FDA Used Invokana Review To Reflect On Diabetes CV Risk Policy

Agency officials viewed the review of canagliflozin – the first program started after the 2008 outcomes trial requirement – as a chance to look at lessons learned since the policy went into effect.

As the first drug approved with full implementation of FDA’s cardiovascular risk requirements for diabetes drugs, the review of Johnson & Johnson’s Invokana (canagliflozin) provided an opportunity for FDA to reflect on how diabetes programs that incorporate risk assessment from the outset should be conducted, rather than parsing data from retrofitted or inadequate trial programs.

FDA’s 2008 guidance requiring cardiovascular risk assessment for anti-diabetics was one of the most seismic regulatory events of the past decade, and it left sponsors grappling with how to add major cardiovascular outcomes trials (CVOT) to already-in-progress clinical programs. Aware of these difficulties, the agency was flexible in working with sponsors caught in the transition.

But now the transition is complete, and with years of experience with various approaches, the Invokana review documents show how agency officials are feeling about the implementation of the policy.

“The canagliflozin clinical development program provides us with an example of how the lessons learned since the December 2008 CV guidance was issued are coming into play today,” FDA Division of Metabolism and Endocrinology Products Director Mary Parks stated in her March 29 decision memo.

“The canagliflozin clinical development program provides us with an example of how the lessons learned since the December 2008 CV guidance was issued are coming into play today,” Parks said.

“Initial programs approved shortly after issuance of the guidance were never considered models for subsequent drug development programs to follow,” Parks said. She highlighted Novo Nordisk AS’ Victoza (liraglutide) and AstraZeneca PLC/Bristol-Myers Squibb Co.’s Onglyza (saxagliptin) as “clear examples of programs ‘caught in the middle’ of a regulatory change.”

That regulatory change meant that sponsors needed to rule out two different thresholds of CV risk, first excluding a high rate of risk before approval and then, after approval, testing a lower threshold. FDA set the risk margins at 1.8 (to exclude an 80% increase in risk for CV events) and 1.3 in post-marketing (excluding a 30% increase in risk).

NDAs for both Victoza, a GLP-1 agonist, and Onglyza, a DPP-4 inhibitor, had already been submitted to FDA when the guidance was issued. “An expectation that their entire Phase II and III programs be modified at that point to exclude a risk margin of 1.8 was unreasonable,” Parks said. “Both liraglutide and saxagliptin have post-marketing requirements to conduct a dedicated cardiovascular outcomes trial to exclude a risk margin of 1.3,” she noted.

“Similarly, linagliptin [Eli Lilly & Co./Boehringer Ingelheim GMBH’s Tradjenta], which was submitted shortly after the guidance was issued, was given certain latitude in its pre-marketing CV risk assessment, but is also required to conduct a dedicated CVOT post-marketing,” Parks said.

This article is part of Pharmaceutical Approvals Monthly’s profile of the Invokana review, which includes articles on J&J’s differentiation strategy and the appropriate size of CV outcomes trials. Contact customer care at 800.332.2181 or [email protected] to see if you have access or to sign up for a free trial.

Canagliflozin, however, was in Phase II when the guidance was issued. “As a result, Phase III trials could be designed to incorporate plans to exclude an unacceptable level of CV risk,” Parks explained. J&J proposed a placebo-controlled CVOT, known as CANVAS (DIA3008), which would provide interim data to be combined in a meta-analysis of nine randomized Phase II and III trials. The strategy provided FDA with sufficient data to answer a wide range of safety questions arising from the drug’s novel mechanism of action (Also see "Invokana Development Directed By Mechanism Of Action" - Pink Sheet, 27 Aug, 2013.), and Invokana received approval March 29 after a single review cycle.

The Value Of Clear Guidance

The value of a clearly understood regulatory policy is shown by the striking contrast between the extensive yet smooth review and approval of Invokana and FDA’s continued grappling with the cardiovascular safety data available for Avandia (rosiglitazone), the GlaxoSmithKline PLC drug that helped put CV safety of new anti-diabetics on the regulatory agenda.

An FDA-requested independent readjudication of cardiovascular events in the open-label, 4,400-patient RECORD trial for Avandia was presented to FDA advisory committees in early June. The flawed RECORD trial was seen to outweigh the meta-analysis that identified an association between Avandia and CV events, and the readjudication eased but did not eliminate the panelists’ concerns (Also see "Avandia Review Endorsement Aided By Good Data, Cultural Shift On Safety" - Pink Sheet, 17 Jun, 2013.). The panelists generally favored a placebo-controlled trial to resolve the issue, but acknowledged that the drug’s REMS-restricted distribution made for poor enrollment prospects (Also see "Avandia Panel Wants Another Trial: Pipe Dream Or Post-market Benchmark?" - Pink Sheet, 17 Jun, 2013.).

The CV event rate in RECORD was in question because the primary endpoint contained subjective elements; it included cardiovascular hospitalization as well as cardiovascular death, and categorization of events varied between adjudicators.

FDA’s strong preference for a stricter endpoint with fewer subjective measures is clear throughout the Invokana review.

FDA’s strong preference for a stricter endpoint with fewer subjective measures is clear throughout the Invokana review.

The agency favors the MACE composite endpoint of CV death, nonfatal myocardial infarction and nonfatal stroke over a “MACE-plus” composite that adds hospitalization for unstable angina, but it acknowledges that fewer strict MACE events are likely to occur, necessitating very large and long clinical trials.

FDA seems content with a compromise that allows MACE+ in the pre-marketing trial to rule out the high risk threshold, but requires MACE for the post-marketing trial to assure definitive evidence of CV safety (Also see "Bigger Is Better, FDA Tells J&J Over Invokana Outcomes Trial" - Pink Sheet, 27 Aug, 2013.).

The same trade off between the statistical rigor of one endpoint versus the speed of the other is arising as FDA develops an approach to CV risk assessment of obesity drugs: MACE+ means faster detection of any CV safety signals, but with the drawbacks of subjective assessments and statistical noise (Also see "Obesity Drugs’ CV Risk Assessment Needs Hard Endpoints, Panel Tells FDA" - Pink Sheet, 2 Apr, 2012.).

Reliance On Single CV Outcomes Trial Carries Risk And Reward

The 2008 diabetes CV risk guidance provides sponsors with flexibility in designing a program that can address the pre- and post-marketing risk assessments. “Many different approaches to demonstrate an acceptable CV safety profile have been proposed by companies,” Parks said. J&J proffered one of the more common approaches, she indicated: “reliance on a single trial to exclude both CV risk margins.”

“Some programs have designed a single trial to demonstrate superiority which if successful would meet the requirements of excluding a 30% excess CV risk,” Parks observed. While such a trial’s “stated objective” would target the 1.3 risk margin, she noted that “an interim analysis of the single trial could be performed to exclude the 1.8 risk margin with appropriate statistical procedures set in place.”

“At issue is maintaining the confidentiality of the interim trial results to ensure the integrity of the ongoing portion of the trial while providing transparency to the public on how FDA has reached its benefit-risk assessment supporting market approval,” Parks said.

With the canagliflozin CANVAS outcomes trial, the sponsor decided to “partially unblind the results of the interim analysis” after finding increases in LDL cholesterol in canagliflozin patients in the core Phase III program, Parks noted. J&J decided to unblind CANVAS without guidance from FDA because “the sponsor felt that having unblinded results would facilitate internal company decision making and that public release of this information would be useful to potential prescribers,” according to minutes of an April 1, 2012 pre-NDA meeting.

“Unblinding of the CVOT and public disclosure of interim results brings up a number of issues that must be grappled with,” Office of Drug Evaluation II Director Curtis Rosebraugh said March 29, especially “whether the unblinding could affect the conduct of the trial so as to invalidate final results.”

J&J’s ultimately unsuccessful attempt to conduct a single CVOT for Invokana should be cautionary for other sponsors hoping to save time and money by using one outcomes trial to address both CV risk thresholds. Unexpected findings, be they statistical artifacts or serious safety signals, can be expected to occur in most, if not all, clinical programs that are powered to detect rare events in large populations.

J&J’s “Hail Mary” Pass Fails, New CVOT Added

CANVAS was initially envisioned as including two sequential study cohorts. The initial group of approximately 4,500 patients (Cohort A) was randomized to placebo or canagliflozin 100 or 300 mg. An interim analysis would be performed about four years in “to determine feasibility of demonstrating CV benefit.” If the independent data monitoring committee so decided, Cohort B would begin enrollment, Parks reported.

However, when the interim results were unblinded, “plans to enroll Cohort B were terminated,” Parks said. J&J proposed that Cohort A, which was completely enrolled, could provide CV safety data “to exclude both CV risk margins of 1.8 and 1.3,” Parks reported. While saying the trial “comfortably” excluded the pre-market CV risk margin of 1.8, she questioned “whether the completion of CANVAS (i.e., Cohort A) can be relied upon to exclude a post-market CV risk margin of 1.3.”

Given the low event rate observed in CANVAS, with its enrollment closed at 4,410 patients, the trial would take too long to produce definitive evidence to rule out cardiovascular risk, she predicted.

“Even under the most optimistic and ideal trial situation, the earliest we can anticipate data from Cohort A would be 2018,” Parks calculated. “Given the length of time to receive this information and the uncertainty on what disclosure of interim results may have done to the interpretability of this trial, I do not believe Cohort A should be relied upon solely to meet a post-marketing requirement to exclude a CV risk margin of 1.3.”

The division gave J&J a choice between expanding CANVAS enrollment or conducting a new trial. J&J opted to continue the CANVAS trial while starting new trial, CANVAS-R, and combining the patients for a CV meta-analysis to rule out the 1.3 risk margin

Rosebraugh saw benefits with both options. “Expanding enrollment allows for a timely answer to the question,” he observed. “However, expanding enrollment will bring in the concern of whether the new subjects are different in some way from the existing trial subjects.” Expanding enrollment “may carry more risk than performing a new trial,” he said. On the other hand, he added, if the publicity around interim analysis is sufficient to corrupt an ongoing trial, then it is also likely to affect enrollment in a new trial.

“Additionally, there does seem to be wasted effort in stopping an ongoing trial that has ≈25%-30% of necessary events when it is unclear if the trial will be corrupted or not,” Rosebraugh said.

He acknowledged that “there is much concern that investigator or participant behaviors could be affected such that bias may be introduced (initial favorable results may drive those on study to seek off-study drug, unfavorable results may cause unbalanced withdrawal from study, corrupting evaluation arms).” However, he pointed out, “there are also not many (or perhaps any) examples to validate those concerns.”

“Things that may make unblinding less concerning are that the CVOT was fully enrolled at the time of public disclosure and the point estimate for MACE was essentially unity,” Rosebraugh said.

By 2017, when J&J is scheduled to analyze the CANVAS and CANVAS-R results, major post-marketing outcomes trials for several other anti-diabetic agents will have completed – and positive findings could reduce the perception of cardiovascular risk for new diabetes drugs in general. The initial reports for the SAVOR trial of Onglyza and the EXAMINE trial of Takeda Pharmaceutical Co. Ltd.’s Nesina (alogliptin) have raised question about whether the policy in place is necessary after all (Also see "Will CV Verdicts For Onglyza, Nesina Cause FDA To Change Its Tune On Antidiabetic Safety?" - Pink Sheet, 9 Sep, 2013.).