Legacy Of Many Sponsors Complicated Review Of Vanda's Fanapt

The multiple sponsors of the antipsychotic Fanapt (iloperidone) produced an inconsistent development program followed by a tumultuous and ultimately dramatic review as FDA worked through a succession of issues surrounding clinical trial design and analysis choices.

The iloperidone development program that Vanda acquired in 2004 was extensive but somewhat piecemeal. Vanda was the third IND sponsor for iloperidone, and conducted only one of the four pivotal trials for the acute schizophrenia indication approved by FDA May 6, 2009.

The one pivotal trial conducted by Vanda, Study 3101, turned out to be the only Phase III study that FDA said "provides convincing evidence of both efficacy for iloperidone and comparable efficacy to an approved antipsychotic agent," in that case, Pfizer's Geodon (ziprasidone).

Office of Drug Evaluation I Director Robert Temple characterized FDA's dilemma in a May 6, 2009, memo: "We were not able to conclude that there was a second supportive study, although several of them were 'close' and their flaws were at least debatable." A "not approvable" letter was issued July 25, 2008, calling for further clinical evaluation.

To get from "close" to "approved" without conducting a new trial, Vanda hired a regulatory consultant to help create a cohesive argument for Fanapt's approval on the basis of the original data set (¹ Pharmaceutical Approvals Monthly July 2009, p. 41). The drug was approved May 6.

To convince FDA, Vanda needed to defend the prior sponsors' choices in clinical trial design - some of which FDA officials flatly called "odd." Notable sticky spots included the inclusion of both schizophrenia and schizoaffective disorder patients in the iloperidone trials and the use of non-inferiority designs in maintenance trials.

Adding to the degree of difficulty, Vanda tried to build in a pharmacogenetic claim, which required extensive genetic testing of patients. The pharmacogenetic aspect proved a bridge too far for FDA at the point of Fanapt approval, but Vanda continues to pursue a pharmacogenetic test to identify responders (see ² (Also see "Fanapt Pharmacogenetic Data Suggestive But Not Conclusive, FDA Finds" - Pink Sheet, 1 Aug, 2009.) ).

Too Many Cooks?

Vanda hinted at the challenge of reconciling the differences in clinical trial design and administration across its Phase III package during the pre-NDA meeting. "The sponsor noted that combining data across these very different programs would be difficult," the minutes of a Feb. 1, 2007, pre-NDA meeting relate.

Hoechst Marion Roussel initiated development of iloperidone in 1990, and conducted 13 Phase I and II studies. Titan licensed the drug from HMR in January 1997, but turned around and licensed it to Novartis before the year was out.

Novartis conducted another 12 Phase I and II trials of iloperidone, three of the four short-term Phase III studies submitted to the NDA (3000, 3004, and 3005), three long-term maintenance Phase III trials (3001, 3002, and 3003) and one study in elderly patients with dementia.

In 2004, Vanda acquired the iloperidone IND from Novartis. The CEO and founder of Vanda, Mihael Polymeropoulos, had come across iloperidone during his tenure at Novartis, as head of the Big Pharma's global pharmacogenetics department.

Once Vanda controlled the asset, the biotech conducted an additional Phase I study as well as the fourth short-term Phase III trial, Study 3101. Vanda submitted the NDA on Sept. 25, 2008, seeking indications for both acute and maintenance treatment of schizophrenia.

Wanted: A Second Well-Controlled Trial

With the success of its own study 3101, Vanda only needed to convince FDA that one of the three Novartis trials was adequate and well-controlled to meet FDA's requirement for two pivotal trials to support approval.

All three of the studies ran into trouble at FDA, however. The clear failure of Study 3000 to show efficacy for iloperidone shifted the focus to studies 3004 and 3005 (see chart for ³ descriptions of all the Phase III trials ).

In its response to the not approvable letter, Vanda proposed a route to approval based on counting 3101 and 3004 as the two adequate and well-controlled trials. FDA's minutes of a Sept. 10, 2008 meeting - the meeting where Vanda successfully made its renewed argument for approval - reveal the shifting strategy: "They now seem to concede that study 3005 was not a positive study overall," and "Vanda has focused again on study 3004 as the one positive study."

FDA, however, weighed the drawbacks of studies 3004 and 3005 differently, and ultimately approved Fanapt with labeling citing studies 3101 and 3005.

Be Sure To Enroll The Right Patient Population

Novartis' decision to enroll both schizophrenia and schizoaffective disorder patients in iloperidone trials was a major stumbling block for study 3004. Schizoaffective patients made up 22 percent of the sample in study 3004, but iloperidone was only found to be superior to placebo in all patients enrolled in the trial. In the analysis of only schizophrenic patients, iloperidone did not show superiority to placebo.

"The sponsor focuses heavily on an argument that we should not distinguish between schizophrenia and schizoaffective disorder," Division of Psychiatry Products Director Thomas Laughren reported in his March 27, 2009, memo. "They argue that this is a difficult distinction in the acute setting and that the treatment of psychotic symptoms in the acute setting is the same, regardless of the diagnosis."

"While these may be true statements, they are irrelevant to the discussion," Laughren countered. "The investigators had some basis for making the diagnoses they made, and it is highly inappropriate for the sponsor or its consultants to discard these judgments that the protocol required the investigators to make."

"The more important facts are that these are considered distinct diagnoses by DSM-IV, however difficult it is to distinguish them acutely, and the outcome is strikingly different for the different subgroups," the division director continued, coming to the conclusion that "the data for that trial are fatally pathological, and one cannot reasonably pool data from the schizophrenic and schizoaffective subgroups."

"It is not that I fundamentally object to pooling data from schizophrenic and schizoaffective patients (we have accepted this approach many times in the past)," Laughren explained. "But for this study, where the positive finding in the schizoaffective patients is by the sponsor's own admission an 'anomaly,' there is no justification for such a pooling."

Prospective Plan Of Analysis Wins Out

Nonetheless, at the pivotal Sept. 10, 2008, meeting to discuss the "not approvable" letter, FDA declared that "we are now willing to consider an analysis of all patients randomized for study 3004, since this was the planned analysis for that study."

The change in course from Laughren's previous position appears partially due to FDA's sense of fairness. After Vanda submitted its response to the "not approvable" letter on Aug. 21, 2008, FDA reiterated its position that both efficacy and safety data were deficient. "We did, however, acknowledge their complaint that they were not informed until the time of the action letter that we would focus on subgroups of patients with schizophrenia in studies 3000, 3004, and 3005," Laughren's March 27, 2009, memo states.

"We felt, however, that our advice to them to limit enrollment to patients with schizophrenia in study 3101 should have been a clear signal that this subgroup would be our focus in analyzing the other three studies as well," he declared.

Vanda's victory was short-lived. When study 3004 was analyzed using the planned all-patient analysis, "the result was driven by a huge effect in the roughly 20 (out of 150) patients per group who were schizo-affective, with almost no effect in the larger schizophrenic group," Temple reported. "Study 3004 remains a negative study in our view."

Schizoaffective disorder continues to emerge from the shadow of schizophrenia, reaching a milestone on July 31, 2009, when FDA approved Lilly's Invega (paliperidone) sNDA for schizoaffective patients based on its effect on both psychotic and affective symptoms. Lilly is touting Invega as "the first and only treatment for schizoaffective disorder," which it characterizes as "understudied" thanks to the difficulty in differentiating schizoaffective disorder from schizophrenia or mood disorders, a difficulty that Vanda referenced as well.

Trouble With Subgroups: The Geographic Version

If 3004 could have been one of the two necessary studies, another problem would be obviated: the disparity in results between the U.S. and non-U.S. sites "would no longer be an issue, since this difference was not observed in study 3004."

The apparent efficacy driven by the schizoaffective subgroup in study 3004 "leaves study 3005 as the only remaining primary source of support," FDA said. But study 3005 had its drawbacks as well, with its efficacy results also driven by a subgroup. The review team listed the geographic disparity in results as a concern, pointing out that "the positive findings are coming almost entirely from non-U.S. sites."

In the U.S. subgroup in Study 3005, Vanda argued that "neither iloperidone nor risperidone was superior to placebo," Laughren pointed out. "Thus, they consider the U.S. subgroup in this study a failed experiment that cannot be interpreted."

However, Laughren himself was "now willing to consider study 3005 as a second primary source of evidence for iloperidone," his March 27 memo declares. "I am persuaded by the sponsor's argument that the lack of efficacy in the U.S. sites for study 3005 should not rule out this study as a source of evidence. As they point out, risperidone also failed in the U.S. sites, and thus, data from these sites is simply uninterpretable. They have other positive data for U.S. sites, in particular, from study 3101."

Non-inferior Is Not Enough For Maintenance

Vanda sought indications for both acute and maintenance treatment of schizophrenia. The long-term claim was based on three active-controlled maintenance trials (studies 3001, 3002, and 3003) conducted by Novartis.

At the Feb. 1, 2007, pre-NDA meeting, Vanda sounded FDA out on the adequacy of the design of the maintenance therapy trials it had inherited from Novartis, inquiring about current division policy regarding non-inferiority design for maintenance studies in schizophrenia.

"We indicated that, although we are still open to considering such an approach, we have not completed the work needed to establish a policy change," FDA's meeting minutes note.

By the following year, FDA was no longer open to a maintenance claims based on non-inferiority trials. "These were active-controlled maintenance studies that relied on findings of no difference between iloperidone and haloperidol," Laughren said. "We have not yet accepted non-inferiority designs as a primary source of support for efficacy claims."

Also complicating matters was the EMEA. "Originally, the primary variable was the change from baseline to week 52 in the PANSS total score," FDA statistician Philip Dinh noted in a review. "During the interactions with the [EMEA] the sponsor was advised that, in order to demonstrate the long-term maintenance effect, efficacy analyses should be based on a time to relapse of schizophrenia/schizoaffective symptoms." The sponsor amended the protocols to include the time to relapse analysis. A pooled analysis of the three maintenance studies, trials 3001, 3002 and 3003, found iloperidone non-inferior to haloperidol on time to relapse.

Dinh, however, found several limitations with those maintenance studies, starting with the non-inferiority design. Also, "the studies did not include a placebo arm, [which] made the interpretation of the non-inferiority more difficult," and the pooling of studies for efficacy is not "current standard practice for the Division of Psychiatry."

"More seriously, by changing the analysis plan from a change-from-baseline analysis to a time-to-relapse analysis, the analysis population was amended. … Randomization may be violated." Dinh said.

In the end, the statistician determined that the long-term non-inferiority claim based on studies 3001, 3002 and 3003 did not have regulatory merit given the designs and analyses of these studies. "Currently, the Division of Psychiatry does not consider a non-inferior, active-controlled study as an appropriate design for the schizophrenia indication." Dinh said.

But Laughren didn't rule the data out completely. While clearly stating that non-inferiority findings from the three maintenance studies should not be relied upon as support for a maintenance claim, the division director commented that "I do believe this additional suggestive evidence of iloperidone's efficacy in schizophrenia can be considered in the overall decision for this drug."

Fanapt labeling states that "although there is no body of evidence available to answer the question of how long the patient treated with Fanapt should be maintained, it is generally recommended that responding patients be continued beyond the acute response" and "periodically reassessed."

Data Integrity Sideshow Fails To Derail 3101

The statistical review of the Fanapt NDA was also the scene of another complication. During the NDA filing meeting on Nov. 9. 2007, statistical team leader Peiling Yang reported that she had received a call from someone "who had some concerns regarding the sponsor's data integrity." On follow-up, the caller "had nothing specific regarding any of the sponsor's study sites."

The caller's "impression was that the sponsor would stop at nothing to get approval," but provided "no hard evidence or promising leads," clinical reviewer Michelle Chuen reported in her June 13, 2008, review.

A 3101 study site selected for inspection by the Division of Scientific Investigations also raised concerns. The investigator, John Gilliam, from Richmond, Va., "had been under investigation by FDA's Office of Criminal Investigation since an FDA inspection in 2003 (on a different NDA and different product) revealed possible falsification of study records. Dr. Gilliam reportedly signed a plea agreement admitting to falsification of study records, but before this could be posted with the court, he died on Feb. 2, 2008."

The psychiatry division noted that its decision on the Fanapt approval did not depend on the data from Dr. Gilliam's site. "Because of the plea agreement(s) from this site on data falsification on a different study, DSI does not consider the data from Dr. Gilliam's site to be reliable." Chuen said.

The Gilliam affair failed to leave a lasting mark on Fanapt. "If we believe this study [3101] is positive despite the unreliability of one site, which is apparently everyone's view, I see little reason to undermine the study in labeling," Temple said in a July 25, 2008, memo.

- Bridget Silverman ( ⁴ [email protected] )