TMI from TIMI: Xarelto and the Trouble with Very Large Trials

With bigger clinical trials come potentially bigger problems. That, at least, is what sponsors of 15,000-plus trials in cardiovascular medicine are finding out. J&J/Bayer’s experience in late May with Xarelto for acute coronary syndrome is a case in point—and it also provides timely context for the surprising setback for another CV project, Pfizer/BMS' Eliquis.

There were two direct, explicit messages to NDA sponsors from the May 23 meeting of the Food & Drug Administration’s Cardio-Renal Drugs Advisory Committee review of Johnson & Johnson/Bayer AG’s Xarelto for a new indication.

The first came from the dean of FDA’s review process, Center for Drug Evaluation & Research Deputy Director Robert Temple.  His worry:  that as Phase III trials get larger and more geographically diverse, FDA is seeing a trend towards patients lost to adequate follow-up.  His advice: watch out for the warning sign of a large number of patients classified as “withdrew consent without further explanation.”

“We are obviously always troubled by missing data,” Temple declared, “and we would like there to be as little of it as possible.”

FDA has “been particularly concerned about what seems to be a new trend which says ‘withdrew consent without further explanation.’ We have a number of documents in which we have urged people not to do that, but to try to find out” the reason for withdrawal.

Temple suggested investigating whether withdrawal of consent is “because you are bleeding, because you are mad, because you can’t afford to come to the clinic anymore. We try to get those things. All too often you get ‘withdrew consent’ without much further explanation. We definitely want more stuff; but I think often, when they [clinical trial patients] go away, they don’t want to be questioned anymore. So, it is a little hard to tell why they withdrew.”

Nissen Urges Large CV Sponsors to Avoid mITT

The second related message came at the end of the meeting from the frequent Cardio-Renal Committee member and active outside commentator on the clinical trials process, Cleveland Clinic Cardiovascular Medicine Chairman Steven Nissen.  His worry: the adoption of too short a follow-up period for the key part of the trial population flaws a trial from the start.  His solution: make sure that investigators know from the start that long-term follow-up is crucial and avoid a foreshortened modified intent-to-treat (mITT) population.

“I want the sponsor [of the pending application, Janssen Research & Development] and others in the audience who may be considering such trials to hear this,” Nissen pronounced.  “The decision to use mITT as the endpoint, I believe, had a profound impact. I think what happens when you say that the primary endpoint is 30 days after stopping study drug you are telling the investigators and the patients that you don’t care so much about what happens later on.”

The large withdrawal of consent rate in the trial under consideration “was pre-ordained by the mITT which is really an on-treatment analysis,” Nissen said. That decision “colored the trial in ways that we can never recover from.”

The most compelling message, however, from the May 23 meeting may have been an implicit and painful warning to all sponsors of very large trials: as impressive and ambitious as some worldwide registration efforts have become, they are dangerous just from the amount of material they generate. The massive amounts of data lead to tiny p values, but also to mountains of case reports—and a corresponding invitation to individual regulatory/drug approval staff members to undertake forensic analyses of the conduct and documentation of the trials.

The trial at the center of the concerns that day was a 15,526-subject, Phase 3 trial conducted for J&J and Bayer – the ATLAS 2 ACS TIMI 51 trial. The objective was to add a significant new indication (acute coronary syndrome) forXarelto (rivaroxaban), the Factor Xa anti-clotting agent that had already been approved for two indications.

Fastidious FDA Reviewer Checked Original Status Report Forms – Found Sloppy Errors

The attack on the ATLAS TIMI trial for the ACS  indication was led by a determined and fastidious FDA reviewer in the Cardio-Renal division, Thomas Marciniak. He is the same reviewer who handled another mega cardiovascular trial (18,421) for AstraZeneca PLC’s Brilinta (ticagrelor) platelet aggregation inhibitor as well as one of the reviews of GlaxoSmithKline Inc.’s RECORD study for information about cardiovascular risks associated with Avandia (rosiglitazone). 

In the case of the Xarelto study, Marciniak dug into the backup material from the Clinical Status Review forms used by the sponsor for contact during the trial with the trial centers (766 study centers in 44 countries). That data source was extensive: J&J described 134,023 pages of CSR reports with five questions on each page relating to the 15,526-patient population. 

Marciniak found 61 patients “counted as living [that] had at least one CSR [form] reporting death.” The company said that following a request by FDA to check back into those status reports  the firm found that “all death events recorded in the CSR pages were found to have been checked ‘yes’ in error by the sites and notes to file were collected.”

The FDA reviewer also found three deaths among patients marked as withdrawing consent. There were 1,294 subjects or 8.3% of the study population who fell into the withdrew consent category.  The relatively small numbers of missing patients and inconsistencies on deaths and causes of death are  important because they can easily erase or undermine effect sizes from the trials.

As Temple pointed out at the meeting, “it is worth knowing even if the missing data is only 4% or 5%, [because that] it is always bigger than the effect size. So if you are worrying about that, then you have to worry whether it is 8% or 4% —even though 8% might be a little worse than others.”

The process of careful digging and questioning described by Marciniak changed the tone and nature of the debate about the supplemental indication before the advisory committee. Marciniak and J&J sparred over the purpose of the CSR forms: an important first-warning signal from the field (Marciniak) or solely as a “worksheet reminder” to site staff (J&J).

Whatever the correct interpretation of the CSRs,  the open debate about their role in front of the advisory committee detracted from the analysis of the reported results and led to a number of committee members commenting about concerns regarding the “missingness” of data.

Harsh Critique of Data Collection Created Harsh Climate for Xarelto

Marciniak questioned the conduct of the trial and the accuracy of information reported to the trial group, the venerable TIMI Study Group affiliated with Brigham & Women’s and Harvard. Referring to similar problems with other big trials run by academic/research groups for sponsors, the FDA reviewer raised questions about the validity and accuracy of the data supplied to the study groups for interpretation.

“Looking at recent trials,” Marciniak said, “I have not seen any significant problem with the adjudication done by TIMI, with the adjudication done by Duke or anybody else based on what they are given….I just have a very poor idea what really was out there rather than what they were given and who they were given it on. “

Those questions did not provide a good foundation for the advisory committee review of the ACS indication;  it was rejected by a split vote of 6-4-1. The no vote was follwed by a “complete response” letter in June. (Also see "Lone Wolfe: Drug Industry Critic Ends Tenure As Advisory Committee Member With Rare Decisive Vote" - Pink Sheet, 25 Jun, 2012.)

Still, there were indications at the meeting that senior review  staff at FDA did not feel as strongly set against the application as Marciniak or the committee majority.

Temple, for example, noted at one point that the discussion was tending to make the drug appear to have little effect on ACS, an interpretation that he seemed less ready to adopt at this point.

“What I was hearing” from the committee, Temple commented, “was that the effect on mortality was sort of trivial. That doesn’t seem quite true.” Temple who was head of the review office in charge of the Cardio-Renal Division for decades before recently assuming a new role as Center for Drug Evaluation & Research deputy director, noted there is “a good question” about whether the effect on mortality “was real.” He pointed out “if it is real, it is the sort of thing that you want to do.”

Other committee members echoed Temple’s suggestion that adding an anticoagulant to ACS treatment could be a significant change in therapy. One member called it a potentially “major leap forward,” another said it would be a “really big deal.” But the size of the study and the openings that the study presented for an aggressive review into procedures for conduct and reporting got in the way of a discussion of the potential advance in treatment.

Temple’s attitude throughout the May 23 meeting indicated an effort by FDA senior review management to make sure that the concerns raised by Marciniak would be heard in public and vetted by the advisory committee. This did not appear to be a prejudgment of the issue by the agency officials who eventually made the final call on the application.

At one point, Temple seemed to try to coax Marciniak back from his sharp criticisms of J&J’s handling of the trial.  He gently lauded Marciniak for raising the broader issue of informative censoring in his examination of patients who were lost for follow-up.

“I think what Tom is addressing,” Temple said, “is what we more broadly would refer to as informative censoring: that is, the people who are leaving are not similar to the people who are staying in. They are more likely to have a cardiovascular event.”

Following that praise, Temple gave Marciniak a chance to soften his assessment of the importance of the lost patients by noting that J&J had tried to track down vital status on many of the lost patients.

“Tom, one of the things that I had asked before, that you might want to comment on,” Temple said, “is that they [J&J] did have follow-up on 400 out of the 1,200 [consent withdrawn patients] and at least in that subset of people who left the trial, they did not look particularly worse. Does that affect your views at all?”

Marciniak was unappeased and not in a frame of mind to go softly on the importance of the lost patients : “I don’t think the follow-up is reflected in any of the statistics that you have seen. I have no idea what was reported as follow-up.” Temple repeated: “They had vital status on about 400 of the 1,200 patients.”  Marciniak: “I have a lot of credibility problems with this particular submission, where I am told one thing one day and another thing another day. “

Temple made one last try to argue more for the importance of determining some level of acceptable effort to collect vital status rather than the fact of drop outing of the study. “If you thought they had good vital status data on a third of the people who were dropped out, would that make any difference to you?,” Temple asked.

Marciniak responded: “No. I think for vital status there should be a very, very simple requirement: that is it should be 99.9% complete. You have the mechanisms to do it.“ He urged FDA and sponsors to take a hard line about doing studies in countries that will not permit follow-up with withdrawn patients.

“If you have a country that says you cannot follow my patients,” Marciniak said, “then you have to tell that country that I cannot do these outcomes in that country – the eight countries that were mentioned.”  Temple said: “I am inclined to agree with that but it is true at the moment that some will not let you look at the data.”

An Advisory Committee Frequent Flyer

The size and logistical complexity of the trials make them a very high-risk regulatory endeavor -- a point driven home by a June 21 Complete Response letter from FDA. J&J did not detail FDA’s concerns contained in the later but said they would continue to evaluate the issues and work with the agency to resolve them.

Exhibit 1

Rivaroxaban Clinical Development: Completed Phase 3 Trials

Indication	Trial Name	Randomized Subjects
Venous Thromboembolism: Prophylaxis after hip or knee replacement surgery (FDA approval July 1, 2011)	RECORD	12,729
Stroke and embolism prevention in atrial fibrillation (FDA approval Nov. 4, 2011)	ROCKET AF J-ROCKET	15,544
Venous Thromboembolism: hospitalized medically ill patients	MAGELLAN	8,101
Venous Thromboembolism: treatment and secondary prevention	EINSTEIN	9,415
Secondary prevention after Acute Coronary Syndrome	ATLAS ACS 2 TIMI 51	15,526
Total (through December 31, 2011)		61,045

Source: Janssen Research & Development

The two existing indications for Xarelto also previously went through advisory committee reviews: DVT prevention in surgery through the Oncologic Drugs Advisory Committee (as part of ODAC’s hematology portfolio) in March 2009 and stroke prevention in atrial fibrillation through the Cardio-Renal Advisory Committee in  September 2011.  Xarelto, in fact, may have the unenviable record of having to go through the most  number of advisory committees for indication reviews in recent FDA history.  

J&J and Bayer – J&J handles the regulatory process in the U.S. – have had a checkered experience with Xarelto at the advisory committees and with subsequent FDA reviews. ODAC voted strongly in favor of the first indication for surgical prevention (15-2) but FDA issued a complete response decision that delayed approval of the indication until July 2011. 

The sponsors took a beating at a tough advisory committee on stroke prevention in atrial fib in early September 2011. At that session, another experienced FDA reviewer, Marty Rose, raised issues about how warfarin was used as a comparator in the large Rocket trials (again with more than 15,000 patients).

The FDA reviewer again dug deep into an extensive database to argue that weak instructions and control at some non-US study sites led to less than optimal use of warfarin, negating it as a valid comparator. One difference to the ACS review, Rose did not openly question the accuracy of the data generated; he focused instead on whether it was meaningful given the less than optimal use of warfarin in some parts of the trial.

The Cardio-Renal Committee discounted those questions on warfarin as an effective comparator and voted 9-2 in favor of approval and the drug was approved on time in early November.

The impact of an open, negative discussion like the one around the afib indication is hard to quantify.

One marker on the impact appears in usage figures for afib cited by J&J in its ACS briefing book. Those numbers suggest that roughly 45,000 patients received Xarelto for afib in the first two months of US approval, November through December 2011. The company reported  92,132 “patient months” of treatment in that two-month period. 

Do Large Trials Lead to Large Commercial Use?

The following data are from J&J’s account of postmarketing experience with Xarelto (rivaroxaban) in the first two indications. The data (through Dec. 31, 2011) was included by J&J in the briefing documents for FDA for the Acute Coronary Syndrome application.

Surgical VTE Prevention (FDA Approval July 1, 2011)

• Worldwide postmarketing exposure to XARELTO on the 10 mg tablet for prevention of VTE following elective hip or knee replacement surgery since the first approval of rivaroxaban in Canada on 15 September 2008 until a cutoff date of 31 December 2011 was estimated at 1,553,600 patients (excluding clinical and observational studies).
• In addition, data were collected from over 18,000 patients (with over 9,000 subjects exposed to rivaroxaban) in a Phase 4 postmarketing observational study (XAMOS study 13802).

Stroke Prevention in Atrial Fibrillation (FDA Approval Nov. 4, 2011)

• In the indication of stroke and systemic embolism prevention in patients with nonvalvular atrial fibrillation, based on one 20 mg or 15 mg tablet of XARELTO per patient per day, the cumulative exposure to XARELTO since the start of marketing approval through 31 December 2011 was estimated to be 92,132 person-months or 7,678 person-years in the United States.

Size Does Not Improve Regulatory Predictability

From the continuing Xarelto tale, it has been difficult to read through from one event to the next how the different steps of the FDA review process plays out:  the tone of the initial FDA review (detailed in pre-committee briefing books) has not been a solid predictor; neither has the treatment of the product at the advisory committee been a clear predictor of when and how FDA will finally act. Each step has been difficult to predict with the preceding experiences offering little guidance on the outcome of the next step in the approval process or the next indication. 

Overall, six Xarelto randomized clinical trials have enrolled over 61,000 patients. While the weight and scope of the trials are clearly impressive to regulators, the amount of material presents an almost irresistible database for close examination by some FDA reviewers.

It is noteworthy that in both tough internal reviews on Xarelto, the skeptical reviewers had partners in the review who were less critical. That acts as a balancing factor but did not keep the product from experiencing the complications of a messy public advisory committee.

If J&J and Bayer felt that the pure weight of evidence in the Xarelto NDAs would assure a smooth road to approval, the results have been less than the size suggested.Nor is it just Xarelto: Several other high profile cardiovascular drugs have had unexpectedly rocky paths through the agency despite massive pivotal trials seeming to demonstrate unequivocal superiority to a standard comparator. The list includes Eli Lilly & Co./Daiichi Sankyo Inc.’s  Effient, AZ's Brilinta and, as of June 25, Pfizer Inc./Bristol-Myers Squibb Co.'s Eliquis.

Eliquis The Latest Case-in-Point

In the latest twist regarding the tortured regulatory pathway for Eliquis, which many drug industry stakeholders viewed as slam-dunk approval given published clinical trial results, FDA issued a complete response letter (CRL) on June 25 asking the sponsors for more information.

Eliquis went from no FDA advisory committee meeting, to a tentative advisory committee meeting coupled with a three-month extension to its priority review user fee deadline, to no advisory committee again. The avoidance of the advisory committee stage seemed to bode well for the drug’s approval prospects, appearing to signal that FDA and BMS/Pfizer had worked out any outstanding issues keeping the drug from reaching the market.

That reasoning turned out to be wrong.

In a press release announcing the CRL, the sponsors said: “The CRL requests additional information on data management and verification from the ARISTOTLE trial. Bristol-Myers Squibb and Pfizer will work closely with the FDA on the appropriate next steps for the Eliquis application. The FDA has not requested that the companies complete any new studies. FDA and the companies are committed to working expeditiously to address the outstanding questions and move the application forward.”

While most drug developers would breathe a sigh of relief that “FDA has not requested that the companies complete any new studies,” the request for “additional information on data management and verification” is more ominous given the recent history of the Cardio-Renal division.

By no means does the delay in the Eliquis review mean that FDA has the same concerns as it did with Xarelto in ACS. But it proves that data management and verification issues are no laughing matter.

During Pfizer’s recent earnings call, Pfizer’s President for Specialty Care and Oncology Geno Germano commented on the delay.

“I think one point which is really important for me is, that it’s our belief that the information requested doesn’t impact the outcome measures of the trial and importantly as [CEO Ian Read] has already said, we haven’t been requested to conduct any additional studies. So our confidence and the results from both the ARISTOTLE trial and the AVERROES trial hasn’t changed.”

BMS Chief Scientific Officer Elliott Sigal added during the company’s earnings call: “I will emphasize [that] the questions raised do not relate to the primary or key secondary endpoints and do not contest the validity of our database. Importantly, the complete response letter does not request that we conduct any additional studies.”

Germano’s and Sigal’s assessments of the situation should allay major concerns over the approvability of Eliquis. And the Xarelto experience demonstrates the perils of going to a committee meeting for a sponsor, something Pfizer and BMS have not had to do—yet.

So, Bristol and Pfizer can take comfort from the fact that they have at least found a way to deflect the downside of too much information from big studies reviewed in the harsh atmosphere of a confrontational advisory committee. At least they have so far.

Despite credible reports on a tentative scheduled date for an advisory committee review in conjunction with the May 23 Xarelto review, the Eliquis sponsors were able to avoid it. They did not, however, escape a complete response letter that, according to Pfizer, focuses on

A delay to resolve "data management and validation" issues isn’t a great outcome. But it is far preferable to  facing irritable reviewers at a public advisory committee. Though, of course, this delay could still lead to an advisory committee vetting when the application is resubmitted.

The advice from Temple (pay attention to withdrawal of consent and collecting vital status on withdrawn patients) and Nissen (don’t cut the follow-up period short) seems like useful advice from the trials of Xarelto.

But the tougher business issue is whether the larger trials raise more regulatory risks than they resolve.