ATDEPARTMENT OF HEALTH AND HUMAN SERVICES

FOOD AND DRUG ADMINISTRATION

CENTER FOR DRUGS EVALUATION AND RESEARCH

 

 

 

 

 

 

 

 

 

 

ANTIVIRAL DRUGS ADVISORY COMMITTEE MEETING

NDA 21-266, Vfend (voriconazole) Tablets

NDA 21-267, Vfend I.V. (voriconazole) for Injection

Pfizer Global Research and Development

 

 

 

 

 

 

 

 

 

 

Thursday, October 4, 2001

8:30 a.m.

 

 

 

 

 

 

 

 

 

 

 

The Town Center Hotel

8727 Colesville Road

Maryland Ballroom

Silver Spring, Maryland

PARTICIPANTS

Roy M. Gulick, M.D., M.P.H., Chair

Tara P. Turner, Pharm.D., Executive Secretary

Members:

John D. Hamilton, M.D.

Wm. Christopher Mathews, M.D.

Sharilyn Stanley, M.D.

Brian Wong, M.D.

Ram Yogev, M.D.

Consultants (voting) Pending New AVAC Members:

Victor G. DeGruttola, Sc.D.

Janet A. Englund, M.D.

Jonathan M. Schapiro, M.D.

Lauren V. Wood, M.D.

Consultant (voting):

Keith A. Rodvold, Pharm.D.

Industry Representative (non-voting):

Eugene Sun, M.D.

Guest (non-voting):

Vicki A. Morrison, M.D.

FDA:

Renata Albrecht, M.D.

Marc Cavaille-Coll, M.D., Ph.D.

Mark Goldberger, M.D., M.P.H.

ohn Powers, M.D.

Rosemary Tiernan, M.D., M.P.H.

C O N T E N T S

Call to Order, Roy M. Gulick, M.D., M.P.H. 4

Introductions 4

Conflict of Interest Statement, Tara P. Turner, Pharm.D. 6

FDA Introductory Remarks, Mark Goldberger, M.D., M.P.H. 8

Sponsor Presentation

Pfizer Global Research and Development

Introduction, Reinhard Baildon, M.D. 10

Invasive Fungal Infections, Management and

Medical Needs, Thomas F. Patterson, M.D. 11

In Vitro/In Vivo Data, Clinical Pharmacology,

Reinhard Baildon, M.D. 17

Efficacy, Helen W. Boucher, M.D. 33

Safety and Conclusions, Reinhard Baildon, M.D. 67

Questions from the Committee 90

FDA Presentation

Rosemary Tiernan, M.D., M.P.H. 128

John Powers, M.D. 136

Rosemary Tiernan, M.D., M.P.H. 162

Questions from the Committee 169

Charge to the Committee 183

Committee Discussion and Vote 186

P R O C E E D I N G S

Call to Order

DR. GULICK: Good morning. I am Trip Gulick from Cornell. I am happy to call this meeting of the Antiviral Drug Advisory Committee to order.

For those of you who were here yesterday, you will be happy to know that a plague of locusts was sighted on Colesville Road; we don't think they are going to be interfering with the meeting this morning.

I would like to remind the committee and all the speakers to be aware of the mikes, to make sure that you are about--I was told six to eight inches is optimal, so that everyone can hear people's comments.

I would like to start the meeting by having the members of the committee introduce themselves. Dr. Sun, we will go to you to start. Please state your name and where you are from.

Introductions

DR. SUN: Eugene Sun, Abbott Laboratories.

DR. MORRISON: Vicki Morrison, VA Hospital, Minneapolis.

DR. RODVOLD: Keith Rodvold, University of Illinois, Chicago.

DR. STANLEY: Sharilyn Stanley, Texas Department of Health.

DR. WOOD: Lauren Wood, National Cancer Institute, Bethesda, Maryland.

DR. MATHEWS: Chris Mathews, University of California, San Diego.

DR. HAMILTON: John Hamilton, Duke University and Durham VA Medical Center.

DR. ENGLUND: Janet Englund, Department of Pediatrics, University of Chicago.

DR. TURNER: Tara Turner, executive secretary for the committee.

DR. SCHAPIRO: Jonathan Schapiro, Stanford University and Tel Aviv University.

DR. WONG: Brian Wong, the West Haven Connecticut VA Hospital and Yale University.

DR. DEGRUTTOLA: Victor DeGruttola, Harvard School of Public Health.

DR. POWERS: John Powers, medical officer, FDA.

DR. TIERNAN: Rosemary Tiernan, medical officer, FDA.

DR. ALBRECHT: Renata Albrecht, Acting Director, Division of Special Pathogen Immunologic Drug Products, FDA.

DR. CAVAILLE-COLL: Mark Cavaille-Coll, the Office of Drug Evaluation I.V., FDA.

DR. GULICK: Thank you very much. Tara Turner will now read the conflict of interest statement.

Conflict of Interest

DR. TURNER: The following announcement addresses the issue of conflict of interest with regard to this meeting, and is made a part of the record to preclude even the appearance of such at this meeting.

Based on the submitted agenda for the meeting and all financial interests reported by the committee participants, it has been determined that all interests in firms regulated by the Center for Drug Evaluation and Research which have been reported by the participants present no potential for an appearance of a conflict of interest at this meeting, with the following exceptions:

In accordance with 18 USC 208(b)(3), full waivers have been granted to Dr. Courtney Fletcher, Dr. Roy Gulick, Dr. Brian Wong, Dr. Ram Yogev and Dr. Lauren Wood. A copy of these waiver statements may be obtained by submitting a written request to the agency's Freedom of Information Office, Room 12A-30 of the Parklawn Building.

Further, in accordance with 21 USC 355(n)(4), Dr. Courtney Fletcher, Dr. Brian Wong, Dr. Ram Yogev and Dr. Victor DeGruttola have been granted waivers that permit them to vote on matters related to today's discussions.

We would like to disclose for the record that Dr. Victor DeGruttola and Dr. Keith Rodvold have interests which do not constitute financial interests within the meaning of 18 USC (208)(a) but which could create the appearance of a conflict. The agency has determined, not withstanding these interests, that the interest of the government in their participation outweighs the concern that the integrity of the agency's programs may be questioned. Therefore, Drs. DeGruttola and Rodvold may participate fully in today's discussion and vote concerning Vfend.

In addition, we would like to note that Dr. Eugene Sun is Division Vice President, Infectious Diseases and Virology Development Department, Pharmaceutical Products Division at Abbott Laboratories. He is participating in this meeting as an industry representative acting on behalf of regulated industry. As such, he has not been screened for any conflicts of interest.

In the event that the discussions involve any other products or firms not already on the agenda for which an FDA participant has a financial interest, the participants are aware of the need to exclude themselves from such involvement and their exclusion will be noted for the record.

With respect to all other participants, we ask in the interest of fairness that they address any current or previous financial involvement with any firm whose products they may wish to comment upon. Thank you.

DR. GULICK: Just to clarify, Dr. Courtney Fletcher was unable to make this meeting. And, good morning, Dr. Yogev, and will you introduce yourself, please, and state your affiliation?

DR. YOGEV: Ram Yogev, Children's Memorial Hospital in Chicago.

DR. GULICK: Thanks. I would like to turn to Dr. Goldberger who will present some introductory remarks from the agency.

FDA Introductory Remarks

DR. GOLDBERGER: Thank you, Dr. Gulick. I would like to welcome you, advisory committee members, invited guests, Pfizer, FDA staff and those in the audience to the second day of this Antiviral Advisory Committee.

Today we will be discussing NDAs for voriconazole tablets and voriconazole for injection. The applicant has actually requested a series of indications: invasive aspergillosis, empiric antifungal therapy of febrile neutropenic patients, treatment of Candida esophagitis, serious Candida infections, serious fungal infections due to Fusarium and Scedosporium, and serious fungal infections in patients refractory or intolerant to other therapy.

In general, we are in agreement with the applicant with regards to the numerical results of the studies supporting these different indications, as well as to the data in the overall safety profile in this drug. We have, after discussions with the applicant, chosen to focus the presentations at this meeting on two of the incations, treatment of invasive aspergillosis and empiric antifungal therapy of febrile neutropenic patients as well as, of course, a complete discussion of the safety profile of this product, including drug interactions.

We have chosen to focus on these two indications first, because as a practical matter, it is very difficult to review detailed data on so many indications in a single day. Secondly, as I noted, we are in general agreement with the numerical results on all these indications and because we feel that in the case of invasive aspergillosis the applicant will be presenting data from a randomized trial for therapy in this indication. It is a type of study that realistically has not been presented before and we think the results are very important, both for presentation, discussion of indications for this product, and possibly for future drug development. So, we think this is an extremely important piece of data and we wanted to have sufficient time to discuss this.

In the case of empiric antifungal therapy of febrile neutropenic patients, we are, as noted, in general agreement with the sponsor with regards to the numerical results of the study. However, the numerical results are unclear as to the appropriate regulatory action that ought to be taken, and we feel that having a full discussion of this indication, with a lot of input from the committee, will be quite useful.

Finally, not surprisingly with a product in the azole class and with treatment of a wide variety of patients with serious underlying disease, there are a number of issues from the safety profile, both safety issues and issues related to drug interactions. We feel it is very important to have an adequate discussion of these topics, as well as an adequate discussion of the drug interaction topics, since they can certainly affect decisions even on the first two indications and certainly on some of the others as well.

With that, I will turn it over to Dr. Gulick again.

DR. GULICK: Thanks, Dr. Goldberger. We will begin with the presentation from the sponsor, Pfizer Global Research and Development. Dr. Baildon will be beginning.

Sponsor Presentation

Introduction

DR. BAILDON: Thank you very much, Dr. Gulick. Good morning.

[Slide]

My name is Reinhard Baildon. I am from Pfizer Global Research and Development. I appreciate the opportunity today to present our data on the efficacy and safety of our new antifungal agent, voriconazole.

[Slide]

This is the order of our presentation today. I will provide some of the background to the discovery and development of voriconazole and describe the in vitro and in vivo activity. We will then focus our discussion on the questions Dr. Goldberger asked.

Before I start on the voriconazole data, however, I would like to ask Dr. Tom Patterson, Professor of Medicine at the University of Texas at San Antonio, to highlight key aspects on the invasive fungal infections and why there is still a high unmet medical need in this population. Those are actually the first slides in the document that you have in front of you. Dr. Patterson?

Invasive Fungal Infections: Management and Medical Need

DR. PATTERSON: Thank you, Dr. Baildon and thank all of you for the opportunity for me to discuss with you today the management and medical needs for invasive mycoses.

[Slide]

This slide outlines the mortality due to invasive mycoses over the last two decades in the United States, from data published this last month by the CDC. You can see significant increases in mortality due to invasive mycoses. Notably mortality due to Candida, in the yellow line, has recently declined perhaps due to better recognition of these infections and earlier antifungal therapy.

However, I think you will also see significant increases in mortality due to Aspergillus, in the purple line, with those rates in mortality rising four-fold over that period. Although not as common, other mycoses, particularly molds and Fusarium, for which limited therapeutic options exist, are associated with even higher rates of mortality.

[Slide]

One of the most challenging of these opportunistic fungal pathogens is Aspergillus. This slide shows mortality associated with invasive aspergillosis. From a review by Lin and colleagues of almost 2000 patients with invasive aspergillosis from 50 published studies, you can see overall mortality rates were 58 percent but were dramatically higher in some groups. For example, in those patients undergoing bone marrow transplantation mortality was 87 percent, while those with leukemia or lymphoma had a mortality of 49 percent.

Extent of infection also correlates with outcome. Patients with pulmonary infection had mortality of 59 percent, while those who developed central nervous system or disseminated infection had mortality approaching 90 percent, suggesting that an early diagnosis and therapy could be a way to reduce the mortality of this often lethal disease.

[Slide]

Recently guidelines were published by the IDSA for management of Aspergillus infections. These published guidelines suggested the importance of a prompt, aggressive diagnosis and therapy to be initiated at early suspicion of infection. It was also recognized that that strategy was often the one employed with the use of early empirical antifungal therapies. Antifungal therapy was suggested to be given intravenously in seriously ill patients, utilizing maximum doses of amphotericin B deoxycholate. Alternative agents at that time included lipid forms of amphotericin which were recommended largely for patients intolerant of standard therapies, and itraconazole for sequential use following initial amphotericin B therapy, although the importance of adjunctive therapies was also recognized, as well as the need for new therapies and diagnostic tools for this disease.

It should also be noted, however, that most of these recommendations are based on clinical opinion as few successful randomized trials have been conducted in this area.

[Slide]

Unfortunately, outcomes using these approaches are currently still very poor. This slide shows data we recently published on the outcomes of invasive aspergillosis. We reviewed case records of 595 patients with invasive aspergillosis and showed response rates overall of 37 percent. However, I will draw your attention to those patients with more severe immunosuppression, those undergoing marrow transplantation or with hematologic malignancies, and those patients' outcomes were favorable in only 28 percent. Importantly, in the subgroups of allogeneic transplant alone, favorable response occurred in only 13 percent of those patients, while in patients with hematologic malignancies the outcomes were 39 percent.

In patients with less severe immunosuppression overall outcomes were favorable in about half the patients. Again, site of infection was very important. Patients with pulmonary infection had favorable outcomes in 40 percent, while those who developed later, more advanced disease and dissemination had responses that totalled less than half that. With central nervous system involvement favorable responses occurred in only 9 percent of the patients, again suggesting the potential role for early intervention.

[Slide]

Unfortunately, clinical trials have been difficult to perform in this area. These issues were recently addressed by a consensus panel led by Dr. John Rex, as outlined on this slide. Limitations would include the relative infrequency of these infections as well as the difficulty in establishing a diagnosis of these diseases. The panel suggested the use of surrogate markers like antigen testing, which is not available here in the U.S., or the use of high resolution imaging like CT scans, the halo finding specifically for Aspergillus, which we, in fact, did demonstrate at last year's ICAAC could be successfully used in clinical studies.

Host factors and trial issues, including the selection and use of approved comparators and accepted definitions and endpoints, further hinder clinical trial development. The panel identified only two studies they classified as large prospective, randomized trials that were really ever successfully conducted for Aspergillus and each of then contained about 100 patients.

[Slide]

Because of these clinical trial limitations and the clear unmet medical needs of this disease, most recent studies evaluating new agents have employed either historical controls or utilized smaller numbers of patients. This slide shows three recent trials and the comparative agents, along with the number of patients in each arm. Historical controls were used to evaluate ABLC for a salvage indication of Aspergillus and more recently used to evaluate the new capsofungin.

You can see, I think, the efficacy of these compounds as compared to control patients who were refractory to or intolerant of standard therapies, although overall response rates were slightly more than 40 percent in these patients with very advanced disease.

In a small open, randomized trial which also allowed the inclusion of suspected infection you will see higher response rates, both with liposomal amphotericin and its comparator amphotericin B deoxycholate, perhaps due to the inclusion of those patients with suspected disease but, again, perhaps due to the earlier initiation of therapy.

[Slide]

Because of these poor responses in established invasive mycoses, the use of empirical antifungal therapy is recommended. This slide shows data from the recent mycosis study group trial, published by Dr. Tom Walsh and colleagues, which shows the clear decrease in fungal infections that occurred with liposomal amphotericin as compared with amphotericin B deoxycholate. However, you will note that Aspergillus infections continue to occur in each arm of the trial, again a very significant concern and a major challenge in clinical mycology.

[Slide]

In summary, the epidemiology of invasive mycoses demonstrates an increasing number of infections due to Aspergillus and other molds which are associated with major causes of morbidity and mortality. Clearly, prognosis is improved with a very prompt diagnosis, although it remains very difficult to establish. Host factors are also critical in determining outcome.

Finally, the use of early aggressive antifungal therapy appears to improve responses but, clearly, new approaches and new agents are needed for this disease. Thank you for your attention.

In Vitro/In Vivo Data, Clinical Pharmacology

DR. BAILDON: Thank you. I will now return to the voriconazole development program.

[Slide]

Voriconazole was first administered to humans in 1991 and has undergone extensive development since then. Clinical studies in the United States were begun in 1955 and last year, in November, we filed the NDA. In June of this year we provided a substantial update, including the results of our global comparative aspergillosis study.

Throughout the development we have benefited greatly from close interactions with the Division and external experts. Through the use of external data review committees and rigorous, standardized assessments of patients throughout the program, we aim to ensure a consistent approach across varying study populations and study designs.

[Slide]

Voriconazole is used in a complex patient setting against a background of serious disease and concurrent use of multiple other medications and invasive interventions. To help us in assessing the efficacy and safety in this complex setting, we have over the years consulted with many leading experts in the field and are fortunate to have those listed on this slide with us in the sponsor section. As needed by the committee, they can be active participants in our discussion later today.

[Slide]

As a high level summary, let me describe our assessment of what voriconazole might contribute to the setting of invasive aspergillosis and other invasive fungal infections. We have seen significantly improved outcome and survival benefit in patients with acute invasive aspergillosis. We think the efficacy seen in infections due to emerging pathogens, like Scedosporium and Fusarium, is very encouraging. The in vitro and in vivo data seen with Candida infections, including infections due to Candida krusei as well as the results in the patients that we have treated to date, make us believe that voriconazole is an appropriate treatment for these infections as well.

The efficacy seen in patients with documented fungal infections, as well as the results seen in our empirical therapy study, led us to conclude that voriconazole is an appropriate option for empirical therapy of neutropenic patients with persistent fever. The large safety data base that we have collected led us to conclude that voriconazole has an acceptable safety profile, given the population targeted for its use, and is better tolerated than amphotericin B formulations.

The availability of an I.V. and oral formulations allows switching between these two, especially as the oral formulation has very high bioavailability. We feel that we have collected sufficient data, and I will discuss this, to provide detailed guidance to the prescribing physician on managing the interactions, and we feel that these are manageable given the populations that are treated. Overall, we think that voriconazole addresses an important currently unmet medical need in patients with invasive fungal disease.

[Slide]

The clinical program included patients with documented fungal infections, aspergillosis, candidiasis and emerging pathogens, and a large empirical therapy study in patients with neutropenia and persistent fever. In addition to the controlled clinical trials, our program included patients treated with compassionate use. These patients are included in our pooled safety database, and some of them are included in some of the efficacy results that Dr. Boucher will be discussing later.

In total, over 2000 patients received voriconazole in the clinical program. As I mentioned earlier, our discussion will focus today on the aspergillosis and empirical therapy.

[Slide]

This slide illustrates the close proximity of voriconazole to fluconazole in its structure. Fluconazole is a well established drug with a very large safety database, however, it is not active against Aspergillus. Our design effort was focused on introducing the broader spectrum against Aspergillus. This was achieved with the introduction of the fluoropymridine group here and the methyl group at this location.

[Slide]

I would now like to turn to the in vitro and in vivo characteristics of voriconazole. The high in vitro potency of voriconazole has been established in numerous published experiments. Dr. Lingroff, from the Medical College of Virginia, recently published an overview of this extensive literature.

As a further illustration, I present here the MICs observed in our esophageal candidiasis study for the 546 Candida albicans isolates and the 87 non-albicans isolates. The MIC-90 for voriconazole was about 60-100-fold lower than those observed with fluconazole, and was well within the reach of plasma exposure seen in our patient populations.

[Slide]

Here, the data from this recent collection of clinical isolates are presented graphically with the MICs of fluconazole on a log scale along the X axis and voriconazole susceptibility along the Y axis. As you can see, the relationship between fluconazole and voriconazole susceptibility is linear, however, it is shifted by about two logs if you look here or here, for example.

[Slide]

Extensive data on the in vitro activity of voriconazole against molds have also been published. Here I present the data on 212 Aspergillus isolates from the global comparative aspergillosis study. This somewhat difficult to read table is also included in your briefing document.

I would just like to point out that across the different Aspergillus species are shown MICs of 1 or lower for voriconazole. For example, maximal MICs observed for Aspergillus flavis is 0.5 and for Aspergillus nidulans 1.25, here. Fungicidal concentrations are usually about twice the MIC and so are well within the range of exposure achieved in the majority of our patients for the majority of these isolates.

[Slide]

To study the in vivo behavior of voriconazole we used the immune compromised guinea pig model as our standard antifungal model. These animals are rendered severely neutropenic by injection of cyclophosphamide, and then are maintained immune compromised by steroid treatment. They are challenged with a fungal load that is sufficient to result in immediate invasive disseminated disease. Outcome is assessed by survival and tissue burden remaining in specific organs. The guinea pig is a well established model for azole activity and the kinetics of voriconazole in this model are similar to those observed in humans.

[Slide]

As an example of the extensive and published data generated in this model, I have shown here disseminated invasive aspergillosis from experiments conducted in San Antonio and published recently. The fungal burden injected in these experiments was sufficient to kill all animals within five days if untreated. Antifungal therapy was begun 24 hours after inoculation and was continued for five days. All were voriconazole-treated; high dose itraconazole-treated animals survived. The results shown on this slide are fungal burden up to day seven after challenge when the remaining animals were sacrificed. Fungal burden is measured as colony count as mean log 10 CFU expressed per gram of tissue on the Y axis and relevant organs are displayed on the X axis.

As you can see, voriconazole, itraconazole and amphotericin B, the fungal agents tested, reduced the fungal burden in all these organs. For voriconazole, in the blue color, and itraconazole, in green, the dose response is also visible. The remaining fungal load at the highest dosages of voriconazole, for example in the liver and in the brain, are very low after treatment, and these organs are important targets for Aspergillus infections.

[Slide]

In summary, we believe that the in vivo data indicate that the high potency and cidality observed in vitro against yeasts and molds respectively translates into an efficacy benefit observed in the severely immune compromised animal model.

[Slide]

I would now like to turn to the pharmacokinetics of voriconazole. I will discuss the results of our extensive clinical pharmacology program, addressing the topics outlined here. Our goal was to be able to provide specific guidance to prescribers based on this extensive investigation into the pharmacokinetic behavior of voriconazole in healthy volunteers and in patients.

Voriconazole is well absorbed and has a high bioavailability of 96 percent. This allows switching between the I.V. and oral formulations, which was allowed in all clinical trials using both formulations. Voriconazole has a high volume of distribution and moderate protein binding, suggesting significant tissue distribution.

[Slide]

This is illustrated in the top part of this slide, here, where we show the tissue distribution in a male rat five minutes after infusion. You can see, for example, that brain concentrations are about those seen in blood. This is further illustrated by the detection of voriconazole in the cerebrospinal fluids of guinea pigs after multiple dosing, where we observed similar levels in the cerebrospinal fluid to those seen in plasma. We also received samples of cerebrospinal fluids from patients, collected in their routine medical care, where we also detected voriconazole. In autopsy samples of brain, kidney, lung and liver tissue, the levels seen in those tissue samples were consistent with the data in the animals that I showed here.

[Slide]

Voriconazole is metabolized primarily by cytochrome P450 isoenzymes 2C19 and 2C9 and, to a lesser degree, by 3A4. The isoenzyme 2C19 exhibits genetic polymorphism, with about two percent of the Caucasian population being poor metabolizers.

The major circulating metabolite in humans is also the major circulating metabolite in our toxicology species discovered by our preclinical program. In vitro data indicate that this metabolite does not contribute to efficacy.

[Slide]

The pharmacokinetics of voriconazole are non-linear due to saturation of metabolism. Increasing dose will result in a disproportionate increase in exposure. For example, our pharmacokinetic model predicts an average 2.5-fold increase in exposure for a 1.5-fold increase in dose from, for example, 200 mg to 300 mg.

[Slide]

Without a loading dose it takes approximately six days to reach steady state. The clinical program used a loading dose regimen of two doses of either 400 mg or 6 mg/kg I.V. twice on the first day and then followed with a maintenance dose.

On this slide, here, you can see the concentration time curve for the first day using the 400 mg twice daily oral dosing, loading dose regimen on day one and then the concentration time curve seen on day ten when this treatment is followed with 200 mg twice daily. As you can see, concentrations that are close to those seen in steady state are reached on day one, and we think this is important in acutely life-threatening disease such as invasive aspergillosis.

[Slide]

We have also investigated factors influencing voriconazole exposure in more than 1000 healthy volunteers. We have also analyzed over 3000 plasma samples from more than 1000 patients. Within subject variability is low, however, between subject variability is high, with a coefficient of variation close to 100 percent. I will discuss specific factors influencing this between subject variability as listed on this slide.

Genotype distribution varies according to race and contributes significantly to the observed variability of voriconazole exposure in healthy volunteers, where we studied this specifically. After accounting for genotype, race did not further contribute to variability in the Phase I population.

In patients we did not determine genotype, but subdividing the clinical trial population by race showed a continuous and overlapping distribution of plasma concentrations. Furthermore, there was no difference seen in the safety profile across the various subgroups of our patients. However, I want to point out that the majority of our patients treated were actually Caucasian. Based on the data we have, we do not recommend dose adjustment based on race or genotype.

[Slide]

We observed higher exposure in female and elderly volunteers in our Phase I program. Analysis of the patient population again showed widely overlapping exposure and no difference in the safety profile. We concluded that no dosage adjustment is needed according to age or gender.

[Slide]

We also studied the kinetics of voriconazole in children. In a single dose and a multiple dose study including pediatric patients at risk for fungal infections, we observed somewhat higher metabolic capacity in children from two years to under 12 years of age. We, therefore, recommend using the higher maintenance dose of 4 mg/kg as the usual dose in this population.

[Slide]

Body weight has some influence on exposure and when using oral dosing we recommend halving the maintenance dose for patients under 40 kg of weight. This recommendation was also used throughout our Phase III program and resulted in similar exposure in the two groups.

[Slide]

Because of the hepatic metabolism, hepatic impairment has a major effect on voriconazole exposure. Guided by the results of a multiple dose study in patients with Child's A or B cirrhosis, we recommend halving the maintenance dose in such patients.

[Slide]

Renal impairment has no effect on voriconazole exposure as such, however, in the I.V. formulation we use a substituted SBECD, as our excipient for the solubilization of voriconazole. The SBECD is excreted exclusively renally and accumulates with renal dysfunction. We, therefore, recommend oral therapy for patients with a serum creatinine over 2.5 mg/dL and, again, this was a recommendation that was used in our clinical program.

[Slide]

I would now like to turn to drug-drug interactions. These were evaluated in 19 studies with 365 volunteers, assessing the effects of nine drugs on voriconazole and the effects of voriconazole on 11 medications. The resulting data were analyzed with a view to whether voriconazole exposure with that of the concurrent medication would change significantly.

From these analyses we concluded that some drugs should be contraindicated; some would require dose adjustment of either voriconazole or the concurrent medication; and for some we could rule out an interaction. I will describe each group of these interactions in turn.

Rifampin is contraindicated when using voriconazole as even a dose decrease of voriconazole could not overcome the enzyme induction due to rifampin. This is also predicted to occur with the use of long-acting barbiturates and carbamazepine.

The interaction with sirolimus results in an 11-fold increase of exposure to sirolimus and co-administration is, therefore, contraindicated. Concomitant administration of voriconazole with terfenadine and the other medications shown here is contraindicated because increased plasma exposure to these medications could result in Q-T prolongation. Ergot alkaloids are contraindicated as voriconazole is predicted to increase exposure to this concurrent medication significantly.

[Slide]

Because of enzyme induction, when voriconazole is used concurrently with phenytoin or rifabutin the dose of voriconazole needs to be increased to 5 mg/kg I.V. or 400 mg orally to achieve adequate exposure. For rifabutin toxicity complete blood counts should be monitored, as is routine, and for phenytoin, phenytoin serum concentrations should be monitored, as is also routine.

[Slide]

Voriconazole acts as a P450 inhibitor and as exposure to the drugs listed on this slide increases, we recommend monitoring of either the plasma level of the concurrent medication or the biologic effect to avoid overexposure to these medications. For some drugs, listed here, the data from our interaction studies provided guidance for dose adjustments of the concurrent medication, as listed. We have included detailed guidance on monitoring and on these dose adjustments in our proposed package insert.

[Slide]

Voriconazole exposure is not affected by co-administration of 3A4 inhibitors such as erythromycin and indinavir. Similarly, non-specific P450 inhibition with cimetidine has no significant effect on voriconazole exposure. This absence of effect of P450 inhibitors on voriconazole exposure is likely related to the multiple pathways open to the metabolism of voriconazole. Likewise, there was no interaction with drugs not metabolized by the P450 system such as ranitidine, digoxin or mycophenolate. The ranitidine data also indicate that gastric pH has no influence of bioavailability of voriconazole. For the drugs on this slide then, dose adjustment is not necessary for voriconazole or the concurrent medication and no special monitoring is recommended.

[Slide]

In summary then, the pharmacokinetic behavior of voriconazole is described by rapid and consistent absorption with high bioavailability. The high volume of distribution and high tissue concentrations observed indicate extensive tissue distribution. There is non-linear elimination, primarily by metabolism via the P450 enzyme system. In hepatic cirrhosis this leads to increased exposure, requiring dose adjustment. The metabolic drug interactions are well characterized. Our in vitro data indicate that voriconazole is actually less potent as an inhibitor of the most important cytochrome, 3A4, than itraconazole or ketoconazole for example. The data we have allow us to provide specific guidance on managing the interactions and we, thus, feel that they are manageable in the population targeted for the use of voriconazole.

[Slide]

Before Dr. Boucher discusses the outcome of our efficacy studies, I want to briefly highlight our rationale for selecting the doses used in the clinical program. In considering dose selection for antifungal agents there are a variety of factors that play a role. I have already highlighted the variability around voriconazole exposure.

Other factors of importance include the variability in pathogen susceptibility. Minimum inhibitory concentrations for clinical isolates were mostly well below 0.1 mcg/mL for Candida and other yeasts and below 1 mcg/mL for Aspergillus. However, they vary greatly between different isolates. Lastly, host factors will be of critical importance in this disease. Patients who do not recover their own immune competence generally experience much poorer outcome following fungal infections.

[Slide]

When deciding on the dose for the clinical studies we considered these factors. Because of the severity of the disease and the extremely poor outcome in patients suffering from invasive fungal infections, as outlined by Dr. Patterson earlier, we intentionally targeted the upper end of the dose-response curve to achieve maximum benefit for patients.

We further aimed to reach plasma concentrations that are above MIC for the common pathogens. We would expect this to result in tissue concentrations above MIC as well, and we realize that these are likely to be a more relevant parameter, however, it is much more difficult to measure.

[Slide]

These considerations led to the regimen as shown on this slide. A loading regimen as an I.V. is 2 doses of 6 mg/kg or orally 400 mg 12 hours apart on the first day. For serious Candida infections and empirical therapy we then recommend a maintenance dose of 3 mg/kg I.V. every 12 hours or 200 mg orally.

For the invasive Aspergillus infections and molds, highlighted with MICs as somewhat higher than for yeasts, we recommend an initial maintenance dose of 4 mg/kg I.V. but then also a switch to 200 mg twice daily. Based on the clinical response, investigators were allowed to escalate dose to 4 mg/kg I.V. or 300 mg orally across our program.

I will now hand it over to Dr. Helen Boucher, one of the lead clinicians closely involved in the program, to discuss the efficacy results.

Efficacy

DR. BOUCHER: Good morning.

[Slide]

My name is Helen Boucher, and I am an infectious disease physician for Pfizer Global Research and Development. Thank you for inviting me to present the efficacy data for the voriconazole clinical program.

[Slide]

As Dr. Baildon mentioned, we conducted an extensive clinical program investigating the efficacy and safety of voriconazole in the treatment of invasive fungal infections.

Most of this morning's presentation will be devoted to the results of the comparative controlled trials in acute invasive aspergillosis and empirical therapy. We will also present supportive efficacy data from the non-comparative aspergillosis study, the contemporaneous historical control study, as well as data on the efficacy of voriconazole in the treatment of infections due to emerging pathogens, including Scedosporium and Fusarium as well as Candida infections. This will provide a framework for an in-depth review of the data from our empirical therapy study.

[Slide]

The global comparative aspergillosis study, a large comparative study in the primary treatment of acute invasive aspergillosis, was completed in April of this year. The goal of this study was to evaluate the efficacy, safety and tolerability of voriconazole compared to conventional amphotericin B in the primary treatment of acute invasive aspergillosis in immunocompromised patients.

[Slide]

Before discussing the design I will provide some history of this challenging large global trial. The global comparative aspergillosis study began with two identical protocols in 1997, study 602 led by the United States and study 307 led by the EORTC in Europe.

From early in the development of these protocols there was a prospectively defined umbrella analysis of the combined interim data. In October of 2000, a consensus recommendation was made by the U.S. investigative group, the EORTC and the sponsor to close enrollment in each of the two studies due to changes in medical practice and choice of comparative agents. This recommendation was shared with FDA and the European regulators.

[Slide]

As I mentioned, the goal of the study was to examine primary therapy for acute invasive aspergillosis in a severely ill patient population. This is shown here in the major inclusion criteria. This was a study of primary therapy. So, prior antifungal therapy was allowed for less than 96 hours. Patients were immunocompromised as a result of malignancy, chemotherapy or transplantation. Patients were required to have definite or probable acute invasive aspergillosis based on modified MSG or EORTC criteria. Again, the goal was to examine the efficacy in the treatment of definite or probable acute invasive aspergillosis in immunocompromised patients.

[Slide]

The statistical plan was developed in order to fully power the study to properly compare the two treatment strategies. A 50 percent response rate was assumed based on prior studies. We intended at least a 90 percent power to exclude a difference of minus 20 percent in success according to the data review assessment at week 12.

The sample size estimate assumed a 25 percent exclusion in the modified intention to treat population. So, a total sample size of 368 patients was required to enroll the necessary 276 patients. Patients were stratified using a central randomization system according to three factors, site of infection, underlying disease and neutrophil count.

[Slide]

A great deal of time was spent on the design of these studies and some of the issues are listed here. High dose amphotericin B has historically been the standard and only treatment indicated for the primary treatment of acute invasive aspergillosis. It is only available in an intravenous formulation and is not well tolerated.

Newer therapies, including itraconazole and lipid formulations of amphotericin which have less associated toxicity, while licensed only for salvage therapy, were being used more and more in the treatment of acute invasive aspergillosis. Our investigative groups, thus, developed the concept of following conventional amphotericin B with other licensed antifungal therapy, or OLAT, ultimately allowing the comparison of treatment strategies for voriconazole as a start-up regimen compared to conventional amphotericin B as a start-up regimen. This also provided the ability to assess response at a fixed time point, here, 12 weeks.

[Slide]

This diagram shows how patients progressed through the study. I would like to take a few moments to discuss details of the study design. All patients initially received their randomized therapy. For voriconazole patients received standard loading doses followed by 4 mg/kg I.V. every 12 hours for at least seven days. They then had the option of oral dosing at 200 mg twice a day.

In the conventional amphotericin B arm patients received a minimum dose of 1 mg/kg/day which was to continue for 14 days. Patients could stay on the randomized therapy through the study. Patients also had the option, at the discretion of the investigator, to switch to other licensed antifungal therapies if they developed problems with toxicity or lack of efficacy. For a few patients it was necessary to withdraw from randomized therapy, for example due to withdrawal of consent.

[Slide]

This diagram shows progress to week 12. Again, the major goal was to ensure that patients could remain on the study for the entire 12 weeks and be followed for as long as possible. So, again, patients could start on randomized therapy and continue through to week 12 on their randomized therapy. For these patients the assessment of response at week 12 is the same as the assessment at the end of randomized therapy.

For our patients who switched to other licensed antifungal therapy, they had an assessment of success at the time of the switch on a randomized therapy and continued in the study on their other licensed antifungal therapy through week 12, when they had a second assessment of success at week 12.

For the few patients who withdrew from the study, for instance due to withdrawal of consent, they had their assessment of success at the time of withdrawal. Even though these few patients withdrew, we were able to collect survival data through day 84 for all patients.

[Slide]

This led to the three endpoints in our trial, which are shown here. First, success in a test for non-inferiority at week 12, the primary efficacy endpoint. This allows comparison of the two treatment strategies.

Second, success in a test for superiority at the end of randomized therapy. This was intended to allow direct comparison of drug to drug, voriconazole with amphotericin B, and was a secondary efficacy endpoint. Third, survival through day 84 was a secondary efficacy endpoint.

[Slide]

Because of the nature of the treatment comparison, only an open design was possible. Therefore, we convened a data review committee and used a central randomization system in order to minimize bias. The data review committee was a global effort of leading experts in the field of antifungal therapy and included infectious disease physicians, hematologists, oncologists, as well as radiologists. We understood that much of the assessment of certainty of diagnosis, as well as assessment of outcome, was based on radiology so we had radiologists participate in the data review committee process. The DRC had two subgroups, a U.S. subgroup selected by the sponsor and a European subgroup selected by the EORTC.

[Slide]

In their review, the data review committee utilized mycology reports, clinical evaluations, investigator responses, as well as digitized radiology studies in a blinded review in order to assess the certainly of infection at baseline, the outcome at the end of randomized therapy, the outcome at week 12 and the cause of death relative to fungal infection in the patients who died.

There was a standard operating procedure that involved the exchange of cases between the DRC subgroups in order to ensure consistency. In all, approximately 20 percent of the cases were exchanged. Review was done after the first 37 cases were exchanged and resulted in 92 percent major agreement, that is, in success or failure. This process and these results were shared at last year's ICAAC by Dr. Patterson and colleagues.

[Slide]

Here we see an example of the classic halo sign. As I mentioned, diagnosis, particularly radiographic diagnosis, was an important element of our study design. I would like to review some key criteria before going on to discuss the results.

This high resolution chest CT is of a 45-year old man with relapse of acute myelogenous leukemia, from our study 602. He was status post allogeneic bone marrow transplantation two months prior to the study and had complications with graft-versus-host disease in the skin and liver.

Here we see his CT scan, a nodular infiltrate with the surrounding opacities in the characteristic halo sign. This patient also had a bronchoalveolar lavage positive for Aspergillus fumigatus. So, he met the criteria for definite acute invasive aspergillosis.

This same patient had had a negative BAO or had not had a bronchoalveolar lavage because of his specific host factors, the allogeneic transplantation. He would meet the criteria for probably invasive aspergillosis based on the halo sign.

[Slide]

I will now present the results of the study.

[Slide]

This table illustrates some of the challenges in conducting a comparative trial in aspergillosis. As illustrated here, of the 199 centers approved to enroll patients, only 95 contributed to the 392 patients enrolled over the 4 years of the study.

[Slide]

This chart displays disposition of patients in the global comparative aspergillosis study. And, 392 patients were enrolled and 11 received no therapy. For two patients investigators had difficulty in assessing the electronic randomization system and inappropriately allocated patients to voriconazole, leaving 194 voriconazole and 85 amphotericin B patients in the intention-to-treat population. For 50 voriconazole and 50 amphotericin B patients the data review committee was unable to confirm the presence of invasive aspergillosis, definite or probable invasive aspergillosis. This leaves 144 voriconazole and 133 amphotericin B patients in the modified intention-to-treat population, a primary efficacy population. In all, 277 patients were assessed by the data review committee as having definite or probable invasive aspergillosis. So, the target of the umbrella analysis was achieved.

[Slide]

Looking at demographic characteristics, this chart shows that patients were well matched according to age, sex and race.

[Slide]

This table presents the baseline characteristics according to the three stratification factors. As we see at the top, most patients had pulmonary aspergillosis, and 25 voriconazole and 21 amphotericin B patients had extrapulmonary aspergillosis.

In terms of underlying disease, approximately 25 percent of the patients in each arm had prior allogeneic bone marrow transplantation. This includes 27 voriconazole and 20 amphotericin B patients with documented graft-versus-host disease. The majority of our patients fell into this middle category, the patients with autologous transplantation or other underlying neoplasms, in this case mostly leukemia.

For the third stratification variable, neutrophil status, patients were well balanced, with 63 and 60 neutropenic patients in each treatment arm. Therefore, patients were well balanced according to each of the three stratification variables.

[Slide]

Looking at certainty of infection which could not be stratified at randomization as it was established by the data review committee, we see in this table that there were more definite infections on the voriconazole treatment arm and more probable on the amphotericin B arm.

Also, there was an imbalance in the individual sub-studies, with more definite infections in the U.S. led study 602, than in the EORTC led study 307.

[Slide]

Before presenting the results, I would just like to remind you of the three endpoints -- success at week 12, success at the end of randomized therapy, and survival through day 84.

[Slide]

Now I will review the actual progress of our patients over time during the study. This stack-bar chart shows the progress of all the 144 patients in the voriconazole arm, the modified intention-to-treat population, primary efficacy population through each of the 84 days of the study.

The height of the blue bars represents the number of patients receiving voriconazole as randomized therapy. The median duration of therapy was 77 days. At week 12, 62 patients remained on voriconazole.

The bars above the blue bars, these red dotted bars which are a little bit difficult to see, represent the patients receiving other licensed antifungal therapy, those patients who switched during the study. There were 22 patients receiving other licensed antifungal therapy at week 12.

The green bars represent the patients in the post treatment follow-up period. These are patients who had completed therapy and had a response assessed but were still being followed in the study. There were 13 patients in the follow-up period at week 12.

The yellow bars represent those few patients who withdrew from the study, for instance due to withdrawal of consent. There were five patients who had withdrawn at week 12.

At the top of the chart, in the grey bars, are the patients who had died, 42 patients died by day 84.

[Slide]

This stack bar chart shows the progress of all the 133 patients in the modified intention-to-treat population on the amphotericin B arm through each of the 84 days of the study.

Looking at initial randomized therapy, here in red -- again, these bars represent our patients receiving amphotericin B and we note a contrast to that seen for voriconazole. On the amphotericin B arm only two patients remained receiving conventional amphotericin B as randomized therapy at week 12. It is important to note that the median daily dose of the amphotericin B was 1 mg/kg/day during the first two weeks and the average duration of therapy was 11 days.

Again, the red dotted bars represent the patients receiving other licensed antifungal therapy, and there were 57 patients receiving other licensed antifungal therapy at week 12.

The green bars represent the patients in the post treatment follow-up period. There were 11 by week 12.

In the yellow are represented the patients who withdrew from the study and there were seven patients who had withdrawn at week 12.

The grey bars at the top represent the patients who had died, and 56 patients had died by day 84.

[Slide]

I will now present the results in each of the three endpoints, first, addressing success as assessed by the data review committee at week 12, a primary efficacy endpoint.

[Slide]

This bar chart shows success in each treatment arm, and 52.8 percent of patients initially treated with voriconazole had success as assessed by the data review committee at week 12 compared to 31.6 of those initially treated with amphotericin B. This difference is 21.2 percent and 2.8 percent when adjusted for protocol. In both cases 95 percent confidence intervals of the difference do not include zero.

[Slide]

The chart on the left again displays success in the umbrella analysis. Recall that the treatment difference seen was 21.2 percent.

The bars on the right show data review committee assessment of success at week 12 for each individual study. The treatment difference observed in the umbrella analysis is consistent in both the U.S. led study 602 and the EORTC led study 307. In both studies the 95 percent confidence interval does not include zero and, therefore, it meets the statistical criteria for superiority.

[Slide]

I would like to take a minute to look further at data review committee assessed success at week 12, the primary efficacy endpoint according to the stratification factors and certainty of infection.

This chart displays the treatment difference along with the associated 95 percent confidence interval in the overall umbrella analysis at the top, and then in our stratification variables. On the right is the success presented in percent.

If we look first at site of infection, you see that whether patients had pulmonary or extrapulmonary disease this treatment difference of just over 20 percent is maintained. Among the patients with extrapulmonary aspergillosis were ten patients who had documented CNS infection, five in each treatment arm. Two of the five voriconazole patients had success at week 12, while none of those treated with amphotericin B had success.

In terms of underlying disease, whether patients had prior allogeneic transplantation, autologous transplantation of leukemia or other cause of immunocompromise, again we see that the absolute treatment difference of over 20 percent is maintained in each arm.

Looking at neutrophil status, we see that whether patients were neutropenic or not neutropenic at baseline the treatment difference was maintained.

Turning back to certainty of infection, we note that whether patients had definite aspergillosis or probably invasive aspergillosis that treatment difference was maintained.

So to summarize for our primary efficacy endpoint, the data review committee assessed success at week 12 the treatment difference between voriconazole and amphotericin B to be maintained in both individual protocols, three stratification groups, and in patients with definite and probable infection.

[Slide]

I will now turn to the second efficacy endpoint, success at the end of randomized therapy. Here success was seen in 53.5 percent of voriconazole and 21.8 of amphotericin B patients. Note that the median duration of randomized therapy was 77 days for voriconazole and just 11 days for amphotericin B. This reflects the well-known problems of maintaining patients on conventional amphotericin B due to its associated toxicity. This difference is 31.7 percent and 31.9 percent when adjusted for protocol. In both cases 95 percent the confidence interval does not include or cross zero. The results were consistent in each individual substudy.

[Slide]

The third endpoint was survival through day 84. Recall that survival data was collected for every patient through day 84.

[Slide]

Looking at deaths as an efficacy endpoint in the modified intention-to-treat population, 71 percent of the patients on the voriconazole arm and 58 percent of those on the amphotericin B arm survived through day 84. This is an absolute 13 percent difference in survival.

This Kaplan-Meier plot demonstrates an early separation which is maintained through day 84 and the associated hazard ratio of 0.6 and 95 percent confidence interval of 0.4 to 0.89.

[Slide]

As I mentioned earlier, the data review committee was asked to address the cause of death, specifically addressing the impact of Aspergillus for all patients who died while in this study up to day 84. This table shows that more than twice the number of patients treated initially with amphotericin B died due to aspergillosis than did voriconazole patients.

[Slide]

In summary, this large comparative study demonstrates, both at 12 weeks and end of randomized therapy, that the data review committee assessed success as higher in voriconazole-treated patients than for amphotericin B patients.

The treatment difference is maintained whether we look at stratification groups or certainty of infection, and includes success for voriconazole therapy in patients with central nervous system infection, allogeneic bone marrow transplantation and documented graft-versus-host disease. Treatment difference is also maintained for each individual study and the criteria for non-inferiority at week 12 and superiority at the end of randomized therapy are met. We further observed a 13 percent absolute survival benefit at day 84.

[Slide]

I would now like to return briefly to our other data supporting the use of voriconazole treatment in acute invasive aspergillosis.

This table displays outcome in a non-comparative aspergillosis study or earlier European study. In this study success was seen in approximately 50 percent of patients in the intention-to-treat and expert evaluable populations. Here, again, success included patients with central nervous system infections as well as allogeneic transplantation.

I would like to draw your attention to this column on the right. These 50 patients from the per protocol population received voriconazole as primary therapy following five days or less of prior antifungal therapy, and this is the group that was used in our contemporaneous historical control study. In this study, these 50 patients from the non-comparative aspergillosis study were matched with patients from U.S. and European centers who were treated during the same time period.

[Slide]

This chart shows success in the non-comparative aspergillosis study and its historical control. At the top we see that success was seen in 26/50, or 52 percent, of the voriconazole-treated patients in the matched primary population, those 50 that we just saw from the non-comparative study. This is compared to success in 23/92 patients, 25 percent of the patients, from the historical control population. These global response data are actually consistent with what we later saw in the global comparative aspergillosis study.

[Slide]

In summary then, voriconazole has been used to treat 476 patients with documented invasive aspergillosis from our clinical program, most of whom were severely immunocompromised and critically ill. We saw superior efficacy and a survival benefit in our large comparative study. This was consistent with what was seen in our earlier non-comparative aspergillosis study which compared favorably with the historical control study.

[Slide]

Before presenting the efficacy data in empirical therapy I would like to briefly highlight other areas of therapy with voriconazole, first in the treatment of emerging pathogens.

[Slide]

Infections due to the more rare fungal pathogens, such as Scedosporium and Fusarium, are increasingly recognized, especially in the era of the increased use of non-myeloablative transplantation and solid organ transplantation. These organisms are poorly susceptible to available antifungal agents, as shown on the MIC table drawn from the medical literature.

Clinically, these infections are associated with high morbidity and mortality. For Scedosporium, mortality reaches 87 percent in the published literature. For Fusarium, attributable mortality approaches 80 percent and is largely dependent on the recovery of neutrophil function.

[Slide]

Across our clinical program we have treated a total of 35 patients with infections due to Scedosporium. Success was seen in 16/27 patients with Scedosporium apiospermum. This includes patients with S. abortia. As expected, success was less, seen in two of eight patients with Scedosporium prolificans. For Scedosporium, successes include patients with central nervous system infection as well as disseminated and pulmonary infections. In the patients with central nervous system infection, these 13 patients at the top, it is important to note that 10 of these 13 patients received voriconazole for greater than 90 days and several of our successes here included children.

[Slide]

In Fusarium, a total of 15 patients from across our clinical program have been treated with voriconazole and six of these 15 had success while on voriconazole therapy. In Fusarium success was seen in patients with eye, sinus, skin and disseminated infections.

[Slide]

Turning now to Candida infections, voriconazole has been studied in esophageal candidiasis as well as in patients with invasive Candida infections.

[Slide]

This chart shows success in our comparative, double-blind, double-dummy study of oral voriconazole versus oral fluconazole in the treatment of endoscopically and mycologically proven esophageal candidiasis. The study was conducted in Europe predominantly.

As shown here, endoscopic success was seen in 98.3 percent of voriconazole and 95 percent of fluconazole-treated patients. The difference is 3.2 percent and the lower limit of the 95 percent confidence interval is minus 1.08, which falls well within the prespecified minus 15 percent limit. Endoscopic successes in this study included patients with advanced AIDS as defined by CD4 counts of less than 50.

I would also like to note here that the side effects seen with voriconazole therapy in this study are consistent with those of other azoles but were seen more frequently in the voriconazole treatment arm. Dr. Baildon will discuss this in greater detail during his safety presentation.

[Slide]

This table shows success in patients with non-esophageal Candida infections who received voriconazole as salvage therapy following greater than five days of prior systemic antifungal therapy. This group includes patients enrolled in clinical trials as well as those from our compassionate use program.

In this salvage population success was seen in 22.43 or 51 percent of the patients. Importantly, most of these patients, 36 as shown here in the top row, received voriconazole following documented efficacy failure to prior antifungal therapy. This includes 26 patients who had received prior fluconazole. Successes here also include patients with infections due to non-albicans species.

[Slide]

In summary, voriconazole has demonstrated efficacy in treating Candida infections. Voriconazole was equivalent to fluconazole in endoscopically and mycologically proven esophageal candidiasis. We have seen reassuring efficacy in cases of patients with invasive Candida infections, particularly in those who failed other therapies including fluconazole, as well as infections due to non-albicans species. This confirms our earlier in vitro and in vivo work where we see potency in clinical isolates and efficacy in standard animal models of disseminated candidiasis, including published studies of disseminated Candida krusei.

Our comparative study of voriconazole in the primary treatment of candidemia in non-neutropenic patients, study 608, is ongoing. As of this week, 256 of our targeted 426 patients have enrolled in that study.

[Slide]

Having presented the data on voriconazole's efficacy in treating documented infections due to yeast and mold, I would now like to turn to a discussion of empirical therapy.

[Slide]

The practice of empirical antifungal therapy is based on historical data from Dr. Pizzo, in the EORTC, who demonstrated that treatment intervention with conventional amphotericin B reduced the number of documented breakthrough invasive fungal infections. More recently, comparative study MSG32 showed that liposomal amphotericin B was as effective as conventional amphotericin B in a composite endpoint and formed the basis for our study.

[Slide]

We conducted a large randomized study in 73 international centers in collaboration with the mycosis study group. The primary goal was to evaluate the efficacy of voriconazole compared to liposomal amphotericin B in a composite endpoint.

In this endpoint success required meeting each of the five criteria listed here: no documented breakthrough of invasive fungal infections during therapy or for seven days following end of therapy; survival for seven days following the end of therapy; no premature discontinuation of therapy due to toxicity or lack of efficacy; sustained defervescence for 48 hours prior to the recovery from neutropenia; and a successful response to treatment of any diagnosed baseline invasive fungal infections.

[Slide]

Patients were required to have persistent fever and neutropenia in the setting of 96 hours or parenteral empirical antibiotics. In this study an additional criterion was that patients had to have an absolute neutrophil count of less than 250 in the 24 hours prior to randomization.

[Slide]

The statistical plan was developed to compare the two treatments in a previously published composite endpoint. The statistical calculations assumed a 50 percent response rate based on the prior study, MSG32. Our intention was to have at least 80 percent power to exclude a difference in success in the composite endpoint of minus 10 percent. The sample size estimate assumed a 10 percent exclusion. So a total sample size of 866 patients was necessary to enroll 786 patients in the modified intention-to-treat population.

[Slide]

Patients were stratified at randomization by a central randomization system, first according to risk of developing invasive fungal infection. High risk patients include those with allogeneic transplantation or relapsed leukemia. Moderate risk patients are those with newly diagnosed leukemia, autologous transplantation or other underlying neoplasm. Patients were also stratified according to the presence or absence of prior systemic antifungal prophylaxis.

[Slide]

A blinded data review committee reviewed all potential invasive fungal infections. Our experts included infectious disease specialists as well as hematologists and oncologists. They used mycology and radiology reports, clinical assessments and investigator responses in order to assess the presence of infection; the type of infection, baseline versus breakthrough; the certainty of infection; as well as the global response at the end of therapy, and a standard operating procedure was followed when assessing all cases.

[Slide]

Now I will present the results from our empirical therapy study.

[Slide]

As shown here, a total of 871 patients were randomized; 22 patients did not receive therapy, leaving 421 voriconazole and 428 liposomal amphotericin B patients in this safety population. An additional 12 patients, six in each arm, did not have sufficient information available to confirm the investigator's assessment. We, therefore, have 415 voriconazole and 422 liposomal amphotericin B patients in the modified intention-to-treat population, a primary efficacy population. In all, 837 patients made up the modified intention-to-treat population so the target of 786 was achieved.

[Slide]

Looking at demographics, patients were well matched according to age, sex and race.

[Slide]

Patients were also well matched according to stratification factors. Of note, seen in the middle row of the table, over 50 percent of patients in each arm received prior systemic antifungal prophylaxis, mostly with fluconazole. This is more than seen in prior studies. Also, approximately 50 percent of patients in each arm underwent bone marrow transplantation, including 18 percent in each arm who had allogeneic transplantation.

[Slide]

Median duration of therapy was seven days in each arm. Importantly, 91 voriconazole patients were able to receive oral therapy in addition to their intravenous therapy.

[Slide]

This table displays the results in our primary efficacy analysis, success in the composite endpoint. Twenty-six percent of the voriconazole and 31 percent of liposomal amphotericin B patients had success. The difference is minus 4.5 percent, and the lower limit of the 95 percent confidence interval is minus 10.6 percent. This is outside the prespecified minus 10 percent limit.

When these results are adjusted for the risk of developing fungal infections, the presence or absence of systemic antifungal prophylaxis and the duration of neutropenia at baseline, success is seen, here at the bottom, in 23.7 percent of voriconazole and 30.1 percent of liposomal amphotericin B patients. In this stratified analysis, the primary efficacy analysis, the difference is minus 6.1 percent and, again, the lower limit of the 95 percent confidence interval is minus 12 percent which is outside the prespecified minus 10 percent.

[Slide]

I would now like to present the results in each of the five components of the primary efficacy endpoint, first examining breakthrough fungal infections.

[Slide]

Eight voriconazole and 21 liposomal amphotericin B patients developed breakthrough invasive fungal infections. These were definite or probable invasive fungal infections as assessed by the blinded data review committee and diagnosed greater than 24 hours following the first dose of study drug.

[Slide]

Here we see a breakdown of these fungal infections and I would like to spend a moment on these. First and most notably at the top, while four voriconazole patients developed breakthrough pulmonary aspergillosis, 13 liposomal amphotericin B patients developed breakthrough aspergillosis, nine involving the lungs; two involving the sinuses; one involving the CNS and skin and another which was disseminated.

Two voriconazole patients developed breakthrough Candida infections while six liposomal amphotericin B patients developed breakthrough candidemia. While two voriconazole patients developed infections due to Zygomycetes, two liposomal amphotericin B patients developed infections due to dematiaceous molds.

[Slide]

Turning to survival, this criterion required survival for at least seven days following the end of therapy. Thirty-three voriconazole and 25 liposomal amphotericin B patients died within seven days of the end of therapy.

[Slide]

This table lists the investigator reported cause of death for all patients who died. I would just like to note that the number of causes is greater than the number of patients as investigators frequently listed more than one cause of death per patient.

In the voriconazole arm more patients died due to progression of malignancy, sepsis and pneumonia. In the liposomal amphotericin B arm patients died due to progression of malignancy, sepsis and hemorrhage.

[Slide]

Looking at discontinuation from study medication due to toxicity or lack of efficacy prior to recovery from neutropenia, we see on this table 41 voriconazole and 28 liposomal amphotericin B patients discontinued prematurely.

[Slide]

There was no difference between treatment arms in discontinuation due to toxicity, as you see in the top row, with 19 voriconazole and 23 liposomal amphotericin B patients discontinuing due to toxicity. It is important to note that this endpoint included patients who permanently discontinued only. It does not include the seven voriconazole and 52 liposomal amphotericin B patients who discontinued temporarily or had dose reductions during therapy.

[Slide]

On the other hand, there was a difference in discontinuation due to lack of efficacy in this open-label study, with 22 voriconazole versus five liposomal amphotericin B patients discontinuing due to lack of efficacy. This difference is primarily due to more patients, 14, discontinuing due to fever alone. These included three patients with bacterial infections; one with TTP; and two with no documented fever.

[Slide]

The fourth component of the primary endpoint was defervescence during neutropenia. Success here required having a temperature of less than 38 for 48 hours prior to recovery from neutropenia.

[Slide]

Here we see the number of patients who met the defervescence criterion. Thirty-three percent of voriconazole and 36 percent of liposomal amphotericin B patients met this criterion. This is less than expected and may be related to the shorter than expected duration of neutropenia seen during our study. The median time to recovery from neutropenia after randomization was 4.8 days in the voriconazole arm and 5.4 days in the liposomal amphotericin B arm. This is less than that seen for liposomal amphotericin B, 8 days, in the prior MSG32 study. So, in our study there was less time during the study in which to defervesce.

[Slide]

The last component of the endpoint is success in treatment of any diagnosed baseline invasive fungal infections, those diagnosed within 24 hours of randomization.

[Slide]

This table shows that six of 13 voriconazole and four of six liposomal amphotericin B patients had success following treatment of baseline fungal infections.

[Slide]

Returning then to our composite endpoint, this graph shows overall success at the top, expressed in treatment difference with the associated 95 percent confidence intervals. In both the raw and stratified analysis the lower limit extends beyond the prespecified minus 10 percent limit. This overall lower than expected success, approximately 30 percent rather than the expected 50 percent, is driven mostly by the failure to defervesce.

However, if we look at one clinically important secondary endpoint, the breakthrough of invasive fungal infections, we see that voriconazole was associated with greater success here. On the other hand, the survival and premature discontinuation data trend in the other direction. We learn that the deaths were more often due to progression of underlying disease, not fungal infections. Further, they occurred among the patients at moderate risk for developing invasive fungal infections.

[Slide]

This led us back to reexamine our stratification factors. When we look at breakthrough fungal infections according to risk we see that the prevention of breakthrough fungal infections was enhanced in the high risk strata, those at high risk of developing invasive fungal infections, allogeneic transplants and relapsed leukemia patients.

Here we see, just to recall, that eight voriconazole and 21 liposomal amphotericin B patients developed breakthrough invasive fungal infections. In the high risk group, shown here at the top, 1.4 percent of voriconazole and 9.2 percent of liposomal amphotericin B patients developed breakthrough invasive fungal infections. Among these, one voriconazole and nine liposomal amphotericin B patients had received prior prophylaxis, mostly with fluconazole. So, again, in this clinically important parameter prevention of breakthrough of invasive fungal infections voriconazole provided protection in the patients at high risk.

[Slide]

Looking at overall efficacy in our composite according to risk, in this unadjusted subgroup analysis the results for overall success in the high risk group are within the prespecified minus 10 percent bounds.

[Slide]

This graph again displays outcome for overall success at the top and for each of our five components in the high risk group, in yellow, and the moderate risk group, in blue. These results are unadjusted for the other stratification variables.

This table shows that the results are consistent in the high risk group stratum and suggest that voriconazole's empirical therapy may be associated with benefits in this population.

[Slide]

Turning briefly to toxicity, infusions were prospectively monitored with daily bedside infusion work sheets. This table shows several infusion-related reactions of interest in each treatment arm shown in the order of voriconazole frequency. In the voriconazole arm abnormal vision, an adverse event frequently associated with voriconazole, was seen in 22 percent of patients during infusion. This will be discussed in greater detail by Dr. Baildon during his safety presentation.

In the liposomal amphotericin B arm we saw the expected infusion-related reactions of chills and fever. We also saw some unexpected reactions including wheezing, tachypnea, chest pain and anaphylactoid reactions in a syndrome of acute infusion-related reactions that will be presented at ICAAC in December.

[Slide]

In conclusion, voriconazole has demonstrated superior efficacy and a survival benefit in the treatment of documented acute invasive aspergillosis, including high risk patients with prior allogeneic transplantation, central nervous system disease, as well as documented graft-versus-host disease.

We have also seen efficacy against infections due to emerging pathogens, including Scedosporium and Fusarium, as well as efficacy in Candida infections. These data suggest that voriconazole is an appropriate option for empirical antifungal therapy in patients with persistent fever and neutropenia, particularly those at high risk.

Thank you very much, and Dr. Baildon will now continue with the safety presentation.

Safety and Conclusions

DR. BAILDON: I would now like to turn to the safety and provide an overview of the investigations into the safety profile of voriconazole as experienced in over 3000 healthy volunteers and patients.

[Slide]

Therapeutic studies of voriconazole included 1493 patients, mainly drawn from the large comparative trials that you have seen in Dr. Boucher's presentation. Including compassionate patients, this provides an overall pooled population of 2090 patients. The numbers shown here for the comparative studies are the safety patient population in these trials, including all patients who received study drug.

[Slide]

We also have experience in 443 healthy volunteers who received only voriconazole, comparing that to placebo. This excludes the Phase I volunteers treated in drug interactions and various special populations.

[Slide]

The 172-year exposure seen in the therapeutic trial population allows us to characterize the voriconazole safety profile. Many of the compassionate patients included in the pooled population were treated for a long time, and we have experience now of more than 500 patients treated for longer than 12 weeks. Thirty-eight patients in total have received voriconazole for more than one year.

[Slide]

In considering the safety profile in the population, I would like to highlight that this is a very ill patient population, frequently severely immunocompromised and undergoing many invasive procedures such as, for example, bone marrow transplant as well as receiving many concurrent medications. The serious illness of these patients, the absence of a placebo control and the many concurrent medications make it more difficult to decipher what adverse events could be attributable to voriconazole as compared to those being related to some other medications received.

[Slide]

I will discuss the frequency of death and discontinuations, and then turn to adverse events as observed in the safety population. We have carefully evaluated the emerging safety profile throughout the development and have also investigated any other special issues arising, and I will address some of these as well.

[Slide]

The severe nature of the underlying illness in the patients studied is also illustrated by the mortality observed. On this slide deaths reported to 30 days post last dose are shown. There was no difference between the treatment groups in the esophageal candidiasis study, at the bottom of the slide, and Dr. Boucher already discussed the results of the empirical therapy study where more deaths occurred in the voriconazole arm. However, these were mostly related to progression of the underlying leukemia in these patients.

In the global comparative aspergillosis study treatment duration of initial randomized therapy varied greatly. So, the numbers on this slide reflect the variable length of follow-up and a Kaplan-Meier presentation is actually more informative.

[Slide]

When discussing the results of the global comparative aspergillosis study, Dr. Boucher outlined the observed mortality difference in the modified intent-to-treat population. This slide illustrates the same presentation for the safety population of all patients treated with either voriconazole or amphotericin B and then followed for 84 days.

As you can see, the difference in the safety population is of similar magnitude as in the modified intention-to-treat population, with an early separation of the curves that is maintained out to day 84.

[Slide]

The overall rate of discontinuations for safety reasons of 18 percent in the therapeutic studies is low in the context of this patient setting. There were more discontinuations in the voriconazole group compared to fluconazole in our esophageal candidiasis study, and the most common reason for discontinuation here was hepatic enzyme elevation.

In the global comparative aspergillosis study, however, more patients had to discontinue amphotericin B therapy than voriconazole therapy, and there the discontinuations were mostly related to renal function in the amphotericin B-treated patients. Please note that in this comparison I am comparing the two drugs as initial randomized therapy only.

[Slide]

Because of the difficulty in deciphering the safety profile of voriconazole in the background of this severely ill patient population, we have also carefully analyzed the adverse events seen in healthy volunteers who received voriconazole versus those receiving placebo.

Of the adverse events shown on this table, visual adverse events are the most prominent adverse event and were reported by 35 percent of volunteers on a background of 12 percent in the placebo group.

[Slide]

This slide shows adverse events sorted by frequency on the voriconazole arm as observed in the global comparative aspergillosis study. In this presentation I am comparing all events on the voriconazole initial randomized therapy arm with those seen on amphotericin B, followed by other licensed antifungal therapy. This results in a more comparable treatment exposure of around 1200 days in both populations.

Again, visual adverse events are the most common and occur more frequently on voriconazole than on the comparator arm. Nausea, vomiting and diarrhea occurred more frequently in the patients treated with amphotericin B or other licensed antifungal agents.

[Slide]

Turning to the ten most frequent adverse events in the empirical therapy study, it provides a similar picture. The frequency of abnormal vision seen in voriconazole-treated patients was 26 percent and chills were more frequent among the patients treated with liposomal amphotericin B.

[Slide]

In the double-blind esophageal candidiasis study abnormal vision was also reported as the most frequent adverse event in the voriconazole arm. Increases in alkaline phosphatase and other hepatic enzymes were also more common and, as I highlighted previously, resulted in more discontinuations in the study on voriconazole.

[Slide]

I would like now to turn to the special safety topics of visual disturbances, hepatic adverse events and skin reactions. As I mentioned before, we also investigated the other topics listed here.

[Slide]

The frequency of abnormal vision is constant across the clinical trial population and in health volunteers, around 25-35 percent. This is also seen in the large comparator studies. Visual disturbances occasionally lead to discontinuations, however, this is a rare consequence.

[Slide]

The descriptions patients provide when asked what they are experiencing are enhanced perception of light, blurred vision, photophobia or changes in color perception. The top category shown here is the most frequently encountered one.

[Slide]

The visual adverse events occur early on in treatment, and this slide depicts time to occurrence of the first event reported in patients. As you can see, after the first week rarely are new events reported by the patients. Each individual event also occurs early on during therapy, usually within the first half hour after initiation of infusion, and typically lasts about 30 minutes.

[Slide]

We have conducted extensive investigations into the mechanisms of these events, and identified reproducible changes in the electroretinogram of healthy volunteers. This indicates that the site of action is in the retina. To analyze reversibility of this underlying electrophysiological phenomenon for the visual disturbances, we conducted a multiple dose visual function study in healthy volunteers, as shown here.

This study used the high maintenance dose of 300 mg twice daily and volunteers received this dose for 28 days. Electroretinograms and other visual function tests were performed as outlined.

The electrophysiological results are summarized here and in my next slide. The Y axis depicts a negative amplitude of the a-wave after white stimulus, scotopic conditions. The a-wave measures photoreceptor activity. The presentation is of the range of values encountered, with a 25th and 76th percentile and the median as the line. The continuous line joins the means of each group.

As you can see, for the placebo patients there is basically no change throughout the treatment period up to day 29 and follow-up at day 43. For the voriconazole-treated patients there is an immediate decrease in amplitude at initiation of treatment, which is maintained throughout treatment and then reverts to normal 2 weeks later.

[Slide]

This slide illustrates the effect on the b-wave amplitude, again measured with white stimulus and scotopic conditions. The b-wave measures the activity of the inner nuclear layer of the retina. As you can see again, there is no change in the placebo-treated patients and on voriconazole there is an immediate decrease of amplitude on day one that is maintained throughout treatment and then reverts back to normal two weeks after discontinuation.

[Slide]

Visual acuity in this study also showed no difference between voriconazole-treated volunteers and those receiving placebo. Color vision and visual field tests suggest an impact of voriconazole on these, however, these also returned to normal after discontinuation.

[Slide]

In the esophageal candidiasis study, where patients were more ambulatory, we also evaluated visual function at baseline and at follow-up. Again as you can see, there is no difference between the treatment groups on visual acuity. Fundoscopy results in the patients also indicate no difference in changes between voriconazole and comparator treatment treated patients.

[Slide]

In our 24-month carcinogenicity rat study the decrease in the cell number of the peripheral retina was observed in female rats when compared to the control group. This difference was not observed in the central retina, nor was it observed in male rats.

In dogs, where we have been able to replicate the electrophysiological phenomenon, as I showed you in human volunteers, we saw no changes in the histopathology of the dog retinas or the visual pathways after 12-month of high dose treatment.

Our assessment of the visual disturbances is that these are functional changes during therapy without long-term sequelae. The underlying electrophysiological phenomenon is reversible upon discontinuation of treatment after one month of therapy. We suggest including appropriate warnings in the package insert, describing these events and highlighting that patients should avoid activities that could cause problems.

[Slide]

I would now like to turn to the hepatic adverse events. We consider hepatic enzyme elevations to be dose-limiting to the use of voriconazole.

[Slide]

In the multiple dose escalation study the usual loading dose of 6 mg/kg twice on the first day was followed by either 3 mg/kg I.V. and then 200 mg orally, 4 mg/kg I.V. and then 300 mg orally, or 5 mg/kg I.V. and then 400 mg orally. I.V. treatment was for 5 days and then was switched to oral. One of 7 volunteers at the middle dose group experienced an ALT elevation, whereas 5/14 volunteers experienced that in the highest dose group.

[Slide]

This slide shows the ALT time curve for the 6 volunteers who experienced a rise in ALT. The patient in green, S39, is a patient from the 3 mg/kg followed by 300 mg oral dose group. That is the highest dose group we currently recommend. The other colors are for the patients of the highest dose group used in this study. Reassuringly, all these elevations returned to normal within two weeks after the end of dosing, which was on day 14.

The results from this study led us to conclude that 4 mg/kg I.V. twice daily and 300 mg orally twice daily are the maximum tolerated dosages for voriconazole.

[Slide]

In the patient population we used a criterion of ALT elevation of three times the upper limit of normal to indicate clinically relevant abnormality, without regard to baseline. This was observed in about 12 percent of the therapeutic study population. However, when we look at the comparator study, the frequency that occurred was similar when compared to amphotericin B formulations but was somewhat higher when comparing to fluconazole.

A conservative algorithm to capture patients who discontinued is used here in counting any patient in this column who discontinued and had this ALT abnormality whether that was a primary reason for discontinuation or not. The percentage of patients discontinuing due to ALT elevations is low.

[Slide]

This slide uses a similar presentation for abnormal notal bilirubin, using a cut-off of 1.5 times the upper limit of normal to indicate clinically relevant abnormality. In this comparison the data comparing voriconazole to amphotericin B formulations or to fluconazole indicate no difference between comparative groups. Again, the percentage of discontinuations due to this is low.

[Slide]

We also investigated hepatic failures leading to deaths across the program. To date, we have received 26 such reports. Of these, 19 patients were treated with voriconazole and seven received amphotericin B formulations. In the comparative trials these patients were equally distributed between the two treatment groups. The global comparative aspergillosis was one comparator and empirical therapy study, five in each arm.

We have also asked an independent panel of hepatic experts, under the leadership of Dr. Maddrey who is in the sponsor section today, to review these cases in a blinded fashion. The deliberations of this panel were recently added to our submission. Of the cases, there were four patients treated with voriconazole where either the investigator or the sponsor could not rule out a potential association with study drug and I will discuss these cases in more detail.

[Slide]

This slides provides the relevant patient details, as well as the most conservative expert assessment of any one of the experts in our panel that we convened.

The first patient was treated for five days for overt aspergillosis. He then suffered an acute hypoglycemic event and voriconazole therapy was discontinued. On day 44 a liver biopsy was performed. Unfortunately, this resulted in intraperitoneal hemorrhage and the patient died. The biopsy showed cirrhosis, and the family later disclosed a history of alcohol abuse in this patient. The impaired liver function explains the high exposure observed in this patient.

The next patient was a 64-year old male with anemia and sepsis who had failed amphotericin B therapy for pulmonary aspergillosis. The patient was treated for 19 days and then discontinued because of hepatic enzyme elevation. The hepatic enzymes returned to normal but the patient continued to deteriorate and died on day 40.

The third patient suffered hepatic failure due to graft-versus-host disease after an allogeneic bone marrow transplant and received voriconazole for 34 days. The patient then deteriorated, suffered from CMV pneumonitis and died eight days post last dose.

The last patient is the only case for which one of the panel members in the blinded evaluation thought that it could probably be related to study drug, and the study drug was voriconazole in this case. The patient had responded well to voriconazole for central nervous system aspergillosis and was treated for 257 days. She then suffered an acute exacerbation of the underlying systemic lupus erythematosus and this flare involved an elevation of hepatic enzymes. Voriconazole was discontinued because of that hepatic enzyme elevation. The patient, however, continued to deteriorate further and died of hepatic failure and multi-organ failure on day 323, about 60 days after discontinuation of voriconazole.

[Slide]

From all the data reviewed, the expert panel concluded that while a contribution of voriconazole in some of these events could not be excluded, there are no features that would clearly point to a causal association. Our conclusion around the hepatic effects of voriconazole are that these are manageable for the patient population targeted for treatment. As a consequence of the potential for voriconazole to induce hepatic enzyme elevations, we recommend monitoring liver function, as is routinely done in this patient population. Patients who develop liver function abnormality should be monitored for signs of more severe hepatic injury and voriconazole should be discontinued if these are observed.

[Slide]

I would now like to turn to skin reactions observed in the clinical program.

[Slide]

Skin reactions in general and rashes in particular are described by investigators in a number of ways. For example, here is a list taken from our case record forms that would then code to rash in our evaluation, or similarly, at the bottom of the slide, to photosensitivity.

[Slide]

In each of the comparator studies the frequency of rash amongst the voriconazole-treated patients or amphotericin B formulations or fluconazole was similar across the comparator agents. The high rate seen in the population may reflect the extensive use of drugs such as penicillins and sulfa antibiotics. The patients also frequently received blood products and have underlying conditions such as graft-versus-host disease. Only a low percentage of patients discontinued while experiencing rash. Again, I used a conservative presentation here, any patient discontinuing while experiencing rash.

[Slide]

Photosensitivity was reported in 41 patients in the pooled population. Most of these events occurred in patients treated for long durations in the compassionate program or the extension protocols. None of the patients in the therapeutic trials discontinued with a photosensitivity reaction, and the voriconazole molecule as well as the anoxide metabolite do not absorb UV light at the spectrum where one would expect a photosensitivity reaction. Nevertheless, from the clinical reports received to date we cannot exclude a contribution of voriconazole to these events.

We have also received two reports of Stevens-Johnson syndrome. These were not assessed as related to voriconazole treatment by the investigator and both abated during voriconazole treatment.

[Slide]

Our conclusions then around skin reactions are listed on this slide. While the frequency of rash is similar between voriconazole-treated patients and those receiving fluconazole or amphotericin B formulations, we cannot exclude the photosensitivity potential at the current time. We have included a warning in our proposed package insert describing the data and providing guidance for patient management.

[Slide]

I would now like to return to the other topics investigated.

[Slide]

We examined cardiac adverse events because of the sudden death of a patient recruited into the empirical therapy study, which occurred about 30 minutes following the first infusion. This death occurred in a 50-year old female patient, with left ventricular dilatation who had recently received idarubicin therapy acute myeloid leukemia. At the time of the first voriconazole infusion the patient was profoundly hypokalemic and had received a potassium infusion as well as a bolus infusion of potassium. During the voriconazole infusion the patient experienced an episode of vomiting. Then, about 20 minutes after the end of the first infusion the patient experienced again nausea, vomiting and diarrhea. The patient was found convulsing shortly after. The cardiologist at the scene observed cardiac arrest with ventricular fibrillation. Resuscitation efforts were, unfortunately, not successful. The autopsy revealed pronounced left ventricular dilatation, and the investigator assessed the conditions as listed as cause of death. The sponsor could not exclude a contribution of voriconazole to these events.

[Slide]

We then examined the preclinical and clinical program for other signals, including an analysis of QTc intervals. This slide presents the results of three separate experiments investigating the potential for voriconazole and compares with ketoconazole to influence cardiac depolarization. The first is competitive dofetilide binding, which is a sensitive assay we use to look at potential influences on the HERG protein. The second is a patch clamp study on HERG inhibition, shown in the blue triangle curves here. Thirdly, we looked at the effects on action potential as measured in canine Purkinje's fibers.

The results of the in vitro tests are reassuring, and ketoconazole differentiates from voriconazole in terms of its effects on HERG inhibition and dofetilide binding at concentrations that would be relevant to those seen in patients.

[Slide]

We also analyzed ECG data collected from 197 participants in several Phase I studies. Using the standard criteria for QT, here corrected using Fridericia's formula, showed no difference from placebo with regard to mean changes and outliers of QTc, presented on this slide. The one occurrence of a QTc greater than 470 msec was in an elderly female volunteer 4 hours post dosing with 6 mg/kg. Her baseline value had actually been 480 msec so there was not much change. There were no volunteers with an absolute QT greater than 500 msec. Using Bazett's formula for correction results in very similar results.

[Slide]

On this slide I have presented all of the QTc data from these volunteers against plasma concentration observed at the time of C-max about 1 hour post single dose. Linear regression analysis indicates a shallow slope, ending about here. The results using Bazett's correction formula look very similar to the plot.

Also, the results of this extensive examination have thus far revealed no clear signal for voriconazole to cause arrhythmias. We continue our investigation into the potential cardiac effects of voriconazole. In the patient population we have observed at least one plasma concentration being outside of this range in about nine percent of patients. We currently have no QT data for that kind of exposure. For this reason, we are undertaking a Phase I study specifically targeted at measuring QT effects across the full range of exposures seen in the patient population. The first attempt at such a study rises the next special issue I would like to address, anaphylactoid reactions.

[Slide]

The design of the QT study involved escalating doses of voriconazole administered intravenously to healthy volunteers. We have twice attempted such a study and have stopped because of the four events shown here. The first two in the first trial and the second two in the second.

In the first study two female subjects experienced events considered to be anaphylactoid reactions and manifested principally by vasodilatation during their second dosing period. Both infusions were stopped within minutes of starting, and abated with only oxygen with no further therapy administered. In that study, one subject, the first one, had been administered our solubilizing agent, SBECD, whereas the other subject received I.V. voriconazole which also includes SBECD.

During the second study, two additional anaphylactoid reactions were observed, again principally vasodilatation, again in two young females. In the first event there was no treatment. Actually, the I.V. infusion continued and the event abated spontaneously. The infusion in that subject was stopped when the second reaction occurred in the second subject. So, both infusions at that time were discontinued. Again, symptoms resolved without any further intervention after about 17 minutes.

A full investigation into these events was conducted. The subjects in the second group both received I.V. voriconazole as SBECD. None of our due diligence activities have identified a probable cause for the infusion-related events seen in the Phase I studies.

[Slide]

The clinical database does contain reports of vasodilatation or flushing at low frequencies. However, as Dr. Boucher had shown you, no anaphylactoid reactions have been reported in our comparative empirical therapy study where infusions were prospectively monitored at the bedside of these patients. The mechanism of the anaphylactoid reactions in healthy female volunteers remains unknown, and we currently intend to include suitable warning statements in the package insert. I just want to highlight that we also still remain committed to the exploration of the potential for acute QTc prolongation with voriconazole across the full range of exposures seen, and we intend to start that study again soon.

[Slide]

Renal function -- this was investigated because of an observation of renal tubular vacuolization in the preclinical toxicology program using our excipient SBECD. Overall, a link to renal adverse events in the clinical trials with either voriconazole or SBECD is considered unlikely; also monitoring of creatinine levels will be recommended.

We investigated sepsis as a result of literature reports of a link between sepsis and the use of azole antifungal agents. In the large comparative studies there was no difference between voriconazole and comparator, neither was there a delay in recovery from neutropenia observed in our empirical therapy study. Overall, we conclude that the voriconazole database does not provide any additional evidence to support the hypothetical link.

Hallucinations, also occurring at low frequencies, show an imbalance in reporting frequency, particularly in the empirical therapy study. Hallucinations have, therefore, been examined as to whether these were confounded by visual adverse events and that is the case in some reports, and we, therefore, overall conclude that our analysis revealed that the role of voriconazole cannot be excluded.

[Slide]

This slide depicts our conclusions drawn from the extensive safety database on voriconazole. The visual disturbances are the most frequent adverse event. They are related to functional alterations without long-term impact. And, the underlying physiological phenomenon appears to be reversible after one month of dosing.

Hepatic enzyme elevations are dose limiting, however, are reversible upon discontinuation and occur with similar frequency as amphotericin B formulations. Monitoring of hepatic function will be recommended in the package insert.

While skin reactions occur with the same frequency as the comparator agents in the clinical program, we currently cannot exclude a potential for voriconazole to cause photosensitivity reactions, especially in long-term therapy.

[Slide]

In summary then, we believe that voriconazole addresses an important currently unmet medical need in seriously ill patients suffering from invasive fungal infections. We have seen significantly improved outcome in acute invasive aspergillosis, with a survival benefit.

We think the data in infections due to emerging pathogens such as Scedosporium and Fusarium are very encouraging. From the in vitro potency and the in vivo efficacy seen against Candida infections, such as Candida krusei, and the results of the patients treated to date, we conclude that voriconazole is efficacious in Candida infections.

The data seen in patients with documented infections, as well as the results of our empirical therapy study and the better safety profile compared to amphotericin B formulations, lead us to conclude that voriconazole is an appropriate option for empirical therapy in patients with persistent fever and neutropenia.

The large safety database leads us to conclude that voriconazole has an acceptable overall safety profile and is better tolerated than amphotericin B formulations. We feel that we have collected sufficient data on interactions to provide specific guidance to the prescribing physician, and that these are, thus, manageable.

Overall, we think voriconazole addresses an important medical need that is currently not met. Thank you for your attention.

DR. GULICK: Thanks, Dr. Baildon and Drs. Patterson and Boucher. At this point I would like to open it up to the committee members to ask questions of the sponsor. Again, I would suggest to people that we stick to questions of information and clarification, and we will take up more of the discussion in the afternoon. Dr. Wood, you are jumping right in.

Questions from the Committee

DR. WOOD: I have several questions regarding safety. Since visual disturbances are most common and were observed not only in clinical trials but also in animal studies, I want to know was there any examination of whether or not it was associated with C-max concentrations of voriconazole or AUC levels in terms of any correlation with specific drug levels that would correlate with visual disturbances.

DR. BAILDON: Let me address your first part with C-max. C-max is about one hour after infusion. The onset of the visual disturbance is somewhat earlier; it is about half an hour and then it disappears within each event. Each individual event disappears after about another 30 minutes. So, the time course is not quite identical.

In our overall analysis of the correlation between plasma concentrations and adverse events there is a correlation between visual adverse events and plasma concentrations, this being somewhat more frequent at the higher end of exposure but it is also present at the lower end of exposure. It is between about 20 and 25 percent.

DR. WOOD: Regarding the visual disturbances as well, most of the time they were reversible within 14 days of discontinuation of treatment. I am curious, the 38 patients whom you have been able to monitor for over a year, whether or not there were changes in their vision over time, or if things remained stable or if they had any chronic visual problems associated with chronic administration.

DR. BAILDON: Let me just clarify that our electrophysiology study addressed reversibility of the underlying electroretinogram changes, the underlying electrophysiological phenomena. What we observe in patients, as I showed you, is that the frequency of reports is highest in the first week.

DR. WOOD: Right.

DR. BAILDON: A number of patients continue to report these events throughout treatment but, actually, many patients do not report them anymore after one, two, three weeks. So, if I were to show you a curve showing time to last treatment, that looks very similar to the one of the first. So, it is somewhat difficult to say what happens in these patients, the 38 patients treated out for over a year. We have not received any reports of significant alterations as adverse event reports that we would pinpoint to the visual disturbance as such.

DR. WOOD: A couple of other questions regarding the cardiac deaths as well as the anaphylactoid reactions. Were drug levels examined at all in those patients, or were you able to obtain drug levels? The reason I am curious is because in the FDA handout, on page 27, one of the things that I was quite impressed by is when you look at the mean AUC for females compared to males, whether it is with single dosing or with multiple dosing, the AUCs for females are almost twice -- excuse me, much more than twice; seven times more for young healthy males in terms of multiple dosing, at least twice as much for single dosing, and very similar -- five times as much for single dosing in young health females and then nine times as much almost in terms of chronic dosing. And, the cardiac death and the anaphylactoid reactions were in females.

DR. BAILDON: Can I address your pharmacokinetic question first? If I could have PK backup slide number 83?

[Slide]

This shows the data that you refer to around the difference in exposure observed in Phase I volunteers by sex and also by age where we see what you highlighted, the higher exposure seen in the female population. And, there was also among males a shift in the Phase I population.

[Slide]

As I mentioned, we have collected over 3000 samples in over 1000 patients, and this replicates that same analysis using our database from the clinical trials. What you can see there is that variability that we saw in the males, young versus older patients, and versus the females we don't observe that anymore. So, given all the other factors influencing the variability on exposure, we cannot decipher that same difference in the patient population. That led us to the conclusion we don't dose adjust.

Now I will turn to your question around the anaphylactoid reactions. These were events that occurred right at onset of infusion. So, we checked blood levels and we could confirm that the volunteers received what we thought they should have had, which was the solubilizing agent in one volunteer and voriconazole I.V. in the three others. But the exposure was extremely minimal. This was a very, very small amount. So, I don't think it is related there.

In that one death I described in the empirical therapy study, which was in a Canadian patient, that was after the end of the first infusion at a time when we would expect C-max but we do not have plasma samples. We are fortunate to have Dr. Jeremy Ruskin here and maybe we can raise that later. It was a patient who had suffered significant cardiac damage in the time prior to infusion, probably related to idarubicin therapy. So, it is somewhat difficult to assess but we cannot exclude the relationship.

DR. WOOD: In the 26 deaths that were associated with hepatic transamination elevation, we went through four in detail. I found it very interesting that one of the four deaths was associated with an undiagnosed history of an individual being at risk for cirrhosis. Did you go back to the remaining 22 patients and see whether or not any of those 22 patients had conditions such as a history of alcohol abuse or chronic hepatitis B or C that would predispose them for maybe having undiagnosed cirrhosis?

DR. BAILDON: We did not find anything that would contribute to an explanation beyond the very brief data I have given you. The first patient I highlighted stands out actually because that patient had a higher exposure. It would be only a very, very few percent of our patients who have exposure beyond 10 mcg/mL. The exposure in that patient was 13 mcg/mL and that is explained by the hepatic impairment. The other patients, where we have plasma samples, actually have exposure around 6 mcg/mL which is well within the range we would observe. So, I don't think that would add anything. I do not have any more specific detail. I think we could get the narrative if you want to review the patient. We have a narrative description of it, but there is nothing that stands out.

DR. WOOD: All right. My last question relates to the poor metabolizers and the alleles in the CYP2C19. In your study you noted that it was probably in approximately two percent of Caucasians. I am curious whether or not anyone from your pharmacology team can comment on what the incidence of poor metabolizers may be in other ethnic groups.

DR. BAILDON: If I could have PK backup slide 81, I can actually show you the data that have been published to date on this issue.

[Slide]

As you can see, what I talked about is the Caucasian population, and that is the majority of our patients. Two percent are poor metabolizers and 73 percent extensive. This is about the same distribution seen in black populations where there is a difference, and Asian populations where 15-20 percent are poor metabolizers. We have not analyzed that in our Phase III studies but what we have done is analyzed again plasma samples, as I showed you, in our Phase III population, subdivided by race, as a surrogate for the genotype.

[Slide]

This shows you the results of that analysis. I highlighted that the majority is Caucasian, 861. This is the distribution observed here. This is the 25th percentile and 75th percentile, with the median as a line in the box, here. If you look at the Asian population, it is somewhat higher. The median is shifted somewhat higher, which would reflect this up to 20 percent incidence of poor metabolizers. However, it is still contained within the range of exposures we observed overall and that is why we concluded that we don't need to dose adjust. Does that answer you?

DR. GULICK: Dr. Hamilton and then Dr. Morrison.

DR. HAMILTON: Given the recognized difficulty of distinguishing the contribution of one pathogen from another, your focus on invasive Aspergillus and the common coo-existence of several pathogens at the same time -- could you elaborate a little bit on what efforts were made to distinguish invasive aspergillosis as a sole pathogen from others? What diagnostic techniques were employed to make a definitive diagnosis, just in general; I am not talking about in detail? And, this is not to question the acumen of the expert clinical panel but it would just be useful for me to know what those elements were.

DR. BOUCHER: Certainly, I would be happy to address that. In our major comparative studies, the global comparative aspergillosis study and the empirical study that we described, we used modified MSG EORTC criteria which have been developed specifically to aid in the diagnosis which, as Dr. Patterson mentioned, is difficult in this setting.

For a definite diagnosis in both studies and across our clinical program, including all the patients that we shared, a definite diagnosis is based on the recovery of the pathogen from a normally sterile site or the recovery of the pathogen by culture, say, from a bronchoalveolar lavage in the lung with a concomitant histopathologic confirmation. So, that would be the criteria for a definite diagnosis.

Now, in terms of ruling out other pathogens that might be mistaken for aspergillosis, a particular concern might by Zygomycetes which are not well covered with voriconazole. In that case, culture diagnosis for confirmation was rigorously applied and I can tell you that in the global comparative aspergillosis study we had, among the deaths not due to aspergillosis, one patient with Zygomycetes in each treatment arm. So, those were patients where a diagnosis was made of Zygomycetes.

In the empirical therapy study we shared we had two breakthrough infections due to Zygomycetes in the voriconazole arm. They were diagnosed as the investigators were rigorously pursuing confirmation.

In terms of the relevance of other pathogens, say Candida when you we are looking for aspergillosis, our data review committee in the global comparative aspergillosis study reviewed all the available mycology, as well as the radiology and the other evidence, and decided in their decision of assessment of certainty of diagnosis that this was definite or probable invasive aspergillosis, and several of our patients had isolates of Candida from bronchoalveolar lavages and aspirates an sputum that were not considered pathogenic.

DR. HAMILTON: How about viruses? I am particularly interested in CMV, and seeing Bob Rubin here, I suspect some serious consideration has been given to the contribution that that might have in these very seriously ill patients.

DR. BOUCHER: Certainly, and that was again rigorously looked for. We had a few patients with CMV pneumonitis, including transplant patients, in our total population -- again, pathologic and virologic diagnoses were sought. Dr. Rubin, would you like to comment further? Dr. Rubin was a member of the data review committee for the global comparative aspergillosis study and has been involved.

DR. RUBIN: Thank you. John, there was an attempt. That is always the problem in these complicated patients in terms of what else could be going on. CMV was looked at seriously with both viremia assessment and BAL assessment and pathologically. It did not seem to be playing a significant role, but if there were patients with combined infections where there was definite or probable evidence of Aspergillus they were carries as Aspergillus and we were looking for an endpoint on the Aspergillus.

In point of fact, that kind of prejudiced -- and there were more of them in the voriconazole as I remember -- that prejudiced because it meant there had to be effective treatment not only of the Aspergillus but of the other pathogens that were there. And, we had an uncomfortable feeling that there were deaths that were attributed to Aspergillus that probably were another pathogen so that the success may have been better but the strength of the analysis was that we were all blinded and had no idea which group we were thinking about.

DR. GULICK: Dr. Morrison?

DR. MORRISON: I have two questions with regard to the use of this agent in patients with renal or hepatic impairment, and then also a brief question with regard to aspergillosis.

First of all, in patients with hepatic impairment you recommended that the maintenance dose would be halved in these patients. How are you planning on defining hepatic impairment? Are you going to be using any specific LFT cut-off numbers?

DR. BAILDON: No, this was, as in our multiple dose study, defined as child pure B cirrhosis so it would be the same recommendation. We have not studied child pure C.

DR. MORRISON: Okay, because from a clinical standpoint, when you are caring for these patients sometimes it is helpful to have some more discrete cut-off. So, you are not planning to do that?

DR. BAILDON: No.

DR. MORRISON: My second question, patients with renal impairment -- you would advise that the oral formulation of the agent be used in patients with creatinines greater than 2.5. In our grey manual, on page 19, it said that in patients who have creatinine clearances of less than or equal to 50 only the oral formulation be used. My question is I have concerns that we are going to see not a small number of patients who will have creatinine clearances less than 50, based either on the fact that they are middle aged; they are receiving concomitant nephrotoxic therapy. So, the question I would have is that in patients with febrile neutropenia you feel the oral formulation would be effective for prophylaxis in this patient population. Do we need to be checking blood levels in these patients?

DR. BAILDON: That opens up a number of interesting discussions actually. Let me first come to the oral form. As you saw, we had over 90 patients in the empirical therapy study who were able to switch to oral therapy. The cut-off of 2.5 mg/dL creatinine is one that was used throughout the program and has not led to problems. We have a few patients in whom, based on an individual benefit/risk assessment, the investigators felt that they needed the I.V. despite having at least temporary higher creatinines. The recommendation of 2.5 just comes out of the clinical program. We did not determine creatinine clearance. We have studied it relative to creatinine clearance and we can say that there is no influence on the voriconazole exposure across the spectrum of creatinine clearance that we studied, but we clearly see that relationship to the excretion of the solubilizing agent. That is why we have that limit.

Now, the other question around blood level monitoring is a totally different question. I am happy to address it but I think it takes a bit longer to explain our approach to that.

DR. GULICK: Go ahead.

DR. BAILDON: Obviously, when you study non-linear drug with that kind of variability and with all the factors influencing this variability and none of them actually being predominant, that is a major issue.

[Slide]

I can highlight what we have done but I do need a couple of slides to explain the story there a little bit.

DR. GULICK: Sure.

DR. BAILDON: This is what we asked physicians to do, investigators in our study -- to collect random samples throughout the treatment period. Actually, it is quite impressive. We have samples from over 75 percent of our patients.

This is a patient from our empirical therapy study who actually developed breakthrough pulmonary aspergillosis and then was treated for a total of 80 days -- successful outcome I am happy to report. This investigator took the effort of sampling quite seriously and continued to sample throughout the 80-day period. For every dot I have highlighted here, this is plotted as a result of that plasma level concentration versus the time of dosing throughout that 80-day period.

So, what you see is that they are all normalized to the time of dosing, and it is impressive and I talked about the low within subject variability and it is examples like that that confirm our modeling exercise and our complex PK analyses. Actually, over the 80 days there is quite low within subject variability.

Then to address the question around what can I do for the benefit of monitoring these patients, we have taken a mean concentrations -- not all patients had this many; it is 3.5 samples per patient on average, but we have taken the mean concentration as a point estimate for each patient.

[Slide]

We have related that point estimate of a patient for exposure to the relevant outcomes, here efficacy in our global comparative aspergillosis study, and we have done that across the other indications as well.

What you can see here is that each patient is now placed in one of these pockets according to the mean concentration observed, and then we looked at the success rate in the global comparative aspergillosis study across that. As you can see, there is no obvious relationship here and that is actually confirmed by the modeling exercises.

The only group that sticks out is these 14 patients who had actually higher exposure, more than 6 mcg/mL, and when we reviewed these cases this does not indicate a kind of inverted dose response, these are 14 patients who were septic; who were extremely ill, already close to multi-organ failure and they actually fail early on. They die or they fail within the first week or 10 days of therapy. So, we think that is more related to impaired hepatic function already at that state and some of these 14 patients received early dose escalations, as I described, which was allowed sometimes too to try to rescue the patient. So, that is not indicative of dose response. We don't see that across our exposures.

[Slide]

We have done the same relating ALT abnormalities, and here we take weekly plasma level with weekly occurrence of ALT abnormality, again across the population. Here, this is our therapeutic studies population. So, we received samples from everybody. Again, each patient is placed within one pocket but now that patient might appear several times because that patient had plasma concentration at week 1 and ALT at week 1 and then again at week four and that patient could contribute two data points here.

Here you can see, and our modeling confirms that, somewhat of a relationship there which has a slight positive slope. So, that would then be the starting point for saying would therapeutic drug level monitoring add any benefit to that?

[Slide]

I want to highlight the approach we used for that. The methodology used is receiver operating characteristics. I am sure some of you are familiar with that. That is a graphical depiction of the false-positive rate versus the true-positive rate, and then you can use several different thresholds to try to determine is my threshold predictive of the outcome. This is an example from a recent publication by Min et al., using that for predicting renal allograft rejection with cyclosporine, and you can see a reasonably good predictive value here of the sensitivity of about 60 percent and a specificity that is definitely beyond 80 percent. If a test is not evaluable you will progress along the line of identity as you take different cut-off levels.

[Slide]

This shows you the result for what I showed you earlier, the determinations in aspergillosis. So, we have taken the ROC approach and taken very low cut-offs. Our detection limit is 0.1 mcg/mL. As we move up in cut-offs we, unfortunately do not see a certain line of identity. So, the plasma level for efficacy doesn't add any value, which is not surprising given the correlation I showed you.

[Slide]

This also holds true actually for the ALT prediction. We can identify in our database a linear or some kind of correlation but the predictive value for ALT abnormality again yields a more appropriate starting at a higher exposure level. Our conclusion, to answer your question however, then to monitor for efficacy we recommend clinical judgment which we expect to be up here. To monitor for hepatic enzyme elevation, which we consider dose limiting, we suggest to monitor the hepatic enzyme elevation which is also more predictive. Does that address that?

DR. GULICK: Thanks. I think that was useful. Another question, Dr. Morrison?

DR. MORRISON: Just a very brief question. With regard to your Aspergillus data, obviously for the right reasons most of your cases are with fumigatus. Do you want to make any comments with regard to treating other non-fumigatus Aspergillus infections with this agent?

DR. BOUCHER: I would be happy to.

[Slide]

As you said, Dr. Morrison, most of the species identified were A. fumigatus. In this slide we depict the success according to species from patients with definite or probable invasive aspergillosis according to the data review committee. So, these are species obtained just from those patients in the modified intention-to-treat population.

At the top we see that, indeed, the majority of patients had A. fumigatus. There was a small number of patients with flavis, terreus, niger and nidulans and we see some efficacy although the numbers are extremely small, but those are the data that we have in this study.

DR. GULICK: Dr. Wong?

DR. WONG: The principal question I want to ask is about the assessment of efficacy in the global aspergillosis treatment trial. The problem I am having is that the patients who were assigned to voriconazole received voriconazole for a median of 70-some odd days, whereas those who were assigned to receive amphotericin B received it for a much shorter period of time. Can you break out patients who were assigned to amphotericin B, for example, who received amphotericin B initially and then also received it for a significant period of time after the 10-14 day range, or received conventional amphotericin B and then liposomal amphotericin B thereafter so we can compare those results to those who received voriconazole? Or, alternatively, can you show us some assessment of clinical efficacy in the voriconazole group at a time point comparable to the period of time that patients received amphotericin B?

DR. BOUCHER: I think I can address your first question most easily.

[Slide]

Here we can look at success in patients, at the top, who just received their initial randomized therapy and then in those who switched to other licensed antifungal therapy. So, if we look first in the voriconazole arm we can see that 5/13 voriconazole patients who received lipid preparations, or 38.5 percent, had success compared to 29.8 percent in the amphotericin B arm.

Looking at itraconazole, 64.7 percent of voriconazole and 50 percent of amphotericin B patients had success. So, that is looking at the success according to other licensed antifungal therapy used.

Returning to your question about a different time period, to answer your question about the time periods, we can't give you success other than the two time points from the study, end of randomized therapy or week 12, because you couldn't be a success until you got there. So, we know that the patients who were a success at the end of randomized therapy is 70 percent. You know, they were on their way to success at two weeks but they did not have a formal assessment.

DR. WONG: I guess another way to look at it is if we see a Kaplan-Meier curve for survival over the course of 84 days, and it looks like maybe there is a separation of those curves early on but it is so small at that point that it is difficult to see. Do you have a form of the Kaplan-Meier curve that really highlights the early events, and did you do a statistical analysis of survival at an earlier point? What I am driving at is that at a point where the patients were still receiving their initial randomization treatments.

DR. BAILDON: If I can go back to my main slides that highlight the survival --

[Slide]

We have not done the analysis you describe at any interim time points. This is the safety population so it is all patients receiving drug. What you can see, if you look at day 14 here, is that there is already somewhat of a difference here which is maintained and progresses. We have never cut this in another way. We could prepare a slide for after lunch that blows up this part. But I think one of the problems in treating fungal disease is that it does take a while to actually see some efficacy.

DR. GULICK: Dr. Rodvold and then Dr. Stanley.

DR. RODVOLD: I would like to follow-up on a couple of the questions that were previously asked and ask you to extend maybe your comments. In poor metabolizers where they do have higher concentrations, did you further break that down by taking a poor metabolizer and separate out gender as well as separate out age, in other words, looking for the worst scenario -- an older female that is a poor metabolizer? Are they even further at risk for higher exposure than, say, a male that is a young poor metabolizer?

DR. BAILDON: We have not broken it out that way. We only have relatively few poor metabolizers --

DR. RODVOLD: Yes, I know.

DR. BAILDON: With two percent of the population it is actually difficult to capture that.

DR. RODVOLD: I would think that that might be worthwhile to either explore or put in further studies to look at just because of the potential of specific Asian populations in certain areas that may get drug in the future. So, that might be something you want to come back to.

The second issue to tag onto that, is that a population that may need therapeutic drug monitoring? In other words, not the whole population but a real specific population? And, I am not saying therapeutic drug monitoring is the ultimate here but is there a real small population, like poor metabolizers, that need it?

DR. BAILDON: Yes, as I showed you in our surrogate race parameters looking at the exposure we actually observed in patients is overlapping exposure and I think one of our interpretations around that is that the genotype status is very important in a Phase I population that receives only voriconazole. Once you go into a patient population with all the other factors influencing exposure, it becomes somewhat less important.

DR. GULICK: Dr. Stanley and then Dr. Schapiro.

DR. STANLEY: Thank you. I am just trying to still get a handle on the abnormal visual toxicities. You say that the vast majority of them occur within the first week and that then they either resolve or the patients stop complaining about it, one or the other. The only hard data you have shown us this morning is on your challenge trial with 28 days of treatment and then showing documented resolution of the changes. When I try to go back into the data that we had ahead of time, I see charts that show me, like in the global trial, 33 percent of patients complained of visual abnormalities; only 1 percent discontinued drug because of visual abnormalities. But I don't see any documentation of any actual testing being done of visual acuity or impairment. There is referral in both the empiric trial and in the esophageal candidiasis trial that they had similar low rates of discontinuation and that visual function testing showed no difference. But your longest drug exposures in your global trial was 77 median days of exposure. Is there any actual testing of visual acuity or retinal changes that you did in patients that have seen drug for that long?

DR. BAILDON: Yes.

[Slide]

This shows you the acuity results I showed you for the esophageal candidiasis study where we observed no change over usually the 14 days of therapy. This is the same presentation for the empirical therapy study, again testing acuity at baseline and at follow-up. One of the problems we have in that patient population is, because they are more severely ill, we have a large number of patients who don't have a baseline value. But if you look at those where we have baseline values, you see there is a very similar percentage of improvement or no change, deterioration of one line or two lines.

[Slide]

We have the same presentation for our global comparative aspergillosis study. Again, it is the same presentation. Again, we have an issue that not all patients are able to either do it at baseline or follow-up but we have tested for acuity and you can again see -- well, here is actually more improvement or no change but it does not differentiate between the comparator agent. It reflects more the underlying disease of the patient. Does that address your question?

DR. GULICK: Dr. Schapiro and then Dr. Englund.

DR. SCHAPIRO: Dr. Boucher, you mentioned the success anecdotally with non-albicans Candida. Do you have a summary of the clinical experience with those?

DR. BOUCHER: I do.

[Slide]

It is in the esophageal candidiasis trial as well as from our series of patients collected across the program to date. Dr. Baildon showed the organisms in the esophageal candidiasis trial. Just to refresh your memory, most of the isolates here were albicans; 56 were glabratu but there are very small numbers of other non-albicans species.

The other thing to note is that these were isolated, virtually all except for two, in the presence of C. albicans and we saw the same efficacy overall in this study, whether you had pure albicans or you had a mix.

[Slide]

Turning then to our series of patients, we can first look at our series of Candida krusei patients. This shows a series of 12 patients with Candida krusei from across the program, six of whom were non-neutropenic and 6 of whom were neutropenic at baseline. Most of these patients were severely immunosupressed with hematologic malignancy and transplantation and in either case the total of eight of ten patients had success. The numbers are small.

DR. SCHAPIRO: Thank you. Can you also show the data on the interaction study with Coumadin?

DR. BAILDON: The warfarin interaction study? Do we have that? I have to ask my PK colleagues which slide number that one is. Maybe I will take another question first?

DR. SCHAPIRO: My last question would be on your summary slide you mention I think that it is better tolerated than the ampho formulations. Was that specifically relating also to the lipid ampho, that it was better tolerated? I don't remember actually seeing data that there was superior toleration.

DR. BAILDON: Dr. Boucher highlighted that in the infusion-related reactions in that study, which we monitored prospectively, there was an imbalance with more infusion-related reactions on the amphotericin B and the liposomal amphotericin B arm, except for the visual disturbances which were more frequent on the voriconazole arm.

DR. SCHAPIRO: So, it is not a global statement that relates to the infusion --

DR. BAILDON: If we would go to renal function we would see some imbalance as well because liposomal amphotericin B also has an effect on the renal function. Do we have the warfarin study?

DR. GULICK: We can have it after the break. We can come back to that. Oh, there you go.

[Slide]

We recommend monitoring the biologic effect and dose adjusting as appropriate. This is for voriconazole exposure 300 mg twice a day. That is our high dose plus warfarin. This is only warfarin so the prothrombin time is higher. That is right. So you would need to dose adjust and monitor biologic effect as I highlighted in one of my interaction slides.

DR. GULICK: Dr. Englund and then Dr. Mathews.

DR. ENGLUND: Yes, I am interested in the Candida susceptibilities to your drug, particularly in your empiric therapy trial and your esophageal candidiasis trial. You have data nicely presented from pre-therapy but do you have data during and even post-therapy, and if you treated them for 14 days, many of them were HIV patients and I am sure it comes back. Do you have longer-term follow-up?

DR. BOUCHER: We do in the esophageal trial. I can come back after the break with a slide -- very few patients with isolates, but I can show you a slide on that. In the empirical therapy study we don't. We have very few patients who had subsequent isolates in that study.

DR. ENGLUND: You had two failures I believe with Candida.

DR. BOUCHER: I will present details of those two when we come back from the break.

DR. GULICK: Dr. Mathews?

DR. MATHEWS: Can you show us some of the outcomes on the 25-28 percent of patients that were excluded from the modified intention-to-treat analysis in the global comparative study?

DR. BOUCHER: Certainly. Just to confirm, the modified intention-to-treat included those patients who met the data review committee's criteria for definite or probable aspergillosis. So, we can look at the intention-to-treat population in the global comparative aspergillosis.

DR. MATHEWS: Well, my question is not about that inclusive group but to break out separately the 102 patients.

DR. BOUCHER: Okay, that is the non-MITT population and we do have those data, if you will bear with me one moment.

[Slide]

These are the 50 and 52 patients excluded because we were unable to confirm the presence of invasive aspergillosis. Again, this is 12-week response assessed by the data review committee. Here, again, we see 44 percent success in the voriconazole arm, 23 in the amphotericin B arm, and a difference of approximately 21 percent.

DR. MATHEWS: And mortality? Do you have that?

DR. BOUCHER: It is similar.

DR. BAILDON: That was included in the safety population. As you can see, the Kaplan-Meiers for the MITT in the safety population look actually very similar. So, the Kaplan-Meier for the non-MITT would be identical to those. The hazard ratio is the same across all three, 0.6.

DR. MATHEWS: A different question, you stated that the drug interaction studies that you had done would provide guidance to the practicing physician, but there occurred to me a number of other interactions of potential importance. For example, I was curious why you picked indinavir to do the protease inhibitor interaction when I would have thought that ritonavir would have been the most obvious one to look at because of its more potent effect on cytochrome P450 system.

DR. BAILDON: That is correct, and we are actually planning a ritonavir study. The point of that was that our early in vitro work indicated that indinavir would not have an interaction and we wanted to confirm that in Phase I so we could offer a treatment choice to physicians treating HIV patients.

DR. MATHEWS: I wonder also about, you know, you looked at two-way interactions but obviously these patients are on multiple drugs that may have more than one inducer or more than one inhibitor, for example, somebody on phenytoin and rifabutin. I would have no way of knowing what would happen with the levels of voriconazole.

DR. BAILDON: Well, I highlighted that the number of concurrent medications, for example in the global comparative aspergillosis study, was 26 in the populations, and I am sure many of these concurrent medications contribute to the variability of exposure I showed you in my plasma concentration slides. The statement about general manageability is more that this is a patient population of close to 200 patients who were able to be maintained on voriconazole for a median of 77 days despite receiving a median of 26 concurrent medications. We have not studied multiple interactions specifically but this indicates to me that in that population at least it was manageable, otherwise these patients would not have been maintained for 77 days on voriconazole. But these interactions would certainly contribute to the variability we observed.

DR. GULICK: Dr. Yogev?

DR. YOGEV: You mentioned that you picked out the maximum tolerated dose, 4 mg/kg, because 6/21 had hepatic function problems, and we were looking today basically at only one who got 5 mg/kg, the rest were 4 mg/kg. So, I was just wondering, taking into account that you didn't show any difference in hepatic toxicity between the amphotericin group and the voriconazole, basically showing that the 5 mg/kg did not increase it, do you have any correlation between the peak level or area under the curve and liver toxicity? Because that would be, to me, much more meaningful if the drug really has any effect on the liver.

DR. BAILDON: Let me just clarify a point. There was one patient in the dose group that we currently recommend as the highest dose group who experienced an ALT abnormality.

[Slide]

That was this patient, in green, who had 4 mg and 300 mg orally. That is our current highest dose recommendation. The other subjects on this slide actually received the one that was the highest in the study, which was 5 mg followed by 400, and we considered that exceeded actually the maximum tolerated dose.

DR. YOGEV: What happened to the other 15 out of the 21? Because this is only 6/21. Were the other 15 on 5 mg or higher?

[Slide]

DR. BAILDON: They did not show ALT elevations. I only picked out those five that I showed you over the time course who had an ALT elevation. The other nine here did not show an ALT elevation.

DR. YOGEV: No, my basic 101 pharmacokinetics is that the MTD is defined as 50 percent of the population are getting that toxicity, or something like that, and I wonder if this is not an accidental finding because of the small number. Do you have any levels in the blood or area under the curve that correlate toxicity or liver elevation?

DR. BAILDON: Well, let me state that for us this was close enough to 50 percent in healthy volunteers.

DR. YOGEV: The reason why I am asking that is because I am confused from your pediatric data that --

DR. BAILDON: Can I just answer your C-max question? I showed you our modelling approach to the PK/PD relationship using the mean concentrations. We used the mean because that gave us use of the full information in the data. We have also repeated that both for efficacy and safety using either the maximum concentration per patient or using the minimum concentration per patient. As far as ALTs, for example, are concerned, it doesn't show any more. There is a relationship in our database but it does not provide us with more predictive value than otherwise.

DR. YOGEV: Because in the pediatric -- correct me if I am wrong, because it is non-linear and everything else, your recommendation is going to be 4 mg/kg as a maintenance dose, and I presume that means you are going to give 8 mg/kg as a loading dose?

DR. BAILDON: No, we use the same loading dose.

DR. YOGEV: So, you will go with six, six and then you will go to four?

DR. BAILDON: Correct.

DR. YOGEV: Yet, you are suggesting that in the adult to go from 3 mg/kg to 4 mg/kg for the Aspergillus --

DR. BAILDON: Yes.

DR. YOGEV: Would one make an assumption then that in pediatrics you would go to 5 mg/kg?

DR. BAILDON: That would certainly be one consideration there. We have not studied that. What we have studied is 3 mg and 4 mg in children and in adults, and what we have seen is that the 3 mg/kg in adults and the 4 mg/kg in children results in very similar exposures. I can show you a slide on that actually.

DR. YOGEV: You showed it. Don't waste your time. My problem is the five on one side you say is toxic, on the other side it seems like a dose for pediatrics -- add to that that you are going to recommend that for less than 40 kg to halve the dose. We have lots of pediatric patients less than 12 years of age or less than 40 kg. Should they get 4 mg/kg or should they get half of the dose of the adult?

DR. BAILDON: There are two questions in there. One is about dose and one is the formulation I use in children. Currently, we are recommending I.V. for very young children because our tablets actually -- currently 50 mg is the smallest tablet size we have. We have actually put a fair amount of effort into developing an oral suspension and we hope that we will be able to use that oral suspension in clinical trials from next year on. And, one of the studies we intend to do then is actually an I.V. to oral switch in children, which would allow us to explore the dose range further. Right now our recommendation is in children under 12 years of age -- we have studied 4 mg/kg and we know that that results in similar exposure as 3 mg/kg in adults. We have treated I think 60-some children compassionately and we have seen efficacy in that population with that recommendation. I have no data going beyond that.

DR. YOGEV: The last question is just that I was impressed by your esophageal candidiasis study that you chose the fluconazole as the comparative drug when it is almost 100 times less potent in vitro than your drug. What was the reason not to choose itraconazole which has the same efficacy? I didn't see any data comparing to itraconazole which, in the in vitro data, seems to be comparable.

DR. BAILDON: We have no clinical comparisons to itraconazole. We chose fluconazole because it is the accepted standard of care in esophageal candidiasis, and you have seen the results of our study looking at efficacy and safety. We would not expect that to change.

DR. GULICK: Are there other committee members who haven't yet had an opportunity to ask questions who would like to ask a question? I have two myself.

In talking about the proposed doses for candidal infections for intravenous you are suggesting 3 mg/kg q 12. Then an increased dose for Aspergillus and other fungal infections. Yet, the oral dose remains the same for the two. Could you comment on that?

DR. BAILDON: Yes, and actually what we see in our database is a decrease of exposure in the global comparative aspergillosis study when we switch from I.V. after about 10 days usually, 10, 14 days. As you switch to oral you see a decrease in exposure. That is intentional. As I highlighted, we wanted to be at the very upper end of the dose-response curve for acute invasive aspergillosis which is acutely life-threatening. I think the experience from the previous versions of amphotericin B dose regimen escalations used show that it is not a good idea in that disease to spend time waiting for the dose to come up. However, as we are at the very upper end of the dose-response curve we felt that it is justifiable then after the two weeks when patients are stabilized to go back down in the oral dose. It is a dose reduction in that sense, and we see that actually in our exposure. But the efficacy results are as you have seen using that regimen. We have about 10, 15 percent of our patients who undergo dose escalation from 200 mg to 300 mg. The efficacy results are similar but it is not a comparative trial of the two dosages.

DR. GULICK: My other question is concerning the trial in empiric treatment of febrile neutropenia, a non-inferiority based trial. I was wondering if you could give us some insight as to why minus 10 percent was chosen as the lower bound for the 95 percent confidence interval, particularly because in the background material we noted that the itraconazole trial in a similar setting chose minus 15 percent.

DR. BAILDON: Yes, our study was modeled on the published MSG study led by Dr. Tom Walsh before, with liposomal amphotericin B versus amphotericin B, and they used the same criterion and we found that appropriate as well in that population.

DR. GULICK: Dr. Wood had some follow-up questions.

DR. WOOD: I have a note here that there was just an increased discontinuation of voriconazole secondary to transaminitis and treatment for esophageal candidiasis. Again, given that many patients who need voriconazole for EC would be individuals with HIV infection, I am aware that data regarding hepatitis B and C status is not available for all patients but I am curious as to whether or not there has been any substudy analysis for those patients in whom you do have a diagnosis of hepatitis B or C, and whether or not there has been an increased incidence of a need to discontinue drugs specifically.

DR. BAILDON: We do actually.

[Slide]

This should show us the maximum total bilirubin observed in our total therapeutic trial population for any patient who had some indication in the case record form of hepatic viral infection. I think that is the subgroup you are talking about. I haven't split it out further than that because it gets to very small numbers then.

What that shows you is the distribution here, baseline bilirubin either as normal or close to normal in this population, or somewhat abnormal or highly abnormal, and then the maximum bilirubin observed during treatment of those that stay in the group, and that is for the vast majority of patients, and that reflects very similarly data we see in the overall population or it shifts out. This is all amphotericin B formulations; it is not separating out the two studies. But it shows you that that looks, if anything, better which again confirms what we have seen in each of the individual studies when we look at bilirubin. If anything, we see less abnormality on voriconazole than we observe on amphotericin B-treated patients. Does that address your question?

DR. WOOD: It does. Would you specifically recommend any alteration in doses if you had an individual who you knew had cirrhosis due to whatever cause?

DR. BAILDON: Yes, we would recommend a halving of the maintenance dose; same loading dose because that is a tissue distribution issue; halving of the maintenance dose. DR. GULICK: Dr. Wong?

DR. WONG: I just have one more question about the empiric antifungal therapy trial. In your table 7-43 in the yellow book you make a distinction between a raw success rate and a stratified success rate, and you give a little bit of information about how that was done but I wonder if you could elaborate a bit. And, was this differentiation between the two success rates specified in the protocol in advance?

DR. BAILDON: Maybe I could ask our senior statistical consultant from Pfizer, Prof. Andy Grieve, to come to the microphone and address that.

DR. GRIEVE: Thank you. Throughout the program a stratified analysis, where appropriate, was the primary efficacy analysis. So, in study 603 the primary efficacy analysis was stratified for the variables in which stratification was based, as well as one prognostic variable as well.

DR. GULICK: Dr. DeGruttola?

DR. DEGRUTTOLA: I have one question about the empirical therapy study as well. For the analyses that breakdown the response by high risk and moderate risk, was that protocol specified, and are the confidence intervals adjusted for the multiple comparisons?

DR. BAILDON: I can answer that, that is a simple one. The population was prespecified at randomization. The subgroup analysis was not. And, the confidence intervals Dr. Boucher showed were not adjusted for multiple comparisons.

DR. DEGRUTTOLA: Thank you.

DR. GULICK: Would any other committee members like to ask questions of the sponsor? Dr. Yogev?

DR. YOGEV: Very quick, you, for some reason, said that you are going to recommend to do creatinine while in the 2000 patients you didn't see any toxicity. What is the logic to do creatinine?

DR. BAILDON: There are two points. For one, that is what we recommended in the clinical program, but the more important one is if creatinine goes beyond 2.5 mg/dL we recommend oral dosing because our solubilizing agent would accumulate in those patients.

DR. YOGEV: I have no problem with the beginning but you don't accept that your drug will cause kidney toxicity. From what you suggested, it sounds like that during giving it you should do creatinine.

DR. BAILDON: We have not seen anything.

DR. GULICK: Let's thank the sponsor for answering questions and the presentations. At this point, let's take a ten-minute break. We will reconvene at 11:35 for the FDA presentation.

[Brief recess]

DR. GULICK: There was a question the sponsor would like to respond to before we go to the agency for their presentation.

DR. BAILDON: Right. There was a question around the 38 patients who had continued therapy for more than one year. In those 38 patients, five subjects experienced a visual adverse event. So, it is actually somewhat lower than we experienced in the general population, but that is likely related, as I said to the fact that they just don't report it anymore. So, 5/38.

DR. BOUCHER: I have the answer for the Candida empirical therapy study question. The question was regarding high MICs in isolates obtained during that study. And, we had two Candida infections.

[Slide]

For one we had no isolate. For the second, this is a 68-year old man with leukemia who was in the empirical therapy study for 34 days and developed grade 2 C. albicans with an MIC to voriconazole of 0.01, and another isolate of less than 0.007. He was dose escalated and had mycological eradication and success at the end of therapy.

[Slide]

We had one other patient. This was the one patient from the study who had isolates and high MIC. This patient had a baseline Candida infection. This was a 36-year old lady with leukemia who had prior allogeneic transplantation and she had received prophylaxis for 20 days prior to the study. She developed a baseline fungemia with an MIC of 4. She was discontinued due to persistent fever and fungemia and was switched to amphotericin B later.

DR. GULICK: Thank you. I would like to now call on Drs. Tiernan and Powers to present on behalf of the agency.

FDA Presentation

DR. TIERNAN: Good morning.

[Slide]

My name is Rosemary Tiernan and I work in the Division of Special Pathogen and Immunologic Drug Products, and today I would like to begin the FDA presentation of our review of NDA 21-266 and 21-267 for voriconazole tablets and voriconazole for injection.

[Slide]

Before we begin the presentation, I would just like to acknowledge the efforts of all of the members of the voriconazole review team who are listed on this slide, and I would like to especially thank Dr. Regina Alivisatos, Dr. Rosemary Johann-Liang and Dr. Edward Cox for their help in this review.

[Slide]

The purpose of this advisory committee is to specifically discuss the indications of treatment of invasive aspergillosis and the empiric antifungal therapy of febrile neutropenic patients. However, it is important to note the other indications submitted in this voriconazole NDA and listed on this slide. We are in general agreement with Pfizer that voriconazole is efficacious in the treatment of Candida albicans esophagitis and in the treatment of serious fungal infections due to Fusarium and Scedosporium spp. Efficacy in these areas should be taken into account when considering the indication of empiric antifungal therapy of febrile neutropenic patients.

[Slide]

Our presentation will cover the following areas, I will summarize our review with study 307/602 and study 304 with the historical control which was submitted to support the indication of treatment of invasive aspergillosis. Dr. John Powers will then discuss study 603 which was submitted for the indication of empiric antifungal therapy of febrile neutropenic patients. Then I will return to summarize specific areas of clinical safety. Finally, Dr. Goldberger will present the questions to the advisory committee.

[Slide]

The basis of evidence to support the treatment of invasive aspergillosis was supported by two trials, study 307/602 and study 304. Study 307/602 was a randomized, controlled, open-label initial therapy study of voriconazole versus amphotericin B, both of which could be followed by other licensed antifungal therapy. Because of the open label nature of the trial, a blinded data review committee was utilized to assess certainty of diagnosis, global response to treatment and cause of death, among other factors.

Study 304 was an uncontrolled study of voriconazole use in primary and salvage patients, conducted in Europe. A retrospectively designed historical control study, study 1003, was used as the comparator for study 304. Cases were obtained both from the United States and Europe. The Division is in general agreement with the applicant's presentation of the aspergillosis data. The goal of this presentation is to highlight the main aspects of the aspergillosis trials.

[Slide]

The modified intent-to-treat, or MITT, population, the primary analysis population, consisted of 144 voriconazole subjects and 133 amphotericin B subjects. These subjects were primarily white males with hematologic malignancies as their underlying disease and pulmonary sites for their Aspergillus infection.

The studies were designed to allow a switch from randomized therapy to OLAT. More amphotericin B patients switched to OLAT when compared to patients randomized to voriconazole. Details of the OLAT regimen for the amphotericin B arm were reviewed and felt to represent adequate antifungal therapy.

[Slide]

Recall that the primary efficacy endpoint was outcome at week 12 as assessed by the DRC, and satisfactory response rates of 52.8 percent for voriconazole and 31.6 percent for the amphotericin B regimen were seen. The 95 percent confidence interval for the difference in satisfactory response rates, stratified by protocol, was 9.6 to 33.6. Since the lower limit of the confidence interval was greater than minus 20 percent, voriconazole is considered to be non-inferior to the amphotericin B regimen, and in this case the 95 percent confidence interval does not include zero and, thus, voriconazole is statistically superior.

[Slide]

The Division performed three additional analyses to assess the robustness of the previous results. In the process plan and operating procedures for the data review committee it was possible to upgrade investigator assessment of response. Consequently, the Division performed a conservative analysis which did not allow the DRC to upgrade the investigator assessment. In the second analysis, modified week 12, the Division treated voriconazole patients who switched to OLAT as failures, with the exception of a few patients. These few patients had completed at least 84 days of voriconazole treatment with a satisfactory response and then were placed on prophylaxis. In addition, response at week 16 was assessed. All of these analyses demonstrate that the response of voriconazole was consistently greater than the response of the amphotericin B regimen.

[Slide]

Survival through day 84 was a secondary endpoint and voriconazole was shown to have a survival advantage compared to the amphotericin B regimen.

[Slide]

In study 304 the expert evaluable population, the primary analysis population, consisted of 112 voriconazole subjects. The experts classified 58 patients into the primary therapy group and 54 patients as salvage therapy. As assessed by the experts, a patient was considered to be on primary voriconazole therapy if they received less than 10 days of adequate antifungal treatment. All other patients were considered to be on salvage therapy.

As seen in study 307/602, these subjects were primarily white males with hematologic malignancies as their underlying disease, and pulmonary sites for their Aspergillus infection. This study, 304, was conducted solely in Europe.

[Slide]

The primary endpoint was the expert's global response at the end of treatment, and the overall satisfactory response rate was 49.1 percent. A satisfactory response of 60.3 percent was seen in the primary patients, and a satisfactory response of 37 percent was seen in the salvage patients.

[Slide]

Since study 304 was a non-comparative study, the Division requested that a retrospective historical control study be performed to act as the comparison group to the primary treated patients in study 304. It should be noted that for this comparison the definition of primary therapy was less than five days of prior antifungal therapy. In order to provide the most comparable population, patients were matched on a 2:1 basis by the prognostic factors of certainty of diagnosis, underlying disease and site of infection.

[Slide]

The best matched, less than five-day prior therapy population consisted of 50 study 304 voriconazole subjects and 92 historical control patients. Satisfactory global response rates were 52 percent for study 304 voriconazole patients and 25 percent for the historical control patients. The probability of survival was 0.554 for voriconazole and 0.417 for the historical control.

[Slide]

Even though the applicant took substantial efforts in the design of the historical control, all of the inherent potential biases were not adequately controlled. Differences in patient populations can impact the success rate of treatment if patient care and support differ across countries. Study 304 was conducted exclusively in Europe, whereas the historical control included both European and U.S. patients. Satisfactory global response and probability of survival were lower in the U.S. historical control population. This may be due to the fact that the majority of the U.S. historical control patients had bone marrow transplants or other underlying diseases whereas the majority of the European historical control patients and the study 304 voriconazole patients had hematologic malignancies as their underlying disease. When these U.S. historical control patients are removed, the difference in global response remains but the difference in survival between the European and historical controls and the study 304 voriconazole group become smaller.

[Slide]

Additional issues regarding the historical control included differences in the total days of treatment, with the voriconazole-treated group having a longer duration of therapy and differences in the inclusion and exclusion criteria which could possibly allow for sicker patients to be included in the historical control. All in all, these differences in study populations could predispose the historical control to lower success rates and the voriconazole-treated group to have higher success rates independent of treatment with voriconazole.

[Slide]

In summary, the historical control trial, study 1003, was a good effort but concerns still persist regarding the comparability of study populations. Study 304 results are being used to support the randomized controlled study 307/602. Study 307/602 demonstrated a non-inferior global response and actually met the definition of statistical superiority as well. In addition, voriconazole demonstrated a survival benefit.

We applaud Pfizer's efforts and success at being able to complete the randomized, controlled aspergillosis trial.

Now Dr. John Powers will discuss empiric therapy of febrile neutropenia.

DR. POWERS: Thank you, Dr. Tiernan.

[Slide]

I would like to discuss with you today the FDA perspectives on study 603, which is the study in empiric antifungal therapy in febrile neutropenic patients that compares voriconazole to liposomal amphotericin B.

[Slide]

I would like to start out first by discussing some scientific and regulatory background in this indication of empiric antifungal therapy in febrile neutropenic patients. Then we will go on and discuss some of the issues around the primary composite endpoint and the statistical definition of non-inferiority used in this trial. Then we will go on to discuss some of the selected issues with the secondary endpoints and the analyses of these that are used in the trial.

As Dr. Boucher presented this morning, the five secondary endpoints here are breakthrough infections within seven days of end of therapy; survival at seven days after end of therapy; discontinuations due to lack of efficacy or toxicity; defervescence prior to recovery from neutropenia; and global response in baseline fungal infections. Then we will make a quick summary of these points.

[Slide]

Neutropenia is one of the major risk factors for development of invasive fungal infections, and the risk of such infections varies with the depth and the duration of neutropenia. The most common infecting organisms in neutropenic patients are Candida species and Aspergillus species.

Autopsy studies show an incidence of somewhere between 12 and 43 percent incidence of invasive fungal infections in neutropenic cancer patients. The actual incidence in patients that do not come to autopsy really remains unknown, and part of the reason for that is the difficulty in premortem diagnosis. It is felt that up to 50 percent of neutropenic cancer patients may have occult fungal infections that are not able to be diagnosed by our current culture methods or antigen or antibody detection methods. Once a neutropenic cancer patient develops a fungal infection, they tend to have quite a high mortality, ranging between 48-80 percent depending upon which study you look at.

[Slide]

Therefore, the idea of empiric therapy was introduced for two reasons. One was to treat these occult fungal infections which would not be diagnosed by conventional means, and also to prevent infections in patients who then remain at high risk during their period of neutropenia.

There are two randomized trials which address this question of empiric antifungal treatment in neutropenic patients. The first was performed by Pizzo and colleagues, at the National Institutes of Health, on patients in the late 1970s and this was published in The American Journal of Medicine in 1982. The second trial was really a compilation of four different trials, performed by the European Organization for the Research and Treatment of Cancer, and this was published several years after the Pizzo trial, in the same journal in 1989. In the succeeding 20 years empiric antifungal therapy in neutropenic patients, after a period of antibacterial therapy if they still remained febrile, has become the standard of practice.

[Slide]

There were two public meetings conducted by the FDA in 1994 and 1995 to discuss issues of study design in antifungal drugs. Some of the important issues that came out of these two meetings referable to empiric therapy in neutropenic patients were that a non-inferiority study design was recommended for future trials in this indication. Amphotericin B deoxycholate was suggested as the comparator as, in 1994, it really was the only effective drug that was available. Amphotericin B deoxycholate is not officially approved by the FDA for this indication but this drug was licensed in 1956, prior to our current regulations on efficacy, and it is also a generic drug and there really hasn't been any interest in submitting an NDA for this particular indication.

It was also suggested at these meetings that for approval in the indication of empiric therapy in neutropenic patients that an applicant would need to submit at least one study showing efficacy in another fungal indication of proven disease, plus one study in the empiric therapy of febrile neutropenic patients. Since many of these patients who are treated may not actually have fungal infections, it is important that the drug have proven efficacy in documented infections with Candida and Aspergillus.

Some of the other points from those two workshops were, again, that idea that it was important to prove the efficacy of the drug since resolution of fever, rather than proven infection, would be used to determine sample size in trials of empiric therapy in febrile neutropenic patients. The lower bound of the 95 percent confidence interval suggested to determine non-inferiority was minus 10 percent, as indicated at these particular workshops.

Just as a brief aside, Dr. Gulick asked earlier why was minus 15 percent chosen in the itraconazole trial, in fact, when one looks through the medical officer's review of that, it never came up for discussion. So, it is important to realize that that was not a conscious decision to make it a minus 15 percent in that particular trial.

The composite endpoint was also recommended in trials of empiric therapy of febrile neutropenia because the primary endpoint of breakthrough infections may result in a sample size that would be so prohibitively large that trials would not be able to be conducted for this particular indication.

Finally, the workshop included that some other things that were important to detect in these trials would be differences in proven fungal infections; differences in mortality between the two arms of the trial; differences in fever within a 10 percent confidence interval; and, finally, differences in safety especially since, again, some patients here will receive treatment that do not have infections.

[Slide]

In a 1997 advisory committee surrounding issues on the approval of liposomal amphotericin B, Dr. Alan Sugar summed up some of the issues in empiric therapy. He stated that empiric therapy is given to patients because some will actually need it. However, others will not. That is, they are treated unnecessarily. At issue was the diagnosis of invasive fungal infections and the difficulties associated with that. Also at issue is the degree of neutropenia. What places the patient at higher risk? Overall consensus is that patients with an absolute neutrophil count of less than 100 cells per cubic millimeter are at highest risk. The duration of neutropenia also contributes to risk. So, this quotation sums up the fact that some patients will have fungal infections; some will have fever for other reasons; and, given our current problems with diagnosis, it really is impossible to tell which group the patient falls into.

[Slide]

Dr. Boucher already went over with you this morning the important points about this trial design and I am just going to highlight some of these things again to refresh your memory.

Study 603 was designed as a non-inferiority trial, with a non-inferiority margin of minus 10 percent. The primary analysis population in this trial was the modified intent-to-treat population, defined as any patient who received at least one dose of the study drug who also had information available on their outcome at least 7 days after the end of therapy. Patients in this trial were stratified by risk of fungal infections, whether they received antifungal prophylaxis, and by duration of neutropenia.

This was an open-label study design for two reasons. One is there is no orally systemically active form of liposomal amphotericin B. The other reason is that the applicant questioned the ethics of giving two I.V. infusions in seriously ill patients that may have problems with their volume status.

Finally, a data review committee blindly assessed the incidence of fungal infections and the outcomes of these patients as well.

[Slide]

There are two agents currently approved in the indication of empiric therapy of febrile neutropenic patients, and the approval for these drugs used a composite endpoint that was similar to that used in study 603. The first one to be approved was liposomal amphotericin B, or Ambisome. The label states that this drug is indicated for empirical therapy for presumed fungal infections in febrile neutropenic patients.

The other drug approved in this indication is itraconazole injection and oral solution. The label for this drug states that empiric therapy of febrile neutropenic patients with suspected fungal infections is the indication. There is also a note in the Sporanox label that states that the overall response rate was greater for itraconazole than amphotericin B deoxycholate in the study used to license this drug, but there were more discontinuations due to lack of efficacy in the itraconazole arm of the trial, and there were more discontinuations due to toxicity in the amphotericin B deoxycholate arm of the trial.

[Slide]

Again just to refresh your memory, the primary endpoint was the stratified overall response rate to the composite 5-component endpoint, again, stratified by risk of fungal infection, duration of neutropenia and receipt of antifungal prophylaxis.

The lower bound of the confidence interval to select the sample size and define statistical non-inferiority was defined as minus 10 percent for this trial based on the 50 percent overall response rate in mycosis study group 32 which compared Ambisome to amphotericin B deoxycholate and, again, that was, in turn, based on the 1994 or '95 workshop recommendations.

The stratified response rates that we see in study 603 show a 23.7 percent response rate in the voriconazole arm and a 30.1 percent overall response rate in the Ambisome arm. This gives us a difference of 6.1 percent in favor of Ambisome with a 95 percent confidence interval that ranges from minus 12 percent to minus 0.1 percent.

We performed our analysis at the FDA, again weighted for those three risk factors, and came out with a 95 confidence interval that ranges from minus 11.6 percent to a positive 0.1 percent. As we noted earlier, in either one of these analyses the lower bound of the 95 percent confidence interval falls below the prespecified minus 10 percent.

[Slide]

Just to make some overall statistical points about the finding of non-inferiority in non-inferiority trials, the lower bound of the 95 percent confidence interval, when it falls below or that is more negative than the prespecified limit, implies that that test drug may not be non-inferior to the control drug. However, defining non-inferiority implies that we have some knowledge already about the efficacy of the control drug over placebo, or in this case it would be no treatment, in a superiority trial. The lower bound of the 95 percent confidence interval that is used to define the non-inferiority margin in any non-inferiority trial cannot be greater than the difference in control drug over placebo or no treatment.

[Slide]

Let me give you a hypothetical example to try to illustrate this. If in a superiority trial the lower bound of the 95 percent confidence interval in a trial which tests a placebo or no treatment versus a control drug -- suppose this comes out to be 7 percent, if then we do another trial, a non-inferiority trial of control versus test drug, if the lower bound of the 95 percent confidence interval then, say, hypothetically comes out to be minus 11 percent, then we have a test drug that may be 11 percent worse than the control but the control may be no more than 7 percent better than no treatment. Therefore, the possibility exists that the test drug may actually be no better than placebo or no treatment.

[Slide]

So, what we do know is that it is clear that neutropenic patients do develop invasive fungal infections. What is the issue when we come to decide about clinical trials in selecting a non-inferiority margin are three issues. Do antifungal drugs prevent breakthrough infections in neutropenic patients? If so, what is the magnitude of this benefit relative to no treatment? Then, thirdly, how does this impact the selection of a non-inferiority margin in clinical trials where we are not using breakthrough infections as the primary endpoint but in which we are using a composite endpoint?

[Slide]

The two trials done by Pizzo and the EORTC comparing amphotericin B deoxycholate to no treatment had some issues that were associated with them. Neither of these trials were adequately powered to determine a difference in breakthrough infections. The Pizzo trial had 16 and 18 patients per arm in that trial, and there were 60 of so patients in each arm of the EORTC trial.

The EORTC trial also used resolution of fever as the primary endpoint and breakthrough infections were secondary analysis in that trial. Also, these two trials included mucosal as well as invasive disease in their descriptions of what they would call breakthrough infections, which is not what we do in the current trials. Neither of these trials included deaths and discontinuations as failures in an analysis of breakthrough infections. The reason that is important to do is that if a patient dies before they had a chance to develop a breakthrough infection one could consider them a failure. The other issue is that since we can't tell who actually has fungal infections that may be occult, given our current diagnostic techniques, a patient who dies and doesn't undergo an autopsy may have had an actual occult fungal infection.

[Slide]

The true difference of amphotericin B deoxycholate versus no treatment in those two trials in the prevention of breakthrough infections may range anywhere from 60 percent better to 8 percent worse than no treatment. Again, this raises the question how does this impact on studies that use our current composite endpoint, not a single endpoint of breakthrough infections? Which then, for the purposes of this trial, again raises the very germane question of what is the clinical relevance of a non-inferiority margin of minus 10 percent in studies in empirical therapy in febrile neutropenic patients?

Finally, how can we extrapolate from these studies, which were done almost 20 years ago, to our current rate of emergent fungal infections given the difference in care that exists now versus those studies?

[Slide]

Well, when the primary endpoint in the trial is not met one can attempt to explain the failure to meet that primary endpoint by looking by secondary and subset analyses. However, one must also keep in mind that these secondary and subset analyses are considered hypothesis generating in the setting of a trial which does not meet its primary endpoint.

The subset analyses of patients was stratified according to the risk of fungal infections and receipt of antifungal prophylaxis in this trial. And, the secondary endpoints were the five individual components of the composite endpoint.

[Slide]

In at attempt to adjust for multiple comparisons for these five secondary endpoints, as Dr. DeGruttola brought up this morning, we opted to use 99 percent, rather than 95 percent, confidence intervals to describe the differences between the study arms and the secondary endpoints. Since these were not the primary endpoint of the trial, obviously the trial was not adequately powered to determine true differences in the secondary endpoints and subsets.

Also, these secondary and subset analyses may actually contain small numbers of patients in each group, and you will see the numbers as I present them. Also, one of the things that is important is to look for consistency in outcomes of these secondary endpoints. In other words, do all the secondary endpoints trend in the direction favoring one drug or the other, or do some of the endpoints go in one direction and others go in the opposite direction?

[Slide]

Well, if we look at the overall response rate as a subset analysis by patients according to risk of fungal infection, you have already seen this morning, as Pfizer presented, that in high risk patients it appears that the difference is 1.7 percent in favor of voriconazole. If we then look at the 95 percent confidence intervals only in high risk patients, they range from minus nine percent to a positive 12.4 percent. Again, this would meet the statistical definition of non-inferiority, however, this is a subset analysis. In moderate risk patients the difference is minus 7.8 percent, that is, 7.8 percent in favor of Ambisome with a 95 percent confidence interval that ranges from minus 15.2 percent to minus 0.4 percent.

The other thing I would like to point out on this is that it appears that the response rate is consistent in the Ambisome arm between the high risk and the moderate risk patients, with 30 percent of patients being judged an overall success in the higher risk Ambisome patients and 31 percent being judged an overall success in the moderate risk Ambisome group.

On the other hand, in the voriconazole arm of the trial there is a 32 percent success rate, similar to that seen in the two stratifications here for Ambisome, but the moderate risk group in voriconazole has a 23 percent response rate.

[Slide]

If then we go on to look at the various secondary endpoints -- the reason that I put all five of them on here is to address the issue of do they all go in the same direction favoring one drug or another? Here again is the overall stratified response which shows the 6.1 percent in favor of Ambisome, with the lower bound of the confidence interval or minus 12.1 percent and the minus 0.1 percent upper bound of the confidence interval as presented by the applicant.

No breakthrough infections showed a difference in favor of voriconazole of positive 3.1 percent. Survival showed a difference in favor of Ambisome of 2.1 percent. No discontinuations. So, the difference in favor of Ambisome of minus 3.3 percent. The defervescence prior to recovery from neutropenia showed a difference of three percent in favor of Ambisome, and the response in baseline infections showed a difference of 21 percent in favor of Ambisome.

I have presented all of these as positives. I will flip some of these around so I can actually show you the number of deaths, but I just wanted to make the point that all the subsequent 99 percent confidence intervals are based on these differences but I put it this way so you can always focus on the lower bound of the 95 percent confidence interval in the succeeding presentations.

[Slide]

There were fewer breakthrough infections in the voriconazole arm compared to the Ambisome arm of the trial. In voriconazole there were 1.9 percent breakthrough infections versus 5 percent in the Ambisome arm. This is a difference of 3.1 percent in favor of voriconazole. The 99 percent confidence intervals for this ranged from minus 0.37 percent to a positive 6.5 percent. As presented earlier this morning, the difference was greatest in Aspergillus breakthrough infections, with four breakthrough infections caused by this pathogen in the voriconazole arm versus 13 in the Ambisome arm.

These breakthrough infections included proven and probable disease, and most of the patients had proven disease, with 6/8 patients in the voriconazole arm and 20/21 patients having proven disease in the Ambisome arm of the trial.

[Slide]

If one wants to perform a sensitivity analysis considering breakthrough infections as the primary endpoint, one would also want to consider patients who die as failures, for reasons that I talked about earlier. If one includes the 30 people who died prior to their end of therapy and the 18 patients who died prior to the end of therapy in the Ambisome arm, this results in a breakthrough infection rate of 9.2 percent in both the voriconazole and the Ambisome arm with a symmetrical 95 percent confidence interval here. This is 95 instead of 99 since this is the primary endpoint here because we are doing a sensitivity analysis using breakthroughs as the primary analysis.

[Slide]

Again, we do a subset analysis by risk of infection and prophylaxis, we can also see some differences occur here. If we look in the right-hand column we see that the difference is greatest in the high risk group, with 1.4 percent of patients having a breakthrough on voriconazole and 9.2 percent having a breakthrough infection in the Ambisome arm.

If we look at the moderate risk group the differences are not as large, with 2.2 percent patients experiencing a breakthrough infection in the voriconazole arm and 2.8 percent experiencing a breakthrough infection in the Ambisome arm.

The other way to look at this is by prior antifungal prophylaxis, as well along the bottom column, here, and here the difference is greater in patients who received prior antifungal prophylaxis, with 0.9 percent of patients developing a breakthrough infection in the voriconazole arm versus 5.2 percent in the Ambisome arm. There is still a difference here of 3.1 percent in the no prophylaxis group in the voriconazole arm versus 4.7 percent in the Ambisome group that did not receive prophylaxis.

Important to note here is that some of the numbers in these individual cells are actually quite small, with only one or two patients included in each group.

[Slide]

If we then move on to look at survival at seven days after end of therapy, there were more deaths in the voriconazole arm compared to the Ambisome arm of the trial, with eight percent of patients dying in the voriconazole arm within seven days of end of therapy and 5.9 percent in the Ambisome arm. This is a difference of 2.1 percent in favor of Ambisome with a 99 percent confidence interval, ranging from minus 6.9 percent to positive 2.7 percent.

The problem in any trial of patients this ill is the difficulty in attribution of death. As Dr. Boucher presented this morning, many of these patients had more than one reason checked off for the reason for which they died. Also, half of the patients who had sepsis also had progression of the malignancy, and half of the patients with progression of malignancy had sepsis, showing that the investigators have a difficult time picking out the actual cause of death.

The other difference between this trial and the global aspergillosis trial is that death here was not blindly reviewed by the data review committee as it was in that trial, and death was attributed by the investigators only in this trial.

[Slide]

If we then discuss discontinuations, there were more discontinuations due to lack of efficacy or toxicity in the voriconazole arm, with 9.9 percent of patients discontinuing therapy on voriconazole versus 6.6 percent in the Ambisome arm of the trial. This is a 3.3 percent difference in favor of Ambisome with a 99 percent confidence interval, ranging from minus 8.4 percent to 1.8 percent.

One of the issues with the composite endpoint is that it combines discontinuations due to lack of efficacy with those due to toxicity. This is really combining an efficacy and a safety endpoint in one and can obscure important differences in outcome. Therefore, it is important to look at the various reasons for discontinuations in these patients.

[Slide]

There were more discontinuations due to lack of efficacy in the voriconazole arm compared to the Ambisome arm or the trial, and 5.3 percent of patients discontinued in the voriconazole arm versus 1.2 percent in the Ambisome arm, which is a difference of 4.1 percent in favor of Ambisome.

However, on the other side, there are fewer discontinuations due to toxicity in the voriconazole arm compared to Ambisome, with 4.6 percent of patients discontinuing in the voriconazole arm and 5.5 percent. This was a difference of 0.9 percent in favor of voriconazole. Since these numbers were so small we didn't present 99 percent confidence intervals because we would have to further adjust them even for more multiple adjustments.

As Dr. Boucher pointed out this morning, more patients temporarily discontinued Ambisome than temporarily discontinued voriconazole, however, the discontinuation was actually specified as people who permanently discontinued in the protocol.

This actually brings up the issue of what happens in an open-label trial. The toxicities of amphotericin B are well known and, also, physicians may have been more likely to stop a patient on a new test drug versus a drug such as Ambisome which people have a more favorable opinion of as far as its efficacy, or at least they are more used to using.

[Slide]

There were more discontinuations with persistent fever as the reason checked off for lack of efficacy in the voriconazole arm of the trial, 14/22 patients discontinued because of persistent fever although two of those patients didn't have documented persistent fever. So, that would be 12/22. In the Ambisome group 2/5 patients were discontinued due to persistent fever.

The issue here is that the failure to become afebrile in a neutropenic patient may indicate the presence of an occult fungal infection and, as we have discussed already, current diagnostic techniques aren't good enough to pick out that a patient who only has fever may still have the presence of an occult fungal infection.

[Slide]

If we then move on to discuss defervescence prior to recovery from neutropenia, there were fewer patients in the voriconazole arm that met the protocol specified definition of prior to recovery from neutropenia. In the voriconazole arm there were 32.5 percent of patients who met this criterion versus 36.5 percent in the Ambisome arm, which was a difference of 4 percent in favor of Ambisome with a 99 percent confidence interval here, ranging from minus 12.7 percent to positive 4.7 percent.

However, these results are highly dependent on the definition of defervescence. Previous trials with itraconazole actually used different criteria for defervescence and allowed the patient to be considered a success if they defervesce at any time prior to recovery from neutropenia. However, in study 603 it required that the patient be afebrile for 48 continuous hours prior to recovery from neutropenia to be considered a success in this particular part of the composite endpoint.

[Slide]

If we look at the results of defervescence prior to recovery from neutropenia in the current study, 603, we can compare that to the two other drugs that are currently approved in empiric therapy of febrile neutropenia. In the Ambisome trial it was 58 percent and 58 percent were considered successes by this particular criterion. Again, there are important differences between these trials. For instance, this particular trial, which is mycosis study group, study number 32, actually included patients all the way down to two years of age. The current study actually includes patients only from 12 years of age on up.

The itraconazole trial allowed patients to defervesce at any time 28 days after randomization. So, it was clearly a different definition, and that is reflected in the different cure rates here -- 73 percent for itraconazole and 70 percent defervesce prior to recovery from neutropenia based on that definition in that trial. You can see the numbers I have already presented here, 33 percent for voriconazole and 36 percent for Ambisome, and you can just compare that to the other Ambisome trial and show that the 58 percent here for Ambisome and 36 percent here in the current trial for Ambisome.

[Slide]

So we performed a sensitivity analysis for the overall response rate and changed the definition of defervescence in study 603 to look more like the definition in the other trials. First we did a sensitivity analysis which changed the definition to the patient being afebrile for 24 hours prior to recovery from neutropenia. Here we see that this raises the success rate to 35.9 percent for this overall success in the voriconazole arm and 40.8 percent in the Ambisome arm, which gives us a difference of 4.9 percent in favor of Ambisome. Again, since we are looking at the overall response here we are using 95 percent confidence intervals that range from minus 11.7 percent to positive 1.9 percent.

If we then change the definition to defervescence at any time prior to recovery from neutropenia, these results look much more like the previous mycosis study group, study number 32, with a 50.1 percent success rate in the voriconazole arm and 56.2 percent success rate in the Ambisome arm.

Although this raises it to the level that is consistent with the prior trial, there is still a 6.1 percent difference in favor of Ambisome here with a 95 percent confidence interval, ranging from minus 13.1 percent to 0.9 percent, which again still would not meet the statistical definition of non-inferiority of minus 10 percent used in this trial.

[Slide]

Finally, to look at baseline infections, part of the exclusion criteria in this trial were that any patient with a known baseline infection was to be excluded from the trial. So, it is not surprising that there were very few of these patients in the trial. The reason that this is included as part of the composite endpoint is to try to get an idea of how good are these drugs at actually determining efficacy in documented infections. Unfortunately, you really can't determine that given the small numbers that are present in this trial. Five of ten patients and two of three in the voriconazole and Ambisome arms respectively had cures of their candidal infections, and there were only two patients in each arm with baseline Aspergillus infections. One patient in each arm recovered from those.

[Slide]

If we also look at some of the differences here between them, these small numbers make it very hard to draw an interpretation. Again, most of these patients are excluded from the trial at baseline and the only difference between these patients and the patients that are excluded is that their culture happened to come back a couple of days later.

Also, two of six patients in the voriconazole arm were considered successes at the end of therapy. However, they developed disseminated candidiasis, one patient at one month and one patient at two months after end of therapy. There was one patient out of those four successes in the Ambisome arm who developed disseminated candidiasis 20 days after end of therapy. So, although we would consider these people cures based on the definition in the protocol, they may clinically be considered failures. Again, that is not a fair comparison because by definition the people that are in this trial remain neutropenic and some of the cure rates of these patients are highly dependent upon whether they recover their neutrophil count or not.

[Slide]

So finally, to sum up, drugs in the empiric therapy of febrile neutropenia should have proven efficacy versus documented Candida and Aspergillus infections given the nature of these trials. The aspergillosis global study, presented here today, shows efficacy in Aspergillus and Pfizer has presented data on Candida esophagitis and study 608 on candidemia is still ongoing. Since therapy in this indication is empiric, patients in the empiric therapy of febrile neutropenia may actually receive treatment and not have fungal infections, which makes it important to look at the safety profile of the drug.

To sum up the points that we have made today, in this particular trial voriconazole fails to meet its statistical definition of non-inferiority as it falls below the minus 10 percent lower bound of the confidence interval that was prespecified in the protocol.

The subset analyses of the overall composite endpoint showed a numerical advantage of voriconazole in high risk patients but an advantage of Ambisome in the moderate risk patients.

The secondary analyses of breakthrough infections showed a numerical advantage of voriconazole, especially in the prevention of Aspergillus infections. However, a sensitivity analysis that included deaths as failures showed no difference between the drugs in the incidence of breakthrough infections.

Finally, secondary analyses of the other four components, other than breakthrough in infections, were in favor of Ambisome.

[Slide]

So, the considerations that we would like the committee to discuss today include what is the clinical relevance of a non-inferiority margin of minus 10 percent in trials in empiric therapy of febrile neutropenia that use this composite endpoint, given what we know about this indication?

Finally, how do we look at secondary analyses in a study which did not meet its primary endpoint? Finally, we need to consider the safety profile of a drug in empiric therapy where some patients may receive the drug who do not actually have infections and, therefore, would derive no benefit from the drug.

Given that last consideration, I will pass the baton back to Dr. Tiernan who will give the FDA perspective on the safety issues associated with voriconazole.

[Slide]

DR. TIERNAN: For the safety section of this I will pretty much focus on the following five areas: ocular, cardiac, hepatic safety, rash and drug interactions.

[Slide]

In the preclinical evaluation of this drug pharmacology-toxicology studies have demonstrated that voriconazole produced dose-related effects in the ERG of dogs exposed to voriconazole. The voriconazole plasma levels which produced these results in dogs were similar to those plasma levels achieved in human studies.

Histopathology results for female rats which received 50 mg/kg of voriconazole, which is equivalent to 8 mg/kg I.V., demonstrated mild thinning of the outer layer of the retina at 24 months. During the clinical trials the incidence of visual symptoms was one out of every three subjects, and the symptoms included decreased vision, photophobia, altered color perception and ocular discomfort. The exact mechanism underlying these visual symptoms is unknown. There is no human histopathology data and ocular biomicroscopy has not detected ocular lesions.

[Slide]

Results from study 150-1004, a study in which voriconazole was given to healthy volunteers at a dose of 400 mg q. 12 hours for one day and 300 mg q. 12 hours for 27 days, are listed as follows.

Effects were noted in the ERG, Mamsworth-Munsell 100 hue test for color vision, and Humphrey Perimetry studies for visual field. Voriconazole has effects on both rod and cone function. Decreased visual function is present on the first day and continued through 28 days of therapy. At 14 days after the end of therapy the visual function returns to normal.

[Slide]

Additional issues regarding the use of this drug include that there is insufficient information to predict what the ophthalmologic effects will be in patients who are either rechallenged or retreated with voriconazole. Ultimately, if this drug is considered for use in children less than nine years of age, we do not have sufficient data to predict the effect of voriconazole on the eye which is not yet fully developed. A careful risk-benefit assessment will have to be made when considering the use of voriconazole in patients with underlying eye disease, such as CMV retinitis. We cannot predict that visual changes will resolve if this drug is used beyond 28 days of therapy.

[Slide]

In vitro studies of this drug demonstrated no major effects for voriconazole in HERG channel studies or in dofetilide studies when compared to ketoconazole. In vivo data demonstrated that in dogs high doses of voriconazole produced arrhythmia, PVCs and prolonged QT intervals.

As already discussed by the applicant, there was one sudden death in the Phase III clinical trials for which a role for voriconazole could not be excluded. In response to this event, investigators were asked to monitor patients with telemetry if the patient had underlying heart disease and was to receive intravenous voriconazole. The applicant planned to further investigate the effect of three intravenous doses of voriconazole at 4 mg/kg, 8 mg/kg and 12 mg/kg on QTc interval in healthy subjects aged 18-65 years. However, this study was terminated because of anaphylactoid reactions, and the applicant has conducted a thorough investigation and no causative factor has been found for those anaphylactoid reactions and the applicant has committed to complete a similar study but utilizing the oral preparation of voriconazole.

[Slide]

In the controlled clinical trials, studies 307-602, 603 and 305, no major differences between voriconazole and controls have been detected in the cardiac adverse event reporting or rate of discontinuations specifically for events such as arrhythmias and congestive heart failure. However, the voriconazole arm in study 307-602 did have more grade 3 cardiac adverse events and a role for the drug could not be excluded.

It is important to note that these clinical studies were not specifically designed to assess the risk of developing arrhythmia in a population who may be on multiple medications, have underlying heart disease and may also be on antiarrhythmic drugs.

[Slide]

Pfizer has already presented information on the hepatic safety of voriconazole and, in general, we are in agreement with the information as presented by Pfizer. Our conclusions are similar in that the data from Phase I/Phase II studies support that there is a positive dose or an exposure response relationship between voriconazole and transaminase increases, and a possible association with alkaline phosphatase increases.

In the Phase III comparative studies hepatic adverse events and abnormalities of transaminases and alkaline phosphatase were more frequent among patients taking voriconazole than fluconazole. The hepatic adverse events were similar between voriconazole and its amphotericin B comparators, with the exception of bilirubin elevations which were more frequent among those treated with amphotericin B formulations. While the numbers are small, serious adverse events were reported more frequently among patients receiving voriconazole in the Phase III comparative studies.

Regarding hepatic failure deaths, Pfizer has already described the findings of the expert panel of hepatologists who reviewed those specific deaths.

[Slide]

We agree with the applicant that rash is a potential hazard associated with the use of this drug. Patients in the clinical studies were often on concomitant medications that could either cause rash themselves, or concomitant medications such as antihistamines and steroids that might affect the type and severity of skin symptoms observed. Conditions such as graft-versus-host disease can also make it difficult to completely assess a causative role for study drug and the development of rash.

[Slide]

In the controlled clinical trials rash was observed in voriconazole study patients at a rate similar to that of its comparators. Overall, rash develops in about 18.6 percent of patients receiving voriconazole. The applicant has already presented several descriptions of the various types of skin examples observed. No specific rash is characteristic of voriconazole exposure, and there is insufficient information at this time to conclude that the rash is due to photosensitivity.

Most rashes were of mild to moderate severity. There were no major differences in the discontinuations for rash between voriconazole and its comparators. However, the applicant, in their briefing package, has already described four non-fatal cases of Stevens-Johnson syndrome which were noted in patients on voriconazole. Two of these cases developed rash; discontinued drug and then re-exacerbated upon rechallenge. It should also be noted that there was a case of toxic epidermal necrolysis reported for amphotericin B in the trial as well in the database.

[Slide]

We will finish with drug interactions. To address the safety implications of drug interactions with voriconazole, the in vitro metabolism studies performed with human hepatic microsomes and genetically engineered cell lines indicate that voriconazole is both a substrate and an inhibitor of three cytochrome P450 enzymes, 2C19, 2C9 and 3A4.

The substrate affinity and inhibition potency of voriconazole is greater for 2C19 and 2C9 compared to 3A4. For comparison, the potency of voriconazole as an in vitro inhibitor of 3A4 appears to be weaker than ketoconazole and itraconazole. The in vitro potency of voriconazole to inhibit the metabolism of 3A4 substrates, and for 3A4 substrates to inhibit voriconazole varies among classes of drugs, including HIV protease inhibitors, non-nucleoside reverse transcriptase inhibitors and immunosuppressant drugs.

[Slide]

The applicant has evaluated representative substrates, inhibitors and inducers of the three CYP enzymes both in vitro and in vivo, however, it is not possible to evaluate every potential drug interaction. To illustrate, representative protease inhibitors and non-nucleoside reverse transcriptase inhibitors were studied in vitro but not in vivo. The exception is indinavir which was studied under both conditions and found not to interact with voriconazole. However, other protease inhibitors and non-nucleoside reverse transcriptase inhibitors are known inhibitors and/or inducers of CYP3A4, and the clinical significance of an in vivo interaction with voriconazole is currently unknown. Therefore, the potential for drug interactions with voriconazole presents a therapeutic challenge for the prescriber when attempting to manage patients on multiple concomitant medications.

The applicant states that these drug interactions are manageable but, please, keep in mind that this is predicated on experience within the setting of a carefully monitored clinical trial and we look to the advisory committee for advice regarding additional drug interactions that may need to be explored.

This concludes the safety presentation. I can take some questions.

DR. GULICK: Thanks, Drs. Tiernan and Powers. We will open it up to the committee for questions of detail or information. Dr. Stanley?

Questions from the Committee

DR. STANLEY: Again obsessing with the ocular abnormalities, did I miss it -- they did have in the briefing document a little bit about some pediatric dosing trials but there was nothing that I saw about evaluation of ocular effects in the pediatric population, or did I miss it?

DR. TIERNAN: I am not familiar with any of the pediatric data, to be honest.

DR. GULICK: Dr. Schapiro?

DR. SCHAPIRO: I think you mentioned that there was retinal thinning in an animal model. I wonder if you could elaborate on that?

DR. TIERNAN: Yes, and I will call on our pharmacology/toxicology staff to answer that, and that was at 24 months I believe.

DR. MCMASTER: Could you repeat the question, please?

DR. SCHAPIRO: Yes, there was mention that in an animal there was a finding of retinal thinning. We hadn't heard before about any histopathology regarding the retinal damage. Could you elaborate on the findings in the animals?

DR. MCMASTER: There were a number of studies that were done -- shorter studies, one year, six-month studies and up to 24 months in rats. The earlier studies did not show findings. The later studies, in particular one 24-month rat study in which the sponsor examined the thickness of the retinal layers, as mentioned earlier, the center of the retina was fine but as they looked to the outer layers, the outer layers were thinner in the female rats on the high doses. This was not found in the male rats and this was not found in the animals that were dosed for shorter durations. So, this a longer-term finding, which is very consistent with the findings that we had in the patients and findings in the dogs.

DR. GULICK: Dr. Wong?

DR. WONG: What we heard from the sponsor was that this seemed to be a functional abnormality that was completely reversible upon withdrawal of the drug. If with prolonged exposure at high doses we have a morphologic abnormality or structural abnormality, that is really different from what we heard about earlier this morning, isn't it?

DR. BAILDON: I could comment on that.

DR. GULICK: Okay, let's go ahead and do that just to clarify this one issue.

[Slide]

DR. BAILDON: This is the 24-month carcinogenicity study. That is actually where this was observed in rats. This is a 24-month study which actually covers the full lifetime of the rats, and these are albino rats that are sensitive to light and actually show retinal thinning as an age phenomenon. That retinal thinning as an age phenomenon is somewhat dependent on light exposure and varies between animal and animal.

What you see here in carcinogenicity studies -- at that time two controlled groups were usually used for these studies -- and if you look at the thickness of the peripheral retina, this is a cell count of cell layers in that retina and you can see there is considerable variability between the control groups, and the same here for female animals where you see some variability, less so in the peripheral retina. The finding that the colleagues at FDA described is in the female who, in the peripheral retina, had a thinning at the 50 mg/kg dose at 24 months compared to the control group. It was not observed in the same animals in the central retina, and it was not observed in the male rats.

[Slide]

Then, this slide shows our findings. The same analysis in dogs treated for 12 months, and my assessment was related to the dogs which display the same electrophysiological phenomenon that we have observed in humans. These dogs were treated for 12 months at 12 mg/kg, which is a toxic dose. It shows abnormalities at that dose. There, we see no difference in central or peripheral retina between dogs treated with voriconazole or control animals.

DR. GULICK: Thanks for that clarification. I don't know if the agency has further comment.

DR. MCMASTER: We agree that there is, in fact, only mild thinning but, if there is an age effect, it would be expected to affect all the animals because they are all at the same age. We maintain that there is a slight change, however, in the context of the very remarkable findings in the patients, and in the context of the findings in the dogs this represents an actual histopathological change which might, in fact, reflect something that could be found when patients are treated for a long time.

DR. GULICK: One other comment?

DR. TIERNAN: Yes, this is Dr. Chambers, our ophthalmologist at FDA, who did the ocular safety.

DR. CHAMBERS: Yes, I am Wiley Chambers. What we observed is that the findings in the animal have shown up before in histopathology findings. We don't have any long-term human findings that would correlate to the same degree as what was seen within the animals. Yes, there is a natural decrease in layers, but what you saw, what was just displayed, is that all the groups go down but there is a statistical difference from the control group in that one set and we don't know what that means, and we have no corresponding long-term human data to match up and tell you whether that occurs in the species or does not. With the other animals, the dog animals, you don't have a comparable long-term history. As was pointed out, this was a carcinogenicity study so it was carried out for the lifetime of the animal. We have nothing comparable in any other groups.

DR. GULICK: Thank you. Other questions of clarification or detail? Dr. Yogev?

DR. YOGEV: Maybe I missed it, but was there any data about the teratogenicity at all?

DR. GULICK: Please come to the mike.

DR. MCMASTER: I missed the question.

DR. YOGEV: Data about teratogenicity.

DR. MCMASTER: Right.

DR. YOGEV: Should this drug be prevented from women who are suspected to be pregnant?

DR. MCMASTER: Well, yes, there is teratogenicity. In fact, we have cleft palates in some animals that were treated, and there are a number of other findings which were not unexpected, this being an azole. This would be a Class C drug and so pregnant women would be cautioned against using this.

DR. GULICK: Dr. Hamilton?

DR. HAMILTON: I am sorry, I may have missed this but I was obsessing about something else while Dr. Stanley was obsessing about the visual thing. Could you tell me the number of people who have been followed the longest on voriconazole, and what the length of that follow-up is with credible ophthalmologic examinations, both physiologic and objective?

DR. CHAMBERS: The one study that you heard is the only kind of sophisticated visual function testing.

DR. HAMILTON: That is how many people?

DR. CHAMBERS: I don't have the number in front of me. It is not a large group.

DR. GULICK: Could we just clarify? What you said was that a group hasn't been followed with the diagnostic testing through 28 days, but the question was how many people have been followed clinically?

DR. HAMILTON: Including with some attempt to assess --

DR. CHAMBERS: Yes, with sufficient testing to be able to tell, and it is only that one trial. Straight visual acuity testing, as was presented, was not done in a sufficient form to necessarily pick up small changes. A lot of the discussions that have gone on during the trials have been are these patients well enough to be able to get to facilities to do the testing you would like to see have done, and have the baseline information earlier on to be able to tell if the changes that you see, are they due to drug product or due to something else. But to go back to answer your question, we essentially have good testing on 28 days of dosing.

DR. GULICK: Other questions from the committee? Go ahead.

DR. HAMILTON: Dr. Tiernan, I don't think I understood well enough to ask the question that I want to ask from slide number nine that you presented.

DR. GULICK: Was that in the first set or the second set?

DR. HAMILTON: First, study 307-602, additional efficacy analyses.

DR. TIERNAN: Yes, that was to corroborate the initial results that we had.

DR. GULICK: Can you speak up? I am sorry, we can barely hear you.

[Slide]

DR. TIERNAN: This was just to corroborate the initial results that we had in 307-602, and these were just attempts to perform a more conservative analysis.

DR. HAMILTON: Using what parameters? I mean, what were the assumptions for these more conservative analyses?

DR. TIERNAN: The first one was that the DRC was allowed to upgrade investigator assessment. So, we did an analysis that did not allow the DRC to upgrade the assessment of the patients and we penalized for that. So, that is the most conservative analysis that we could have done in that situation.

Then, in the second one if the voriconazole patients switched to other licensed antifungal therapy we actually penalized them for that. We considered them failures. So, we felt that was a very conservative approach. Then we looked at four weeks after the end of the primary endpoint, which was 16 weeks, to make sure that there wasn't a problem with relapses.

DR. HAMILTON: I see. Thank you.

DR. GULICK: Other questions for the agency from the committee? I have two short ones. We have heard twice that two patients with persistent fever actually didn't have persistent fever. Can you explain that?

DR. POWERS: The way this works is that in the case report form the investigators are allowed to check off why the patient discontinued. So, in two cases an investigator checked off fever. When you go back and you look through the logs there is no fever. So, it was inadvertently checked off on the case report form.

DR. GULICK: Okay. My second question was, if I understood you correctly talking about the lower bound for non-inferiority, did you say that the minus 15 percent picked for the itraconazole study actually wasn't prospectively chosen?

DR. POWERS: No, it was prospectively defined. I guess what I am trying to say is if you look at these three trials, the MSG32 trial to test Ambisome versus amphotericin B deoxycholate, there was discussion about what the lower bound should be. There was discussion about what the lower bound should be for this trial. For the itraconazole trial there was no discussion. The protocol came in that way and that is the way it stayed, and there was no active discussion about what it should be.

DR. GULICK: So, it was arbitrarily selected and not really focused on?

DR. POWERS: I could actually show slide 40. Let's see how long it takes us to pull this one up.

[Slide]

This is actually a comparison across these three trials. Again, I want to caution people about making comparisons across these because there are a lot of differences between these trials as far as demographics and other issues about the trials.

In the Ambisome versus amphotericin B deoxycholate trial, the top one which is MSG32, we can see that this trial had a prespecified lower bound of minus 10, and we see 50 percent response rates -- about a similar number of patients that there are in this trial, but they meet their lower bound. They come out to be minus 6.8 percent.

The itraconazole trial had a prespecified lower bound of minus 15. You can see the differences here in the cure rates, but they met their delta of minus 1 percent of the lower bound, with an upper bound of 20 percent. Again, the caveat went into the label for this drug that there were more discontinuations due to lack of efficacy in the itraconazole arm and more discontinuations due to toxicity in the amphotericin B deoxycholate arm.

Then we get down here and we look at this trial, 603, and here we have again almost 400 patients. The reason why there are fewer patients in the itraconazole study is because it was designed with a lower bound of minus 15 which allows you to set a sample size that is smaller. But you can see here that even though we use a prespecified lower bound of minus 10, in either the raw or stratified analyses it does not go above minus 10.

DR. GULICK: Thanks. Any further questions from the committee for the agency?

[No response]

I have about 12:45 and we will break for lunch for an hour. So, we will reconvene at 1:45, right on the dot.

[Whereupon, at 12:45 p.m., the proceedings were recessed, to be resumed at 2:00 p.m.]

AFTERNOON PROCEEDINGS

DR. GULICK: There is no one who has officially asked to speak at the open public hearing, who made that known ahead of time. Is there anyone who would like to speak that didn't sign up ahead of time? Seeing none, we will close the open public hearing and we are now four minutes ahead of the agenda.

[Laughter]

As one point of clarification, there was a lot of discussion towards the end of the last session about the ophthalmologic findings. I would like the sponsor, who has offered to present us a couple of clarifying slides, to start off with that and then we will proceed to the charge to the committee.

DR. BAILDON: Thank you very much. I just do want to share the clinical data we have.

[Slide]

This is the slide on the acuity changes observed from baseline to follow-up in our esophageal candidiasis study, where we could investigate visual effects a little better than in some of our other studies. The median treatment duration in this study was close to two weeks, which is a similar duration you would see in an empirical therapy study and as I have shown you before, there were no changes in acuity between groups.

[Slide]

This shows the same population and the investigation we did here is contrast sensitivity testing. So, this is testing contrast sensitivity and, again, we see no difference in changes from baseline to end of therapy between fluconazole and voriconazole.

[Slide]

We have then done the same test for color vision where we look at color vision changes and, again, we see no differences between the two groups.

[Slide]

This is mostly an HIV-infected population and we also looked at abnormal fundoscopy at end of therapy, about two-week therapy with either fluconazole or voriconazole and, again, there is no immediate apparent difference between in the two groups.

[Slide]

As Dr. Chambers quite rightly highlighted, we don't have very much good data on long follow-up and I want to highlight that we are committed to a long-term safety study, ocular safety study in patients where we are recruiting patients with Paracoccidioides infections who require at least six-month therapy and who are also ambulatory. Of the 42 patients targeted, we have recruited 26 patients and 4 have completed our protocol. That is an ongoing protocol where we are doing more extensive visual testing, along the lines suggested by Dr. Chambers, and actually in the break we agreed on a few extra measures that should be part of that to provide the detail that we discussed earlier. So, we hope that this study will then provide at least 6-month data. Patients are treated for at least 6 months and then longer. If necessary, they can go on longer, and they will undergo this extensive visual testing at baseline and then at the last follow-up. That was all I wanted to add.

DR. GULICK: Thanks for that clarification. Are there additional questions just to follow-up on these specific points? Dr. Wood?

DR. WOOD: Is there any plan to try and capture ocular toxicity, or at least assessment of ocular toxicity in pediatric patients?

DR. BAILDON: Well, I discussed these 38 patients who were treated for more than one year. We actually saw a lower reporting frequency of visual adverse events than in the overall population.

DR. WOOD: And those were all pediatric patients?

DR. BAILDON: They were not all pediatric but a good part of that population is actually pediatric patients. Dr. Boucher highlighted that early on in the program we have seen very good efficacy in central nervous system disease, and we allowed compassionate treatment of children, in particular if they had failed other agents and fatality is around 100 percent in that, of CNS disease. So, these children, treated for a long time, are actually more young children. The problem we have there is that the lower frequency of reports partly is due because they don't report it after a long time anymore but young children don't report it anyway.

I think Dr. Thomas Walsh, in one of our studies, had a child who said it was pretty cool when she experienced this, but it is a bit more difficult to test that and we have not done any visual real testing on the children.

DR. GULICK: Thanks again. Dr. Goldberger will now review the charge to the committee.

Charge to the Committee

DR. GOLDBERGER: Thank you.

[Slide]

Question number one, is there sufficient information to support that voriconazole is safe and effective for the treatment of invasive aspergillosis? If not, what additional information would be needed to support this indication?

Unless the committee has any questions, we felt that this was a pretty straightforward question.

[Slide]

Question number two, is there sufficient information to support that voriconazole is safe and effective for empiric antifungal therapy in febrile neutropenic patients? If not, what additional information would be needed to support this indication?

I would make a couple of comments about this. One is that unlike a separate question on study design for aspergillosis studies, for future aspergillosis studies, we didn't include a separate question for study design for empiric therapy, I think in part because we expected that during the discussion there would be some comments about endpoints and other issues which would be helpful. Although, certainly, if the committee later on wishes to talk more about study design in this area, that would be fine.

The second thing is there was quite a bit of data presented this morning about analyses of how different subgroups perform, different components of the overall endpoint in empiric therapy performed. Should the committee have an interest in thinking about this indication rather than simply yes or no but, rather, attempting to define a group for whom the drug might be indicated, or should the committee want to provide some specific advice about caveats that should be included in any indication, that would be fine.

What we would obviously like though is advice that is sufficiently specific so that it can be included in product labeling and be useful and comprehensible for treating physicians. So, in other words, a lot of data was presented and either the sponsor or ourselves can review some of these analyses. But should you wish to do that, and we are in no way recommending that you do but a lot of time was spent discussing this, we would like advice, and sufficiently specific, so that it could be helpful ultimately in allowing physicians to make appropriate decisions in how to use the drug.

[Slide]

Question number three, what additional Phase IV studies would you recommend? Obviously, one thing here we are certainly interested in to a large degree but by no means exclusively is any additional advice you have about drug interaction studies. Obviously, a fair amount was presented about drug interactions. The company certainly has made a substantial effort. There are still some unresolved questions so that specific advice about this would be most welcome, as well as any other advice about Phase IV studies.

[Slide]

Finally, what additional advice does the committee have regarding the design of future studies needed in the development of therapeutic agents for the initial therapy, and therapy of patients refractory or intolerant to other antifungal therapies, in patients with pulmonary and/or disseminated aspergillosis?

We thought that we would ask this question again, in part because this was a question we asked last January when another product came before the advisory committee for an aspergillosis claim, although a somewhat different claim than this one, i.e., limited to salvage therapy. And if, based on the data you have seen today, you have any other observations in this area they, again, would be most welcome. In addition, should you have any other comments regarding design of trials for empiric therapy, they also would be most welcome. That is basically it unless there are questions.

Committee Discussion and Vote

DR. GULICK: Thank you. Just to inform the committee, I think the way I would like to do this is to take the questions one by one. We will start with some general discussion and then at the conclusion of the discussion we will take a formal vote among the voting members of the committee; the same for question number two. Questions number three and four are really additional advice from the committee. We won't be taking formal votes on those particular questions. So, let's start with question number one.

Is there sufficient information to support that voriconazole is safe and effective for the treatment of invasive aspergillosis? Dr. Wood, would you like to start?

DR. WOOD: I think the answer is yes. The data is pretty straightforward and I don't have any issues or concerns about invasive aspergillosis.

DR. GULICK: Dr. Mathews?

DR. MATHEWS: I agree.

DR. GULICK: Dr. Hamilton?

DR. HAMILTON: I basically agree as well, I would just like to make a few comments, however. Probably not everyone has overlooked the very important facts that we are dealing with both very serious underlying diseases and consequent very serious infections. So, we have to view any decisions we make in light of that. This is not a trivial circumstance by any means. I think that is important both in terms of the assessment of the urgency of new treatment options and in the assessment of whatever safety considerations there may be and in light of what our current therapeutic options are which are, in my view, imperfect, at best, as will this one be. We are not talking here about a therapeutic intervention that is going to solve disseminated aspergillosis. We hope it has some substantial impact but it is not going to eliminate that problem. So, we hope others will be searching for even newer and more effective and potentially less toxic interventions as well.

As to that broad array of diseases that we have left out of the analysis for today, because I am sure there had to be just a whole lot of other serious problems that were encountered that have not been focused on today, some of which fall into an infectious category, I would guess that the sponsor probably has substantial information about that population of illnesses, and I would hope they have plans to pursue a formal analysis of those to see whether their drug or potentially others might have interventions that would help. It is a pretty desperate population, I would say. So, I would encourage rapid assessment of that.

Although there remain some issues of safety here, again in the context of the problem is we are dealing with, I, for one at least, am willing to deal with the uncertainty of absolute safety of this drug to proceed, but I strongly recommend that the sponsor pursue what avenues should be pursued in light of what has been said about the visual changes, etc. I vote yes.

DR. GULICK: Just to clarify, we are not taking a formal vote yet, but appreciate people's opinions. Dr. Schapiro?

DR. SCHAPIRO: I would also agree definitely that the answer to the first question is definitely yes. I think to stress a little bit what additional studies would be done, I think that, again as we have discussed in the past, looking at the benefit/risk ratio for the different indications, I think it is an easy answer for us for invasive aspergillosis. Based on what Dr. Hamilton said now, that is definitely a situation where with what we have now I think it is definitely justified to approve the drug and give it to patients.

As we maybe slide down to less life-threatening, urgent conditions --

DR. GULICK: Can I suggest that you hold -- finish your thought.

DR. SCHAPIRO: Just one point. I think that that should not stop us. The fact that we consider it safe for one indication -- safety is relative, and it is relative to the urgency. So, although I think that for that indication we definitely have seen data, it is not a matter if the drug is safe or not safe. I think for that indication for that serious infection, yes, it is safe.

DR. GULICK: Thanks. Other committee members wish to ring in her? Dr. Wong?

DR. WONG: Sure, I mean I guess I want to begin by saying that I really want to congratulate Pfizer for putting this together. I had my doubts before I saw the data that we saw here today that anyone was going to be able to show us such powerful data as a mortality difference between a standard therapy and a new therapy for aspergillosis, and I am delighted to see that and I think you are really to be congratulated. So, the answer to the question is, yes, it is safe and, yes, it is effective.

DR. GULICK: Dr. Englund?

DR. ENGLUND: I agree, yes; yes.

DR. GULICK: It sounds like we are reaching some consensus here. Others want to make a statement? Dr. Morrison?

DR. MORRISON: I agree with what has previously been stated. The one question I have is does any caveat need to go into the package insert or information regarding this compound with regard to the non-fumigatus Aspergillus species? The fact that the response rate in these may not be as good, but we really don't know that for sure because the patient numbers are small.

DR. GULICK: Barring further comments, clearly the consensus of the committee is we are delighted to see a randomized study for this particular disease. As Dr. Wong mentioned, the results are quite impressive in terms of demonstrating a survival benefit. As others pointed out, assessing a risk/benefit ratio in this particular population is extremely important -- a very compromised group; a disease that is quite severe, and the current treatment options are suboptimal.

So, the consensus seems to be that we find the drug safe and effective. I would like to take a formal vote so each voting committee member can go on record as stating what they believe. So the question again, is there sufficient information to support that voriconazole is safe and effective for the treatment of invasive aspergillosis? Dr. Rodvold, we will start with you.

DR. RODVOLD: I vote yes, and I again endorse that this is a great study.

DR. GULICK: Dr. Wood?

DR. WOOD: Unequivocally yes.

DR. GULICK: Dr. Mathews?

DR. MATHEWS: Yes.

DR. HAMILTON: Yes.

DR. YOGEV: Yes, but just at the bottom, if yes, what additional studies?

DR. GULICK: We can come back to that.

DR. YOGEV: If you are going to come back to it, I will talk about it later but I think it is important to realize that this is a devastating disease. There was breakthrough on the dose which was recommended and it is not about 50 percent, and I would encourage the company to go with a higher dose compared to this dose to show that although in an animal model there is something that suggested the same rate of reduction in the amount of the fungus, we might get an additional benefit without increasing the risk.

DR. GULICK: So, can I suggest that we come back to this point when we talk about Phase IV commitments and further studies that need to be done? Is that okay? We still need to record your vote though, Dr. Yogev.

DR. YOGEV: I said yes.

DR. ENGLUND: Yes.

DR. SCHAPIRO: Yes.

DR. WONG: Yes.

DR. DEGRUTTOLA: Yes.

DR. GULICK: And the chair votes yes. That is unanimous, ten votes yes; no votes no. So, with your permission, Dr. Yogev, we won't answer the second part of that question.

Let's move to question number two, which should generate some discussion. Is there sufficient information to support that voriconazole is safe and effective for empiric antifungal therapy in febrile neutropenic patients? There are a number of issues to consider in answering this question. Dr. Schapiro is going to start.

DR. SCHAPIRO: I think it is a difficult question to answer. I think we saw a very nice presentation from the sponsor. I think the analysis done by the agency was very helpful, especially for us who are less schooled in the statistics. I think it is very clear that this is a complex question and it is not one necessarily that has a yes/no with the statistics.

Going back to the issue of risk/benefit, I think that there are definitely a lot of open questions with the toxicity. They have been well addressed and well discussed. I think there is still concern regarding the visual disturbance. I think that we have no data. The patients who died in the study were not investigated with histopathology from the retina of those patients, which might have given us some insight, and I think that there is not a lot of data, not any long-term data and that is concerning.

I think that for the hepatic toxicity it is a pity we weren't able to delineate some markers that would help us identify the high risk patients. Although rates of Child's A or B cirrhosis are conditions which would have a dose reduction, as I think was brought out earlier, many patients that you see, you don't know that. You have some clinical markers but we weren't given any guidelines that a clinical can use and often we don't have time to make a diagnosis of cirrhosis. So, that is of limited help.

Of course, therapeutic drug monitoring -- there was some data in the background that in some of the early studies it looked like there were thresholds at 5, maybe 0.5-5 but, of course, in the analysis that the company did it didn't pan out. I think maybe in the future if more work does define some good clinical markers and maybe a drug level, I think that would greatly help us reduce some of the concerns with the toxicity. It may also allow us to increase the doses if we were able to monitor how high they are going. So, I think that is also of concern when we look at that.

I am pretty much convinced that that is an alternative therapy. I don't know if we have wording where we can say that, you know, we would use it if Ambisome was contraindicated, if there is that type of wording that can be put in. That was sort of my feeling from this.

I am also not convinced -- the summary statement that this drug is better tolerated than Ambisome, I wasn't convinced by the data we saw that it is a more tolerated drug. I thought maybe there is data going both ways. But with what we have seen now, the current drug looks a little bit better.

DR. GULICK: Dr. Wood and then Dr. Yogev.

DR. WOOD: Why don't you go ahead?

DR. GULICK: Okay, Dr. Yogev first.

DR. YOGEV: My answer will be definitely no. I think all the complexity is part of the issue, but as the data were presented it is not. If I take the Aspergillus out I have even more concerns. For some bizarre reason, this drug did less good in Candida than one would expect. This is a population where 1/20 really need it, and some people say it is higher. With all the toxicity you mentioned, I think this drug should wait for much better data to convince me.

Now, if you tell me there is Aspergillus, we already answered that in number one. But if you don't know, is it empiric, my answer is no.

DR. GULICK: Others want to make comments? We will take one at a time, Dr. Rodvold and then Dr. Wong.

DR. RODVOLD: Well, I guess one of the caveats that struck me was that there was a category of high risk patients that did better. They had bone marrow transplant people and another group there, and I just wondered if language is needed for alternative therapy for them. When you take them away from there and you get into some of the others, the moderate risk group and things like that, then the data, you know, goes swinging the other way. So, the high risk group was something that either needs discussion or whether or not that can be incorporated into a label underneath this indication.

DR. GULICK: Dr. Wong?

DR. WONG: I guess if I had to give a yes or no answer, have they demonstrated the efficacy of this drug for empiric therapy, I would say not yet. But I want to make a couple of comments to put our interpretation of these results in the context of what we have seen earlier.

I wasn't on this committee back in '94 or '95 when it was decided that these sorts of studies should be analyzed using this composite endpoint that has now been used three times in a row. But I was here the first time we heard data like this when we looked at Ambisome versus Fungizone. I believe I commented at the time that I thought that using an endpoint like that was really going to hurt people down the line because it mixes too many things. It combines efficacy and toxicity, which I think is clearly a big mistake. It also gives the local investigators preconceived ideas about whether or not a drug is going to work, who is not blinded. It gives that person the opportunity to really determine the outcome in an individual patient by withdrawing his patient from that study drug. It makes efficacy assessment in a study like this almost impossible. That is what I thought a few years ago when we saw the Ambisome versus Fungizone results and I think that is precisely what has happened here.

I think the study, as designed, was negative, but I also believe that this drug is probably going to be found to be useful as empiric antifungal therapy for neutropenia patients with persistent fever once we get past the idea that we should be trying to make up a fuzzy endpoint to get around the problem of insufficient power, and start counting what we really care about, which is how many people develop fungal diseases and how many people don't die during the course of their antifungal prophylaxis. When we do that my guess is that this will be just as good as anything else.

DR. GULICK: Other thoughts? Dr. Morrison?

DR. MORRISON: Just a brief addition to Dr. Schapiro's comments. I likewise have a little bit of difficulty trying to determine who has hepatic impairment because I don't know the Child's classification right off the top of my head. I wonder whether within the sponsor's database of those patients if there is any way they can go back -- maybe they have already done this -- and look at the actual transaminases, the bilirubin levels, and try to determine if there might be some more discrete laboratory cut-offs that could be used for knowing when you need to have the dose.

DR. GULICK: Dr. Hamilton?

DR. HAMILTON: I guess I have a version of Dr. Wong's opinion about this. In this population I don't think we do all that well with anything really, which is very discouraging for those who take care of these patients. And, progress in that regard comes in nanometers, it seems like. So, I am actually not that impressed with any deficiencies that have been identified in the voriconazole approach to the neutropenia host. I think they are about the same. I mean, you can do the statistics if you want to but I don't see that there is a whole lot of difference.

With that background, my expectations are not incredibly high. I can tell you that if that drug is available and I am on the unit and my hospital administrators are asking why it costs me so much to treat all these patients with liposomal amphotericin, among the first questions I am going to ask is how much does this stuff cost. Now, I don't know how much this is going to cost. I am sure nobody is going to tell me --

[Laughter]

-- but I think cost will become an issue here in terms of the practicality when this becomes available, which I have no doubt it will be with it is today, tomorrow or six months from now.

DR. GULICK: Dr. Englund?

DR. ENGLUND: Well, I agree that we need something in our bone marrow transplant patients, and I would focus on those. I would agree that we have better data on the more severely immunosuppressed than we have on the moderately immunosuppressed. We need something, and we need something so we don't have to suffer through this amphotericin liposomal, and then what do you do?

I am concerned about the indiscriminate use of it and, as a pediatrician, I am concerned that there is no data that I have seen on retinal changes in kids over time, and those kids are going to get things a long time. You have some patients and I would like to see that before I would recommend using an empiric therapy in children with developing eyes. I don't know that much about their eyes but when we treat these kids we are treating them for a cure. You know, they are getting a bone marrow transplant for a cure and they are going to be living a long time, and we want to make sure that we are not doing any harm. So, I would be very concerned about empiric therapy in children.

However, for the adults I really like the idea of having an alternative, as a clinical, and if, in fact, we are going to do that I think we need more resistance data. So my two things are safety data in kids, and I would like resistance data. What are you going to do? You have got to be telling me a little more when we are treating people with voriconazole for esophageal candidiasis. What is happening two months down the road? Are we then, again, knocking out all our azoles so that I don't have any other therapy to give to them? Maybe there is not enough resistance data out there but I haven't seen anything and I have experience with patients that have not responded and you don't have much left after that.

DR. GULICK: Something that is clearly influencing the committee is the non-inferior study where the lower boundary for the 95 percent confidence interval and minus 10 was crossed. I wonder if we would like to comment on that in particular since that has come up several times today, about how appropriate the cut-off is. That is based upon recommendations from committee meetings from 1995 and '95. Dr. Mathews?

DR. MATHEWS: Well, you know, I think that is a clinical judgment but there is a way to kind of translate it into other terms that clinicians might understand. It would be the number of patients you would need to treat to observe a benefit for a single patient. If you look at the point estimate for the empiric therapy indication, if you compare the response rates of the voriconazole group to the amphotericin group, it is 16 patients who would be treated with voriconazole for one patient who would die that wouldn't have died if they had gotten amphotericin. In the Aspergillus study, where there was a benefit shown for voriconazole, it is five patients treated for every life saved by voriconazole.

You know, I am as troubled, as everybody else is, about why there is a clear-cut positive, unequivocal response in the people with definite disease and I would say a negative response, based on the prespecified criteria, in the empiric therapy group. But one thing we have to remember that I think was talked about in the FDA background is that in non-inferiority trials any misclassification in outcome assessment is going to tend to make the drugs look more similar. And, despite the fact that the expert review panels were blinded, there had to be some misclassification. So, that is going to tend to attenuate the differences between groups. I think it is a very different setting than if this was set up as a superiority trial.

With regard to the issue of the composite endpoint, you know, I understand Dr. Wong's point on that and we have debates about composite endpoints here and elsewhere, but the fact is, at least as I understand the history of this, that one of the motivations for the composite endpoint is that the ascertainment of definite infections is incomplete, at best. So, there is a significant proportion of patients enrolled in these studies for whom the outcome would be unclear because definitive diagnosis wasn't made.

So, the composite endpoint, with all its faults, I think is a serious attempt that would seem to work well in two other trials that we saw presentations of their data here, and I don't think the rule should be changed to rely on post hoc analyses of efficacy in this setting.

DR. GULICK: Dr. Wong?

DR. WONG: I guess I disagree, Chris, with that last point. We were both here for the Ambisome presentation. My impression of the data that was presented when the Ambisome versus Fungizone study was presented was that on the basis of the composite endpoint the two treatments gave similar results. But the number of breakthrough fungal infections that occurred with Fungizone was higher than occurred with Ambisome. Here again today, we have seen a comparative study in which I believe the composite endpoint basically gave us misleading results.

Now, they can't be ignored because that was the criterion that was set out in the protocol and you can't change the protocol after the results are analyzed. But I believe that the results that we see today based on that composite endpoint are actually misleading because what we really want to know in a study such as this is how many patients developed fungal infections and how many patients are alive at the end of the treatment. In this study the answer to those two questions was the same number, and the answer in the Ambisome versus Fungizone study was that the Ambisome was better.

So, I think in both cases trying to craft a multi-component endpoint that takes into account safety, efficacy and investigator bias in an unblinded study is really the wrong way to go. I am afraid that if the agency feels committed to using a study design like this the same thing is going to happen over and over again until, you know, the third or fourth time this happens and people just throw up their hands and say let's start counting what we really care about, which is fungal infections and deaths.

So, I think we can't apply that post hoc to this study and say that we really got a positive result when the result was negative, but the next study I think should be designed very differently from this one and should really be designed to show what we care about most rather than surrogates such as, you know, fever or whether or not the investigator felt comfortable enough to continue the study drug. I mean, when an investigator says no, I want to switch, that doesn't mean the drug has failed. That means that the investigator's comfort level has been reached. That may or may not have any relationship with whether or not the drug is working. So, in an unblinded studies I don't think that should ever be allowed to be an efficacy endpoint. But -- well, that is all I have to say.

DR. GULICK: Dr. Wood?

DR. WOOD: While I think the overall answer to question number two is no because of the combined endpoints that we talked about, I would like to concur with Dr. Englund in the sense that when we look at high risk patients who are at great risk for Aspergillus and breakthrough infections, there definitely was an advantage with voriconazole and I think I would like to make that available to patients for whom we really don't have anything good. I don't think that that component of the data, in terms of that subset analysis which was consistent both when done by the sponsor as well as by the FDA, should be ignored since our ultimate goal is to try and make available efficacious as well as safe treatments for people who need them, particularly if our treatment options are very limited.

DR. GULICK: One thing that we have only touched on tangentially today is how much the standard of care in this particular field has changed even since '94 and '95, and some of the points in the background section were that fluconazole prophylaxis is used quite routinely today; also that growth factors often reduce the time of absolute neutropenia in patients. Actually, we saw some data to show that that was true of the study, that the time that patients spent with absolute neutropenia was much shorter than in prior studies.

I wonder if we could call on our expert consultants in the field maybe just to make some comments about that. Dr. Rodvold?

DR. RODVOLD: Well, I don't know if I am expert on that one. But those are all true, I mean, every one of those points. When I read this document before this meeting I found that that is what was throwing me a little bit in how to interpret the study. These days, you even treat people more on an outpatient basis that are less severe, and trying to keep them out. Then, when they do come in you are pushing them out faster. So, all those factors come up.

I think that tells you that you almost have to come back and re-look at this negative 10 percent on the bottom because there are new factors even since this decision was made. So, I would think that the agency is probably going to have to come back and maybe readdress that whole document again a little bit to account for that, or at least give opinions on that because I think it is influencing the outcome to these type of studies.

I agree with what was said by Dr. Wong, that you could run into this time after time in future studies. So, despite that it is not that old a document, it almost needs to be updated already.

DR. GULICK: Dr. Morrison or Dr. Wong, anything more to say about the changing standard of care?

DR. MORRISON: I think the points you make are good and I think one certainly needs to keep these aspects in mind, specifically not only did the receive antifungal prophylaxis but the type of antifungal prophylaxis and the issue of whether growth factors were used or not. Those two facets need to be built into and considered for future trials. I don't know how easy it will be to backtrack in this study to look at those aspects. I think it will be difficult.

DR. GULICK: Dr. DeGruttola?

DR. DEGRUTTOLA: I have a comment going back to your question about is 10 percent the right criterion for non-inferiority, and I think that question is obviously related to the issue of endpoint, what is chosen and how rare or frequent the endpoint will be. I agree with Dr. Wong's comments about the need to distinguish between safety and efficacy endpoints.

My concern about using an absolute 10 percent for your definition of non-inferiority is that it has different meaning depending on what the response rate is. If the response rate is 30 percent for the control drug, as in this case, then what you need to exclude is 20 percent. In that case you would be talking about one arm having 50 percent more success than the other arm. So, that is quite a considerable difference in relative terms. Whereas, if you are at about a 50 percent response rate, then the 10 percent rule would require greater than 40 percent, which is, in relative terms, less important.

So, I think that rather than using consistent 10 percent rule across the board, regardless of what the endpoint is or what the response rate is expected to be, to have a rule that is either in relative terms or at least takes into account not only what the expected response rate will be but what you do if it actually turns out to be less than you expected. I mean, you know, if you had 15 percent response then, obviously, five percent could not be considered equivalent. So, I think that the whole issue of endpoint -- how it is going to be, and then deciding what non-inferiority means and what is appropriate will need more consideration.

DR. GULICK: Dr. Schapiro?

DR. SCHAPIRO: Going back to your question regarding the changes in prophylaxis, the duration I think was remarkable. It was far shorter, and the fact that patients are being given prophylaxis. I think what it does impact, of course, is what degree of safety are we looking for. Going back to what Dr. Yogev and Dr. Mathews mentioned, how many patients are actually needing to be treated and how many are we treating that actually don't need it? I think it pushes us to be more careful. I think that was also in the directive of the agency. We will be more cautious regarding interactions and toxicities and I think that will push us to looking for a higher degree of safety and more knowledge of interactions before we would use a drug.

DR. GULICK: Maybe we can consider some of the questions that Dr. Goldberger elaborated on. Would the committee be in favor of defining certain subgroups, based on the data that we have seen today, that might be appropriate for this particular indication? Did the data support that? Or, would people be in favor of breaking things down? Dr. Mathews?

DR. MATHEWS: You know, a number of people have taken note of the observation that in the high risk group there seemed to be a benefit but, again, you know, that is a post hoc analysis although it was a stratification variable, if I am correct in my memory. I would have to ask what is the biological plausibility of that observation. For that reason, I would have problems saying that it wasn't -- we couldn't say that it was effective overall according to the primary endpoint analysis but then make a statement that it was effective in a high risk group. I mean, why wasn't it effective in the moderate risk group, unless it was a problem of ascertainment -- under-ascertainment of outcomes tending to attenuate that? I don't know.

DR. GULICK: Dr. DeGruttola?

DR. DEGRUTTOLA: I don't know if I am allowed to ask a question of either the agency or the sponsor about whether there was ever a formal test of whether the category of high or moderate risk -- if there was an interaction between that and the randomized therapy on the outcome. Obviously, in one case the lower confidence bound excludes 10 and in another case it doesn't, but that doesn't imply that there is actually an interaction. I mean, these estimates and confidence intervals are going to bounce around a little bit by chance. So, the question is was there a formal test for interaction between high and moderate risk and the randomized therapy?

DR. POWERS: We didn't do such an analysis. I don't know if the sponsor did or not.

DR. BAILDON: We did that; we are checking for the results. The point was made about the biologic plausibility. I would just like to highlight that the period at risk at time of neutropenia varied greatly between the patients who were in the high risk category and the patients who were in the moderate risk category. That is something we have. There was quite a big difference and one of the points that is well known is that there is a good correlation between patients remaining neutropenic and their increasing risk of a fungal breakthrough infection as they remain neutropenic. As soon as you go out towards ten days or longer, then the risk of having a fungal infection goes up considerably. So, I would say that there is some biologic plausibility for that difference in treatment effect, which would be higher in the group of patients who are at higher risk of fungal infection, be it occult or open or documented.

DR. GULICK: Thank you. Just let us know and we can display the data. Dr. Englund?

DR. ENGLUND: I just think that the difference reflects the fact that we aren't diagnosing true cases, and when we combine safety and efficacy we are seeing something because in the moderate risk group the true incidence of aspergillosis is really quite low and, therefore, to approve a drug for a population in which it is quite low and I don't have all the data I want would make me uncomfortable. Now, breaking up the group into two groups statistically, I leave that -- I would like other people's opinions.

DR. GULICK: Dr. Wong?

DR. WONG: I guess I would look at it in a little bit different way. I don't think I would try to break the groups up, especially since that wasn't specified before. But I found that the FDA's slide number 21, breakthrough infections sensitivity analysis, where voriconazole and liposomal amphotericin B were compared more breakthrough infections plus protocol-specified failures in the voriconazole group, if someone switched for so-called effective, giving equivalent results tells me that these are almost surely equivalent therapies. And, I am very sorry that the study wasn't designed to ask this question rather than the question that it was designed to ask, which I think was really muddled and confusing. Because this is what I would really want to know and here they came out the same. So, I don't think I would try to massage it by subgroup analysis. I think that the question was the wrong question and the data actually to answer what I wanted to know is here.

DR. GULICK: Another question that Dr. Goldberger asked us to consider was whether approval for this indication might be appropriate with some caveats. Maybe the committee could entertain that approach also -- or not. Do people have thoughts about that? Dr. Englund?

DR. ENGLUND: I just want to say that I don't think there is enough pediatric data to approve it with caveats. So, for the children at least -- not for treatment but for empiric therapy without data, some data at least on the eye findings -- I think that shouldn't be a caveat; we shouldn't do it.

DR. GULICK: Dr. Morrison?

DR. MORRISON: Do we know how many people with calculated or actual creatinine clearances of 50 or less got oral drug out of these studies?

DR. GULICK: Either the agency or the sponsor, could you help us with that question?

DR. BAILDON: Could you give us just a minute?

DR. GULICK: Sure.

DR. RODVOLD: One of the problems, to support what I think she is asking is that of making dosage adjustments on creatinine clearance, not creatinine. I know they used some creatinine as a cut-off here but creatinine clearance is a little bit better because you account for age and you account for gender, which are factors in this drug's disposition at times. So, I guess creatinine clearance comes back to be a player for me to think through who should be either on or off, or dose adjustment of the drug.

DR. GULICK: Dr. Hamilton?

DR. HAMILTON: While they are looking for this additional information, maybe Dr. Goldberger could maybe expand a little on what the breadth of the caveats might be. Are we talking about caveats that just take the form of verbiage in the package insert, or are we talking about requirement for additional analyses, additional data? What?

DR. GOLDBERGER: To answer the second part of your question first, if you don't think there should be an approval now, the kind of things we would be interested in are specific recommendations about additional studies, additional data, additional analyses, etc.

When I spoke about caveats in the labeling, caveats can include -- and there are always issues with these, one caveat can include -- and I think someone on the committee has already suggested this, recommending the product when, for instance, other products cannot be used, or refractory or intolerant to other approved products. There is such a thing as a second-line indication, and that is one possibility. Obviously, one issue here is that is not really how the drug, of course, was studied. It was studied as initial therapy and, therefore, one would have to believe that there was sufficient -- on balance, looking at all the data -- sufficient biological plausibility.

Another caveat gets down to sort of what I also talked about, about subgroups. That is, is there a group for whom the product could be recommended and, therefore, obviously other groups which could be specified for whom the product might be recommended? In other words, a little more specificity instead of simply saying this drug is approved for empiric therapy, this drug is approved in such-and-such. In other patient groups, you know, the drug was not shown to be effective. Issues like that.

Now, the reason I would even bring this up is that so much data was presented this morning about the different subgroup analyses, and all, that I think it is an inevitable question that people would ask, and such things can be done. The last concern about them, obviously, is what many people on the committee have already talked about, not all of these analyses were obviously clearly prespecified. Some were prespecified in terms of stratification; others were prespecified perhaps as secondary analyses, etc., and that raises some concerns as well.

In the end, of course, what we are asking you to do is to look at the totality of information that is available, recognizing that the drug did not meet the agreed upon primary endpoint, but that there are other pieces of information that suggest the drug could potentially be useful using all that information as well as what you have heard about the safety profile, giving us a recommendation basically about what you think in this situation. I mean, there are ways to craft wording in the labeling or defining indications but we would ask you to be, you know, as specific as possible because we would have to turn a recommendation into labeling that would be useful to clinicians so they would understand what we were trying to say and could use the product accordingly.

DR. GULICK: Have you found the information that you were looking for?

DR. BAILDON: Thank you.

[Slide]

Now, this denotes creatinine shifts. If you look on the left side, we have classified patients by calculated creatinine clearance, and the bottom group would be those patients who were considered with a severe renal impairment at baseline with a calculated creatinine clearance of less than 30 mL/minute. Then we looked at what the baseline creatinine was and how they did they shift, did the maximum creatinine during study increase significantly, by the categories moving to the right there. That is our total NDA pooled population so this is the voriconazole-treated population looking at creatinine shifts.

[Slide]

Another way to look at it, this analyzes creatinine levels, median change in creatinine over time for the two groups, voriconazole and Ambisome, in the empirical therapy study. As you can see there, voriconazole is an absolutely flat line. There is no influence on voriconazole and there were patients with preexisting renal impairment in that study, and with Ambisome you see the shift that you would expect.

DR. GULICK: Thank you. Dr. Wood and then Dr. Yogev.

DR. WOOD: In reviewing the trial design analysis, it was stratified by risk of fungal infections. So, I don't know exactly how that risk was defined but that was done prospectively before the trial was conducted and analyzed. So, I think the issues that we are seeing in terms of a benefit in terms of breakthrough infections in high risk individuals versus moderate risk individuals is legitimate because there was the stratification according to risk prospectively as part of the trial design.

The other question, and I don't believe this data was presented but I would be very interested in seeing it in terms of in the empiric trial what were the outcomes of voriconazole versus Ambisome in patients who had ANCs less than 100? So, that would clearly be a very high risk for invasive disease. Then, those with prolonged ANC durations of greater than 10 days, if there was any difference there.

DR. BAILDON: Sure. Do you mind if we --

DR. GULICK: No, no, please go ahead and respond.

DR. BOUCHER: I will try to answer the duration of neutropenia question first, and just to clarify, the prespecified risk strata specified that high risk patients were patients with allogeneic transplantation as well as relapsed leukemia. I just wanted to clarify, in our comparison of our other studies this morning that Dr. Powers presented the important point I think in comparing the recent paper published on itraconazole was that that study did not include allogeneic transplantation at all in the study. So, another point about the design of our trial was that patients had to have an ANC below 250 for 24 hours prior to randomization. So, these were profoundly neutropenic patients.

[Slide]

I just want to clarify again about the duration of neutropenia because, again, it has been a little confusing in a way when we discussed it. We know in our patients the median duration of neutropenia prior to randomized, which you see in the top row, during therapy and then in total. What we know is that in the high risk patients the median duration of neutropenia was 17-18 days compared to 12-13 days.

[Slide]

If we could go back to slide 10 and we go back to our breakthrough infections looking at them, as Dr. Powers also presented, according to risk, we have nearly 10 percent emergence of invasive fungal infections in this high risk group, which many would agree is an unacceptably high rate of emerging fungal infections given the mortality, as Dr. Patterson shared with us this morning in this setting. Does that help with the neutropenia?

DR. WOOD: It does.

DR. BAILDON: I want to come back to the other question that was asked previously about the statistical analysis between the two groups. If we do logistic analysis on a two-factor interaction model of treatment and risk, that does not turn out to be significant, a significant difference.

DR. GULICK: Dr. Yogev?

DR. YOGEV: I think that looking for language, as you said, I would consider it as preliminary data, at best, in the high risk group, and if you look at the breakdown of what type of fungus Aspergillus makes the difference and, you know, from personal experience in patients who have bone marrow transplants, we see a little bit more Aspergillus than Candida and I think it would be just fair to put in the preliminary data to suggest that if no other therapy is available voriconazole should be, but I would discourage at the same time using it in others because there are no data to support it. That, by the way, would cut also the number of patients by almost 60, 70 percent, those who are going to be exposed to the drug who are in other categories. So, look at a balance in between.

DR. GULICK: Dr. Wong?

DR. WONG: Maybe I can propose some wording. I mean, as I look at these data voriconazole was equally as effective as liposomal amphotericin B in preventing breakthrough fungal infections and death in febrile neutropenic patients. Although that wasn't the protocol-specified primary endpoint, in my opinion that is the most important question. So, I think I could support an approval with working like that.

DR. GULICK: Dr. Yogev?

DR. YOGEV: We are saying the same thing, just if you take the high risk out from that statement it won't be accurate. On the other hand, if you take the high risk what you are saying is right just for the high risk group. That is where we need it. That is where I think it would be unfair not to, but in the moderate you increase toxicity just by the numbers and you don't get really any difference, or even worse if you do the calculation -- it is even worse.

DR. GULICK: Dr. Hamilton?

DR. HAMILTON: Isn't the distinction here between high risk and moderate risk somewhat artificial? I mean, bone marrow transplant recipients and people who relapse with leukemia -- isn't that what you said?

DR. ENGLUND: I think there is a huge difference actually in multiple diseases --

DR. HAMILTON: I don't disagree with that. I am saying there are some people who likely fall into those very high risk individuals who have been excluded because they didn't happen to have that disease. I mean, there must be a range --

DR. ENGLUND: I am sure there is an overlapping range but when you look at just all the other diseases from, you know, CMV to RSV, bone marrow transplant is still the worst.

DR. HAMILTON: I don't disagree with that, but what I am trying to do is homogenize the rules here so that you can, you know, pick out those people who truly are and if all the bone marrow transplant patients fall it, great. They probably will.

DR. ENGLUND: I think we do have to be a little careful to make it somewhat simple; you can't ever --

DR. HAMILTON: I am asking the question can we do that in a way that is other than just disease category.

DR. YOGEV: But that is all the data you have.

DR. HAMILTON: Oh, I don't think so --

DR. YOGEV: I mean from what we have here. The way the high risk was defined, there is a difference where voriconazole has a place. In the other it doesn't. So, to take what we have, that is why I would use it as the preliminary data suggest and not that it is an indication.

DR. GULICK: One more comment?

DR. ENGLUND: I just think this would be a perfect way of introducing a Phase IV study in this patient population to get bigger number and safety and toxicity, a Phase IV study post-approval in bone marrow transplant patients, with ocular and ophthalmology follow-up, to really assess it, and I would say that that should be a condition because I really think you need to see what is going to be evolving over time as you treat these patients. I mean, what are we going to be seeing once we put large numbers of patients, bone marrow patients on this therapy?

DR. GULICK: Dr. DeGruttola, last comment?

DR. DEGRUTTOLA: I just wanted to reiterate the point that if any information from this study is used as a justification that the language should state that it is post hoc analyses of a completed study and that they weren't primary results because the post hoc analyses are never going to carry the weight, obviously, of the protocol-defined analyses, and I would hope that that would be communicated in perfect wording to physicians.

DR. GULICK: Dr. Goldberger?

DR. GOLDBERGER: We can certainly --

DR. GULICK: Can you speak up?

DR. GOLDBERGER: We can certainly put a statement like that in labeling. Keep in mind that basically it would still, from the practical point of view of promoting it, etc., it would obviously still be approved for that indication. We could include it in some detail in the clinical study section, however, that would be part of the labeling.

The other issue is that I just want to make sure this impression is correct, the great bulk of the discussion over the last few minutes in this area has concentrated on interpretation of the efficacy results. I just want to make sure that on balance people, therefore, are reasonably comfortable with the safety profile with regard to potential approval for this indication. We haven't talked a whole lot about it. We have talked a lot about the safety profile, but is this primarily an issue of efficacy? Or, how much does the safety also play a role? It would just be helpful for us to hear a little bit about that.

DR. GULICK: Yes, thanks for helping us focus on that. So, let's consider safety for this indication. Dr. Mathews?

DR. MATHEWS: Well, in addition to the unresolved issues regarding the ocular toxicity, I think that there is insufficient knowledge of drug interactions, that we have already talked about earlier, which creates uncertainty in my own mind about the potential toxicity on both sides of the equation, both in terms of efficacy of the levels of voriconazole achieved in the presence of inducers, for example, and also what would happen to the levels of the other drugs which have not been studied either in two-way or, preferably, in three-way interactions. So, I think the toxicity issue is an important part in my own decision-making about this.

DR. GULICK: Dr. Wong?

DR. WONG: I think with respect to the eye disease what we have seen so far and what has been proposed so far is really not enough, especially with respect to long-term consequences. The patients who have already received voriconazole, even in the course of the clinical trial, could be called back for new eye examinations so that you can answer the question are there remote effects after stopping therapy. And, it seems to me that a study, presumably conducted in South America, of 42 subjects treated long term is just not enough to satisfy the question are there going to be important eye complications with extended use of this drug. So, those studies I think need to be expanded both in breadth and also in time.

DR. GULICK: Dr. Schapiro?

DR. SCHAPIRO: Jut to refocus on the question, I think safety is very important here. I think probably my major consideration is the fact that I am more comfortable with Ambisome than I am with what we have seen for this drug, and even if I am convinced that they are similar, I still think that we have unanswered questions and I have difficulty defining the patients that are high risk for toxicity and, therefore, I still would have a problem with toxicity.

DR. GULICK: Dr. Yogev?

DR. YOGEV: I thought that is why I am going to narrow it to the high risk. That is Aspergillus. That is where your Ambisome is going to fail. For me, I have problems in this disease situation. I agree we need to pursue it and it has to be part of what you do, but to suggest that because of that in such a life-saving procedure to wait for more -- that is why I would like to say it is preliminary data because, to be honest with you, I saw more patients than I would like to failing on Ambisome and this one is telling me you have another percent, or whatever, better. So, for me, yes, side effects are important but that is why I would agree not to do it in other populations.

DR. GULICK: Dr. Wood?

DR. WOOD: I concur with Ram's statement because in the end patients who are dead of Aspergillus don't have to worry about their eyesight, and I think that is a major issue and concern, that we have significant failures with Ambisome and I think there is some data that is significant, that is not debatable, that patients at very high risk for invasive aspergillosis do appear to have a benefit from having voriconazole, and I would like to be able to make that benefit available to patients, with the understanding that the ocular studies and the toxicity monitoring in terms of the breadth, duration and the populations, kids and adults, needs to be done.

DR. GULICK: Dr. Schapiro?

DR. SCHAPIRO: A brief response, Dr. Yogev. If this were the study design to look at that, I think we could answer the question. We still have in mind what percent of patients in the high risk group actually did have the Aspergillus infection. That would have to be the analysis. And, if we would really call that group Aspergillus, not neutropenic patients. I think that would focus it. I think if we did have a group and we found that there is a very high attack rate of Aspergillus in the population -- I think we saw very convincing results for Aspergillus that is a biologic plausibility and I think that would be a good indication. The question is, is that so? I mean, if those patients are impending Aspergillus, that is a good indication.

DR. GULICK: Is this a response because other are waiting?

DR. YOGEV: Yes, unfortunately, it is a response. Don't throw away the baby with the water. Look at what happened to the Ambisome. This is a progression that was randomly assigned and still you have a huge amount of Aspergillus breakthrough, 13 I think versus 4. And, we know from other data in bone marrow transplants with Aspergillus, we are getting more and more of them and, interestingly enough, in the recurrent leukemia which I don't understand, but those are effects that you have higher with this specific fungus.

DR. GULICK: Dr. Rodvold?

DR. RODVOLD: I agree. I think once you put the high risk into it you are willing to take safety a little bit with you, but with that, I think the slide that the FDA presented about additional issues regarding the ocular issues of the rechallenge people, people that have underlying eye diseases, what we don't know is what are risk factors or what characteristics you would want to be careful with. You are totally in the blind, so to speak, on the eyes and so that is the issue in all of these populations but even more here, going on to identify those people -- and this drug is going to be repeated a bunch of times and you are going to have, like it or not, other antifungals on board. I mean, I have already seen that in my own institution with the last compound you approved. They are doubling up on it for not necessarily good reasons, and that is going to happen with this one as well in the wrong patients. So, we need to know what patients we don't use it in, or we do something else and that is not in here as far as I am concerned.

DR. GULICK: Let's see if I can draw some consensus in what we said. The committee was challenged by question two. Let's state that up front. But, first of all, it is important that we did affirm with question one that we feel this drug has proven efficacy in other fungal infections, particularly aspergillosis, and we saw some information about Candida that was part of the 1994, '95 recommendations for an appropriate agent for febrile neutropenia -- the empiric treatment was that it did demonstrate activity against those two organisms. So, we did agree with that.

With regard to safety issues, the points that were made were that the risk/benefit ratio in this particular population is somewhat different from those who have proven or highly suspected aspergillosis, and we talked about that. A fact that several people mentioned was that many patients who don't have fungal infections at all will receive a drug given for empiric treatment of fungal infection. Therefore, the safety issues are clearly important and maybe somewhat different in terms of the risk/benefit analysis in this population.

I think the committee feels that we would like to see more safety information; that there was perhaps not enough safety information, particularly with regard to the visual changes, and that we truly want to see additional data. Also, on drug-drug interactions some information is available but a lot of information is not available. So, those did concern, I think, the committee in general about the safety of using this drug in this clinical setting.

Perhaps even more thorny was the efficacy issue, and I think the fact that the 603 study did have a negative result strongly influenced the committee, with many caveats. Looking with 20/20 hindsight, we are frustrated by the design of this study, in retrospect. The fact that the minus 10 percent lower limit of the confidence interval was chosen based on previous recommendations, as Dr. DeGruttola pointed out, a different response rate might have dictated a different number there. In actuality, the response rate that was predicted for this study was more on the order of 50 percent and it was seen to be more on the order of 30 percent.

And, we were frustrated with the composite endpoint that was recommended by the previous committee. Mixing elements of safety and activity into one endpoint makes it difficult -- "misleading" was a word that Dr. Wong used in trying to really associate what are the activities of the individual agent.

We struggled with an analysis of the secondary endpoints and the subset analyses. We saw some interesting signals that this drug may have more benefit in the high risk group; may be efficacious in terms of breakthrough fungal infections; mortality, but there were also concerns that many of these were not prospective analyses but post hoc assessments of the same data and that the study really wasn't designed to look at those questions.

We mentioned changing standard of care in this particular population. It is more of an outpatient disease; fluconazole is routinely used as prophylaxis, and the use of growth factors and the decreased time of neutropenia, which strongly influenced one of the parts of the composite endpoint for this particular study.

We considered labeling and subgroups that might benefit from this therapy, and there was general disagreement, some people citing the fact that most of these were small retrospective analyses of the data, not prospectively designed; others saying that really those are the endpoints that we should be looking at. So, there was some disagreement about that.

Also, thinking about caveats that one might use, for example, one that was suggested was using the drug if the first-line agents in a patient who is refractory were intolerant to that. There was some support for that. On the other hand, as people pointed out, this particular study would not support that indication in that it was an initial treatment study. There were many calls for additional studies that might help us sort this question out.

With that very clear response of the committee --

[Laughter]

-- we do need to take a formal vote on this question. So, again, is there sufficient information to support that voriconazole is safe and effective for empiric antifungal therapy in febrile neutropenic patients? Again, we will start with you, Dr. Rodvold.

DR. RODVOLD: I have to answer the question this way without a caveat, right? Is that what you are saying?

DR. GULICK: Yes, we have to say yes or no basically to this question and then we may make give some other considerations.

DR. RODVOLD: I would say no.

DR. GULICK: Dr. Wood?

DR. WOOD: It hurts to say no.

DR. GULICK: Dr. Mathews?

DR. MATHEWS: No.

DR. GULICK: Dr. Hamilton?

DR. HAMILTON: Yes.

DR. GULICK: Dr. Yogev?

DR. YOGEV: No, but... we will talk about that.

DR. GULICK: Dr. Englund?

DR. ENGLUND: No, but...

DR. GULICK: Dr. Schapiro?

DR. SCHAPIRO: No.

DR. GULICK: Dr. Wong?

DR. WONG: Yes.

DR. GULICK: Dr. DeGruttola?

DR. DEGRUTTOLA: No.

DR. GULICK: I would also vote no. So, for those keeping score, there were two yes votes and eight no and two "no, but" votes. We are going to combine those into no.

Given that result, we really need to consider the second part of this question, which is if we do not feel that there is sufficient information, what additional information would be needed to support this indication? Dr. Schapiro?

DR. SCHAPIRO: Going back to the little discussion we had in this corner, the box as defined, it is a very high rate. The total number of patients was 141 and it was 13. In the other group it was 2/143. So, I would agree that for that group the risk/benefit does justify it. So, we can discuss this as a caveat or as a study, but I think a study for that group, from this analysis, would suggest we would have results which would very much justify that indication.

DR. GULICK: So just to clarify, you are talking about the high risk subgroup?

DR. SCHAPIRO: Yes, the high risk.

DR. RODVOLD: That is what I agree with too. I brought that up originally, the high risk group, as well as Dr. Wong's comment about the other endpoints. When you throw those two in and that population, my vote goes swinging 360 the other way, to yes. So, that is where you get to can you language this, or do you go pursue another study to beef this up?

DR. GULICK: I look to Dr. Goldberger. Do you want us to consider formal wording for another indication here or would just our advice be sufficient?

DR. GOLDBERGER: Actually, I think you have worded that so well as to make it very easy to answer. I think the latter part, more your advice about whether we should at least consider the possibility of an indication, perhaps more restrictive or more limited than simply saying it is indicated for febrile neutropenia. I don't think you have to necessarily provide us with the exact wording, rather, the sense of what the committee thinks would be sufficient.

DR. GULICK: So, I would suggest that our vote, particularly the "no, buts" should show that the committee was conflicted about this with regard to some of the things that we said. So, we are not going to take another vote. That eases the tension a little.

DR. YOGEV: I think the problem is in the question and not in our answers and we are defined, unfortunately, by the question and that is why at least some of us, me especially, said no, because of everything that we discussed before. But if the question would have been "if not" and "if yes" I would say no to the majority but yes to this one. And I would discourage asking the company for another study. I agree with all the faults that this study had. It was designed like that but I am familiar with many things which we are doing in practice by ad hoc analyses that have become more and more acceptable. I don't want to mention here the meta-analysis that many of us are working by, which is even worse this one.

So, I would encourage the agency to change maybe the question a bit or to accept this variation, and I would be more than happy to suggest to the chair yes to vote on this high risk group because I think it is so important; we need to care about the patient.

DR. GULICK: I would suggest that the agency will benefit from hearing opinions. Our votes are recommendations. They are not binding. So, I think what you just said and what others have said is getting the message across without having to come up with wording and take another vote, which may not really be our responsibility, unless you think differently.

DR. GOLDBERGER: No, I think that you don't have to certainly provide exact wording. We have already gotten just from the discussion, I think, a reasonable amount of advice about some of the possibilities that people have raised, and it is a matter now of thinking about that. So, I think from that perspective we are okay.

The one thing that perhaps is not entirely clear, and members may want to comment -- I don't know if it is necessary to have a formal vote or not -- how strong the desire is that we would attempt to fashion some sort of more limited indication with what information is available now versus the idea of asking for some additional data and deferring that decision about a limited indication, or a more straightforward indication, until the additional data is available. That is actually the one remaining issue. We have certainly gotten sufficient suggestions about what a limited indication might be. On balance, does the committee prefer that approach or getting some more information and then making a decision at that point.

DR. GULICK: Just so I understand, we have taken a vote and said the information that we have today we didn't feel demonstrated the proposed indication as stated.

DR. GOLDBERGER: Right.

DR. GULICK: Possibilities are that a limited indication be recommended from the data we have today, or another indication might be made based on data --

DR. GOLDBERGER: Well, in other words, basically what we would then be thinking is, rather than another indication, that you would recommend some additional studies, the primary purpose of which would be to support the indication that is up there. In other words, there is a belief among at least some members of the committee that based on what we know about the drug it ought to be effective in this indication and that perhaps, for instance, the wrong endpoint was selected and, logically following from that is the idea that a study utilizing a different endpoint might end up being a better test of the drug. That, of course, is a substantial investment by the sponsor. So, it is helpful to get advice as to how useful that would be, for instance, as one type of information.

DR. GULICK: Again just so I am clear on what you want us to do --

DR. GOLDBERGER: What we would like to know, in other words, you voted basically that there is not sufficient data to support the indication of empiric antifungal therapy of febrile neutropenic patients. Two committee members voted "yes, but" which is could. We might include that as an option in subsequent voting --

DR. GULICK: It was actually "no, but" --

DR. GOLDBERGER: Now the question is whether or not or how actively we should pursue some sort of more restricted or limited indication based on what we have now. Keep in mind there is one basic rule that I do want to emphasize. We cannot put something about this study in product labeling unless, in essence, we are giving some type of indication. I do want to make that very clear. We cannot put a description of a study in the labeling but not say that it is indicated. De facto, if it is in the label it should be considered an indication. So we can, of course, be specific about the indication and say for high risk patients or as an alternative in certain situations and the describe the study, and that is fine. We can't just put this in for people to have access to and say nothing about an indication. That is the one option that we cannot do.

DR. GULICK: It sounds like what you would like us to consider is, is there a more limited indication that the committee would support.

DR. GOLDBERGER: Should we attempt to do that with the sponsor, or should we work on defining what type of additional information could be used to support the indication that is up there right now?

DR. GULICK: So, two topics we want to discuss -- is there a limited indication with the data we have and that we have seen today, or is there more data that we would suggest to support the indication in the question.

DR. WOOD: Do the questions have to be "or" or can it be an "and" and an "and?"

DR. GOLDBERGER: You can basically state you would like to see a limited indication now but you also really would like to see some additional data which, of course, might support a broader indication later on. That is perfectly all right.

DR. GULICK: Let's consider them separately. So, let's consider do we feel that there is enough information here to propose a limited indication based on the data we have seen today, and what would that limited indication be. Dr. Schapiro?

DR. SCHAPIRO: For the question of an indication for patients with bone marrow transplant, I would say that I would agree to that limited indication and would not require a study.

DR. ENGLUND: Bone marrow and relapsed leukemia, the two that they had in the study.

DR. SCHAPIRO: For those two indications I do not think we need another study. I think we should approve that limited indication. That would be my feeling.

DR. GULICK: Others have thoughts?

DR. RODVOLD: I agree with that part. To broaden that you would need more data to be able to come to labeling like this. So, the immediate is to include high risk, put language in for that, and move that indication in without necessarily more data. Then, if the sponsor and agency wants that to be broader than the high risk group, then you will have to pursue whatever it takes to get that.

DR. GULICK: So, again, I really would like to consider one at a time. So, we are still considering is there a limited indication that we would feel comfortable approving, or recommending approval today based on what we have seen?

DR. WOOD: I would feel comfortable recommending approval for a limited indication in high risk patients, as defined prospectively, being those patients undergoing allogeneic bone marrow transplant and relapsed leukemia. I believe there is sufficient data there.

DR. GULICK: Dr. DeGruttola?

DR. DEGRUTTOLA: I don't have any comment about recommending one way or another but I do think it is important to remember the analysis that the sponsor just presented saying they didn't see an interaction between the level of risk and the randomized treatment. So, while there may be a suggestion of a difference between the voriconazole and the control drug among the high risk patients, it wasn't demonstrated that, in fact, there was any difference between the low and high risk groups. So, I think people would have to be relying on their other medical knowledge which, obviously, people are for making this consideration. Even as a post hoc analysis there is not strong evidence, as I understood, for an effect of risk on the treatment difference.

DR. GULICK: Dr. Yogev, if you have some questions, it would help everybody to ask them.

DR. YOGEV: Can you repeat it? Maybe I missed it because I thought when you do the high risk by itself there is a benefit.

DR. DEGRUTTOLA: Well, there is no evidence of an interaction -- there may be a point estimate. The point estimate may be a little bit better favoring voriconazole for the high risk and a little bit worse favoring amphotericin for the low risk -- the point estimates, but when a formal test was done, as I understand it, to see whether level of risk had an effect on that treatment difference, it did not. In other words, the question is was one better and the other worse by chance, or was there a real proven statistical difference between the two? And, as I understand it, there a trend there that things look better in the high risk group but it isn't demonstrated that risk interacts with the randomized treatment. So, I think in order to support the notion that there should be an indication in the high risk group, I think people should bring whatever other medical knowledge they may have that tends to strengthen the data by itself but, as a statistician, is not strong enough to support.

DR. WONG: I really agree with that. I personally believe that trying to split the population is the wrong approach. I think that the sponsor's data has demonstrated to my satisfaction that breakthrough fungal infections and deaths are prevented by voriconazole as well as they were prevented by the liposomal amphotericin B in the study we saw today, and I wouldn't try to break out subgroups.

DR. GULICK: Did the agency want to speak to Dr. DeGruttola's point?

DR. POWERS: I am translating. Our statisticians say they agree completely with Dr. DeGruttola's remark.

DR. GULICK: So just to clarify again, although there was a trend towards an improvement in the higher risk subgroup, statistically there was no difference between the two groups.

DR. POWERS: The other point I would like to make again is that even though patients were stratified this way, when you divide it into high risk and low risk and then look at the results in those, that is still a secondary analysis. It is not prespecified ahead of time.

DR. GULICK: Would we like to also consider safety given that we are talking about particular groups? Dr. Englund?

DR. ENGLUND? I would just like to amplify Dr. Schapiro's statements of the whole thing, which is risk/benefit. We don't have, to my mind, great long-term safety data and so by limiting it to the people who have the chance to benefit the most because they are at the highest risk of having serious consequences not based -- well, based on the data from the study and based on general epidemiology, I would vote yes for partial approval.

DR. GULICK: Others? Dr. Schapiro?

DR. SCHAPIRO: Victor, just my take on this, if we didn't have any logical explanation why this would happen you would say the statistics are not saying that this is so. But we say that this is a group where we expect more Aspergillus and we think this is a drug which is better for Aspergillus and, therefore, we are not surprised that these results are the results, that would be the kind of reasoning and not that the statistics are telling you that you have identified a subgroup.

DR. DEGRUTTOLA: Exactly. In other words, what you could say is these results are consistent with the way I would expect it to go given my medical knowledge, but you can't say that they support or really show that result.

DR. GULICK: Dr. Goldberger, do you recommend we take a formal vote on the limited indication of have you heard enough?

DR. GOLDBERGER: I think, personally, I guess I have heard enough.

[Laughter]

Let me just say this, I recognize how painful this has been for the committee members but I will tell you that it greatly facilitates the subsequent discussions we will have with the sponsor on this issue. Although Pfizer is certainly pleasant to work with, labeling negotiations, by their nature, tend not to be very pleasant and having this type of interaction is very helpful in just thinking of a framework for our discussions with them, and it will save both us and them I think a lot of time. So, it was actually quite useful to do it now since the issue would have undoubtedly come up in subsequent labeling negotiations, which is why we wanted to get your input now.

DR. GULICK: So, short of a vote, maybe again I can just try to say what we felt as a group. There is a difference of opinion here. I think people are responding to the fact that the high risk group perhaps is in the most need of therapies and that we have seen at least a signal that there is a demonstration of activity here, and that the risk/benefit in a group that you can define as high risk clearly is different from other groups that may have received the therapy in the overall study.

Along with that is the biological plausibility that we would like to see the drug work in this group that needs it the most. However, we have been reminded that the statistically analyses don't clearly show a difference between these groups, and we have been reminded that the analyses that we did see were actually retrospective; they were secondary analyses, not primary analyses.

We need to come back to what additional data we would recommend in support of a primary indication. Things that people have said so far -- with regard to safety, we would like to see longer term description of the visual changes; that we would like to see more drug interaction information. In terms of efficacy, people have suggested that looking at fungal breakthroughs alone, perhaps in a second study and/or confining a second study to a subpopulation which would be more at high risk, are options that have been brought up in the discussion. There may be others. Other suggestions? Dr. Hamilton?

DR. HAMILTON: I would think the sponsors possibly could look at their data to see if they could define who actually constitutes high risk patients. This would be, obviously, a retrospective analysis but could address concerns that I expressed relative to the use of disease classifications as opposed to biologic measurements that have some objectivity to them.

DR. GULICK: And I think you also made the important point before that there is a large amount of data available now that could be looked at or could be follow-up on, and then another way to approach this is to come up with a completely new study, and that both approaches may have their benefits here.

DR. WONG: One idea that comes to my mind from the Aspergillus study is that you have this outside expert panel that reviews cases in a blinded way, but for this study the efficacy of treatment was not assessed in a blinded way. Am I correct there?

DR. BOUCHER: There was a blinded data review committee in the empirical therapy study that both assessed diagnosis of infection and certainty of infection, as well as outcome at the end of therapy.

DR. GULICK: Just to be clear, they were blinded as to therapy.

DR. BOUCHER: Yes, yes.

DR. GULICK: Other thoughts?

DR. RODVOLD: The sponsor has done a lot of pharmacology work and I compliment them for it. In fact, it is amazing how much they have done. But in this area, and it is not often thought of in the labeling, but drug interactions with cancer drugs is still out there. There are a lot of protein binding issues; there are CYP issues. So, with or not that is influencing the disposition of this drug, I don't know. I know they showed average concentrations and average value across a lot of populations, but maybe just going back and tweaking out these transplant patients, leukemics in other populations that will be in this audience to make sure that the dosage they are using is the right dose based on the pharmacology in this audience because they are complex patients. Other drugs have been characterized with different PK issues, and just to make sure we haven't overlooked that. If they have already done it, that is great but I want to ask to make sure they know the PK in the audience of these neutropenic patients and different types of neutropenic patients, and then also is there any influence of drug-drug interactions from anti-cancer drugs which is often not studied and is only studied with basically specific groups. St. Jude's, for example, has done it with some of the kids. So, I would recommend some of those caveats down the line.

DR. GULICK: I think people are getting tired here. Let's move to the next question, which actually segways nicely onto this question, what additional Phase IV studies would you recommend? Dr. Wong?

DR. WONG: With respect to aspergillosis, I think that the data are very clear that voriconazole worked, but what we don't know is whether voriconazole is as good as, not as good as or better than several alternative drugs and I think that those studies should definitely be done. I would recommend that voriconazole be compared to liposomal amphotericin B and capsofungin in the same sort of clinical study design that was shown here today.

DR. GULICK: Dr. Morrison?

DR. MORRISON: At least in the setting of aspergillosis, clearly, the cure rates aren't as high as we would like although they are better than what we have previously seen. And, I think that raises the issue of looking at combinations of therapies in future trials, possibly either combinations of antifungal drugs but also one could consider combinations of these drugs with agents such as growth factors.

DR. GULICK: Let me mention some of the other things that have come up earlier today in terms of Phase IV things. In terms of safety, again, we wanted to see longer term safety on visual changes. I guess there is very little human histology data. Clearly, it is difficult to get such data.

We wanted to see information on QTc, prolongation of QTc interval at higher concentrations of the drug. We would like to see more description on the anaphylactoid reaction that came up. Other patient populations that we have suggested over the course of the day -- clearly, pediatrics are very important, particularly with regard to visual changes. Several people mentioned those with preexisting eye disease or specifically retinal disease; those with hepatic insufficiency and/or renal insufficiency, and there was a plea that those insufficiencies be defined in terms that clinicians know how to use in particular.

Drug-drug interactions that we have mentioned over the course of the day, specifically the HIV non-nucleoside reverse transcriptase inhibitors Efavirenz and Nevirapine and, as was mentioned earlier today, ritonavir as being, of the protease inhibitors, the most potent one that interferes with CYP3A4. Nelfinavir should probably also be added to that list.

Let's see, we requested earlier today information on activity of voriconazole against other fungi that we didn't really consider today or that we saw limited data on. There was a plea for more resistance data. Then, most recently, comparative data, as Dr. Wong suggested, of voriconazole against some of the other agents that are used commonly, liposomal amphotericin B and capsofungin being two examples. And, as Dr. Morrison reminded us, combination therapy of the antifungal agents would be an appropriate place to go from here.

Others to add to that list? Dr. Yogev?

DR. YOGEV: Well, when you say pediatrics, I just want to make sure the company is not feeling comfortable with up to six months candidiasis in the newborn, and after that kids sick less than six months -- so, I hope both the agency and the company will work on less than six months in pediatrics.

I noticed when I read this very thin yellow book that in vitro data shows, which is surprising, that cryptococcus is very sensitive to this specific material and might be in many populations, especially in AIDS patients if not in the United States and other places, a good one. Then also histo and blastomycosis, again, we are starting to see more and more of them and it will be very important to look into that.

DR. SCHAPIRO: A very minor issue, actually, nelfinavir is important to study. If I am not mistaken, of all the protease inhibitors, it utilizes 2C9 and not 3A4. So, actually although we are sort of fixated on this drug we should actually go from nelfinavir maybe for throughout others. So, it is a little bit different than our classic 3A4.

DR. GULICK: That is right and that is why I threw it in there. Dr. Hamilton?

DR. HAMILTON: Implicit in one of your suggestions, Dr. Gulick, is that the activity against other fungi be explored as well. This points out I think one of the major difficulties of the kinds of studies that we just reviewed, that is, knowing exactly what is going in people who are febrile and neutropenic. If we knew at that moment X person has Candida or Y person has Fusarium and Z person has crypto we would be a long ways ahead. Now, it occurs to me that in the course of this study a substantial amount of specimens must have been collected, and I don't know what happens to specimens in these circumstances but I know what happens to them in circumstances in which I have been involved. They sit in my freezer forever and nothing gets done with them unless I do something. But there are people out there who might want to actually utilize those specimens in a productive way to help identify diagnostic techniques that would be very useful in the acute stage of these infections. That would be very helpful I think in focusing therapy. So, I would encourage general collaboration with your ideas.

DR. GULICK: Thanks. Dr. Wood?

DR. WOOD: I would just like to reinforce a comment that Jonathan had made earlier this afternoon regarding trying to take advantage of autopsy specimens for any histopathologic examination of eye findings because these were huge studies with a huge number of patients, many of whom died and many of whom underwent autopsy. So, just being able to get the retinal autopsy findings of everyone who was exposed to voriconazole would get us some information fairly quickly and rapidly that we really feel is lacking regarding human toxicity in the ocular findings.

DR. GULICK: Let's move to our last question. This really follows up, as Dr. Goldberger mentioned, on several previous meetings. What additional advice does the committee have regarding the design of future studies needed in the development of therapeutic agents for the initial therapy, and therapy of patients refractory or intolerant to other antifungal therapies, in patients with pulmonary and/or disseminated aspergillosis?

Although we have talked a lot about it, we may want to expand that to the empiric febrile neutropenic patient population also. Dr. Wong?

DR. WONG: At the last meeting that we had to discuss an aspergillosis application I think I said that I didn't want to see another presentation that depended on historical controls, and I am delighted to see that today we didn't have to depend on an analysis of historical controls. I hope that the agency has now had demonstrated to it that a proper controlled trial for aspergillosis can be done, and this should be expected in the future. Indications that companies get based on analysis of historical controls alone I think should no longer be allowed.

The second plea that I would make is to put a stake into the idea of this composite endpoint that incorporates both efficacy and toxicity in the same measure. They should be analyzed separately and I don't think that there is any reason not to do that.

DR. GULICK: Dr. DeGruttola, maybe you could comment on the use of non-inferiority studies in general and what limits might be suggested.

DR. DEGRUTTOLA: Well, as I believe it was Dr. Hamilton who said -- my memory may be incorrect -- that the definition of non-inferiority has to be a clinical definition. It is really I think the clinicians that need to say what they consider non-inferiority in a particular setting. But I think that we need to keep in mind that if you are using absolute differences it does matter where you are in terms of the expected response, and that relative differences may be more useful to consider just because you don't have to have as specific an idea of what the response will be.

DR. GULICK: Would others support the continued use of non-inferiority studies in this setting, as was recommended in '94, '95 by the expert committee?

DR. WOOD: I would prefer that the committee or that the FDA consider the modifications that Victor had suggested earlier this afternoon in terms of taking into account the relative response rates when looking at inferiority.

DR. GULICK: One of the things we were asked to consider is initial treatment versus refractory treatment, and that has come up several times today. Do people have comments? If we focused on refractory patients in terms of study design itself, clearly a group that is in need of therapies? Good idea?

DR. ENGLUND: Sure.

DR. GULICK: Okay. It is like pulling teeth at this hour!

DR. YOGEV: let me give you one tooth --

DR. GULICK: Thank you.

DR. YOGEV: I think one of the problems of this specific study and many that I see is because we expect to have difficulty enrolling patients we shy away from trying to define groups in different studies, or stratify as this is called. We are terrified by the N, and this is the best example, if they would be prospectively done, the data we had today that we are struggling with so much, would probably be completely different. And, what always bothers me in studies coming both from industry of even multiple sites from the NIH is that we are opening it too much without realizing where it is going to end. So, I think a better way to define it -- for example, the way the question was put over here, I think it has already put us in trouble because it is only pulmonary and disseminated in one center and it should be different because it is a different disease. That goes to one of your suggestions, "or refractory to therapy." But those have to be defined up front, and the N has to be defined better. It anyhow takes a long time and we need to make sure can we do it or not. I don't think we spend time in devising the protocol and how it would work. So I would encourage to write the protocol in a different way and we will then go into those issues in a more specific population prospectively.

DR. GULICK: Do people have other comments? We struggled a lot earlier with subgroup analyses and retrospective versus prospective looks at the data. Do people have other comments about that issue in particular? Dr. DeGruttola?

DR. DEGRUTTOLA: No, I think actually that the sponsor did a very good job of specifying what were prospectively defined analyses and what were retrospectively analyses and subgroup analyses, and I think the important thing is just keeping in mind those distinctions when you review data.

I think one other issue actually that I would comment on is that if you are doing subgroup analyses, I believe it is useful to do formal tests of interaction rather than say, oh, there was a significant effect in one group and there wasn't a significant effect in another group. We have to raise the issue of was there adjustment for multiple testing; how were these categories defined; and so on. I think if you do formal tests for interaction they can be useful in this setting as well, as we saw in this case.

DR. GULICK: One part of the composite endpoint for the study we looked at was fever in relationship to the time of neutropenia. As we pointed out earlier, the time of neutropenia is shortening in general. Do people have comments about that particular endpoint? Dr. Wong?

DR. WONG: I think the trouble here is that how long someone is febrile is not really the question. So, the study should be designed to ask the most relevant question that matters, which in this case is how many patients developed breakthrough fungal infections and how many patients died. How many days of fever is really much less important, I mean an order of magnitude less important and I am afraid what drove the results in this study, which was a shame.

DR. GULICK: Further comments or issues? To Dr. Goldberger and the agency, have we done what you want us to do? Do you need anything else from us?

DR. GOLDBERGER: No, I think that is fine. I was trying to think of maybe three or four other questions --

[Laughter]

DR. WONG: Each with five sub points!

DR. GOLDBERGER: Yes! No, we thank you. I think this discussion was very helpful. Your comments about febrile neutropenia in particular were very helpful, both in terms of thinking about it with retard to this product and thinking about the broader issues of study design, etc.

DR. GULICK: Great! So, I would like to thank the members of the committee, particularly the ones who are the long-term survivors and lasted to the end of the two days. I appreciate that. Thanks to the sponsor for your presentations and for the follow-up clarifications, and to the agency for organizing the day, and the audience out there somewhere. Thanks very much.

[Whereupon, at 3:55 p.m., the proceedings were adjourned]

- - -