Pfizer Chantix Review Shows Pitfalls Of Patient-Reported Outcomes

The debate over analysis of clinical trial endpoints in FDA review documents for Pfizer's Chantix (varenicline) provides a window into how the agency's view of smoking cessation trial design is changing as new therapeutic classes emerge.

"The development program for varenicline was initiated and conducted during a time of evolution in the Agency's thinking about the design of smoking cessation trials," medical team leader Celia Winchell, MD, declared in her May 3 supervisory review.

Discussion of trial design elements specific to smoking cessation was matched by debate over broader issues, such as use of patient-reported outcomes, abuse liability assessment and analysis of multiple endpoints, that are the subject of both FDA and industry efforts to establish standards.

Pfizer submitted the NDA for its smoking cessation agent varenicline (then Champix ) on Nov. 10, 2005; the selective nicotinic receptor partial agonist was approved as Chantix on May 10, 2006. Labeling claims superiority to GlaxoSmithKline's Zyban (bupropion); the superiority data had earned Chantix a priority review (¹ Pharmaceutical Approvals Monthly December 2005, p. 14).

Grace Periods & The Definition Of Abstinence

"Success should be defined as complete abstinence" from smoking, Winchell observed. However, "if the effect of the treatment is not expected immediately…then efficacy ascertainment may take place over a period following a 'grace period.'"

"Historically, six-week trials were considered conventional, and the ascertainment window was set at weeks 2-6, somewhat arbitrarily," she noted. "This trial duration was appropriate for nicotine replacement products, which were assumed to treat a self-limiting condition (acute nicotine withdrawal)."

"More recently, the clinical approach to smoking cessation treatment has been moving toward a viewing of the disorder of tobacco dependence as a chronic, relapsing condition," Winchell continued.

"The division's thinking has evolved as longer and longer treatment durations have been proposed, sometimes resulting in a grace period of many months and an efficacy ascertainment period of four weeks."

"Ultimately, the concept of 'abstinence throughout treatment following a pharmacologically-justified grace period' was articulated in a White Paper."

The varenicline pivotal study protocols specified an ascertainment window of the "last four weeks of treatment," she noted. Given the 12-week trial duration, that translated to an eight-week grace period.

"The two-month grace period in the varenicline trials is not fully justified," Winchell declared.

Complicating the issue was the possibility that varenicline possesses multiple modes of action. The "partial agonist properties of varenicline might result in blockade of exogenous nicotine, allowing for an extinction approach to behavior change," Winchell said, while "on the other hand, its agonist properties would also predict early efficacy as it provides some relief for acute withdrawal symptoms."

FDA therefore analyzed the trials using grace periods as short as two weeks. "The efficacy of varenicline is apparent even without the prolonged grace period," Winchell concluded.

"There were quite a few subjects who required several weeks to initiate abstinence," she added. "Thus, the antagonist properties of the drug and an extinction mechanism of action may play an important role."

Due to Chantix' nicotine agonist activity, "an additional analysis of importance to the division was the proportion of subjects who initiated abstinence by Week 3 and then maintained that abstinence," Division of Anesthesia, Analgesia & Rheumatology Products Director Robert Rappaport, MD, said in an April 8 memo.

"This analysis was consistent with the division's current analytic approach used to assess other nicotine agonist products," Rappaport explained.

Measuring Quit Rates

The Chantix primary endpoint - the four-week continuous quit rate over the last four weeks of treatment - was acceptable to FDA, although the agency did suggest that the endpoint "loses relevance as treatment period lengthens," according to minutes of a Dec. 9, 2002, end-of-Phase II meeting.

Pfizer's proposed relapse prevention claim ran into methodological problems. The long-term quit rate measure, which differed from continuous abstinence in that it allowed up to six days of smoking during long-term follow-up, is "conceptually…u nobjectionable," Winchell said, but the methods for capturing smoking history were not detailed enough.

As an alternative definition of relapse, LTQR "does not appear to provide evidence that varenicline treatment has a specific effect that allows former smokers to lapse without relapsing," Winchell remarked.

Winchell's interest in alternative analyses extended to labeling discussions. She found a bar graph Pfizer had proposed to present clinical data to be too limited, arguing that measures beyond the primary endpoint should be communicated: "There are any number of ways of looking at the abstinence rates, and there is no compelling reason to choose 'weeks 9-12' other than its designation as the 'primary efficacy endpoint.'"

Designation of a primary efficacy endpoint "is a statistical necessity," she said. "It is not necessarily an endorsement of a particular analysis as the best, most meaningful, most accurate or most illustrative."

"There are few precedents in labeling for this therapeutic category," Winchell observed.

"The nicotine replacement products…did not illustrate a specific mean abstinence rate per treatment arm; instead the range of rates across centers was shown," she noted. "The Zyban label presents quit rates from the beginning of the fourth week of treatment…t hrough various time points in the study through post-treatment follow-up, giving a mean value and confidence intervals."

Regardless of any debate, FDAers commented on the solid primary efficacy data. Office of Drug Evaluation II Deputy Director Curtis Rosebraugh, MD, who had signatory authority on the application, said the data "clearly demonstrate efficacy." He also pointed to "the impressive fashion to which the efficacy advantage over placebo is sustained after drug therapy is stopped" in his May 10 memo.

Patient-Reported Outcomes

"While the secondary outcome measures were generally supportive of the primary outcome measure results, the subjective measures were consistently less supportive," Rappaport said.

Secondary outcome measures included longer-term measures of continuous abstinence from week nine through weeks 24 and 52. The subjective secondary endpoints were based on the Minnesota Nicotine Withdrawal Scale (MNWS), the Brief Questionnaire of Smoking Urges and the Smoking Effects Inventory.

Pfizer "used unacceptable methodologies in evaluating some of these patient reported outcomes such as selecting unvalidated subscales for analysis," he said.

FDA had indicated the problematic nature of PROs during Chantix' clinical phase. At a June 9, 2005, pre-NDA meeting, FDA agreed that PRO label claims were possible and urged discussion post-filing, but the minutes also report FDA stating that "discussion of PROs from a battery of secondary outcome measures, in product labels and package inserts is discouraged."

Pfizer had proposed claims regarding Chantix' effect on "craving," "withdrawal symptoms," and "reinforcing effects of smoking" based on PROs.

The division requested a review of the patient-centered endpoints by the Office of New Drugs' Study Endpoints & Label Development (SEALD) team in December 2005. SEALD focused on an expert report on the PRO instruments, by Mapi Values and the MEDTAP Institute, included in Pfizer's NDA.

SEALD is part of a larger FDA project considering PROs, which are listed on the Critical Path opportunities list. FDA issued a draft guidance in February 2006; a final version is expected by autumn.

Chantix' PRO-based symptom claims were part of Pfizer's moves to position the drug as "a more effective and better-tolerated smoking cessation therapy…because its mechanism of action reduces reinforcing effects of smoking as well as the symptoms of withdrawal and craving," SEALD team leader Jane Scott, PhD, wrote in a March 10, 2006, memo.

"'Urge to smoke' describes the concept measured more accurately than 'craving,'" Scott said. "'Craving' is a value-laden term that is not part of the psychiatric definitions of nicotine withdrawal syndrome."

Scott said that inclusion of "craving" is "based on recommendations by a work group that has not been confirmed by the larger scientific community." The "symptoms of withdrawal" statements also reflect recommendations of the Society for Research on Nicotine & Tobacco (SRNT) Work Group on Assessment of Craving & Withdrawal in Clinical Trials, she noted.

"SRNT recommends that craving and specific symptoms of withdrawal" be used as endpoints, Scott reported. However, "it is not clear that the SRNT work group recommendations are appropriate to define endpoints that will support product claims," she said.

In FDA's view, "to support claims related to withdrawal, it would be important to show that the results for all domains of the MNWS withdrawal measure improve with treatment." Pfizer relied on scores on four domains of the instrument.

While noting that "symptoms of withdrawal appear to be a clinically meaningful concept," Scott continued: "it is not clear what symptoms must be measured."

Pfizer selected the "reinforcing effects of smoking" concept from literature, proposing that Chantix "can limit the beneficial effects of smoking on withdrawal symptoms," Scott said. "This is an exploratory concept that is not well defined."

Pfizer Plays Defense

Meeting records show Pfizer in something of a defensive crouch regarding FDA requirements, including multiple requests for FDA's stance on the PRO measures.

Perhaps seeking to pre-empt the agency, Pfizer submitted a risk management plan with the varenicline NDA - even though FDA's Office of Drug Safety did "not identify a specific safety concern for which a RMP to minimize risk would be normally associated," ODS reviewers wrote in a March 29, 2006, memo.

"The measures proposed by the sponsor seem reasonable but would appear to be routine given the potential risk," the ODS team concluded.

Pfizer requested a meeting on abuse liability "based partly on their experiences with the Agency in determining abuse liability of another Pfizer drug, pregabalin [ Lyrica ]," according to minutes of the Aug. 18, 2005, pre-NDA meeting.

"Pfizer expressed their concerns that, like with their drug, pregabalin, there were no fixed rules or common understandings of outcomes in planning and conducting abuse liability studies," the minutes state.

"This was of particular concern to them given that 30% of their portfolio are drugs with CNS activity."

Pfizer Calls For Collaboration On Abuse Liability

At the pre-NDA meeting, Pfizer proposed a cooperative industry research effort to generate a consensus, because "assessing abuse liability is a common concern [for] sponsors of CNS-active drugs."

"Pfizer expressed the concern that concerned companies may wish to pool research efforts to see if patterns or methods emerge from various studies that would indicate an effective measure for determining abuse liability," the minutes report.

Pfizer observed that FDA's Critical Path initiative "specifically mentions abuse liability assessment as one of the projects," the minutes add.

The pre-NDA meeting evolved into a "general discussion on the merits of multivariate endpoint analysis and/or composite endpoints versus a minimal number of primary endpoints," minutes note. "Pfizer expressed concerns that the criteria for determining abuse liability could not be trusted to accurately judge the degree of abuse liability for a given drug."

FDA statistician Ling Chen, PhD, suggested that "the statistical evaluation of a human abuse liability study is similar to an efficacy study where all co-primary endpoints need to show efficacy individually."

Multiple Endpoints

Pfizer questioned FDA's refusal to support analyzing the four endpoints of the varenicline abuse potential study as a composite endpoint.

FDA stated that "multivariate analysis is not appropriate for drug abuse potential studies," the minutes stated. In this case, "the four co-primary endpoints measure different aspects of abuse liability."

Pfizer took the position that "multivariate tests involving various components or combinations of components could provide a scientifically sound method," according to the pre-NDA meeting minutes.

FDA statistician Yi Tsong, PhD, said a single composite endpoint could be "reasonable" but such endpoints have not been validated for abuse liability.

Chen noted that a Multiple Endpoints Expert Team at PhRMA is "doing research in this area," but "no concrete solution for this problem has been proposed from the team yet."

Guidance may soon be forthcoming from FDA. On July 10, FDA Deputy Commissioner for Medical & Scientific Affairs Scott Gottlieb, MD, told the 2006 Conference on Adaptive Trial Design that a guidance on "how to look at multiple endpoints in the same trial…is currently being drafted and we hope to be able to disclose that work as soon as January" ( ² (Also see "FDA To Issue Five Guidances On Adaptive Clinical Trial Design – Gottlieb" - Pink Sheet, 15 Jul, 2006.) ).