He added that it's important to note that the reanalysis by the ROCKET executive committee is independent from the one done by the sponsors, which was provided to both US and European regulatory agencies.

Reexamining Subgroups

In their research letter, the investigators point out that the FDA's recall said the device INR readings may be lower in patients "with certain specific medical conditions," including abnormal hematocrit levels, bleeding or unusual bruising, or conditions linked to higher fibrinogen levels.

So for their new analyses, they identified study participants with any of the listed "recall conditions" and then compared outcomes in the treatment subgroups. A total of 37% of the patient population had at least one of the recall conditions, which were distributed almost equally between those receiving each treatment.

When they looked at the entire patient population (primary results), those with no recall condition, and those with any recall condition, there were no differences in rates of stroke or embolism.

The subgroup with no recall condition also had safety end points consistent with the primary results, with similar rates of overall bleeding and lower rates of fatal and intracranial bleeding among patients treated with rivaroxaban but a higher rate of gastrointestinal bleeding (hazard ratio [HR] 1.47, 95% CI 1.04–2.06).

Bleeding rates were higher in both treatment arms for those with any recall condition, "with a trend toward a higher relative risk of major bleeding with rivaroxaban than with warfarin" (5.53 vs 4.79 events per 100 person-years, respectively)—something that wasn't found in those without any recall condition (2.34 vs 2.68 events, respectively).

"This result runs counter to the concern that a potentially malfunctioning INR device in patients with relevant conditions would have led to incorrect low reading and inappropriate increase in warfarin dose, resulting in a higher warfarin bleeding risk in these patients," said Patel.

"Given the totality of the findings, we concluded that we do not find a statistically significant interaction between the device in patients who may have had a condition and the outcomes of those patients," he summarized. "Therefore, we think the original trial conclusions are still held."

Still, Patel added that the investigators are doing further analyses "to make sure no other concerns remain." A statement sent to heartwire from Duke added that the trial's executive committee "will continue to discuss options for additional analysis of the data, including review and replication by other groups."

Debate Continues

"Although I congratulate Duke for starting in this direction, I don't think they settled the questions. They took one approach to try to test whether this defective device disadvantaged the warfarin arm; but I think there are many other ways you may want to look at this," said Krumholz.

"This is just a start. And I was a little surprised that they felt that this confirmed the results."

He added that he's skeptical as to whether the device was mismeasuring only certain groups of patients and "would like to see the data on that." Also, "I'd like for someone to really explain why this device is defective. What is the mechanism?"

In addition, the BMJ report quotes former FDA clinical pharmacologist Bob Powell as saying "a comparison should be made between the defective point-of-care readings and the two sets of 'gold-standard' central lab readings" that the investigators conducted at 12 and 24 weeks.

"This is information that the investigators have," said Krumholz. "I just don't know why they didn't share it. My hope is that the data show that there's no difference [in the two types of readings] and then we're all set. I would like to not be uncertain about this drug."

Also in the BMJ, Krumholz noted that further investigation into the trial is needed "by an independent group of experts to quickly determine whether there are grounds for retraction."

Even with the new release of data, he still thinks an independent investigation is needed. "These questions cannot be easily answered by a single analysis. And there should be a public dialogue as to whether the findings are truly valid."

And he reiterated his call for the release of more data, noting that Johnson & Johnson told him it would allow access to all trial information. However, Bayer refused. The BMJ reports that a Bayer spokesperson said the company shares data only on studies for new medications approved after January 2014.

Device Transparency Needed?

"Currently, there is little public information about which . . . devices are used in any of the direct oral anticoagulant trials," writes Cohen. "They are not named in the published phase 3 trials." She reports that it was only after examining European regulatory documents that she discovered that ROCKET AF used one of the faulty devices and adds that Alere told her the faults date back to 2002.

Cohen notes that questions that remain unanswered include whether there were any investigator complaints about mismatched readings and whether the device had been validated any time before or during the trial.

As the new debate continues about the rivaroxaban trial, what changes (if any) should clinicians make in the meantime to the way they prescribe the drug?

"Patients shouldn't immediately stop using this drug. There's no imminent danger that's been raised. But I'm not sure what clinicians should do," said Krumholz. "It's hard to say we know for sure it doesn't work as well as the trial said. But we're certainly less certain of the results than we were before we heard this news."

Cohen reports no relevant financial relationships. Patel reports grant support from Johnson & Johnson, grant support and personal fees from Janssen, and personal fees from Bayer during the study. Outside the submitted work, he has received grant support from Heartflow, the National Heart, Lung, and Blood Institute, and Maquet; grant support and personal fees from AstraZeneca, Medtronic, and CSI; and personal fees from Genzyme and Merck. Disclosures for the coauthors are listed on the journal website. Krumholz reports having a contract with Johnson & Johnson to distribute trial data. Wallentin reports receiving research grants from AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, GlaxoSmithKline, MedCo, Merck, Pfizer, and Schering-Plough.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.