UNITED STATES OF AMERICA

FOOD AND DRUG ADMINISTRATION

 

+ + + + +

 

NEUROLOGICAL DEVICES PANEL

of the

MEDICAL DEVICES ADVISORY COMMITTEE

 

+ + + + +

 

Seventeenth Meeting

 

+ + + + +

 

TUESDAY

JUNE 15, 2004

 

+ + + + +

 

The Panel met at 8:00 a.m. at the Holiday Inn Gaithersburg, Walker/Whetstone Rooms, Two Montgomery Village Avenue, Gaithersburg, Maryland, Dr. Kyra J. Becker, Chairperson, presiding.

 

PANEL MEMBERS PRESENT:

 

KYRA J. BECKER, M.D., Chairperson, University of

Washington School of Medicine, Seattle, WA

ANDREW K. BALO, Industry Representative, DexCom, Inc.

San Diego, California

JONAS H. ELLENBERG, Ph.D., Voting Member, Westat,

Rockville, Maryland

LAURA J. FOCHTMANN, M.D. Deputized Voting Member,

State University of New York, Stony Brook, NY

ANNAPURNI JAYAM-TROUTH, M.D., Voting Member, Howard

University College of Medicine, Washington, DC

RICHARD P. MALONE, M.D., Deputized Voting Member, MCP

Hanneman University, Philadelphia, PA

IRENE E. ORTIZ, M.D., Deputized Voting Member,

University of New Mexico, Albuquerque, NM

MARY LEE JENSEN, M.D., Director of Interventional

Neuroradiology, University of Virginia


PANEL MEMBERS PRESENT: (continued)

 

PHILIP S. WANG, M.D., MPH, Dr.PH, Deputized Voting

Member, Brigham and Women's Hospital, Boston,MA

CRISSY E. WELLS, R.T., MBA, MHSA, Consumer

Representative, University of Virginia Health

Sciences Center, Charlottesville, VA

 

ALSO PRESENT:

 

DELIA WITTEN, Ph.D., M.D., Food and Drug

Administration, Division Director, General

Restorative and Neurological Devices

 

SPONSOR PRESENTERS:

 

RICHARD L. RUDOLPH, M.D., Vice President, Clinical and

Medical Affairs and Chief Medical Officer,

Cyberonics

A. JOHN RUSH, M.D., Principal Investigator, University

of Dallas Southwestern MC, Dallas, TX

ALAN TOTAH, Vice President, Regulatory Affairs,

Cyberonics

 

FDA PRESENTERS:

 

CHANG LAO, Ph.D., Statistical Reviewer

CARLOS PENA, Ph.D., Neuroscientist, VNS Studies,

Efficacy Reviewer

MICHAEL SCHLOSSER, M.D., Neurosurgeon, Safety Reviewer

 

PUBLIC SPEAKERS:

 

MARNA DAVENTORT, patient in the study

CHARLES DONOVAN, patient in the study

COLLEEN KELLY, patient in the study

LYDIA LEWIS, President, Depression and Bipolar Support

Alliance

KARMEN McGUFFEE, patient in the study

IRVIN J. MUSZYNSKI, J.D., Director of the Office of

Health Care Systems & Financing, American

Psychiatric Association

LAURI SANDOVAL, patient in the study


I N D E X

 

AGENDA ITEM PAGE

 

Call to Order 4

Conflict of Interest and Deputization

Panel Introductions

Update since February 23, 2004 Meeting

Theodore Stevens, Chief, REDB

 

1st Open Public Hearing 13

 

Sponsor Presentation 63

Alan Totah, Cyberonics, Inc.

A. John Rush, U. of Texas Southwestern

Richard L. Rudolph, Cyberonics

 

FDA Presentation

Carlos I. Pena, PhD 149

Michael J. Schlosser, M.D. 160

Chang S. Lao, Ph.D., Division of 178

Biostatistics

 

Panel Deliberations 197

Philip S. Wang, M.D., MPH, Dr.P.H.

 

2nd Public Hearing 368

 

FDA and Sponsor Summations 369

 

Panel Vote 385

 

Adjournment 433


P R O C E E D I N G S

Time: 8:05 a.m.

MS. SCUDIERO: Good morning. We are ready to begin now. Sorry for a little delay.

I am Jan Scudiero. I am the Executive Secretary of this panel and a reviewer in the Division of General Restorative Neurological Devices.

The usual housekeeping matters: If you haven't signed in at the door, please do so, and pick up agendas and other meeting related information.

Before I turn over the meeting to Dr. Becker, I am required to read three statements into the record. There are two deputization of temporary voting member statements and a conflict of interest statement that were prepared for this meeting.

The first: Pursuant to the authority granted under the Medical Devices Advisory Committee charter dated October 27, 1990, and amended April 20, 1995, I appoint the following person to be a voting member of the Neurological Devices Panel for the duration of this meeting on June 15, 2004: Laura Fochtmann, M.D.

For the record, she is a Special Government Employee and is a consultant to this panel or another panel under the Medical Devices Advisory Committee. She has undergone the customary conflict of interest review and has reviewed the material to be considered at this meeting. Signed by Daniel G. Schultz, M.D., Acting Director, Center for Devices and Radiological Health, on June 9th of this year.

The other: Pursuant to the authority granted under the Medical Devices Committee charter for the Center for Devices and Radiological Health dated on October 27, 1990 and amended August 18, 1999, I appoint the following individuals as voting members for the Neurological Devices Panel for the meeting on June 15, 2004: Richard P. Malone, M.D., Irene E. Ortiz, M.D., Richard S. Wang, M.D. MPH, Dr.Public Health.

For the record, Doctors Malone, Ortiz and Wang are members of the Psychopharmacologic Drugs Advisory Committee of the Center for Drug Evaluation and Research. They are Special Government Employees who have undergone the customary conflict of interest review and have reviewed the material to be considered for this meeting. This is signed by Peter J. Pitts, until recently the Associate Commissioner for External Relations, on June 4th of this year.

The conflict of interest statement: The following announcement addresses conflict of interest issues associated with this meeting and is made part of the record to preclude even the appearance of an impropriety.

To determine if any conflict existed, the agency reviewed the submitted agenda for this meeting and all financial interests reported by the Committee participants.

The conflict of interest statutes prohibit Special Government Employees from participating in matters that could affect their or their employers' financial interest. However, the agency has determined that the participation of certain members and consultants, the need for whose services outweighs the potential conflict of interest involved, is in the best interest of the government.

Therefore, waivers were granted to Doctors Kyra Jo Becker and Laura Fochtmann for their interest in firms and issues that could potentially be affected by the panel's recommendation.

Dr. Becker's waiver involves an imputed interest, a contract to her institution for the sponsor's study in which she has no involvement and is uncompensated. Dr. Becker's waiver allows her to participate fully in today's deliberations.

Dr. Fochtmann waiver involves a contract to her institution for the sponsor's study in which she has no involvement and is uncompensated. Dr. Fochtmann waiver allows her to participate fully in today's deliberations.

Copies of these waivers may be obtained from the agency's Freedom of Information Office, Room 12A-15 of the Parklawn Building.

We would like to note for the record that the agency took into consideration certain matters regarding Dr. Mary Jensen. She reported an interest in a firm at issue but not in matters related to today's agenda. The agency has determined, therefore, that she may fully participate in all discussions.

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 participant should excuse himself or herself form such involvement, and the exclusion will be noted for the record.

With respect to all other participants, we ask, in the interest of fairness, that all persons making statements or presentations disclose any current or previous financial involvement with any firm whose products they may wish to comment upon.

I wish to announce that the August 5th and 6th tentatively scheduled meeting for this Panel was canceled, because there is no agenda for a meeting. The remaining tentatively scheduled Panel meeting for this calendar year is October 28th and 29th.

Please remember that this is a tentative date, and monitor the CDRH Panel website for updated Panel meeting information.

I would now like to turn the meeting over to our Chairperson, Dr. Kyra Becker.

CHAIRPERSON BECKER: Good morning. My name is Kyra Becker, and I am the Chairperson of this Neurological Devices Panel. I am a neurologist, and I practice at the University of Washington in Seattle.

At this meeting the panel will discuss, make recommendations and vote on a recommendation to the Food and Drug Administration on the approvability of premarket approval application supplement P970003/S50 for the Cyberonics Vagus Nerve Stimulation Therapy System.

The System is indicated for the adjunctive long term treatment of chronic or recurrent depression for patients who are experiencing a major depressive episode that has not had an adequate response to two or more antidepressant treatments.

We will have an open public hearing and the sponsor and FDA presentations before lunch. After lunch, the Panel will deliberate on the approvability of the PMA. Before the Panel votes, there will be another open public hearing and a time for the FDA and sponsor summations.

Before we begin this meeting, I would like to ask for the Panel members, who are generously giving their time to help the FDA in the matter at hand, and the other FDA staff seated around this table to introduce themselves. I think we are going to start at this end of the table, and I would like you to state your name, your area of expertise, your position and your affiliation.

DR. ELLENBERG: My name is Jonas Ellenberg. I am a biostatistician. I am on staff at Westat, a social services private research firm. I am a Vice President and Senior Biostatistician.

DR. JAYAM-TROUTH: I am Annapurni Jayam-Trouth. I am the Chair of Neurology at Howard University, Washington, D.C.

DR. FOCHTMANN: I am Laura Fochtmann. I am a Professor in the Department of Psychiatry at the State University of New York at Stony Brook.

DR. WANG: I am Philip Wang, a psychiatrist and epidemiologist at Harvard Medical School.

DR. JENSEN: I am Mary Lee Jensen. I am Director of Interventional Neuroradiology and a Professor of Radiology and Neurosurgery at the University of Virginia.

DR. ORTIZ: I am Irene Ortiz. I am a geriatric psychiatrist at the University of New Mexico and with the Albuquerque BA.

DR. MALONE: I am Richard Malone. I am a psychiatrist and professor of psychiatry at Drexel University, College of Medicine.

MS. WELLS: I'm Chris Wells, and I am the Consumer Representative on this Panel.

MR. BALO: I am Andy Balo, the industry rep, but I am Vice President of Regulatory Clinical at DexCom, Inc., in San Diego.

DR. WITTEN: Celia Witten. I am Division Director of the reviewing division for this product at FDA.

CHAIRPERSON BECKER: Thank you. I would like to note for the record that the voting members here at the Panel constitute a quorum as required by 21 CFR Part 14.

Next Mr. Theodore Stevens, Chief of the Restorative Devices Branch, will update the Panel on several matters deliberated on at the last meeting of the Panel on February 23, 2004. Mr. Stevens.

MR. STEVENS: Okay. I don't seem to have slides going here on the screen.

Hi. I am Ted Stevens. I am the Chief of the Restorative Devices Branch, and I will be giving a very brief update on the devices that were reviewed by this Panel previously.

At the February meeting there was advice given to the reviewing branch on the concentric medical Mercy device that remains under review in the General Surgical Devices Branch at FDA.

We have also recently published a draft guidance and a Federal Register notice for availability of vascular and neurovascular embolization devices. The comment period for that draft guidance ended on May 25th.

Finally, several devices have been cleared that are reviewed under the Orthopedic Devices Panel, but because they are devices that are also used by neurosurgeons and interventional radiologists, I thought it would be appropriate to mention them here.

We have recently cleared several PMMA cements for use in pathological fractures of the vertebral body. These 510k's were cleared based on the reclassification of PMMA bone cements. That included pathological fractures in general for the specific indication of vertebral body fractures. These 510k's included clinical data from the literature.

That concludes the update.

CHAIRPERSON BECKER: Thank you, Mr. Stevens. At this time we will proceed to the open public hearing portion of the meeting. We ask at this time that all persons addressing the Panel speak clearly into the microphone, as the transcriptionist is dependent on this means of providing an accurate record of the meeting.

Ms. Scudiero will now read a statement prepared for the open public hearings.

MS. SCUDIERO: Both the Food and Drug Administration and the public believe in a transparent process for information gathering and decision making. To ensure such transparency at open public hearing sessions of the Advisory Committee meeting, FDA believes that it is important to understand the context of any individual's participation.

For this reason, FDA encourages you, the open public hearing speaker, at the beginning of your statement to advise the committee of any financial relationship that you may have with the sponsor, its product and, if known, its direct competitors.

For example, this financial information may include the sponsor's payment of your travel, lodging or other expenses in connection with your attendance at the meeting.

Likewise, FDA encourages you at the beginning of your statement to advise the committee if you do not have any such financial relationship. If you choose not to address the issue of financial relationships at the beginning of the statement, it will not preclude you from speaking.

CHAIRPERSON BECKER: Prior to the meeting, there were seven requests to speak in the open public hearing, and each person has ten minutes to address the panel. They will speak in the order that the FDA received their requests to present to the panel.

Ms. Colleen Kelly, who is a patient in the study, will be the first speaker, if you want to come forward.

MS. KELLY: Hi. Is it possible for me to sit?

CHAIRPERSON BECKER: Certainly, yes.

MS. KELLY: You can hear me okay? Thank you. I am Patient 012 from Site 050 of the D-O1 study for Vagus Nerve Stimulation for the treatment of major depressive disorder. I am here unsolicited to ask you to approve the VNS in its application to depression.

I was not approached by my study site nor by Cyberonics to speak to this panel. I am here upon my own accord, inspired by my own experience, and driven by the necessity that viable alternatives must be offered to persons suffering from treatment resistant depression.

I have secured and paid for my own travel here. In fact, it would behoove me financially if this panel stalled he approval process. I have had the device for four and a half years and, based on my parameter settings, replacement surgery is imminent.

As long as I am in the study and as long as the study lingers, Cyberonics will be responsible for the cost of replacing my generator battery. So, actually, I am shooting myself in my financial foot by encouraging you to approve this device. Upon approval, I would become completely responsible medically and financially.

That said, my case and testimony may be of particular interest to this panel. I went into the study on no medications, remained off medications throughout the study, and am still on no medications.

There is no other explanation for my response other than the device itself. It is good science. Turn the device up; I respond. Turn the device down; I decline.

Let me assure you that my stoicism with medications is not noncompliance. Neither is it an indicator that I am only mildly affected with the illness and, therefore, can exist without medications. In short, I had exhausted them as well as every other treatment currently available to people with illness at this level, as did so many of my peers in this study.

I realize that most on this Panel and most in attendance here are experts in their field, be it neurology, psychology, psychiatry, or mental health administration; but there is one thing I can offer you about which you may know nothing. That is first hand knowledge of what it is like to have severe recalcitrant depression.

To do so, I ask you to imagine the unimaginable, to think the unthinkable, to experience second degree emotional burns with third degree prognosis. All you experience is pain, but with no cure. In fact, there is no viable treatment.

You can attempt to salve it. Only death solves it. But the medical community does not accept death as a cure. It asks us to continue to hang on and to continue to live, yet offers us no viable treatments. Trust me, it is not that we don't want to live. It is that we don't want to live like this.

Our illness is embedded in our physical bodies, ourselves. We are prisoners there, and our sentence is life. Menacing insomnia, isolation, fear, anxiety, sadness, hopelessness, general malaise, malingering fatigue, physical exhaustion, apathy, lack of motivation, concentration and focus, absence of pleasure, amplification of pain, agitation, sensitivity to criticism, thoughts that life isn't worth living -- You are all familiar with this short sheeted laundry list of symptoms.

Now imagine having them all at once. Imagine passing from one room to another in the house of pain where some symptoms are more prevalent than others, sometimes exacerbated by the very medications that were meant to alleviate them.

I will not bore you with the details of the pharmacopoeia that I have tried and then have failed, not to mention the acupuncture, homeopathy, herbal remedies, extreme dietary changes and supplements, light therapy, counseling, yoga and, of course, religion. What god would let a child suffer like this?

Then comes the inevitable, electroconvulsive therapy, ECT, a therapy so beyond the vernacular that it doesn't even pass an automated spellcheck. I would stop at the word, too, but as a person with treatment resistant depression, I could not stop.

I relented to this FDA approved treatment as a last resort. Average session: Three to five treatments. I had 33 nearly consecutive treatments. I lost retrograde 20 years of my life through memory loss, a dismal blessing. At least I could not remember the horrific pain that preceded it for years.

I asked you to imagine the unimaginable, to think the unthinkable. I also mentioned that you are experts in your field. What if, after battling an abusive lifelong illness, after enduring medication where side effects trumped minimal benefit, after relenting to a final last resort -- what if all that is wiped out? You are etherized on a tabula rasa, alive, yes, but no more FDA hearings, no more status, no more career. No more. Someone helps you remember where the soap is kept and helps you into the shower.

I am not here to blast the approval of ECT. Obviously enough, for better or worse, I am still here. I am here, however, to say that I am here today because of the VNS, and I ask you what do you offer someone after the last resort? What do you offer someone for whom ECT has failed?

I had an experimental study. What will the next guy get?

I would be remiss if I did not mention the epileptic community who has blazoned the way for efficacy and safety of the VNS device. For that, I am grateful.

They have shown us the risks and side effects involved since the FDA approved the device for them in 1997, and my personal heartfelt thanks go to those who engaged in experimental implantation even prior to that date. I offer my gratitude, and I share their cautious hope.

I would also like to thank them for their honesty and candor, which I found on the Cyberonics bulletin board. Without their shared experience, I know for a fact I would not have attained the success I have had with this device.

I followed their histories. I tracked and evaluated their settings and results, and I searched and researched my side effects and remedies through their personal experiences. Their trial and error resulted in my trial and success.

Because of what I learned on the chat board, I was eager to boost my amplitude to higher than 1.0 in the first three weeks of the initial depression study, and keep my settings above what later became known as efficacy threshold. As a result, I was one of the first to respond at my study site, and I continue to reap benefit from the device.

I encourage Cyberonics to reopen their bulletin board. I understand the risks involved and the abuses that were made to it, but I challenge the company to research and execute a safe environment where patients can exchange information regarding this very new modality of treatment, especially in its application to depression.

It was critical in my success. I know it would increase the success of others exploring the device as a possible treatment.

I am not going to idealize nor sentimentalize the device. I know I am one of a third who have responded to it. I know there are others who continue to suffer the burden of treatment resistant depression. I see it in their faces as I sit in the lobby and wait my turn at the site.

I know that pain. I suffered it prior to the VNS. Still, I have windows of it now and again. I am passionate, yet realistic, about the device. I do not romanticize its results for me nor dismiss its lack of results for others. I am well aware of its side effects, shortcomings, and its current experimental status. But I do know one thing. We need viable treatment options for those with recalcitrant depression, and VNS has worked for me.

In closing, I encourage this Panel to let the mental health community, both administrators and recipients, review the data and search the testimonies to decide if this is an appropriate treatment for their patients, their loved ones, themselves.

This is an option only you can provide by approving the device for its application to depression. Give us a choice, a possible key, a parole to our life sentences of depression. Thank you.

CHAIRPERSON BECKER: Thank you, Ms. Kelly.

Our next speaker will be Ms. Lydia Lewis, who is President of the Depression and Bipolar Support Alliance.

MS. LEWIS: Good morning. I want to thank the Advisory Panel for this opportunity to talk about the critical issue of treatment resistant depression. I am Lydia Lewis, and I am here as President of the Depression and Bipolar Support Alliance, a national patient-run advocacy organization representing the more than 25 million people living with depression and bipolar disorder.

I did not receive any financial support nor do I have any financial arrangement with any company for my work on behalf of patients with mood disorders. The Depression and Bipolar Support Alliance does, however, receive financial support in the form of program grants, honoraria, consulting fees or in-kind donations from Cyberonics, Inc. and a number of pharmaceutical companies. My travel to Maryland today was not paid for by Cyberonics, and I am not here to advocate for any particular therapy, including VNS, but rather for the critical need for new therapies.

While I have suffered from treatment resistant depression for my entire life and I have taken more than 20 different medications over the past 36 years, I am not here to tell my story. I am here to represent the millions of people with mood disorders who can't get well, people who desperately need better medications and treatment modalities other than pharmaceuticals.

Just about 4 million people directly contact DBSA every year. Connecting with millions touched in some way by Depression and Bipolar Support Alliance -- by depression and bipolar support make us particularly qualified to speak on their behalf.

As we all know, more than 30,000 people in the United States take their lives every year, not because they are weak or because they have a character flaw. They take their lives because they do not respond to any of the treatments currently available for depression or bipolar disorder.

At DBSA we know that new drugs and nonpharmacologic treatments are desperately needed. Far too many people are dying or living lives of quiet desperation, because they can't get sufficient relief form their symptoms of depression. The more treatments the FDA makes available, the more lives that will be saved.

Today I want to put a human face on the tragedy of treatment resistant depression. We can talk science all we want, but science will never drive home the devastating consequences our illnesses can have, if they are not treated.

The human face I want to share today is Barbie's, and I have given you all a picture of Barbie. It is on the last page of my testimony, and it is important, if you would, to look at this while I speak.

Barbie has three kids, two girls and a boy. She works as an oncology nurse. She is very patient, very smart, very kind, very funny, and very gentle. She is the nurse of choice, requested by more doctors when they admit a patient onto her unit.

People report that her care made the difference in their fight with cancer. Barbie was finally diagnosed with bipolar disorder after having been treated for depression for more than 15 years. She has been hospitalized five times for her depression.

The first two times her insurance covered some of the costs. The last three times it covered nothing. She has run through all of her retirement savings. Her last hospitalization was at a state facility and, although many of them are quite good, this one was so horrible and so scarring that she refuses to ever be hospitalized again.

She runs through the insurance money for her medication within the first two months of every year. Her parents use much of their retirement savings to help pay for her meds, and her siblings contribute as much as they can.

She has been prescribed a variety of medications. Sometimes she shakes. Sometimes she drools. A time or two she has found herself somewhere with no idea how she got there, because her mind was so fuzzy from her medication.

She balloons up in weight even though she eats next to nothing. She has no appetite at all. It is hard to find a nurse's uniform that fits. She gets very little sleep, 15 minutes, an hour at a time, no more. But she also can't concentrate enough to watch TV, read a book or anything else. So she lies there worrying, thinking about the patients she is trying to care for, even though she herself is very, very ill.

Once a week she summons up all her courage and drives, shaking and drooling, to a therapist more than an hour away. There is none closer. Every month it takes even more courage to drive two hours away to see her psychiatrist for 15 to 20 minutes. There is none closer.

They do the best they can to treat her at reduced fees, but she can still barely pay for their time, and nothing helps. Quite simply, her life is hell. She can't sleep. She can't eat. Her phone rings with creditors. She feels bad and ashamed about what her family has been forced to go through with her.

Her medication barely touches her symptoms, and because of her illness, everything she does takes tremendous courage. Let me tell you, it is exhausting to never get better. When no treatment works or you can't afford something that does, you can reach a place where there doesn't seem to be any exit.

If you haven't been there yourself, count yourself lucky. I have been there, and it is dark and full of pain and hopelessness. It is joyless and a place where nothing matters.

You can be young or old, beautiful or plain, rich or poor, erudite or illiterate. Everyone who ends up in this place feels the shame. It's hell on earth.

Barbie found herself in this place, because no matter how hard the doctors tried, nothing helped; and regardless of her kids, her husband, her wonderful parents and siblings, and the job she found so important, the despair of that place was too overwhelming.

Barbie isn't just any patient to me. She is my co-worker's sister, and I am sorry that we never met, because I know I really would have liked her. Barbie took her life three months after this picture was taken. She was 49 years old.

It is a profound honor to speak for Barbie who can no longer do so, and I hope her death will have some purpose. No one should have to live with pain like Barbie's. No one's family should have to suffer like hers, impotent to help through the long journey of one failed treatment after another.s

The suffering doesn't end when someone's life ends. Barbie's loved ones still suffer, because even in this day and age, there was nothing that could keep her from that dark place.

Tragically, aspects of Barbie's story still ring true for so many. No one's life should be wasted or ended because efficacious treatment isn't available. This is why we are all here. This is why we all must remain here.

It is too late for Barbie, but it is not too late for us to learn from her life and from her death. Her struggles and loss should inspire all of us to work tirelessly to bring better treatments to the millions of Barbies still suffering.

That is why I am here today, and why I am asking the FDA to remember those of us who desperately need better treatments. It is a race against time, and I ask you on behalf of the millions of people suffering with treatment resistant depression to please do everything you can to help us. Thank you.

CHAIRPERSON BECKER: Thank you, Ms. Lewis. The next speaker will be Ms. Laurie Sandoval, who is also a patient in the Cyberonics study.

(A SHORT VIDEO WAS SHOWN.)

MS. SANDOVAL: First of all, I have had no financial involvement with Cyberonics except travel and hotel accommodations.

Someone once said depression is like being a prisoner of the mind. It could not have been said better except in that prison I was in solitary confinement.

The video you just saw was me five years ago. No longer able to keep my depression at bay, I had just resigned the job of my dreams and lost any hope of living. At that time I was under the care of Dr. Lauren Marengel of Baylor Medical College for treatment resistant depression in which medications, therapy and, in some cases, electric convulsive therapy had not worked.

Dr. Marengel explained an experimental study for depression that Baylor was undertaking with Cyberonics using a vagus nerve stimulator device, and would I be interested in participating. Sign me up yesterday, Doc.

Living in Nevada and without -- and having the VNS device five years now, life had been wonderfully normal, without suicidal depression. That is until three months ago.

I started feeling down, rarely leaving the house, turning off the phone, and only getting out of bed to let the dogs out. Normally, when the VNS device goes off, it causes a tickling sensation in the throat, but latterly there was no sensation at all, and I prayed the battery was dead.

Calling Dr. Marengel's office, they couldn't say for sure if the device wasn't working until the battery was tested. My appointment wasn't scheduled for another month, and I wondered if i could hold on that long. I was once again a prisoner of the mind.

In Houston, I told the doctors I felt desperate, and my only hope was that the depression was being caused because the VNS battery died. Scared, I checked myself into Methodist Hospital psychiatric ward.

The good news is two days later, thanks to the incredible teams at Baylor and Cyberonics, I had surgery for a new and improved VNS device. This battery is expected to last eight to ten years.

Every day I wake up with a bead of joy in my heart, or maybe that's the pulse of my new battery. Anyway, I have no doubt that, if it were not for Cyberonics' innovation and Baylor's tireless dedication, I would not be here today. Thank you.

CHAIRPERSON BECKER: Thank you, Ms. Sandoval. I want to apologize up front if I mispronounce the next speaker's name, Mr. Irvin Muszynski, who is the Director of the Office of Health Care Systems & Financing of the American Psychiatric Association.

MR. MUSZYNSKI: Your pronunciation is perfect. Thank you.

Good morning, and I, too, would like to thank you for the opportunity to offer a few remarks here this morning about your considerations about the Cyberonics application.

As indicated, my name is Irvin Muszynski. I am the Director of the Office of Health Care Systems & Financing at the American Psychiatric Association. In that capacity, my key job responsibilities are ongoing liaison on behalf of the psychiatric community and its patients, with third party patients in both the public and private sector. That includes employers. It includes Medicare and Medicaid, insurance companies, both health and disability.

In that capacity, I would simply indicate to you that the question of chronic mental illness conditions, particular the depressive disorders, is a recurring issue in questions I get -- or I get questioned about all the time, and what can the psychiatric community do, and so on.

In fact, part of my monitoring of the evolution of the vagus nerve stimulation approach was stimulated in large part by interactions with disability insurers who face the burden and the consequences of recurring or treatment resistant depression on an ongoing basis.

So what I wanted to do here with you today is just briefly highlight the problem and its prevalence from our point of view and, secondly, to talk a little bit about the burdens and costs associated with major depressive disorder, treatment resistant depression, however we want to characterize or define that.

That is from the patient, the payer, the purchaser point of view. I think all three -- Rather than qualify it every time, let's just suggest that it is on behalf of all three, albeit they all have different kinds of interests, and I think you have heard some compelling stories about the patient point of view and, I think, what becomes a self-evident need or an overwhelming need to address the need for new treatments to begin to better help manage chronic conditions in a way that people's functioning can be restored and/or recovery is, in fact, completely enabled.

Respecting disclosure, I have absolutely no current or previous financial relationship or interest in Cyberonics. I think probably my travel is not paid for here, and so on and so forth. I think probably the only thing that would be maybe appropriate to indicate in the context of the conflict of interest is Cyberonics advertises in APA journals and publications, as does a number of its competitors that would be on the psychopharmacology side of the house. But other than that, there is no direct interest.

So briefly, what I want to do is highlight the problem, and some of this you may well be aware of. So I don't want to be overly redundant. But as you know, depression is a diagnosable mental medical condition.

The American Psychiatric Association has developed the DSM-4TR. The criteria which surround the nature and diagnosis of depression are well established, continually refined, and furthered by the Association and the medical community at large. So there is nothing new about that.

Its prevalence, as you probably may well be aware, I will rely on National Institute of Mental Health indications of prevalence which show that seven to ten percent of the United States population at any given time suffers from a diagnosable depressive disorder.

As you know, many patients are treated successfully at first line attempts. Depression is a eminently treatable disease, but the issue on the table for us and under consideration here is the significant proportion of people who fail to reach acceptable levels of functioning and wellbeing. That is the population at issue.

I will not pretend to tender -- I've reviewed literature on this subject, and I think I can say conclusively, there is no one definition of treatment resistant depression. Maybe it is failure of one or two psychopharmacological interventions within a period of six months, and so on.

I think the folks who spoke before me give you much more operational definitions of what major depressive disorder or treatment resistant depression is. But in any case, treatment resistance, from my point of view, refers to the absence of an acceptable clinical outcome; that is, sustained remission, defined in terms of depressive symptoms, severity or daily function to one or more prior adequate treatments.

So when we subdivide the entire U.S. population of seven to ten percent that have a diagnosable depressive disorder, of that group there are various ranges of who really would be defined, if you will, as the treatment resistant population, and the estimates range anywhere from 10 to 50 percent, depending on the literature review.

Let's say the reasonable man standard would be somewhere in the middle. So 20 to 30 percent of those with a diagnosable depressive disorder simply don't respond, do not have sustained remission, and the human and economic consequences of that, I think, are kind of self-evident, and I will speak to them a little bit more here in a moment.

Let's look at the consequences then from the human, the patient, point of view. I won't elaborate on that. I think you just heard it, but also from the economic point of view. Here, I want to again reemphasize, my job, my role as an advocate in the psychiatric community both for psychiatrists as medical clinicians who treat this disorder, but also based on the interaction I do with those who underwrite health insurance and disability insurance, and also employers as purchasers who want their employees back at the job functioning, or those who are in the public sector, whether on Medicaid and/or Medicare or dual eligible population, who are interested in some restoration of ability to resume some kind of reasonable life in the community.

You know, the mortality has a tremendous range with depression. It can start with suffering and anguish, but it moves to job absenteeism, mistakes, loss of job, subsequent financial distress. I think you have heard personal testimonies to this, and it contributes to familial/marital discord and community discord in a number of ways.

The consequences of this from an economic point of view or burden point of view are kind of striking. The average annual cost of folks who are mildly resistant to treatment is easily double those for those who are not resistant to treatment, and those who would be categorized by the health services research literature who are severely resistant, if you will, to treatment -- their average annual costs sometimes are fourfold those who are not resistant to treatment at all.

This is an extraordinary clinical challenge. It is an extraordinary economic challenge, since half of the annual costs associated with treating depression are accounted for by this population.

Not only is health care utilization higher -- we can measure that by prescription costs. We can measure it by hospitalization rates. We can measure it by outpatient visits and so on -- the impact is significant from a fiscal point of view.

Then there are also all sorts of indirect effects. The type of workdays that are lost by individuals in the workplaces is on average double those who are not diagnosed with a depressive disorder. The preponderance or the prevalence of individuals on short and long term disability with major depressive disorders, the cost of which is borne directly and indirectly by employers and others, is significantly higher.

So in sum, depression not only has a negative economic burden, but there is no way to quantify the burden on the patient, their family and the community at large.

As indicated with respect to disability, you may be familiar that the World Health Organization now counts depression as the fourth leading cause of disability worldwide. The trend is severely upward, and the projections, if all things hold constant, is that within roughly or so the next ten years it will become the number two or one cause worldwide of disability.

The United States in that respect is not a statistical anomaly. The trend is essentially -- It tracks the world trends.

So it is clinical and economic opportunity loss. It hampers physicians to the extent that first, second line, third line attempts cannot work, and the personal tragedy of suicide and loss of life is untold.

I think what these findings underscore is the need for early identification and effective long term management of treatment resistant depression. And given the prevalence and the reality of nonresponse often to first line treatment, the cost and burden and the residual symptoms of nonremitting depression, we think there is a significant need for new treatments.

I am not here to opine on the science. That is not my job, but I do think, from where I live in a day to day world and the kinds of folks I deal with in the public and private sector, whether insurers or purchasers or human resource individuals, that the development of new treatments for this subset of the population diagnosed with a depressive order would be a significantly welcome development, and I have no doubts at all in thinking that the cost/benefit to all of us will be quite positive in the end. Thank you.

CHAIRPERSON BECKER: Thank you, Mr. Muszynski. Our next speaker will be Charles Donovan, who is also a patient in the Cyberonics study.

MR. DONOVAN: Good morning. First of all, Cyberonics did pay for my airfare here, in coach, I might add, and for my lodging. However, I was not solicited by Cyberonics or the study investigators to come. I volunteered, and as a matter of fact, to volunteer I had to go through a third party. The study investigators would not allow Cyberonics to contact me. I was not allowed to contact them.

What I would like to do, if it is okay, is read a letter that I sent to the FDA in support of the application and make a few comments. This is a letter that I wrote before I had any idea that I would be here, let alone reading it out loud in a room full of people. It is dated May 10th, and it is addressed to the Executive Secretary, Janet Scudiero.

"Dear Doctors: I am a patient in the D-02 study, and I am writing to urge you to do everything possible in your power to unconditionally approve the vagus nerve stimulator as a treatment for chronic depression. It not just saved my life, which I really didn't care about. It changed my life.

"In the spring of 1980 I graduated from Georgetown University -- ironically, less than 30 minutes from where the panel meeting will take place. I also had my first major depressive episode that spring.

"I had accepted a job in the management training program of what is now J.P. Morgan Chase and moved to New York City after graduation. I bounced back during that summer, but the depression was always there, lurking in the background, and depressive episodes became more frequent over time.

"To keep this letter short, let's fast forward 15 years. In November of 1995 I eventually ended up in a lock-up unit of a hospital. I have very little memory of the details surrounding that hospitalization. I can only assume that my family must have had the hell scared out of them, and they didn't know what to do.

"In 1998, I was a 39-year-old man sobbing uncontrollably, hugging my parents in the doctor's office after the psychiatrist recommended shock treatments. I had a series of 15 or so. The ECT treatments did not work nor did any of the countless antidepressants I tried.

"In late 1999, I was no longer able to work. I don't know how I was able to continue to work as long as I did. I put up a long, hard fight, and it was a big mistake. The physical toll and mental toll that it took on my body was agonizing, and I am still recovering physically.

"In 2001 I was implanted with a vagus nerve stimulator. In my final depression rating interview just prior to implant with Raymond Tate, Ph.D. of St. Louis University, I simply told him that I hoped I would die on the operating table. Given the relative simplicity of the procedure, it was an irrational thing to hope for, but dying would have been the ultimate escape.

"I have no idea when the device was activated or how or when it was ramped up. As recently as a few months go, the study investigators told me that I may never know. All I can say is that my life is full of genuine happiness and joy. I don't have to fake it anymore.

"I have lost 30 pounds in the past 18 months. I exercise regularly, swimming, Pilates, running. I am not ashamed to go to a shopping mall or other public places for fear of being noticed.

"Last Saturday night I attended a small dinner party that 18 months ago I never would have gone to. I am also working on several different projects. This is the most productive I have been in many years.

"The improvement in mood occurred very gradually over many months, but every morning I wake up, and I still ask myself the same question: Is it the depression back? And by the time I get in the shower, the answer is no.

"Because the mood improvement was gradual, there was no dramatic epiphany. So at this point, I would have to say that the most remarkable thing is the staying power of the therapy. It's like the Energizer battery. It keeps going and going.

"With the information learned from the studies and the benefit of stimulation strategy that new patients would have that I did not have, many desperate patients could be helped. Unless you have personally suffered from chronic depression, you cannot truly understand it, but it is brutal.

"Again, I encourage you to approve this relatively straightforward procedure for an extremely gruesome disease. Please give the option of vagus nerve stimulation therapy to those suffering patients who are still searching for an answer, just as I had searched. Some desperate patients stop searching forever."

Last night as I was reading this letter out loud to kind of time it, terms like chronic depression and treatment resistant depression are really slang for an incurable disease or a disease that is becoming curable. I think that it diminishes the seriousness of the level of the disease. It diminishes the need for alternative therapies.

I am guilty of it in this letter. I don't adequately relay the day to day grind of white knuckling it. It is exhausting. And again, there is a casualness about the term chronic depression, but the casualness is, in effect -- It is chaos. There is chaos in the family. There is chaos between brothers and sisters and parents and husbands and wives.

I know there was chaos in my family, but I don't know the half of it, because there were certainly countless secret meetings amongst my family members saying, what are we going to do with this guy?

It takes a lot of work before you get the label, treatment resistant depression; and when you get there, you are stuck.

In 2001 I was implanted with the vagus nerve stimulator, but in November of 2000 I had my first meeting with the study investigators, and I remember it was a small office, and the psychiatric nurse was next to me, and the lead doctor was in the office.

I said to the doctor, "What are the chances of this thing helping me? You know, what are the odds?" And he was very measured and cautious in his response, and he basically said there is an inkling that there may or may not be something to this device that could improve mood. And I said to myself, inkling? I'll take it.

But when you think about it, what was I supposed to say? What were the options? Nothing. So the litmus test was "inkling," and I think that is probably true of the 4 million patients that suffer from chronic depression.

You know, there's 20 million people with depression in the United States, and there are 4 million that have chronic depression. You're talking the bottom of the barrel, and the therapy that you are deliberating about today is a therapy for people that are at the bottom of the barrel.

I just want to quickly mention, I encourage you to improve this relatively straightforward procedure. During the first three months after implant, it was clear I had absolutely no benefit from the device, but it was a very odd time, because that was when my hoarseness was the worst.

People would ask me, aren't you mad? And the answer was no. Somebody was telling me to stop. How long have I been talking?

CHAIRPERSON BECKER: A little over 10 minutes.

MR. DONOVAN: The answer was no. I mean, I wasn't happy about the hoarseness, but the hoarseness was completely subordinated to the hopes that the device would work.

Very quickly in one minute, some desperate patients stop searching forever. March 14, 2003: I found the body of my closest friend from seventh grade, dead. Thankfully, his brother, an M.D., was with me. His body was laying between the bathroom in the hallway, and his brother found on his bureau a bottle of antidepressants, but his search stopped. He ended his search.

A year before that, another classmate -- we only had 40 -- put a bullet to his head, a psychologist. And finally, at the same time of the psychologist's death, the wife of my brother's boss, the mother of seven children -- I spoke to her husband, Tony, last week. And unlike Barbie that Lydia talked about, this family is a very prominent family in St. Louis, and they had access to absolutely every possible thing out there. Money was no object, and they went out on a nationwide search for help for this beautiful, stunning woman, and they saw the freight train coming, and there was nothing they could do about it until one of the seven children found her hanging in the family home, dead at age 38.

So I've gone over. If there's any questions that you want to ask me as a patient, please do.

CHAIRPERSON BECKER: Thank you. Our next speaker will be Marna Daventort, who is also a patient in the study.

MS. DAVENTORT: After hearing everyone else speak, it is really difficult to come up here and talk about this clinically, because I had really a lot more prepared, but they have said it all. They are telling my story. Their lives have been my life.

I want you to know that I am a person who can meet any challenge. I fly airplanes. I have a Ph.D. I ride beautiful horses, and I ride them fast. I can meet any challenge, but I couldn't beat depression, not with all the knowledge that I could gain from studying it, and not with all the energy that I could put into it.

I can go over all the treatments I have been through, and I have been through it all, every kind of therapy that you can think of and every drug that any of you could think of prescribing to me, in experimental dosages often and in weird combinations that sometimes even the doctors were afraid to try. But we had to try anything, because the alternative is death, and people like me don't want to die.

I have a wonderful life now. I have a wonderful family. I never had drug or alcohol problems. I never had weight problems. I had no excuse whatsoever to be depressed, and yet I was depressed.

I think that that is what happens. A lot of times, we look at people with depression, and we think that somehow they could just do better. And I know that you can't just do better.

One reason I am here today, and Cyberonics paid my plane fare up -- It won't compensate for the salary I have lost today, but I am here because I think it is -- The single most important thing that I can do today is tell you that I was implanted three year ago.

At first there was really no dramatic result. My family says that they saw results right away. I didn't, but over the next year, especially when it became about the 18 month mark, I began to be the person that my family used to know. My Dad said he had his daughter back.

I think that, with all else I could say, all I can say to you today is that this worked for me, and the alternative for me, to put it bluntly, was to blow my brains out.

So I think that it would be unconscionable for you not to offer this treatment option to people in my position. It has never been proposed that it would be a first line of defense. It has never been proposed to be a replacement for the drug therapy or for the resolution of psychological problems.

All we are asking for is that you make this an option to people who have no options remaining. So I am not going to say everything else I had to say, other than I do want to point out that in May, since I have a compromised voice as a result of the surgery -- it comes and goes -- we did turn the device off in May to see if I would get some relief from the hoarseness for my voice; and I became depressed again.

I began to have these feelings of impending doom, and I just thought to myself, ah, you know, we can't go there again. So I am turned back on, and nobody will turn this device off again unless someone forces me to.

You have to give this opportunity to other people. That's the bottom line, and that's why I am here today. Thank you.

CHAIRPERSON BECKER: Thank you, Ms. Daventort. The final speaker will be Karmen McGuffee, who is also a patient in the study.

MS. McGUFFEE: While they are cuing the video, I would like to state that Cyberonics has accommodated me, paid for my accommodations and paid for my flight. However, I did start a new job last week, and they were not pleased that I would be gone for a couple of days. So it has not benefitted me financially at all. The video.

(A short video was shown.)

MS. McGUFFEE: That was me a little over five years ago. For me, it is very hard to define my first episode with depression. From the time I was three years old, I was a worrier. I worried about things that a toddler doesn't need to worry about: Would my sister die during the night?

In kindergarten I drove my teachers crazy. When my parents wanted to take a short trip, I insisted that they send books along with me so that I wouldn't fall behind. When I was nine, I couldn't sleep. I had nightmares constantly. I would go for days without sleeping. My mother says I would just lie in bed and cry and twist the sheets in my hands until one, two, 3:00 a.m. in the morning.

My mother tried everything that year, from traditional medicine to holistic medicine. We made a lot of adjustments in the family. It was not easy on them. All the doctors said there was nothing wrong with me except my blood sugar.

During my teen years, my family and friends thought that they were dealing with an overweight hypoglycemic, and those were my two problems. But then by the time I was 18, I graduated school with honors from high school. I had a good job as a computer graphic artist and a quality control coordinator.

I had seen a psychotherapist, and a psychiatrist had prescribed Prozac. Six months later, I decided I was cured, and I stopped. Shortly thereafter, I came home from work quite late one night, and my mother found me in my room in the fetal position on the floor, and I had scratched myself on my legs to the point of bleeding.

After a trip to the emergency room, I was admitted to the psych unit. I stayed there for six weeks. By the time I left, I was max'ed out on the dosage of Prozac that was safe for my weight, but I also had to take a booster, Tofranil.

Over the next few years I took Parnate, an MAOI, Zoloft, Effexor, Paxil, Xanax, Halcion and probably a few that I don't remember. There's a lot I don't remember. Drugs would work for six months, if I was extremely fortunate, maybe a year, and then they would have to spin that roulette wheel and come up with a new combination.

I was hospitalized several times. It was like being in a very dark, downward tunnel. I was sliding down, and I could not get out. As I mentioned, the impact on my family and friends was very high. Friendships were always very strained. I never had more than one friend at a time, and I always chased them off. I have difficulty accomplishing the smallest tasks every day.

I required constant reassurance from everyone around me. My family said they felt as though they were walking on egg shells. There was no telling what they would say that would cause me to just crumple. Sometimes I could not be left by myself for any amount of time.

When I started dating my now husband, Jason, we addressed my depression, and he told my parents, well, I've been around Karmen when she is sick; I know what I am in for. And he admitted later that he had no idea.

Sometimes my husband's only goal in the morning was how do I get Karmen up? How do I get her out the door, because if I could pick out my own clothes, I was doing good.

At work I could still function, but my performance was extremely unpredictable. My absenteeism was high, and I had a caustic temper. It caused me to be fired twice.

Just prior to VNS implantation, I was taking five medications daily, seeing a psychiatrist weekly. I was in weekly group therapy, and ECT was next on the list. From the drugs, I had a dry mouth constantly. The MAOI -- I once had a drug interaction that sent me to the emergency room. I have had memory loss, and I experienced massive weight gain.

Every time I would relapse, it was very dangerous and very scary. Each pit was worse than the last, and I always feared they would run out of drugs to try on me. Every suicidal thought that I ever had was when the drugs quit working and I was back in that pit.

I continued in anxious moments to scratch myself to bleeding. I have many scars on my body from that. In psychotherapy -- I mainly found that very frustrating. Everyone there had a reason for their depression, and I had a wonderful family, a healthy family, a good support system.

I used to tell my family, I wish that something bad had happened to me -- I had been kidnapped and abused. At least then I would have a reason to feel so bad.

So about two weeks after that video I had the surgery. I was home that same afternoon, and I returned to work the same week. My husband and mother first noticed an improvement within three to six weeks. At first, they thought they were being overly optimistic.

My mother said she felt like she was not looking into the eyes of a dead person anymore. She has told the doctors, I don't care what scale you use to measure depression; I can tell by looking in Karmen's eyes. My husband said that the dark funeral veil over my eyes was lifted, and he could see my eyes, and they twinkled.

Today I feel great. The only regular side effect I experience is the hoarseness in my voice, which I don't even notice. Whereas, I was taking five medications a day for two and a half or three years, I was taking only Wellbutrin. I was laid off of work about ten months ago, and they added Lexapro, because I wasn't dealing with that very well. It was not on my terms. But prior to my unemployment, I was even talking to my doctor about coming off of Wellbutrin.

I do not believe that VNS therapy has cured me, but it has helped in ways that words cannot express. I constantly worry about will the depression return.

I had a baby 17 months ago, and I was told at my first pregnancy appointment that I was at very high risk for postpartum depression. I had about 11 days of it, more like the blues, and today I enjoy my daughter thoroughly. I've gotten to see her first steps, and I remember them.

Thanks to VNS, I have clarity of mind. I can read books again, whereas I hadn't in years. I don't sleep away my weekends. I was able to research gastric bypass surgery, and had that a little over three years ago. I have lost 226 pounds. I have a joyful and peaceful life, and my family does, too, now.

In closing, people ask me why would you cut yourself open and have something foreign put in your neck and in your chest? My response to them is always, I had nothing to lose.

Please approve this therapy for treatment resistant depression, because it will give both patients and their families something that they have probably lost, and that is hope.

CHAIRPERSON BECKER: Thank you, Ms. McGuffee.

At this point I would like to note for the record that three patients and family members have written the FDA requesting that the agency approve the Cyberonics VNS system, and a patient also wrote to the agency asking that it not approve the VNS system.

Is anyone else here who would like to speak to the Panel now? If so, raise your hand, and come forward to the microphone. I would just like you to state your name and affiliation when you come forward, and whether or not you have any financial interest in Cyberonics.

MS. BARRETT: My name is Mary Barrett. I've been back there trying to write this in a hurry. I am in the D-02 study. I have no financial interest in Cyberonics or its competitors, and no one paid for my trip here today.

I originally had not planned on speaking, but I felt like, because I do have the device and it is important to me that you consider it for approval, I just wanted you to hear my story.

I have been treated for major depressive disorder for over 20 years. I have tried almost every antidepressant drug on the market and combination of meds. Only one medication helped relieve my depression for about four years until I developed an intolerable side effect and was no longer able to take the drug.

Other medications were of no help, and again caused intolerable side effects. I was implanted in February of 2001. It wasn't until the device parameters were turned up to a higher level that I felt consistently better for the first time since I was forced to stop taking the medication that worked.

I volunteered for the study, because it was a last resort. I can only reiterate what previous speakers have said about the pain of depression. It permeates every cell of your brain. It affects every aspect of your life, and it affects the people around you.

This past spring I have had surgery for breast cancer, radiation treatment, and a major automobile accident, things that would cause people without depression to become depressed. I know, had it not been for the VNS, I would have never been able to get through these life traumas.

The VNS has helped me and others I know that have the device get their life back, and I only ask that this device be made available as a choice for others with this horrific illness. Thank you.

CHAIRPERSON BECKER: Thank you. is there anybody else who would like to address the Panel now?

If not, I think we will proceed to the sponsor's presentation on their vagus nerve stimulation therapy system.

This system is indicated for the adjunctive, long term treatment of chronic or recurrent depression for patients who are experiencing a major depressive episode that has not had an adequate response to two or more antidepressant treatments.

I would like to remind public observers at this meeting that, while the meeting is open for public observation, public attendees may not participate except at the specific request of the Panel.

We will begin with the sponsor's presentations. The first Cyberonics presenter is Mr. Alan Totah, Vice President of Regulatory Affairs. He will then introduce the other Cyberonics presenters. Mr. Totah.

MR. TOTAH: Good morning. On behalf of Cyberonics and people in the United States living with treatment resistant depression, we thank you for meeting with us today to review the proposed depression indication for VNS therapy.

My name is Alan Totah, and I am the Cyberonics Vice President of Regulatory Affairs and Quality. I will begin today's sponsor presentation with a brief overview of the agenda, today's available presenters, VNS therapy system and regulatory history.

Dr. John Rush, Professor and Betty Jo Hay Distinguished Chair, Department of Psychiatry, UT Southwestern Medical Center, D-01 study investigator and D-02 principal investigator, will then summarize depression, treatment resistant depression or TRD, and the unmet need for an FDA approved effective and tolerable long term treatment for TRD.

Following Dr. Rush, Dr. Richard Rudolph, Cyberonics Vice President of Clinical and Medical Affairs, who prior to joining Cyberonics played a key role over a 16-year period in the development of the Effexor family of antidepressants at Wyeth, will then present study design and analysis plan, effectiveness and safety analysis, and the risks and benefits of VNS therapy. Dr. Rush will then provide closing remarks.

In addition to Dr. Rush, six other outside experts are with us today, who will be available for Q&A, representing psychiatry, biostatistics, and mechanism of action.

They are Doctors Harold Sackheim and Philip Ninan who are psychiatric investigators representing D-01, D-02, D-04 and D-05 depression studies; Doctors Phil Lavori and Sonia Davis who are providing expertise in biostatistics; and Doctors Tom Henry and Mark George, who are providing mechanism of action expertise and neuroimaging, and specifically VNS therapy PET and fMRI imaging.

In addition to the outside experts, we have a number of Cyberonics medical directors and directors here representing other disciplines to answer your questions.

We are here today to present to you the data that supports the safety and effectiveness of the VNS therapy system as an adjunctive, long term treatment of chronic or recurrent depression for patients over the age of 18 who are experiencing a major depressive episode that has not had an adequate response to two or more adequate antidepressant treatments, with specific definitions of chronic and recurrent depression and failed adequate treatment.

There are very few devices or drugs that are indicated specifically as adjunctive, long term treatments, and there is no FDA approved safe and effective long term treatment specifically for this level of chronic or recurrent treatment resistant depression.

The VNS therapy system, the programming parameters, and the implant technique used in depression are the same as those approved and used in epilepsy. The generator is implanted just like a simple bradycardia pulse generator, and a bipolar lead is simply tunneled under the skin from the left vagus nerve where the lead electrodes are wrapped around the nerve and then down to the generator.

I am pointing out to you the generator. This is the typical implant site. We have a picture of the lead going up, and you can see this exploded view of the electrodes which are wrapped around the left vagus nerve, and we have the other elements or components of our system illustrated for you.

Typically, the device is programmed on for 30 seconds and off for five minutes on a 24/7 schedule. The typical outpatient surgery often lasts approximately one hour and is a low risk implant procedure, and surgical complications are minimal.

A magnet can be used by the patient to temporarily control side effects such as voice alteration during public speaking or singing, if necessary.

From an historical perspective, FDA's Neurological Devices Panel unanimously recommended epilepsy approval in June of 1997, and the VNS therapy system was approved by FDA on July 16, 1997. The epilepsy safety number shown on this slide are at the time of the application.

Today, over 29,000 epilepsy patients have been treated with the VNS therapy system, and we have accumulated a total of 72,000 patient years of experience.

Depression studies were started, because clinical observations from epilepsy use and findings from epilepsy, preclinical and human neuroimaging mechanism of action studies suggested that VNS had a potential antidepressant effect. Depression studies began in 1998 following IDE approval of a D-01 pilot study protocol.

Several significant regulatory historical dates that followed are: In July 1999 when FDA granted expedited review status; European CE Mark and Canadian commercial approvals were granted for the depression indication in March and April of the year 2001, based upon the D-01 results; and in January of 2002 the D-02 acute 12-week study results were unblinded and analyzed.

The primary endpoint did not reach statistical significance. However, the results did show a positive trend in favor of VNS, and a key secondary endpoint was statistically significant.

After consideration of the acute results, the proposed indication for use and the existing D-02 and D-04 protocols, a revised D-02 long term and D-02 versus D-04 standard of care prospective analysis plan was submitted to the FDA in September of 2002. The FDA notified Cyberonics that no manufacturing site inspection would be required, due to Cyberonics good compliance history.

The PMA Supplement was submitted and accepted for filing by the FDA on October 27, 2003. Since then Cyberonics has completely responded to FDA's deficiency letter regarding this PMA application.

That brings us to today's Panel meeting that is occurring just two weeks shy of the seventh anniversary of the original epilepsy panel in June of 1997. During today's meeting Cyberonics is prepared to address FDA's Panel questions and any questions the Panel members may have.

These are the studies that comprise the six-year depression program. Dr. Rudolph will present details in his presentation. There was one important revision during the program that you will hear about today.

For clarity, let me describe the reason for the revision and what was changed. When the acute study endpoint did not reach statistical significance despite a favorable trend, and significance on secondary endpoints, we decided to provide a more definitive evaluation of long term effectiveness by adding an active control for the D-02 outcomes.

The sole change indicated on this slide by a double checkmark added a one-year comparison of D-02 patients treated with adjunctive VNS plus standard of care treatment with D-04 patients treated only with standard of care treatment.

The existing D-04 protocol, which had been previously intended as a comparison with D-02, was then formally added into the statistical plan as a long term active control. This revised plan provided the FDA with comprehensive one-year clinical data and analysis on 460 patients with treatment resistant depression.

Cyberonics and its team of outside clinical, statistical and regulatory experts deemed the revised analysis plan the most appropriate for the determination of safety and effectiveness for the proposed indication for use, after careful consideration of the urgent unmet need for a long term treatment for TRD, which you certainly heard about from our patients today.

They consist of the following: First, the nonsignificant yet encouraging results from the acute, sham control phase of D-02; second, the increasing response rates seen over time in D-01 patients; thirdly, the adjunctive, meaning VNS would be added onto currently available standard of care treatments, long term proposed indication; fourth, the majority of PMA device and neurological device approval precedents did not include randomized controlled trials consistent with 21 CFR 860.7 of the regulations; fifth, D-04 study which started in the year 2000 and was originally designed to be compared with D-02 patients provides a valid, prospective, active standard of care control consistent with the proposed indication; and finally, the infeasibility and limited value of alternative long term study designs in treatment resistant depression patients relative to D-02 versus D-04 comparative analysis.

In conclusion, allow me to once again thank you for your time today, and mention that we are all here primarily because of the significant unmet need for an FDA approved informed use, safe and effective long term treatment for treatment resistant depression.

FDA initially recognized this need in 1999 when expedited review status was granted. Four and a half years later, FDA reconfirmed the continuing unmet need in their December 2003 PMAS filing letter. Please see the noted quote on the slide.

I now invite Dr. Rush to help us better understand treatment resistant depression and the significant unmet need for an effective long term treatment for TRD. Thank you.

DR. RUSH: Thank you very much, and good morning. I am John Rush. I am a full time employee of UT Southwestern, Dallas. I have provided consultation to Cyberonics and received fees for that consultation.

I am going to very briefly review for you probably much of what the Panel is quite familiar with, and certainly what you heard this morning already from the several patients. That is the importance of treatment resistant depression, its impact from a public health significance and from a personal significance impact, and provide you some sense of the kinds of patients that we are talking about that entered into the VNS studies.

So to recap what I think is quite familiar to most of you, a very common syndrome affecting 16 percent of the individuals in the U.S. in their lifetime. This is not TRD. This is major depressive disorder. Two-thirds are female, 9.5 million treated annually and, as mentioned previously, substantially disabling, second most disabling condition in the U.S., fourth worldwide currently, the most disabling condition for women in the United States presently.

It is associated with a marked increase in mortality due to suicide. Thirty thousand suicides per year have been mentioned. Eighty percent of those are attributable to depression, and increased mortality as well due to worsening of the outcome of a number of general medical conditions; for example, cardiovascular disease, but others as well clearly investigated and studied.

It is well known that depressed patients, for obvious reasons, are high utilizers of not just mental health services but general health services.

Very briefly, these slides are, I am sure, presenting you with what you know. So I'll just go very quickly over them. We 30 years ago had the stigma of depression. We didn't even recognize it much as an illness. They were seen as troubled individuals. Obviously, that is not the case. These are individuals suffering from a syndrome defined by a variety of biological abnormalities.

Thirty years ago we thought of these depressions as situational adjustment reactions, basically brief, time limited, and of modest impact on individuals' lives. When I was at the University of Pennsylvania, I was taught to treat these individuals with medication for four to six months to facilitate psychotherapy, and the need for long term medication was not recognized and was not part of training.

We now know that 60 percent of people in the first depressive episode will go on to have either a recurrent, subsequent episodes, or chronic course.

The next level of stigma is that we have now accepted it as an illness, but I think in many people's minds it is a very benign illness. In fact, it shortens life span due to the causes I previously mentioned, suicide, increased mortality from general medical conditions. It is massively disabling. I reviewed the data with you.

The third level of stigma: So now we recognize an illness. We are beginning to recognize just how profoundly severe and disabling this illness is, and costly on a human suffering basis as well as an economic basis. But the third myth that we deal with is, well, the treatments we have are really pretty good.

The fact of the matter is the treatments we have are good, but they are not good enough, and some of that is described here. So our current medications are effective. Fifty percent of symptomatic volunteers, individuals who have uncomplicated non-treatment resistant depression -- those are the individuals that typically enter randomized efficacy trials for regulatory purposes, by the way -- do respond to the first medication, and within that group of responders the substantial majority achieve a remission, virtual absence of symptoms. But that only gives us 35 to 40 percent of the uncomplicated, nonchronic, non-treatment resistant patients achieving remission, the goal of treatment with the first drug.

What about the second or the third drug? We had some discussion about that earlier. This is being evaluated, but at the moment the estimates are that somewhere between 15 and 20 percent of patients will not achieve remission with two or even three medications.

The other element in treatment is sustaining that benefit, if achieved, and I will describe that and discuss that in just a minute.

I will show you some of the data that indicates that, even in non-treatment resistant depressed patients, a substantial proportion having achieved a benefit in acute treatment and continuation phase -- so three months plus four more months, seven months -- do in fact suffer a return of the episode.

Treatment resistant depression: The field has begun to coalesce around an accepted definition. This is obviously on a continuum. There have been various staging systems to define treatment resistant depression, but studies have been launched that accept this definition as certainly a reasonable one. I think a consensus of experts would agree.

It is the lack of an adequate clinical response after at least two well delivered treatments. Looking at symptomatic volunteers from the efficacy trials that we have abundantly, as I mentioned, 50 percent respond. About 35 to 40 remit, no symptoms after the first trial.

It is also known that, if you go to the second treatment -- these are largely open trials, but there are quite a number of them -- about 20 to 25 percent of the original sample respond to the second treatment, having not responded to the first. So that gives us roughly 75 percent of the original sample.

That leaves us with about 20 to 25 percent of individuals who will not achieve the goal of response, which is short of remission, after two treatments.

So what do we know about TRD? It has actually become a focus of research over the last several years. It is quite clear that treatment resistant depression is clearly associated with worse function, worse prognosis, higher health care costs, health care utilization, increased risk of complications, including general medical problems and substance abuse, as we have already heard on an individual basis and has been shown in studies, high risk of family burden, high risk of suicide, and as you know, 8 to 15 percent of previously hospitalized depressed patients do go on to commit suicide, and the worsened mortality we talked about in terms of general medical conditions.

Importantly, treatment resistant depression has a very low response and very high relapse rate, and I will show you a little bit of data to justify that statement in a minute.

I really don't have to spend very much time on this issue, because you have heard the clinical picture of treatment resistant depression from the patients. But these individuals are exactly as described, tearful, suicidal, hopeless, desperate, hanging on by their knuckles, their fingernails, frequent users of hospitalization, emergency room, seeing psychiatrists frequently, often failing to be fully employed or even being unemployed, and a remarkable percentage on ongoing full time disability.

I would point out that in the Texas Public Health System, one-third to 40 percent of the individuals served by the Texas public sector have depression. They outnumber the individuals that are still substantial in number who have schizophrenia. This is a very serious condition.

These individuals depend heavily, as you heard, on families and others, and it really is a different kind of depression. This is not the kind of depression that enters typical efficacy trials, and I will show you a little more data.

These people have long standing disabling illness, rarely achieve sustained remission even spontaneously, have very modest responses to medication, but they are grateful to have even that. They are much more akin to congestive heart failure, chronic renal or lung disease, the kinds of chronic disabling general medical conditions, by the way, for which patients do not take their lives. Eighty percent of suicides are due to depression. This condition is so bad that people kill themselves because of it.

What about utilization? Just very briefly, one slide. This is a study we recently completed. This is the number of different medications, changes that the individuals went through, and this is an estimate of, in this case, cost but, obviously, then frequency of utilization of inpatient/outpatient, pharmaceutical and total health care utilization costs.

The point is the greater the degree of treatment resistance, the greater the use of all of these services, in, out and pharmaceutical.

Clinical management of TRD at the moment: Basically, we do not have an FDA approved treatment. Multiple medication is very common, as you heard. Which treatments, however, are best or what combinations are best is really not known. Is there a preferred series of treatment steps? What treatment to give first, second, third, and when to go to combinations -- that is really not known.

In fact, I am the principal investigator on an NINH sponsored trial called the Sequence Treatment Alternative To Relieve Depression Trial. We have just completed enrollment with over 4,000 patients, and it is a nested, multiple randomized controlled trial effort to see what to do, what is best, if the patient does not respond to the first treatment.

So they are randomized to four different switches or three different augments in the second stage, and the third stage there is again randomizations to different switching and augmenting treatments.

So we hope to have, really for the first time, randomized controlled evidence for what to do after the first treatment doesn't work, and certainly after the second treatment doesn't work.

At the moment -- and I would point out that this trial was launched in October of 1999. It is going to cost $35 million, and I think it represents the importance of TRD now in terms of the public health agenda. This has come on the radar screen.

In 1990, if you talked about this condition, people would deny it existed. But every clinician knew about it, because those are the patients we are treating.

ECT is our best treatment right now for treatment resistant depression. It does work very nicely acutely, but it does not -- it can't be used in a sustained, long term maintenance basis for easily for most patients, because of some cognitive side effects. And when you stop the treatment, as I will show you, the outcomes with ECT are not good, the treatment being discontinued.

The management of TRD involves side effects and adherence difficulties due to multiple medications, and it is known that the greater level of resistance is associated with lower response and higher relapse rates.

The reality of treatment of TRD today is keeping the patient alive, and basically you heard that from the patients. Let me give you one example. There is a case of a young man, 29-year-old graduate student, been in graduate school since age 21. He could not take the full course load. He was taking two courses, seeing a psychiatrist two to three times a week. He had been depressed for 10 years, in an episode for 10 years.

He had been on multiple medications. He had finally settled on a combination of Resperidone and Prosac 80 milligrams. That was the best that he could do. He was having panic attacks, but he was able to stay outside the hospital. He was preoccupied regularly with suicide and suicidal ideation, basically living by himself in the dorm room, totally dependent on his family, clearly ashamed of his life and what he had not become, given his peers with whom he had graduated from college, and this was one of the first patients that we actually entered into the VNS study.

That is very typical, as we have heard from the other patients, of treatment resistant depression.

What about long term outcome with what we have now? Well, here is the results with medication. pay attention to this column. This is from John Greden's recent publication. He looked at long term recurrence rates, comparing placebo and medication.

Indeed, the medication does provide a benefit, but look at the recurrence rate, and this is non-treatment resistant depression. The recurrence rate over a year is up to, in some studies, 50 percent, obviously depends on the population.

All of these individuals had responded acutely and stayed well for four more months of continuation treatment. So in non-TRD long term outcome is not as we hope.

This is another slide of the same question, a different population. This is a population that we recently treated in the Texas Medication Algorithm Project in the public sector. They were treated for a year under algorithm based conditions. So we used an algorithmic sequence with augmented resources.

The algorithm based treatment did very well as compared to treatment as usual. So this is the best outcome, and this is a measure of depressive symptoms analogous to the Hamilton, and what you see is the sustained response rates. That is, response at nine and 12 months out, 14 percent sustained remission, five percent -- and it doesn't matter, really, if it is observed case or LOCF. These are very, very remarkably low figures, much lower than you expect from, of course, RCTs with non-TRD patients.

A couple more slides on long term outcome, and then I will turn over the podium to Dr. Rudolph. This is work from Dr. Sackheim's group. These individuals that we are showing you here had a successful acute phase response to ECT, our most effective treatment for treatment resistant depression at the moment.

He then randomized patients to placebo, nortriptyline plus placebo, or the combination of Lithium and nortriptyline, followed the patients over subsequent six months. As you can see, the relapse rates in patients who had done quite well with ECT and were given the best treatment, still 40 percent were relapsed within six months, and this is under research conditions.

Eighty-four percent of those individuals relapsed with placebo. Not all these people were treatment resistant. Of course, many were, because they received ECT. If you look at the effect of treatment resistance at baseline on the long term outcome following successful ECT, you see this here.

Again, work from Dr. Sackheim's group, patients had done well with ECT, and they were designated at the beginning, blind to this outcome, whether or not they had been medication resistant -- this is one or more medication failures in the current episode -- or had not been so exposed.

You notice, in the people that had medication resistance, two-thirds of these individuals actually relapsed over the subsequent year, after successful treatment with ECT. And this is still an ongoing treatment. This is not in placebo. So these individuals have a very difficult long term course.

Finally, just to provide you with a sense of who the patients are that you are going to hear about. Let me just talk a little bit from a clinical perspective.

What this does is this is a comparison of individuals represented in the community ECT sample developed by Dr. Sackheim from multiple New York metropolitan area hospitals. These are just individuals that are in the community, are receiving ECT.

We then compare them to the individuals, all of them that enter the D-02 and D-04 studies. This is the number of adequately delivered treatments which were scorable by the antidepressant treatment history form, Dr. Sackheim's scale.

The first thing that you see is that level of treatment resistance in the VNS patients, noted in red, is far higher. Almost half, in fact, are not even included within patients who receive ECT in the community. Put the other way, half the patients receiving ECT in the community do not have the level of resistance that we are talking about with patients in the D-02 trial.

That makes some sense. Fifty percent -- Over 50 percent of these patients in the trial had already had ECT in their lifetime, and over a third in the current episode.

The number of hospitalizations in the VNS group nearly twice that for the ECT community group. So we are dealing in these studies with a very, very, very difficult, hard to treat, at the end of the line almost, depressed patient population.

I can tell you, I have done trials for 30 years. I have never ever come close to putting this level of difficult patient, difficult disease in any trial. These patients would not even come ever close to getting into a pharmaceutical trial, because the pharmaceutical trials typically exclude people that have failed on more than one treatment in the current episode.

The other comment is about what is the meaning of these numbers. I just want to help convert. This is a research definition that is extremely conservative. That is, Dr. Sackheim's scale only rates drugs that have been demonstrated to be effective in randomized controlled trials.

Many things that we do clinically with the treatment resistant depressed patient have not been subjected to randomized controlled trials. An example is a typical anti-psychotic augmentation has been subjected to one trial that has been published, and it was widely used practice.

We think it is effective. That would not count in Dr. Sackheim's ratings. To give you a sense, if in his scale there are two to three ATHF failed trials, what is the actual number of clinical trials these patients failed on? Twelve, twelve clinical trials, and I am not talking about all the combinations used, 12 medicines.

When you get to four to five, the number of clinical trials is 16. When it is six or greater, the number of clinical trials is 20. These are extremely resistant patients who are really at the end of the line.

So let me just briefly summarize. I am sure you are convinced that TRD is highly disabling. It affects a large number of people, 20 percent of people with major depression. It is a clear unmet need.

These patients have a high suicide risk. They have very low response rates, high relapse and recurrence rates with our current treatment, high utilization of health care services. They are really analogous to any of the very severe psychiatric or chronic general medicine conditions.

We have no FDA approved treatment that is effective, safe in the long run for TRD. ECT is excellent, but difficult for the reasons I have outlined previously. Multiple medications are used in combinations and in sequences for which there is virtually no evidence. Side effects and adherence, especially with multiple medications, is a huge problem, and we clearly need a treatment for these desperate but important and substantial in number depressed patients. Thank you.

Let me turn the podium over to Dr. Richard Rudolph, Vice President, who will go through the D-01 to 6 studies.

DR. RUDOLPH: Thank you, Dr. Rush. I want to start by thanking the Panel Chair, Panel secretary, Panel members and the FDA for this opportunity to provide an overview of the clinical data supporting safety and effectiveness for VNS for the indication that you are considering today, treatment resistant depression.

This is an outline of the presentation that I am going to give this morning. I am going to provide some general background, and then spend some time going over design and analysis considerations to help you better understand the effectiveness data that I will then present. Then finally, I will move on to safety data and a short set of conclusions.

Well, why did Cyberonics become interested, in the first place, in developing the VNS therapy for depression? There were a number of initial considerations that led us to believe VNS therapy might be useful for this indication.

Those consisted of anecdotal reports of mood improvement in patients in our epilepsy trials where the improvement in mood seemed to be out of proportion to the improvement in the seizure counts that the patients were experiencing.

Also, knowledge that the use of anti-convulsants have been used in the past and currently are used as mood stabilizers and augmenting agents in the treatment of depression, and the observation that electroconvulsive therapy has both antidepressant and anti-convulsant actions.

Subsequently there were some additional considerations that provided a biological rationale for the use of VNS for this indication. These included: A more formal analysis of mood changes in the epilepsy studies, which confirmed the anecdotal reports; a variety of neuroimaging data, some of which I will show you shortly; effects on neurotransmitters which showed that VNS does have effects on norepinephrine and serotonin, the normal transmitters most closely implicated in the action of antidepressant drugs; and most recently, activity in an animal antidepressant model called the 4-SWIM test in which VNS had similar effects to desipramine, a standard antidepressant drug, and clearly distinguishable from placebo.

As I indicated, we have done a variety of neuroimaging studies, and our findings from the neuroimaging studies have shown us that VNS affects a widespread array of autonomic, reticular, and limbic structures within the brain.

The immediate effects of VNS on the central nervous system implicate brain areas known to be primary and secondary vagal projections. So just what you would assume that happens.

Longer term effects of VNS on the central nervous system implicate limbic and paralimbic brain circuits associated with depression and mood regulation. An example of that is shown on this slide. These are PET images three months after the initiation of vagus nerve stimulation.

What one finds is effects, modulation in areas such as the orbitofrontal cortex, L. insula, and the Mid-Cingulate Gyrus. These are areas that are implicated in the regulation of mood.

Our six-year development program for VNS for the TRD indication consists of the studies on this slide. The most important studies, the ones I will be spending the most time on in my presentation, are the ones above the yellow bar.

Those are the D-01 study which was an open-label feasibility study, the D-02 study which had two parts, an acute phase which was a double blind randomized control of sham stimulation and active VNS therapy, and the second phrase, a long term phase in which the sham treated patients crossed over into active treatment. So all patients continued out to one year and beyond in an open label fashion.

Then the D-04 study, which is a prospective observational, which is a prospective observational study of treatment resistant depression patients treated with standard of care therapies, which we used as a control for the long term D-02 outcomes.

Other studies that were in our submission that I won't be speaking too much about were the C-03 study which is actually still enrolling -- it is a European open label study; the D-05 study which was per se not a study but a video tape assessment used to ascertain the inter-rate of reliability of those that were performing the ratings for the D-02 study; and then the D-06 study, which is a study, a pilot study in a very different population. This is an open label feasibility study in rapid cycling bipolar disorder.

Next I would like to go through some design and analysis considerations to help facilitate your understanding of the effectiveness data that I will be presenting subsequently.

There are several lines of evidence to support the effectiveness of adjunctive VNS therapy for TRD indication. The most important evidence comes from a comparison of the 12-month outcomes in the D-02 patients, and again these are patients in the long term that are receiving adjunctive VNS therapy, in comparison with the 12-month outcomes from patients in a separate study, D-04, which were enrolled as similar patients to receive only standard of care therapy.

Other evidence comes from a comparison of the D-02 acute treatment outcomes that compared VNS with sham control, and then longer term outcomes, both from the D-02 patients and form the D-01 patients in the feasibility study, particularly with a focus on the durability of their response.

This slide gives a schematic of the D-02 study design. Patients in the study were qualified during an initial 45-day period, and those that met protocol entry criteria were then implanted and randomized.

Following that, there was a two-week period during which stimulation was not turned on for any patients. It was a period for recovery from the surgery. At that point, the group that was randomized to the active VNS therapy group had their stimulators turned on, and for two weeks underwent a period of stimulation adjustment.

Whatever parameters were obtained and optimized at that period were continued for an additional eight weeks of therapy.

Meanwhile, the sham stimulation group underwent all the same procedures, but never had the output current turned on for their stimulators. So they served as the control.

Following the end of that period, the patients in the sham treatment group had the opportunity to cross over to active therapy and underwent the same procedures, and then both groups of patients continued into a long term open label phase.

During the acute phase, medications had to be fixed. So whatever medications the patients came into the study on had to remain the same. During the long term study phase, medications could be added or increased at the discretion of the investigators.

Here we have some more details on the study design. I have already covered the top part of the slide. There were 235 patients that were implanted in the study at 21 different study sites. The main inclusion criteria to be enrolled in the study were that you could be a male or female between the ages of 18 and 80 years of age. You had to have a current diagnosis of being in a major depressive episode with a background history of having chronic or recurrent depression.

Most importantly perhaps, you had to have failed at least two adequate treatment trials in the current major depressive episode as measured by standardized scale that Dr. Rush referred to before, the antidepressant treatment history form; and patients had to have a minimum score of 20 on their baseline Hamilton reading.

Here are some similar details for the D-04 study design. D-04 was a 24-month prospective observational study in which patients received standard of care treatment but no VNS. So if you think about it for a moment, essentially what we have in the D-02 study is a group of patients receiving VNS long term plus various medications at the discretion of the physicians taking care of the patients, and in the D-04 study we have a similar group of patients, but they are receiving medications only, no vagus nerve stimulation.

The D-04 study enrolled 127 patients at 13 total study sites, 12 of which were overlapping sites with the D-02 study. The main inclusion for the D-04 study were identical to the inclusion criteria for the D-02 study.

So why do we think this nonrandomized D-04 serves as an appropriate control for the D-02, a long term study? Well, for several reasons. First of all, it was a prospectively designed study for comparison with D-02 outcomes. It just wasn't randomized with D-02.

It does represent a clinically relevant control in that it is an active treatment control that corresponds to the proposed indication for VNS, which is adjunctive long term VNS therapy.

The study was done primarily at overlapping sites, as I have already indicated, and it did use the same principal enrollment criteria. Moreover, the study was conducted over a similar time period, which is important, because it helped ensure that patients would have access to the same types of treatments in terms of their standard of care therapy.

Finally, the D-04 represents a large sample size which, of course, facilitates statistical comparisons.

Now for the benefit Panel members that may not be that familiar with research methodology in depression trials, I have included some slides to provide some additional background. The first slide tells you how we measure effectiveness outcomes in depression studies.

In depression studies, effectiveness is measured by standardized validated rating scales. These may be either multi-dimensional scales -- that is, scales that cover different aspects of the depressive syndrome, and examples of this would be the Hamilton rating scale for depression, the inventory of depressive symptomatology self-report, and the Montgomery Asberg Depressing Rating Scale -- or the rating scales may be a Licher type scale like the Clinical Global Impression Scale.

The scales may be either clinician or patient rated. For the multi-dimensional scales, the way the scales are analyzed is to total all the individual items, obtain a total score, and then analyze the total score. Higher scores on these scales indicate a patient who is more severely depressed. So as you improve, your scores go down on the scales.

Here are some examples. I'll just provide some further detail on the Hamilton rating scale for depression. On the lefthand side of the slide, you see the various domains that are assessed by the scale: Mood, feelings of guilt, suicide, sleep, work, activities, psychomotor retardation and agitation, anxiety, somatic symptoms and weight loss.

A sample item from the scale is up here, and a clinical interpretation, and this is only a rough clinical guideline. This is not standardized through research, but here is a rough clinical interpretation of how you interpret the total scores and equate it to how severely ill the patient is.

Here is a similar representation for the Inventory of Depressive Symptomatology Self-report. It assesses some of the same symptoms and some symptoms not assessed by the Hamilton scale. Here is a sample item, and here is the clinical interpretation of the total scores.

One thing to note as I start to show the effectiveness results from the VNS studies is you will find that the baseline scores for the patients entered in our trial fall -- as a mean fall into the severe range on both scales.

Broadly speaking, the types of analyses that you will be seeing fall into two categories, either continuous measures or categorical outcomes analyses. The continuous outcomes analyses measure -- Probably the most prominent we used was a repeated measures linea regression. These continuous measures generally measure a change from baseline.

The categorical outcomes measure discrete categories of outcome. Commonly, these include response. Response is generally defined in the field as a 50 percent or greater improvement on the multi-dimensional scales or on the CGI, the Licher type scale, response is defined as a one or two, which corresponds to a clinician rating of Very Much or Much Improved from baseline.

We also use categorical outcomes of complete response or sometimes called remission. This equates with a patient who is well or almost completely well, and those are defined by absolute cutoff scores on the scales.

Finally, because we were concerned with this particular population that these standard definitions might underestimate the true benefit for the patients, we also included a categorization of clinical benefit derived from the literature, which categorizes different levels of improvement from the Hamilton scale. I will be presenting data from this particular categorical outcome mostly in the form of looking at the durability of response for patients in the D-01 and D-02 studies.

Now, of course, with the multiplicity of scales, it is important to identify one single primary outcome, and this slide shows you the primary analyses that were prespecified in our various statistical plans for each of the important studies I will be talking about today.

In the D-02 acute study the primary analyses were response rates after 12 weeks of therapy determined from the Hamilton rating scale. So that is the 50 percent or greater improvement.

For the D-02 long term study the primary analysis was the repeated measures linear regression analysis of the Hamilton scores over 12 months, estimating the change over time.

For the D-02 versus D-04 comparison, the primary analysis was a repeated measures linear regression analysis of the IDS scores over 12 months, estimating the monthly difference between the D-02 and D-04 patients, in other words a linear study effect.

You may be wondering why we had this transition through different scales and different types of analysis. So let me explain that up front.

When we had the opportunity to revise the statistical plan, which was necessitated by the finding in the acute study that there were trends and some positive findings on secondary outcomes, but the primary outcome failed to reach statistical significance, we then moved to a repeated measures rather than a categorical outcome, primarily because in prior communications with the FDA they had expressed some preference for that as an outcome, and also because it is a more sensitive measure for finding differences between treatment groups.

Then when we moved on to the D-02 and D-04 comparison, having committed ourselves to the repeated measures linear regression approach, we were kind of forced into using the IDS as the primary scale, because the D-02 study only had a baseline and a 12-month measurement on the Hamilton, and the repeated measures approach requires multiple observations over time, which were present for the IDS but not for the Hamilton. Therefore, we chose the IDS for those particular set of analyses or at least for the primary analysis.

So with that as background, let me move on to a review of the primary data that supports an effectiveness claim for VNS for the TRD indication. I am going to start with the most important evidence. That comes from a comparison of the D-02 results versus the D-04 results over 12 months of treatment.

Let's start by looking at the flow of the D-02 study participants through the long term phase. You will recall that I said 235 patients were initially implanted in the D-02 acute study; 233 of those patients continued into the long term phase and constitute our long term safety population.

Our statistical plan prespecified that analyses would be done primarily using an evaluable efficacy subset of patients, and that included 205 of those 233 patients. The reason for excluding 28 patients are shown in the middle box here on the slide.

The majority of those patients are: 21 were patients in the sham control group that were excluded, because after the sham period, their Hamilton score was no longer above 18, which was a prespecified criteria.

Now I would like to point out -- I know in the FDA review material that you received, there was a mention that 20 percent of the patients had a placebo effect, and I want to distinguish that from a placebo response, that it was an effect based on patients falling below 18, but that should not be confused with a placebo response which would require the definition that a patient improve 50 percent or more. In fact, only 10 percent of the patients improved to the extent that they could be called a placebo responder.

The other seven patients were excluded from the treatment group, and they were excluded either because they didn't have long term data or because three did not meet acute phase continuation criteria that were also prespecified in the statistical plan.

For the D-04 study there were 127 patients that were enrolled. Three were excluded from the evaluable efficacy analyses for the reasons shown on this slide.

When we analyzed the patients for their baseline characteristics, we found that they were quite comparable. This is just one of several slides. All told, we analyzed about 20 different baseline characteristics. Only three of them were statistically different between the two groups, and those are shown on this slide in yellow highlighting, along with some of the additional 19 characteristics which I thought would be of most interest to the Panel members.

So let's start with the ones that were statistically different. The first one was ethnic distribution. There was a higher percentage of Caucasians in the D-02 group, but this is probably clinically irrelevant, since as you can see in both groups, they were at least 90 percent Caucasian.

The second difference was in the number of lifetime episodes of depression. The D-04 group had a higher proportion of patients in the category of more than 10 lifetime episodes.

Then the third area of difference was in the percentage of patients that had had exposure to ECT, and both in the current episode and lifetime there was a higher percentage of patients with ECT exposure in the D-02 group.

So one take-home message from this slide and the other information I have given you is that these patients are very comparable at baseline. Also a take-home point from this slide is some indication of just how extraordinarily severely ill and treatment resistant these patients are.

For instance, you can see in terms of the average duration of illness over the lifetime and in the current episode, these are very lengthy illnesses. Patients as a mean had been sick for at least 25 years in their lifetime, and at least four years in the current episode. In fact, fully two-thirds of the patients were actually in a chronic major depressive episode, defined as an episode lasting continuously two or more years.

The primary analysis for comparing the D-02 and D-04 outcomes was a repeated measures linear regression of the IDS scores. That is illustrated on this slide. Now for point of clarity, I should say that the actual graph is drawn from actual raw scores and not from the repeated measures model. I did that for the sake of presentation clarity, but the statistical comparison comes from the primary repeated measures model.

What you will note is the D-04 patients shown in the light blue dotted line improved very little during the course of 12 months of treatment with access to every accepted therapy, every legal therapy, and a lot of churning through therapies.

By contrast, the patients in the D-02 group receiving adjunctive VNS, shown in the solid burnt orange color, improved to a greater degree. That improvement increases. That is the difference between not only the absolute improvement but also the difference between the two groups improves as time marches on through the year, and the comparison is highly statistically significant with a p-value of less than 0.001.

We did a series of alternate methodologic approaches to the data to test the robustness of the data. These included doing the primary analysis on the intent to treat population rather than this efficacy evaluable population. So all patients were included in this analysis, all 235 D-02 patients, all 127 D-04 patients, and that analysis retained statistical significance at the less than 0.001 level.

We also did an analysis where we looked at just the overlapping sites, so just the common sites to both D-02 and D-04, patients from those sites. Again, statistical significance was retained, in this case at a level of 0.002.

The results from the primary analysis were confirmed by a variety of secondary analyses. Here we are looking at the secondary analysis that examines the change in Hamilton scores from baseline to 12 months. Remember, we just had a baseline in the 12 month scores available on the Hamilton.

What one observes is that in the D-02 group there is about an eight-point decrease in the Hamilton score over the course of a year. Again, decreases signify improvement, versus about a five-point improvement in the D-04 patients. That result is statistically significant.

On a variety of secondary outcomes looking at response rates and complete response or remission rates, we find the following. Results from the IDS scale are shown here, from the Hamilton scale here. First we look at response, and then we look at complete response.

So response based on the IDS scale was 22 percent for the D-02 group and 12 percent for the D-04 group. Complete response was 15 versus 4 percent. on the Hamilton scale, the response in the treated group with adjunctive VNS was 30 percent. For the D-04 group it was 13 percent. In terms of complete response it was 17 versus 7 percent. All these comparisons are statistically significant.

One more, using the Clinical Global Impressions as a measure of response where a 1 or 2 corresponding to a clinician rating of Much or Very Much Improved equates with response, we found almost a threefold difference between the groups, with 37 percent of the D-02 patients and only 12 percent of the D-04 patients reaching the response criteria, a result that was statistically significant at a robust level.

So in summary for this section of slides, what we found were comparable, highly treatment resistant groups at baseline, a statistically significant result favoring adjunctive VNS therapy on the primary analysis, statistical significance in both evaluable efficacy analyses and ITT analyses, and statistical significance retained in the subset of patients that come only from the overlapping sites enrolling both D-02 and D-04 patients, and we found that statistically and clinically significant differences were confirmed by secondary analyses using multiple outcome measures, the categorical outcomes being a more appropriate way to assess clinical outcome.

Now because our control was a nonrandomized one, we were very concerned about potential sources of bias or other explanations for the outcome other than the VNS was contributing to the better improvement in the D-02 patients. This slide in a picture way tries to give you what we were most concerned about.

We were certainly concerned about the influence of baseline differences on patients, the influence of medications and electroconvulsive therapy, and I am going to deal with those three right now and show you why those are not the explanations for why the D-02 patients are getting better, and then i am going to address the issue of placebo response a little later in my presentation.

So let's start with baseline characteristics. Baseline characteristics do not explain why the D-02 patients are doing better. Why is that? Well, first of all, there were few significant differences between the D-02 and D-04 on baseline characteristics, as I have already demonstrated. However, we did take an additional measure that was prespecified in our statistical plan.

That was to incorporate a propensity adjustment strategy to provide additional insurance that potential bias associated with the imbalance of measured baseline covariates was removed. For nonstatisticians, such as me, let me try to give you a one-slide lesson on what propensity is all about, because it may be a new concept for some of you.

It is a technique that is particularly suitable for adjusting nonrandom treatment assignment. So it is particularly suitable for this comparison with the D-04 group.

In this particular strategy, you calculate a propensity score, and that score represents a conditional probability of assignment to a group, given a set of measured covariates. So it is a way of encompassing a whole large variety of different characteristics in a single score.

The way we used it was to incorporate it in all analyses of effectiveness and, when we did so, we found that the propensity score did not contribute to the primary repeated measures analysis' statistical significance.

Now the limitation of propensity analysis is that it can only address measured covariates or measured characteristics. It cannot address those that are unmeasured. We do not think, however, that unmeasured covariates are likely to account for the differences either, and the reasons for that are that, first of all, all or nearly all of the covariates that have a well established literature behind them were things that we measured and accounted for.

Furthermore, we think that, to the extent unmeasured covariates might be present, they are very likely to be equally distributed between the D-02 and D-04 groups, because the sample sizes for both D-02 and D-04 are quite large or, if they are not equally distributed, that would probably be because they are so rare that they would be unimportant in terms of affecting outcome.

The second issue that I would like to address is the influence of medications and electroconvulsive therapy. Obviously, this was a major consideration for us, because the way the trials were set up is patients in both groups could have access to virtually every therapy that was legally marketed.

That raises the question of whether, in the end, in fact, the treatment that the patients received as adjunctive treatment for the D-02 and the standard of care treatment for D-04 are indeed comparable.

So we have done a number of analysis to address that issue. The first thing we did was simply to look at the use of new treatments during the 12-month outcome. That is, the addition of a new medication or significant increase in an existing medication, based on those ATHF criteria.

What we found was that, among D-02 responders, 56 percent of the patients added or increased a medication during the 12 months. In contrast, 77 percent of the nonresponders and 81 percent of the D-04 responders did so, differences both of which were statistically different from the D-02 responders.

So this is highly suggestive that the benefit the D-02 responders are deriving is coming from VNS, because they are actually using less medication as time goes on.

That wasn't enough for us, however. We wanted to address this even further. So we undertook a series of censored analyses which we felt would be a rather conservative way to address this potential area of bias.

The censored analysis that I will be sharing with you this morning was the most conservative of a set that we did. In this analysis the D-02 patient scores are censored at the first significant increase or addition of an antidepressant medication.

At that point, what we do is drag forward the last score prior to censoring for those patients into the subsequent observation periods used in the repeated measures in your regression analysis of its scores.

This has the effect of truncating the VNS benefit. So it is somewhat unfair to the VNS. Even in the absence of medication, it is unfair to the VNS group, because it truncates any ongoing or increasing benefit they might obtain from VNS to, in this case, an average of seven months out of the 12 months of treatment.

At the same time, the D-04 patient scores are uncensored. They get the benefit of the full 12 months of treatment with unlimited treatment changes.

This slide shows you the results from that censored analysis. First of all, in the blue line were the results we saw before on an earlier slide for the D-04 patients on the repeated measures linear regression analysis of its scores. The bottom burnt orange line are the scores or the line that we saw before for the D-02 patients uncensored.

The censored line for the D-02 patients is shown in the line in the middle in the yellow color. So not surprisingly, once censored, the D-02 scores aren't as good, and yet there still is a good amount of change.

You can see as a change per month from baseline, uncensored is here, and for the D-02 censored scores it is still a good amount of change, and as a change from baseline both are statistically significant.

More importantly, if we look at the average difference per month versus the D-04 groups, here are the uncensored values that we looked at before. The average change per month on the IDS was a difference of .397 which, as we saw, was statistically significant.

Censored, actually somewhat to our surprise because we didn't expect this in this sensitivity analysis, didn't reach statistical significance, but it came awfully close at .052.

So we conclude that differences in outcomes between the D-02 and D-04 patients are not attributable to baseline characteristics. In fact, the results were the same with and without propensity adjustment, nor are they attributable to concomitant antidepressant medication or ECT, and I should mention that in the censored analysis medication changes always preceded an attempt at ECT. So they are accounted for in the censored analysis.

So the differences in outcomes between the two groups are also not attributed to concomitant antidepressant medication or ECT. The D-02 responders had fewer medication changes, and the D-02 patients, even censored for concomitant treatment changes, still improved more than the D-04 patients, uncensored, but didn't quite reach statistical significance.

Additional evidence for the effectiveness of VNS therapy comes from a number of datasets indicated on this slide. First, let me talk about the findings from the D-02 acute study. That was the randomized control of VNS versus sham treatment. The primary outcome measure was a response on the Hamilton. That is a 50 percent improvement.

There was a numerical trend for the treated group shown in orange to be better than the sham group of 15 versus 10 percent. This numerical trend held up through all the analyses, but it rarely reached statistical significance.

It did not reach statistical significance on the primary outcome. It did occasionally reach statistical significance on secondary outcomes such as the response from the IDS where the slightly larger differential of 17 versus 8 percent in response rates was statistically significant.

In the long term, you won't be surprised if I tell you on the repeated measures of the Hamilton scores compared to baseline, that was statistically significant. Here I have chosen to display the longer term results in terms of the categorical outcomes. What you will note is that, regardless of what scale is used, whether the IDS, the Hamilton scale or Montgomery Asberg scale, there is an accruing response over time as the patients continue from three months out to one year.

AS I indicated before, for these very treatment resistant patients these traditional research definitions may understate the true benefit to the patient. That is consistent with other very chronic and intransigent disorders such as obsessive compulsive disorder or schizophrenia. We sometimes accept a lower threshold for response. So we did use the clinical benefit categories that I showed you on an earlier study.

If you do that, in addition to the 30 percent of patients that were responders based on the Hamilton scale using the traditional research definition, you can pick up maybe another 25 percent of patients that fall into this category of a 25 to 49 percent improvement in the Hamilton, which could be meaningful in terms of producing some significant benefit for the patient, even in terms of functional outcomes.

So all told, when you add all those categories together, maybe up to slightly more than half of the patients do achieve some at least meaningful benefit during 12 months of VNS therapy.

Now, of course, those types of analyses, particularly when you are going from three to 12 months, do beg the question of which are those patients in 12 months? Are they the same patients at three months, and you are just adding more patients to it or is it a total different group of patients?

Obviously, what we would most like to see is that we are adding patients, and the patients that do benefit initially continue to benefit. This is a very important point with VNS therapy, as you have already probably come to appreciate from Dr. Rush's presentation and from hearing from some of the patients.

In this TRD population it is very unlikely that patients are going to respond, but even more stunning it is extremely unlikely that, once having responded, they are going to retain it. So while not controlled, I think these are some of the most persuasive data as to VNS's long term effectiveness.

So for instance, using the categories on the previous slide we found at the end of three months in the D-02 study there were 56 patients that fell into the extraordinary, highly meaningful or what was labeled meaningful clinical benefit, those patients that had at least a 25 percent improvement on the Hamilton score.

So after an additional nine months of therapy, what happens to those patients? Well, 41 of those patients continue to be maintained in one of these categories, and only 15 patients fall out of that category. So 73 percent of the patients all told maintained at least a meaningful clinical benefit from three to 12 months with continued adjunctive VNS therapy.

We can use these same type of analyses, which we refer to as SHIF tables, and this is just a pictorial form of that, to ask what happens to the patients that don't benefit after three months. You already heard from the patients that in some cases it takes a long time to derive benefit, and that is illustrated here.

There were 118 patients that after three months did not fall into those more desirable categories of extraordinary, highly meaningful or meaningful clinical benefit, and after an additional nine months of therapy 56 of the 118, or nearly half, did transition into at least the meaningful category of clinical benefit.

So there is some value-- There appears to be some value in continuing VNS therapy even beyond three months in patients that don't initially respond.

Then finally for the effectiveness data, I want to end with results from the D-01 feasibility study, so we don't shortchange that. The primary outcome identified in that protocol was response rate on the Hamilton.

After three months of therapy you see that 31 percent of the patients were responders, Fifteen percent were in complete response, and at the 12 month point 45 percent of the D-01 patients were responders, and 27 percent were complete responders.

Again, we can do that type of SHIF table analysis, and here are the results for the D-01 study, again using the same categories of clinical benefit. There were 30 patients after three months that were in those desirable categories, and after an additional nine months of therapy 23 of those 30 patients maintained at least a meaningful clinical benefit, and that was 77 percent of the patients.

I promised before that I would address the issue of placebo response. For those of you that are very familiar with depression studies, you know this is a major issue in doing clinical trials in drugs, at least with the more common type of depression. Hopefully, you are already getting an appreciation that it is not as much of an issue with treatment resistant depression, and I will show you why.

There are a number of reasons why improvements in the D-02 patients are not readily attributable to a placebo effect. First of all, I personally think the most persuasive is that we did show statistically significant differences in the D-02-D-04 comparison where we are actually comparing two active treatment regimens.

Moreover, just some general considerations lead us down the road that placebo response is a very unlikely explanation for the D-02 patients' outcomes. First of all, published literature tells us that placebo response rate in this treatment resistant group, unlike more common depression, is very low and the numbers that are cited in the literature are generally between zero and ten percent.

I think more compelling is that placebo response by nature is not usually sustained for 12 months, even in more common depression. There is a good literature now characterizing the pattern of placebo response, and what that literature tells us is that placebo response tends to occur early and is not persistent or sustained.

Then, too, consider that as you saw in some of the data that Dr. Rush presented, even in active treatment maintenance of patients -- and he showed you data from the ECT responders. Even when those patients are not on placebo but in active treatment, their maintenance of response is very poor, and this is another view of data that Dr. Rush showed you earlier, just a simpler presentation using bars rather than survival analysis curves.

Again, to remind you of the findings, following successful ECTs -- these are all ECT responders -- and then following patients in Dr. Sackheim's study out to one year in a naturalistic setting, 68 percent of the patients who had a prior history of medication resistance, and that could be as little as resistance to one drug -- 68 percent of those patients relapsed over the year, which was twice as high as the relapse rate in patients without an adequate medication trial prior to ECT.

We used those observations to do one exploratory analysis that I would like to share with you. There is no statistical values here, because it was just an exploratory analysis, and we didn't do statistical testing. But we asked ourselves what would happen if we just looked at the chronic subset of patients from D-02 and D-04?

Remember, two-thirds of the patients were in a chronic episode, and that is defined as a continuous episode of two or more years. These are the patients you would expect are the least likely to be subject to some type of placebo response. So we wanted to see if their overall response was similar to the total group.

That is indeed what we found. Here you see the overall response on the Hamilton scale was 29 percent for this group and 10 percent for the D-04 group, and complete response was 14 versus 3 percent. So percentages very similar to the overall group.

So in summary for this section, we found consistent numerical advantages for acute VNS therapy over sham control. Although it didn't reach statistical significance on the primary outcome, it was statistically significant on a few of the secondary outcomes. So it lends at least supporting evidence for VNS's effectiveness.

We saw increasing improvement of responders and complete responders over time, and probably most importantly, we saw improvements during adjunctive VNS therapy that were sustained at a high rate.

By contrast, placebo response in depression studies or depression in general tends to occur early and is not sustained. And even in medication resistant ECT responders, relapse is very high, even during active continuation therapy.

Next I would like to provide a very quick summary of the safety data that was in our application. Our safety database consisted of a pool of patients from the D-01, D-02 and D-03 groups or studies. That encompassed 342 patients and, I think, importantly, 689 total patient years of exposure.

The common adverse events that were obtained in the depression studies are similar to what we experienced in the epilepsy studies and epilepsy clinical use. They are listed here on this slide as defined by those events that occurred at least at a five percent incidence during acute treatment in the D-02 group and at a rate at least one and a half times that in the sham control, a sort of convention that helps sort out treatment related side effects from those that aren't treatment related.

The most common side effect that we observed was voice alteration in 68 percent of the therapy group, and the other very common effects, cough increase, shortness of breath, and swallowing difficulties, as well as some discomfort at the site of stimulation, whether that is pain or peresthesias, are all commonly known to occur with VNS therapy when it is used for the treatment of epilepsy.

These different side effects are generally mild to moderate, and most often, particularly the ones on the top of the list, are effects that only occur when the stimulator is actually on. So they may only be experienced by the patient during the 30 seconds that the stimulator is typically on, and then they don't experience them for the five minutes that it is off.

Also, the events tend to decrease over time, or at least the reporting of the events decrease over time, as illustrated by this analysis. Here we are looking at a cohort of patients that report these more frequent adverse events during the first three months of therapy, and then have continuous observations over 12 months of therapy so we can track the persistence of disappearance of that event over 12 months.

So for example, if we just look at the first one here, cough increase, what we found was that there were 55 patients -- you probably can't see the numbers too well, but there were 55 patients that reported this particular adverse event in the D-02 study during the first three months of stimulation, and then over the course of the next nine months their reporting of that side effect decreases, so that by the period nine to 12 months only 11 of the original 55 patients are still reporting that adverse event.

You see a similar pattern throughout all these common side effects, although some decrease to a lesser extent than others, like there still is a fair degree of persistence on the voice alteration. Not too surprising, since it is a direct effect of stimulating the vagus nerve.

Overall, I think very importantly, these adverse events are very well tolerated, and that is as evidence by this slide. Here we are looking at adverse event related discontinuation rates from the D-01 and D-02 studies at the time of our data cutoff for the submission.

That encompassed at least two years of experience for all the D-01 patients and at least one year for the D-02 patients. You will observe that only three percent in each of those two studies had discontinued during that time period specifically related to an adverse event.

This, I think, compares very favorably with what you see in typical drug trials where maybe 10 or even more than 10 percent of patients will discontinue for adverse events even over a short term trial lasting eight to 12 weeks.

In our review of safety, we were particularly focused on some issues that might be specific to depressed patients or the disorder of depression, and the one that we were most focused on was suicide.

Now probably all the Panel members are very sensitive to this, because there has been a lot in the public press recently about concerns that antidepressant drugs may very rarely provoke suicidal type thinking in patients, particularly pediatric patients, which has been the recent focus.

So we looked at this very carefully. First, here is the results for the pool of the D-01, 02 and 03 studies, 342 patients. We have the incidence of suicide attempts per patient year here. That works out to 3.5 percent, and actual suicide 0.4 percent.

The first thing we did was compare that to published literature. There is a nice review by Khan and Co-Workers. It is a very large review, as you can see. It encompasses almost 20,000 patients. Those patients are derived from the FDA summary bases of approval for seven different antidepressant drugs.

What Khan and Co-Workers found in a less treatment resistant group was suicide attempt rates of 2.9 percent, and actual suicide of 0.8 percent per patient year in the combined active treatment groups, and here are data also for the placebo group.

So we think the rates for the VNS group compare quite favorably with this. We also had the ability to look specifically at suicide ideation in the form of the third item of the Hamilton scale, which measures suicidal ideation.

What we were able to do here is use a standard definition that the pharmaceutical manufacturers or sponsors use, which is to look for the percentage of patients that have a two-point increase from baseline in their third score on the Hamilton.

We compared the experience in the VNS group to two control groups. First, in the acute D-02 trial, we compared the VNS group with the sham control group, and you can see similar rates. Two percent of the actively treated VNS patient and three percent of the sham treated patients had these two-point increases.

Then for more long term exposure we were able to compare the D-02 long term experience at 12 months with the D-04 experience at 12 months when we did have that Hamilton rating. Again you see similar rates, three and two percent respectively.

We should not forget that VNS therapy has been on the market for seven years. We have accumulated a lot of safety data from the epilepsy experience, most of which, obviously, is directly relevant to the depression experience also.

As you heard earlier, we now have more than 22,000 implanted patients and over 56,000 patient years of experience. In that experience, we found the VNS implant procedure to be a relatively low risk procedure.

The most important and common adverse associated with the procedure itself are infection necessitating explant in about one percent of patients, nerve damage in about less than half a percent of patients, either in the form of damage to the vagus nerve or the facial nerve, and a phenomenon of transient asystole in the operating room when the device is first turned on for testing that occurs in somewhere between one and two patients in 1,000.

As with the depression data, most of the adverse events in the epilepsy clinical experience and the clinical trials has proved -- most of the adverse events have proved to be minor and stimulation related.

There are few serious simulation related events associated with VNS therapy, and patients -- As a measure of how well tolerated the therapy is, patients in our pool of epilepsy trials continued therapy at a very high rate, about 72 percent after three years.

So in summary, our safety data from the depression studies has shown us that adverse events are mainly stimulation related and not troublesome. There is a low rate of treatment related discontinuation.

There is no signal for treatment related emergence of suicidal ideation or behavior and, obviously, we have to hasten to acknowledge that this is at best a very rare event. So the ability of our dataset to actually detect such a signal would be rather limited, but at least what we can say is, based on the data that we have looked at, there is no signal for emergent suicidal ideation or behavior.

Also data that I didn't show you this morning, another potential adverse event peculiar or specific to depression that we looked at was the rate of emergent mania or hypomania.

About ten percent of the patients in the D-02 and D-04 studies were bipolar patients, and in this group you do worry about the emergence of mania, which can be a side effect either of treatment, and in fact, it is taken by most clinicians to be a sign they have an effective treatment, or it can be due to the underlying disease.

So we looked at that one carefully, too, and we found that the incidence of emergent mania or hypomania was in the range that you would expect for an effective antidepressant. We have data we can show you later, if you are more interested in that.

So overall, VNS therapy was very well tolerated and safe in our clinical trials.

There are also some unique benefits for VNS therapy, as it is a unique approach to treating depression. So from this device based approach, some benefits that you wouldn't get from drugs and, in some cases, from ECT.

For example, Mr. Totah alluded to early in his presentation that patients can acutely disable the device if necessary, to temporarily stop side effects. That is a unique advantage of this therapy.

Also, published data that I didn't present this morning shows that there is an absence of cognitive and psychomotor effects with VNS, and as you have already heard, cognitive and psychomotor effects can be a significant problem with drugs and especially electroconvulsive therapy.

Of course, as a device based approach, there is an absence of overdose toxicity, which is a major problem with antidepressant drugs. Additionally, since it is not a drug therapy, you have the ability to add VNS therapy to other drugs without a concern for a drug-drug interaction.

Not to be overlooked, because VNS doesn't require active participation by the patient in the form of taking a pill every day, treatment compliance is obviously high with this particular therapy. And as you are all aware of, treatment compliance is a major, major problem with all chronic disorders, but particularly psychiatric disorders.

So a lot of patients never get their prescriptions filled. if they get them filled, they don't take them. If they take their pills, they don't take all of them. So this alone, I think, is a significant benefit for VNS therapy.

In conclusion, data that I have shown you this morning shows us that VNS effects on brain structures and neurotransmitters associated with mood regulation provide a biological rationale for the use of VNS therapy in treatment resistant depression.

Adjunctive long term VNS therapy was more effective than a standard of care treatment alone, and the p-value for that was less than 0.001 in the primary analysis.

The differences versus standard of care are not explained by differences in the baseline patient or disease characteristics, concomitant treatments or a placebo effect.

The improvements observed with adjunctive VNS therapy are largely sustained during long term treatment. VNS therapy is well tolerated and safe in depression clinical trials and clinical use in epilepsy, and finally, VNS therapy has additional device related benefits versus standard of care.

At this point I would like to invite Dr. Rush to come back up and just give a few closing remarks, and we will try to put these clinical data into a clinical perspective.

DR. RUSH: Thank you, Richard. I will be very brief. You have had to listen to a long series of presentations.

I want to just address the data and the information from the point of view of a clinician. I have been doing clinical work for 30 years and trial work for the same period of time.

It is important to put into context again, as we tried to do at the beginning, to remind you who it is that we are talking about that we are treating. These individuals, more than half had previously had ECT, and it had not produced a sustained benefit. Many had just not even had a response to it at all.

So these are really the most treatment resistant, the most difficult and disabled depressed patients that I have ever put into a trial anywhere and, as I mentioned earlier, half would not have -- half of the ECT community sample would not be eligible for our study.

To give you a sense of how we recruited patients for this, it might give more of a clinical feel. At least in Dallas, we went out to the well known psychopharmacology masters, if you will, the people that do advanced, complex medication management where treatment resistant patients go.

We asked them each to give us two, your very two worst, most difficult depressed patients. That is how we recruited the patients. So these are really very, very difficult patients for whom we don't have an alternative.

The second question is: Are the clinical effects meaningful? I think, to gauge the value of the clinical effects, you have to keep in mind three things. One is the population itself.

This, as we have discussed, is a population where we really don't have much going, and especially in the long run. So if we can help one in four or one in five to actually achieve a response or better -- and notice that some of our responders actually hit remission, and some of the patients today are in remission -- that is a home run in a patient population where we just don't see it in the long run. It just doesn't happen.

So given the severity of the illness and its treatment resistant nature, and the standard high threshold for benefit, 50 percent, we are looking at 37 percent versus 12 percent using the CGI. So you are looking at a number needed to treat of 4 or a number needed to treat of 5, if you use the Hamilton.

Those are very significant benefits, especially when we are not looking at short term. We are looking at the end of the year, when we should, in fact, have lost territory, if you look at all the other treatments that we have.

We should have done better in the sort run and worse in the long run. In fact, we did not so great on the short run, but really terrifically with this population in the long run. That is the "Duracell bunny keeps on working" for most, not for everybody, as you saw from Dr. Rudolph's presentation, but for most people there is a benefit that largely is sustained at a year.

The good news is some people who aren't benefitted early seem to come up with a benefit later. We have looked at that even over two years, and that seems to be a fact.

Could this really be a placebo? I think, looking at the population just per se, this is so unlikely it would be a miracle to have a placebo of this magnitude that works for this long and that consistently over time. It just would, I think, be looking at it as a clinician, very, very unlikely.

Typically, placebos, when they work, work early. Well, we don't have as much early as we have later. So the timing is all wrong. Secondly, they wear off. Well, we seem to have an increasing benefit over time. That would be a very unique placebo. Finally, those that benefit seem to have a sustained benefit, by and large. That also is unique, would not be easily attributed to placebo.

Finally, the induction of hypomania: While it was an adverse event here, it was more common, as you heard, in bipolar disorder patients. We often look at that clinically in antidepressant trials as an indicator that we have antidepressant activity.

Then finally, we have the experiment in nature, not conducted in any of the D series, but several patients you heard who lost efficacy when their battery ran down and achieved a recapture when the battery was replaced. Not an experiment that we could have done early on, but one that is going on by nature.

Just to put a final face on it and a comment on public health significance. I'll be quiet. I told you about this graduate student at the beginning. That individual entered the VNS study, and he did very well. Within about two months, he achieved remission. So this is unusual. It is one of the earlier responders.

He has stayed in remission now for five years. He finished his graduate school, got married, and has a child.

The other comment I want to make is about the category of response. We are very used to the 50 percent, because it comes from the nontreatment resistant world. Some of the patients mentioned that any benefit to some degree is better than what we have now, and you saw Dr. Rudolph talk a little about the 25 to 49 percent group, and I just want to give you one patient in this regard.

This is a lady who actually was not a responder, but she was a lady who was full time employed, in her mid-forties, married with two children, one just about to graduate from high school. She had had depression, really, since her early twenties, largely on, sometimes off, and the last several years more time -- roughly an eight-year episode, last episode.

She had received 80 ECT treatments. She was in maintenance ECT. She was brought in by her husband who said, the ECT is really helping her; it's the only way we can keep her out of the hospital, but she is having these long term cognitive difficulties and I really -- I can't allow her to go on and she, too, is complaining of it.

So we gave her the VNS treatment. It took a while, probably about three to four months. We kept her medication. So it's past the end of -- Well, at the end of the study, in the three months where medications were fixed, she had a 48 percent reduction in her Hamilton.

Then we followed her out, medications largely fixed. We were very loathe to change medicine. She had been so fragile. She had been in and out of hospital many, many times in the prior several years. So we left her where we were, adjusted the parameters. She never hit response. She never hit a 50 percent reduction. She was always in the forties..

So she comes up as not a beneficiary in the classical definition. Good news is she had been fired from her job because she couldn't function as a computer program or information technology person about a year and a half before we started. She wasn't able to be rehired there. She worked for a large IT company, a famous name you know. But she started a business in her own home and was partially employed.

The most important thing she said was, I got to see my son graduate from high school. Excuse me. So even though we don't call them responders, there are a number of people that actually really do quite well with this. Not perfect, but a lot better than what they had.

Then finally, just let me make one comment about the public health perspective. I did these numbers, and I thought about it, and it's so shocking to me, I thought I would share it with you.

Thirty thousand people per year commit suicide. Eighty percent are due to depression. That is 24,000 people. Treatment resistant depression is known to be the most lethal form of depression. Let's say half of those individuals that commit suicide from depression have treatment resistant depression. That's 12,000 suicides a year. That is 1,000 suicides a month. That is one suicide every 45 minutes. That means we lost four of these individuals in the last two and a half hours due to treatment resistant depression.

This is not a panacea, obviously, but it is a high need and, if we can help one out of five of these people with this treatment, I think it would be a tremendous contribution. Thank you.

CHAIRPERSON BECKER: Thank you. I would like to thank the sponsor for their presentation.

Given the fact that we all have been sitting here for quite some period of time, I think maybe it is appropriate to take a break now and reconvene in 15 minutes, say at ten after eleven. Thank you.

(Whereupon, the foregoing matter went off the record at 10:56 a.m. and went back on the record at 11:14 a.m.)

CHAIRPERSON BECKER: If I could get everyone to take their seats, we'll get the meeting restarted. Alright, thank you. It's now 11:15 and we'll resume the meeting. And we'll start with the FDA presentations on this PMA. The first FDA presenter is Carlos L. Pena, Ph.D. Dr. Pena?

DR. PENA: Good morning panel members. My name is Carlos Pena, and I am here today from FDA to present to you the PMA application for the Vagus Nerve Stimulation Therapy System proposed to treatment of resistant depression. I'm accompanied by Dr. Schlosser, medical officer, who will be sharing with you safety data contained in the application, and Dr. Lao, statistical officer, who will be sharing with you statistical data contained in the application. And I'll be providing the regulatory history of the VNS Therapy System, an overview of VNS studies including efficacy data, and a closing summary.

The sponsor has described the VNS Therapy System in some detail, which includes an implantable pulse generator, lead, and external programming system. The sponsor seeks commercial approval for the injunctive long-term treatment of chronic or recurring depression for patients over the age of 18 who are experiencing a major depressive episode that has not had an adequate response to two or more antidepressant treatments. VNS has previously been approved for use as an injunctive therapy in reducing seizures in patients refractory to epileptic medications.

Regarding the mechanism of action, no definitive mechanism of action has been reported for the proposed indication for the injunctive long-term treatment of chronic or recurrent depression.

I will now discuss the regulatory history of the VNS Therapy System. Following FDA approval for epilepsy in 1997, anecdotal reports of mood alteration were noted for some epilepsy patients. And the sponsor conducted a 30-patient, later expanded to 60-patient, pilot study called D01. The pilot results led to the development of the D02 study. The D02 pivotal study included an acute, randomized, placebo-controlled phase -- the only randomized placebo-controlled portion involving VNS studies discussed today -- as well as long-term follow-up. The sponsor un-blinded the acute phase of D02 in 2002, and found that the study failed to demonstrate a statistically significant difference between responders in the treatment arm and sham treatment control arm, the study's primary efficacy endpoint.

Despite the failed outcome, the sponsor claimed a pattern of increasing treatment effect over time, and suggested that the full antidepressant effect of VNS therapy might take longer. The sponsor proposed to use a non-significant risk study, D04, as a reference group for comparing to D02, long-term clinical data. And FDA advised the sponsor of the serious concerns regarding the ability of this comparison to demonstrate safety and effectiveness of their device due to lack of a randomized subject data set. The sponsor submitted their application in October of 2003.

In all, there are six studies that will be discussed today. During the first part of FDA's presentation I will focus on the first three studies, called D01:, the pilot study, D02: the pivotal study, and D04: the observation of control study. Other trials include D03 and D06, both of which will be discussed further by Dr. Schlosser, and D05, which was a videotape assessment of D02 study subjects only to ensure interrater reliability in assessments. Which takes us to a description of each study.

In the D01 pilot study, this study was an open label, non-randomized, single treatment arm, multi-center study. The primary efficacy endpoint was the proportion of subjects that responded to therapy, response defined as a 50 percent or more decrease reported as improvement in the HAM-D score at post-treatment compared with the baseline. The HAM-D, the Hamilton Rating Scale for Depression, is a clinician's tool to rate depression.

Out of a total of 71 subjects, 11 discontinued prior to implantation, 60 were implanted, and 59 completed the acute phase. Across various times during the pilot study, several subjects had concomitant treatment changes. Six subjects had changes in concomitant treatments during the four weeks prior to their first visit post-implantation, twelve subjects had changes in concomitant treatments during the acute phase, and three subjects received ECTs during the long-term study, and a total of 77 serious adverse events were reported. And Dr. Schlosser will discuss these events shortly.

At the acute phase exit, 18 of 59 patients were responders, 25 of 55 patients were responders at one year, and 18 of 42 patients were responders at two years. Response defined by a greater than 50 percent decrease in the HAM-D score compared with baseline.

The pilot results led to the development of the D02 pivotal study. And the D02 pivotal study was comprised of two phases, including a randomized controlled 12-week acute phase, and a 12-month follow-up evaluation period. During the randomized controlled acute phase, subjects were required to maintain a stable medication regimen, and during the long-term phase, changes to the mood disorder treatments and ECT were allowed, and no concomitant treatment criteria was provided in the clinical protocol. The primary efficacy endpoint of the acute phase was the proportion of subjects who had greater than a 50 percent decrease in the HAM-D at acute phase exit compared to baseline. And responders in the treatment group were compared to the proportion of responders in the control group.

A total of 266 subjects enrolled, 31 discontinued prior to implantation, leaving 235 patients that were implanted. The second yellow box, green box on the left-hand side. And 222 patients were considered evaluable for efficacy analyses. Of the 222 evaluable subjects, 112 were randomized to treatment and 110 were randomized to sham treatment control. And regarding the long-term phase of enrollment, of the 235 subjects who were implanted, 233 subjects were identified as the safety population, 205 were considered evaluable subjects, and 177 subjects were 12-month completers.

During the acute phase, nine subjects had changes in concomitant treatments and noise to use was reported. During the long-term phase, changes in concomitant treatments were allowed, and 169 patients of the 205 evaluable subjects added or increased antidepressant medications. In addition, 14 subjects received ECT. Of those 14 subjects, eight subjects were 12-month completers, four subjects were categorized as responders, and two subjects were categorized as complete responders. And I would like to remind you that one of our panel questions is related to the use of concomitant antidepressant treatment changes, permissible in the treatment group, over the course of 12 months.

The sponsor reported implantation related adverse events, stimulation related adverse events, and other events from the D02 study. These results will also be discussed by Dr. Schlosser shortly.

The primary efficacy endpoint failed to show a significant difference between treatment subjects, those who received VNS, and sham treatment control subjects, those who received regular care for treatment-resistant depression. In other words, the amount of improvement for patients with VNS was not statistically significantly greater than the amount of improvement when receiving standard care. And other psychiatric measurement tools reported similar outcomes during the acute phase.

Despite the outcome of the acute phase, the sponsor submitted a revised statistical plan with new primary efficacy endpoints, and employed an observational control study for comparison. And the revised statistical plan introduced the D04 control study.

The design of the D04 study was to collect long-term clinical quality of life, productivity, and health care utilization data on patients with depression. The D04 study began towards the end of the D02 study, and was a non-significant risk study conducted under local IRB jurisdiction. Up to 130 patients enrolled, and standard of care was defined in the clinical protocol as whatever treatment strategy the physician and the subject chose to follow.

Which takes us to the D02/D04 comparison. The objective of the D02/D04 comparison was to demonstrate that there is a difference in the improvement of patients with VNS therapy plus treatment compared to treatment as usual. The design, schedule, sample size, concomitant treatments have all been described previously. There was a total of 22 sites enrolling patients in either D02, D04, or both. And of those 22 sites, eight sites enrolled patients for D02 only, and one site enrolled patients for D04 only. And Dr. Lao will discuss statistical outcomes associated with the use of overlapping and non-overlapping investigational sites.

The majority of D04 subjects enrolled after D02 was closed. And sites that enrolled both the D02 and D04 patients usually screened and offered patients enrollment into D02 prior to enrollment into D04, because D02 offered a new treatment as opposed to standard of care, and sites were more focused on the treatment study rather than a naturalistic observational study, D04.

During the six months of overlap between D02 and D04, 83 percent of the patients who met the enrollment criteria for the D02 study enrolled into D02. And 17 patients who met the enrollment criteria for the D04 study enrolled into D04. After D02 closed, clinical sites had a pool of subjects interested in D02 that were also eligible for D04. And subjects that could not enroll in D02 typically enrolled into D04. And I would like to remind you that another panel question you will discuss is related to the enrollment outcomes of an investigational study versus an observational control study.

The primary efficacy endpoint was a repeated measure of linear regression analysis performed on raw IDS-SR scores of D02 and D04 patients. The IDS-SR is a self-assessment tool for depression. And the HAM-D, the primary efficacy assessment tool for D02 during the acute and long-term phase, was only included as a baseline D04 assessment, and therefore was not adequate for D02/D04 comparative analyses.

The D02 study also collected safety data. However, the D04 study did not prospectively or systematically collect any safety data while studying treatment-resistant depression, and is an issue determining whether the VNS Therapy System is safe for the proposed indication.

Evaluable patient baseline demographics between D02 and D04 were for the most part comparable. However, there were significant differences in baseline demographics between D02 and D04, including those patients who received ECT during their lifetimes, patients who received ECT during the current major depressive episode, and patients in the control population with greater than 10 lifetime episodes of depression. In addition, there are several patient variables for which no information was collected, and have been reported in published literature to influence treatment responsiveness. And unmeasured patient variables are also an issue that we have provided as a question for your deliberations later today.

Now, if we turn to the primary analysis comparing long-term outcomes between D02 and D04, a statistically significant difference was observed in the estimated IDS-SR raw scores per month between D02 and D04 at 12 months.

To further evaluate the permissive use of concomitant treatments during the long-term, the sponsor performed a second analysis, not specified in the original or revised clinical protocols, and compensating for concomitant treatment use. Namely, if a subject added or increased antidepressant treatment, and their subsequent IDS-SR scores prior to the change in concomitant treatments, last observation carryforward approach was used. The difference observed in the primary efficacy analysis was not statistically different from improvement observed under standard of care. In other words, there was no improvement difference between patients who improved with VNS and those patients receiving standard of care. And this is another issue that we have posed to you in the form of a panel question.

Aside from the statistical numerical outcomes of one analysis over another, more importantly, the overall permissive use of concomitant treatments during the long-term study is an issue in determining the effectiveness of this device.

And now I'd like to turn the presentation over to Dr. Schlosser, who will be discussing with you the safety data and the PMA.

DR. SCHLOSSER: Good morning. I'm Dr. Michael Schlosser. I'm a medical officer in the division reviewing this device. And I'm going to briefly talk about some of the safety data and the PMA.

I'm going to start by clarifying some of the terms. The sponsor purported the data, the safety data, looking at adverse events, and then also looking at treatment-emergent or stimulation related adverse events. And so I'm going to talk about those two categories. And then as a third category there were these serious adverse events. And so, throughout the slides I'm going to be talking about each of those three different groups, and I'll try to explain exactly which group we're looking at at the time.

Because a lot of the events we're talking about were stimulation related adverse events, I wanted to talk briefly about the stimulation parameters that were used in the protocol, just to start with. Specifically, I'm just going to focus on the current output. Current output was limited to 0.25 to 3.5 milliamps by protocol in the IDE. The adjustment protocol called for increasing this output by 0.25 milliamps in steps until a maximum tolerable level was achieved. Programmers were specifically instructed to warn patients that higher level of stimulation and stimulation related adverse events did not necessarily correspond to higher efficacy. In the protocol, it was listed that this was based on experience with the epilepsy patients, and therefore the programmers were instructed to tell the patients not to tolerate events that they really would normally not tolerate because they thought it was going to improve their efficacy.

Stimulation was to be decreased any time a patient reported that there was a painful or troubling adverse event. And the programmers were instructed to continue to increase the current during the programming phase in order to try to reach that maximum tolerable level during the two-week programming phase of the acute phase of the study. It was also noted in the protocol, and in the instructions to the programmers that any individual patient may have very different responses to different current levels, and they even went as far as to say that there may be some patients for who 0.25 milliamps would be the maximum tolerable setting, and that that should be expected and not an area of concern.

In April, 2005, Cyberonics sent a letter to the D02 investigators instructing a new stimulation protocol. This was to be implemented in patients who had a HAM-D score of greater than 10, which was their definition for non-responders at that point in the chronic study. This protocol specified a ramp-up period of six weeks during which, and this is a quote from the letter, several attempts should be made to increase output current to a level of 1.5 milliamps. It was also recommended that patients undergoing the ramp-up procedure be seen more frequently, every two weeks, but as frequently as every week, during the ramp-up period. And it was additionally recommended that if patients couldn't tolerate 1.5 milliamps, that an adjustment be made to -- I'm sorry, an adjustment be made to the pulse width in order to decrease it to 250 milliseconds, which may facilitate an increase in current levels.

So, now moving more specifically to the safety data. I'm just going to go through the different studies that we've already heard about this morning and talk about the safety data presented for each. There were 60 subjects in D01. Every subject reported at least one adverse event. The most common adverse events, which were reported in at least 10 subjects, were device site pain, headache, incisional pain, neck pain, dysphasia, increased cough, dyspnea, and voice alteration. These are kind of the common adverse events that we're going to see though the rest of the slides. They're also similar to the events known to occur during stimulation in epilepsy patients.

There were 77 serious adverse events reported across 38 of the patients. So greater than 50 percent of the patients had a serious adverse event. The most common being 12 suicide attempts or overdose events, and 34 cases of worsening depression. These are incidences of events, not number of patients. There was one death as a complication of surgery due to rectal prolapse.

In the acute study, now looking just at the acute D02 phase, 235 implanted patients. There were 233 reporting an adverse event. So, again, nearly every patient reporting adverse events. The events were classified in the PMA as mild, moderate, or severe. Mild was defined as easily tolerated and transient. Moderate as caused discomfort and interrupted usual activities. And severe, considerable interference with usual activities.

As you can see, there was a very large number of adverse events in both groups. It's important that we obviously look at the adverse events in both of these groups since they both had the surgical procedure and the implant. They were both exposed to risk. There was 61 severe adverse events in the treatment group and 73 severe adverse events in the sham control group. The difference between the adverse event rates to the treatment in sham control group probably relates to stimulation related adverse events, which I'm going to come to.

This is now just looking, again, at just all adverse events in the D02 acute phase. So these are not graded in any way as to their relationship to the device or to stimulation. And we can see the most common adverse event was clearly voice alteration, which was very common. Eighty-one patients, 68 percent of the treatment group; 44 patients, 37 percent in the sham control group. And then again, these are those adverse events that I read in that first slide that are going to be the ones we're going to see over and over again: device site reaction, device site pain, incisional pain, dysphasia, incision site reaction, cough, dyspnea. Similar events that you would expect given the direct stimulation to the Vagus nerve.

This slide now kind of focuses in a little bit on events. These are treatment-emergent adverse events that were rated as possibly, probably, or definitely related to stimulation. So this was a cut made by the investigators in the study. When they reported the events, they would then report whether they thought the event was related to stimulation. So you can see, it's the same type of list. It's really the same adverse events. There are very few of these events listed in the sham control group, which makes sense because these were events that the investigators determined were related to stimulation. So it's obvious that the investigators were able to pick up which events were related to stimulation and which patients weren't getting stimulation.

We've heard this morning that these events are similar in their frequency and in their nature to the events seen in the epilepsy study. I'm going to come back to that in one of my last slides. But I'll just make the point at this point that in this situation, we're comparing the adverse events and risks seen in patients with depression, to the benefit of this device in depression. And the safety of the device in another population doesn't necessarily mean that that device is safe in this different population. And the risk/benefit ratio must be looked at separately.

Moving on to serious adverse events. Just as a reminder, the definition of a serious adverse event is an event that resulted in death, a life-threatening event, hospitalization, prolongation of a current hospitalization, or a persistent disability. There were 39 such events. Nine occurred prior to implantation in the acute phase. So there were a total of 30 adverse events occurring between implantation and acute phase exit, 16 in the treatment control group and 14 in the sham control group.

If we look at these events individually, again, we have the most common is depression. I'm going to mention some things about the depression as an adverse event in a study of depression in another slide. But that was the most common serious adverse event. There was one suicide in the treatment group, none in the sham group. There were two cardiac events of note, an asystole and a bradycardia, both of which rose to the level of a serious adverse event. And then one wound infection. And then you can just go down the list, noticing that these numbers are small and there's one or two patients on either side. So there really are no statistical differences to be examined between treatment and sham control groups. And again, both groups were exposed to the device, and so, really, adverse events in both groups should really be included in the safety profile.

If we look now at the chronic phase of the D02 study, again, this is just serious adverse events. We see seven suicide attempts in six patients, four episodes of syncope in three patients, and then gastrointestinal disorder, convulsion, one episode of sudden death. I'm going to come back to sudden death at the end. And then by far and away the most common reported serious adverse event in the chronic phase was depression, 62 instances in 31 patients. The sponsor in the PMA explains that this probably represents a lack of efficacy of device rather than a true adverse event of the device.

Moving on to the D03 study. We haven't looked at the D03 study in the FDA presentation yet, so I'm just going to review the clinical protocol quickly. This was an open label, non-randomized, single arm, longitudinal study. It's the post-market study in Europe. It actually began before the CE mark, but the majority of the study has actually occurred afterwards. So it became a post-marketing study. Antidepressant treatment changes were allowed in this study, and the primary efficacy endpoint was a portion of subjects with a 50 percent response on the Hamilton rating scale at 12 weeks compared to baseline.

Safety data was collected in this study, and provided to us. Looking at just, again, the serious adverse events, 47 subjects implanted, 14 serious adverse events in those 47 subjects. Again, the most common, four cases of worsening depression. There were two suicides, which were also the only two deaths in the study. And then down the list, bacterial infection, accidental overdose, accidental injury, one case of syncope, and then gallstones, kidney stones, and kidney pain.

I won't go over specifically what the stimulation related non-serious adverse events were for D03. But as per the sponsor's submission, they were similar to those seen in D02 and D01 and in epilepsy studies.

The D06 clinical protocol. This was a pilot study of safety and efficacy in rapid cycling bipolar, as we heard. Standard bipolar disorder patients were included in D02, but rapid cycling patients were not. This was a separate study looking at that population by itself. It was again an open label, non-randomized, single arm study, so it was not designed to be an efficacy study, but did collect safety data, which we have. There were 11 subjects enrolled. Only seven actually implanted. And two subjects had one-year follow-up only, so we don't really have long follow-up on these patients.

We do have safety results. Again, I'm focused on serious adverse events. There was one suicide, three suicide attempts, one prior to implantation, two cases of worsening depression, and one case of manic reaction. So seven events in seven patients, though two subjects reported two events each.

Now we heard this morning about the specific focus on suicide. Obviously, published literature and recent experience has taught us that in very rare cases, antidepressant medications can precipitate suicidal behavior. There were 12 suicide attempts or overdoses in D01. The D02 acute phase had one suicide in the treatment group, none in the control. The long-term phase of D02 had seven suicide attempts in six patients. That was by the cutoff date for submission of the PMA. Since then there have been two more attempts in two additional patients. D03 study had two suicides. We don't have the report of attempts. And D06 had one suicide and two attempts in seven patients. So enough to make us concerned that there might be something to precipitation of suicide by this device, or at least to look at it more carefully. Again, a reminder that safety data was not collected in D04 for comparison.

This is kind of a busy slide, but what we're looking at here is the D02 acute phase on the top, and then comparative data, which includes a combined D01, D02, and D03 group on the bottom. So treatment group versus sham control group. Very small n here, only one suicide in the treatment group, none in the sham group, and no attempts. So tough to make a comparison, though, 4.3 percent versus zero percent. But the numbers are small.

In the larger comparative data we're looking now at a larger group, 342 patients in the combined analysis. And we see, as the sponsor showed this morning, if we look at instances of suicide per year, 0.4 percent. This is, again, comparison to this Khan Study, which was this very large meta-analysis of drug studies used for approval. And there we had a 0.8 percent rate in the treatment group, and a 0.4 percent rate in the placebo group. So similar numbers of instances of suicides per year and suicide attempts per year between the D01, D02, D03 combined and the meta-analysis.

Now, I mentioned I would come back to the epilepsy data. This is a chart that represents the stimulation related adverse events in the E-O5 study, which was the pivotal study for VNS for treatment of epilepsy. And again, you see this very similar list of adverse events that we've seen on all the previous slides: cough, dyspnea, hoarseness and voice alteration being the common events. This was a comparison between baseline and then high levels of stimulation. So these are actually patients compared to themselves. And we see, you know, they all reach levels of statistical significance. So these are all adverse events that are related to the stimulation, and they can be bothersome to the patient.

But again, I'll just restate that the determination of safety is not done in a vacuum. It's done as a risk/benefit analysis, comparing to the benefit or the efficacy of the device. And so the safety in the epilepsy population does not necessarily mean safety in the depression population. The adverse events must be weighed in relationship to the benefit shown by the efficacy studies.

And then finally I'm just going to finish by talking about cardiovascular events. I mentioned the one sudden death in the D02 study. There were also, I believe, two cases of sudden death in the epilepsy studies, and cases reported in the MDRs as well, very rare incidence of sudden death. There is a concern that this might be due to cardiac events due to the direct vagal nerve stimulation. Could this be causing a cardiac event that led to sudden death?

So we looked just kind of specifically at what cardiovascular events were seen. The events in red are the serious adverse events. This whole column is in red, even the zeroes, because I have not included the non-serious adverse events for the chronic phase. There were many, and in many cases they were the same patient reporting the same adverse event at each visit. For example, bradycardia. But there was no action taken. There was no action needed. And so those numbers were very large, but not necessarily meaningful. So the serious adverse events in the chronic phase: four cases of syncope, one case of dizziness. And then when we look at the acute phase, the case of asystole and bradycardia which I've already mentioned, and then several other cases of arrhythmia, hypertension, 10 cases of palpitation, 21 cases of dizziness. I should mention that of course there are many causes for dizziness, so while we lump this under cardiovascular events, there obviously can be other reasons why people can be dizzy. Vasodilatation and syncope.

So there were cardiac events. We don't have any evidence of sudden death due to a cardiac event from vagal nerve stimulation, but it's just something to keep in mind in terms of the safety profile of the device.

Now I'm going to turn it back over to Dr. Pena for a review.

DR. PENA: So in review, regarding safety, the absence of systematically collected safety data in the observational control study for comparison to the investigational study is an issue in determining whether the clinical data in the PMA provides reasonable assurance that the device is safe for the proposed indication.

And regarding efficacy, FDA has identified the following issues, including first, the chief limitation that the long-term D02/D04 comparative analysis is not derived from a randomized subject data set, but rather a comparison of outcomes from an investigational device study and observational control study. And a propensity adjustment strategy used to reduce potential bias in the comparative analysis is not able to address the problems of potential bias due to other unmeasured patient variables.

I would also mention that the sponsor noted in correspondence to FDA that both the D02 and D04 population would not differ on measured factors upon submitting their revised analysis, and that if they did, one could not be as confident in their statistical adjustment for baseline differences. FDA's uncertain whether one can reconcile the sponsor's statement in their own submission concerning there were relevant measured patient variables found to be significantly different between groups.

Second, FDA's concerned with the potential placebo effect rates for patients with VNS. The sponsor has discussed in some detail reasons why long-term outcomes from VNS patients are not due to placebo. Data provided in the submission identifies a placebo response rate of 10 percent, as defined by the clinician's measurement scale HAM-D, which persisted to the exit of the acute phase, namely 12 weeks.

Also, although both D02 and D04 were available to enrolled subjects at similar time periods, almost all D04 subjects enrolled into the study after D02 was closed for enrollment. Only 10 D04 subjects enrolled into D04 while D02 was open. And the sponsor has indicated that sites were more focused on the treatment study rather than the naturalistic observational control study.

And third, moving past the insignificant numerical outcomes upon censoring scores of VNS patients with concomitant treatments, and using their last observation carried forward, FDA is concerned that concomitant medications an ECT use were not standardized in either the D02 long-term study or the D04 observational control study. And I would also mention that when a patient adds or increases treatment, one can reasonably expect that patients are not responding, or poorly responding to their current therapy regimen. And one would be unsure of the cause of the patient's improvement to subsequent additions or increases in antidepressant treatments.

At this time, I would now like to turn it over to Dr. Lao, who will be presenting the statistical data contained in the PMA submission.

DR. LAO: Good morning, my name is Chang Lao, Division of Biostatistics, FDA. Today I am going to present a comparison between D02 and D04. D02 is the VNS plus standard care. D04 is standard care only. And the primary/secondary efficacy endpoint is HRSD-24, is the Hamilton Rating Scale for depression, 24 items. Maximum score 74. IDS-SR is maximum 84.

In the multi-center study, 22 sites, and overlapping Site 12. I'm going to talk overlapping sites later on. And then talk about propensity score analysis, try to test covariates in pairings. And then repeat a measure in the concordance study, try to predict an HRSD from IDS-SR. Then there is a statistical conclusion.

This is a D02, a brief summary on all 22 sites combined. In the three-month actual study, which is double-blind, randomized, VNS was a sham control. Primary HRSD, parameter, and IDSS-SR is secondary. And primary endpoint is the comparison to response proportion. The final result based on three-month actual study, the significant difference was found for the IDSS-SR, which is 17 percent, VNS was 7.5 percent sham, 0.03 based on psych test. No significant differences were found for the HRSD.

Then after three months, every patient was switched into VNS. So the primary endpoint for D02 only is HRSD. Try to estimate a slope for every radar change, scope must -- we try to estimate a mean response score when you study. The slope average amounts of change in HR score 0.45 per month. Which is standard endpoint of 0.05/95 company's interval.

D02 and D04, long-term comparison, the endpoint is average rate of change, which is slope estimate average mean score per quarter by repeated measure and integration. And the longitudinal data for the HRSD, D02 yes, D04 no. Because it only had a baseline and 12 months data only. So the sponsor switches to IDSS now because they had both longitudinal data. And to do the repeated measure of lineal regression.

Secondary endpoints is a proportional response based on the 50 percent reduction in score from baseline. This is a sample size table for the -- all the 22 sites, overlapping sites. Overlapping site means some site had both D02 and D04. As you can see, this sample size here, in the overlapping site the sample size is much smaller than the other 22 sites combined. By the way, sample size was based on the secondary endpoint. Compare the proportion of the response between the two groups. Not based on the primary endpoint.

This chart illustrates the patient by study and center for the D02 and D04 comparison. As you can see, about nine sites which had a D02, but no D04 study, and one site had D04 and no D02 studies. So overall, 10 sites out of 22 sites, which are roughly about 45 percent, no comparison group for those sites. The study design is incomplete and unbalanced. And it's hard to evaluate a true homogenous cross center. A true center interaction effect cannot be evaluated.

Propensity score analysis, which is when you have many, many confounding covariates present. Then the propensity score actually is overall composite scalar, sure to intend to reduce by comparison between D02 and D04. It took the best covariate only. Propensity score is a condition of probability of individual patients receiving D02. Condition even a set of IS patients, of best covariate, XI. We have total 17 covariates. Before after propensity score adjustment, use of logistic regression, which is a logit. Logit is a proportion of success to no success, probability of no success. Logit, log of that, actually is a probability assigned to D02 conditioned on a set of covariates. Vector X for the IS patient. The furnishing of the vector for a covariate for IS patient. So some basis of logistic regression.

Primary effectiveness analysis and repeated measure analysis, which its purpose is to estimate every radar change per month, or to estimate the mean response at 12 months, which is general mean response mode. We are interested in only comparing the mean, scope, from baseline between D02, D04, at 12 months, not individual patients performance, which were required for random effect mode. So the dependent variable here, IDSS, independent variable is the baseline IDSS-SR, treatment in either D02 or D04, or time, four quarters. And the PS quantum five level group by the Propensity Scale for each individual patient. Nine pool sites from 22 sites. And measured by time interaction and the special power correlation which allowed the correlation to change over time. And so missing random. Probably we are missing data in future. All absent data is independent, missing data independent from the future of this issue. Next.

Concordance study shows how good are the ideas of particular HRSD. And the statistics uses a correlation of linear regression. Outcome is a correlation coefficient intercept slope R-square. R-square is a percent variable, about a mean of HRSD, explained by the future regression model. How much can explain by the independent variable, IDS-SR?

A range of R squares, zero to one. Zero is the worst fate. One is perfect fate. So if R squared equals one, that means the IDSS-SR can predict HR very well. If zero, then no prediction at all.

This is a value, the top number is the repeated measure linear regression result. All the 22 sites combined. And top of the graph, the number there is each quarter observed predictive value of mean score, at each quarter for D02 and D04. At the bottom of the chart, there, you see the difference. D02 minus D04, observed at -6.6, which is not adjusted for any patient covariate, individual patient baseline data, or propensity score.

The predicted value after one year is minus 4.8, which is a reduced improvement at one year. This is for all 22 sites combined. If we look only for the 12 overlapping sites, the 12 have both D02 and D04. Then the improvement, at the bottom of the chart you can see that at one year, the improvement -- difference about 1.4 points. And the 95 confidence even before that difference, D02 minus D04, minus 3.82, minus 0.5. That's based on 12 overlapping sites.

This is a result for the concordance study, as I talked before. R-square, the position of R-square by this histogram. You would like to see the R-square equal close to one as possible best prediction. If close to zero, no prediction at all. The mean of this distribution of 235 evaluable patients, D02 study, of 0.55. Which means about 55 percent variability, and above the mean of the HRSD data, explained by IDSS-SR. So I would say it is, R-square is kind of in the middle. So concordance study, also you can look at the correlation coefficient and the slope. The mean correlation is about 0.70. And undefined of 0.67 to 0.73. And the slope, we use the average rate of change for the HRSD, for every unit change of IDS-SR.

Second endpoint is a definition of IDS-SR or HRSD score larger than 50 percent reduction from baseline. Statistics have several concerns here. Treat it like all the 22 sites as one site because the sponsor combined all the responses and non-response by D02, D04, combined together from those 22 sites. So it looks like the data all comes from one site. So no treatment in the baseline interaction was considered here. And supposed to talk about logistic regression for the response rate comparison to response proportion. But I don't see it. The covariate only appear in the linear regression analysis. So no pooling of the data, no modern approach, no covariate adjustment, like a baseline IDSS or HRSD site ECT RX used.

One reminder. IDSS is a baseline which was highly significant in the repeated measure regression. And also pool site is highly significant. So the unclear why to compare two proportion responses.

Summary. Three months double-blind randomized provided most for VNS. You know, this provisional study kind of weak in variable patients, all 22 sites. HRSD and the PY 0.3 second IDS-SR, which is 0.039 second. Based on two-sided extract test.

So the switch from primary HR to the two second endpoints, IDS-SR, in long-term study. Summary. In a way, the D02/D04 study is non-randomized, un-blinded. Propensity score were used for balance of a measured covariate only. Approximately the balance seemed to have achieved. Look at the difference of propensity score between the D02 and D04. Reasonable balance in PS quintiles. PS balance individual patient covariate, which is good. But, PS propensity score cannot balance unmeasured covariates selected by -- not accounted for by covariate, regression to the mean providing a placebo effect.

Summary. Data one can follow. If antidepressant resistant ratings score increased, and/or ECT was applied, the prior IDS-SR HRSD score was carried forward to the place subsequent. Now we're missing observation. Unclear effect of analysis in preparation of these for censored analysis, which we based on definition above. It's kind of not easy.

Now, over to D02/D04 site. Nine sites they had at D02, one site at D04. No D04. Only 12. Reduce the sample size by one-third. PS used nine pooled sites in the repeated measure. So it only partly accounts for the imbalance. And also used only 12 over the sites. Show a last effect.

This is the difference D02/D04 in repeated measure regression for IDS after one year. We have different end result here. Covariate -- observe just for covariate, based on raw data. The average difference after one year, about 6.6. Ninety-five confidence interval, about -10 to the -3.2. And it's just by covariate. All the 22 sites are about 4.8, with a 95 confidence interval. But if we use the overlapping site, we censor data. Then implement only cut it down to 2.1 points. And the 95 confidence interval minus 3.842, minus 0.54.

Always think of the 95 confidence interval rather than P value is more meaningful. Because P value only tests low hypothesis. The true difference equals zero. Equals zero doesn't equivalent -- may not be equivalent to the true clinical difference.

Summary. Unclear concordance study. Ask why 235 patients, about 0.55, which is kind of, you know, it's hard to say. It is not perfect predictor, anyway. And the way we estimate a mean difference IDS-SR, D02 minus D04, minus 2.1 unit to 4.8 unit, depending on which data you use, minus 2.1 with baseline of 12 overlapping sites of 4.8 points with all 22 sites. This improvement is clinically meaningful.

So this is based on the second endpoint, based on 12 months proportion of response based on 12 overlapping sites only. The only significant difference for the nine site data and all the 22 sites was 0.001. The main comparison source of data for the 12 overlapping sites didn't show any significant difference between the D02 and D04, for the second endpoint.

This ends my talk. Thank you.

DR. PENA: Panel members. The management of treatment-resistant depression is a therapeutic challenge to clinicians, and many such patients continue to lack adequate treatment options. However, any proposed device would need to have balanced scientific evidence to establish reasonable assurance of the safety and effectiveness of its use. Thus FDA has drafted five panel questions for your discussions in the afternoon session. At this time, FDA has completed its presentation, and we wait for instructions by the chairperson.

CHAIRPERSON BECKER: Thank you. We have about 45 minutes until the lunch break, so I think I'd like to open the session up for questions by the panel to the FDA presenters initially. Does anybody on the panel have a question for the FDA presenters?

DR. ELLENBERG: Yes, I wonder if you would mind bringing up the slide again for the covariate adjustment. I believe it was the third from the end in the statistical analysis. I'm afraid I didn't understand the presentation.

DR. PENA: Which slide did you refer to?

DR. ELLENBERG: It's the third from the end. It's titled Difference D02 minus D04 in IDS-SR improvement by one year.

DR. LAO: If you don't observe -- it's just based on observed only. And the sponsor's presentation shows after one year the difference, about 6.6 point difference with 95 confidence interval -10 to the -3.2. That's didn't use any statistics, just use the raw data, observed data. But --

DR. ELLENBERG: So that is not with any propensity score adjustment? Is that what you're saying, that's the first row?

DR. LAO: Yes. You are right.

DR. ELLENBERG: And then the second row is?

DR. LAO: Use the repeated measure linear regression, which use the propensity score adjustment. Also, individual patients' IDS-SR score. And the ninth --

DR. ELLENBERG: Wait, wait. So the second one is -- they're both done through the IDS-SR?

DR. LAO: Yes.

DR. ELLENBERG: Both rows?

DR. LAO: Yes.

DR. ELLENBERG: Alright. And the second row includes the PS adjustment. I believe the sponsor indicated that in the linear regression modeling, that the propensity score adjustment was not significant.

DR. LAO: That's right.

DR. ELLENBERG: Is that clear? Okay. So, in the second row you are presenting this data, and you're doing it for all 22 sites, and then for the 12 sites that are overlapping.

DR. LAO: Yes.

DR. ELLENBERG: So you give two results. And what is the point you're making?

DR. LAO: Making is because total 22 site. Ten site incomplete, missing, that didn't have the both D02/D04 study. So there's no comparison can be made for those 10 sites. So I would think you're using 12 sites which you had both D02/D04 study can do a meaningful comparison with each site.

DR. ELLENBERG: So you believe that using the overlapping sites only is a more legitimate comparison than 22 sites?

DR. LAO: Yes.

DR. ELLENBERG: Okay. I understand that. And then, in terms of differences between what the sponsor presented on the first row and what you're showing for the 12 overlapping sites, they're both below zero. The confidence interval are both below zero. So, are you making a claim that there is a difference in the differences here?

DR. LAO: You have some difference here. But the confidence interval leaves some clinical decision here.

DR. ELLENBERG: Right. So you're just stating that the company's interval is different from the interval you get with the 12 overlapping sites.

DR. LAO: Yes.

DR. ELLENBERG: And also -- I'm not sure why that's important if the propensity score wasn't statistically significant.

DR. LAO: The propensity score is only one of the covariates used in the linear regression. There are some other nine pooled sites, or individual IDS-SR baseline data which is highly significant.

DR. ELLENBERG: I'm sorry, I'm not following that at all. Start again?

DR. LAO: Patient baseline data, IDS-SR.

DR. ELLENBERG: That was not included in the propensity score?

DR. LAO: That was included in the repeated measure linear regression.

DR. ELLENBERG: In addition to the propensity score?

DR. LAO: Yes. Yes.

DR. ELLENBERG: And so this analysis with the 12 overlapping sites includes those baseline scores?

DR. LAO: Yes.

DR. ELLENBERG: And the first row for the sponsor submission did not include those covariate baseline scores?

DR. LAO: First row -6.6 didn't use any covariate adjustment, didn't use any repeated measure linear regression, just --

DR. ELLENBERG: Okay, thank you. I understand.

CHAIRPERSON BECKER: Are there any other questions for the FDA presenters?

DR. ELLENBERG: Yes, I'm sorry. An additional question. Another statistical question. Can you explain to panel why it's important that if you're using all of the sites in an analysis as was done by the sponsor, you thereby, since you don't have complete overlap, are eliminating the possibility of looking at any differences in the results by site. In other words, the more technical term, the site by treatment interaction can't be estimated.

Can you explain to the panel why for this particular PMA looking at the site by interaction would be extremely important, moderately important, very important. Are their hints in what you've seen in the data that it's really critical to look at the site by treatment interaction, or are you just making a statement that one always looks at the site by treatment interaction because there may be differences and we would want to see that?

DR. LAO: The idea that you would like to see the treatment effect D02 minus D04 difference is homogenous across center. So then you can try to use the statistical model to pool data, get overall evidence, summary evidence from all the sites. But if you have the 10 sites without a comparison group then the statistical model would be very difficult to the pooling of the data.

DR. ELLENBERG: I understand that, but is there something in the data that you've seen for what the D02 minus D04 difference is looking at them by site that would hint to you that this is something we should be concerned with?

DR. LAO: Yes. Hopefully they go the same direction. D02 is always superior to D04 for most of the sites. Not necessarily for all the sites, but for most of the sites. If you see the opposite direction, if you saw within some site D04 superior than D02, or other way around, then you wonder is it a correlative interaction there. You wonder how can you combine the data. Is something going on there that's not only due to treatment effect, maybe due to site effect.

DR. ELLENBERG: But are you presenting a hypothetical, or are you seeing the data in a way that indicates --

DR. LAO: No, I didn't see the data.

DR. ELLENBERG: Alright, so let me restate what I think your point is. In general, there could be an interaction between the treatment and the sites, which in lay terms simply means that the treatment effect might be different by sites. The treatment effect could be different by sites in two major ways. One way is that it's -- the treatment effect is considerably less in some sites than in other sites. Another way is that the treatment effect is entirely different in some sites than other sites. So it's useful to, at a minimum, review the site differences, and at a maximum be able to model through using a technical term, an interaction term, in the model. But there's no evidence that you've seen from the data that indicates that there is a treatment by site interaction of either sort. Has that made your point correctly?

DR. LAO: Yes, you are right.

DR. ELLENBERG: Thank you.

DR. LAO: The difficult point is here this is not a randomized trial. So the sample size for some sites is very small. You really cannot make a meaningful comparison for those sites with small sample size. Not enough power.

CHAIRPERSON BECKER: There's another question for the FDA presenters. Dr. Wang?

DR. WANG: What I'm interested in is your possible reasons for that discrepancy in the acute phase D02 results depending on which measure is used. Maybe actually if you can go back to Dr. Lao's slide, it's the one, D02 Brief Summary. It shows the acute phase results. There it is. What's your -- you've shown us that the correlation is moderate at best between the HAM-D and the IDS. What is your suggestion here? Is it that the -- someone suggested the IDS is maybe a more relevant modern measure of the depression constructs. Or potentially this is a self-report measure and you could imagine maybe being more prone to bias or a placebo effect? What's your sort of reasoning here?

DR. PENA: Well, we have concerns regarding the various psychometric tools that were used to measure effectiveness during the acute phase. So, while the IDS-SR demonstrates statistical significance between the viable patient population, there were other psychometric measurement tools that did not demonstrate statistical significance. So that makes us wonder if the outcomes, really how strong those outcomes are across different measurement tools.

Some of those measurement tools include the HAM-D, the BDI, the SF-36, and the MADRS, Montgomery Ashberg Depression Rating Scale.

DR. WANG: What is your sort of suggested or proposed weakness of the IDS-SR?

DR. PENA: Well, you would want -- we haven't arrived at a conclusion. We have serious concerns, though, with only one scale able to demonstrate statistical significance in acute phase period.

DR. LAO: If you look at the histogram for the R-square, the prediction varies from patient to patient. That's the point there. Some patients predict very well. Some patients no prediction at all.

DR. PENA: I'd just like to add one comment. And it's one of the reserve slides that we do have. During the acute phase, in both the treatment group and the sham treatment control group, there were four subjects that were categorized as a responder, but neither were in either the HAM-D or the IDS-SR. They were responders in one, not the other. So it's further concerns regarding the concordance and the outcomes of one tool over another, and the strength of the data.

DR. WITTEN: I think you also might want to note that it's not -- there's not the same totals. If you look at the slide that you were referring to for Dr. Lao, there's not the same total of patients. It's a relatively small number that aren't in both groups. But it's a small number that's a difference. Numerically, they were successful in both groups.

DR. WANG: The non-responders might be an extreme.

CHAIRPERSON BECKER: Any further questions for the FDA? Dr. Malone?

DR. MALONE: On the D02 acute, I think you did have a slide that said it was a failed study because the primary efficacy measure was not met? Is that right?

DR. PENA: Right. The acute phase data, the outcome of that acute phase failed to reach its primary efficacy endpoint. So it failed to reach that prospective outcome.

DR. MALONE: So my understanding would be if you want to do all these other tests, you have to start doing corrections for them. Because you're doing multiple tests. I mean, but regardless of that, it's a failed study. Is that right? If you take the primary outcome measure?

DR. PENA: Correct.

CHAIRPERSON BECKER: If there are no further questions for the FDS, perhaps we can start taking some questions for the sponsor at this time before we break for lunch.

Actually, I'll get things rolling by just asking a very simplistic question. If the sponsor wants to take the table. I was wondering how you chose the 12-week time frame for the acute study, which seemed not to be effective, but in the long term it was. Why did you initially choose that acute time frame? And when you saw that the study was negative, why didn't you continue in a randomized sham controlled fashion?

DR. RUDOLPH: The 12-week period is pretty standard for a drug trial. So we were following the pattern that's been used for drug trials. They're typically nowadays eight to twelve weeks. So that's the explanation for why it was set up that way initially.

I think the second part of your question was why didn't you just extend that longer. Well, by the time of course we un-blinded the results for that acute phase, the patients were all beyond the acute phase, so there wasn't an opportunity to continue it as a double-blind randomized trial.

CHAIRPERSON BECKER: There wasn't the opportunity just to continue stimulating one group and not stimulating the other group?

DR. RUDOLPH: Well, by the time that we saw the results, the sham treatment patients had already crossed over into the continuation long-term VNS stimulation.

CHAIRPERSON BECKER: Okay. Further questions for the sponsor?

DR. ELLENBERG: Yes, if I may. In reviewing the material and in hearing the superb presentations today by the sponsor, it seems to me that the analyses that have been presented in detail make every attempt to cover the issue of potential bias in making comparisons between one group of patients who have the essentially standard of care plus the VNS versus another group of patients who had basically equivalent standard of care. But patients that based on the FDA testimony just finished might not overlap totally in time, patients that might not be coming from the same centers, and most importantly patients that were not randomized to the two treatment arms.

At the end of the day, I wonder if you could respond to a reasonably direct question, whether the analyses that you have presented today make the presentation and our job in terms of making a recommendation to FDA based on a non-randomized comparison an okay decision. Can you try and help me to understand why there is not a need to do a new randomized study to make this comparison based on the initial findings, which looked extremely promising, especially after all the analyses that you presented today. Can you just help at least me, if not the rest of the panel, in arguing the case?

DR. RUDOLPH: Certainly.

DR. ELLENBERG: Thank you.

DR. RUDOLPH: One of the very first things we did when we saw the results and understood the results from the acute study was consider the possibility of doing another study. Ultimately we decided that the use of the D04 group as a control would give -- could potentially give high confidence in a determination of effectiveness. But we were also influenced by the fact that every other study design that we envisioned and discussed, both internally and with external consultants such as Dr. Rush, had significant limitations. And Dr. Rush can walk you through some of our concerns about doing another study.

DR. RUSH: Could I have the slide up, please? Let me try to take you through the thinking in the pattern, because the timing is very relevant here. Obviously it would be wonderful to have a randomized control trial. We began the entire evaluation of VNS in depression at a time when there, in fact, were no short or long-term randomized control trials in TRD short of acute trials with ECT that a few investigators had conducted. In fact, the natural course of TRD had not been mapped out by anyone, under routine treatment conditions for a 12-month period. And of course there was no long-term safety. In fact, in the initial study there as no short-term safety of Vagus nerve stimulation in depressed patients at all.

So the plan, which you saw as the D02 trial, was to conduct based on the epilepsy model. Part of the reason for the 10 weeks was the epilepsy model. It was an acute trial, 10 weeks long, following two weeks recovery after implantation, and it would compare sham versus VNS with very tight controls. So no medication changes within that three-month period. If you found a positive difference between the two groups, you could uniquely and absolutely attribute cause to VNS, and that would be the best evidence for efficacy. Even if it were modest in the short run, you could say it was the VNS.

And then the plan actually agreed to with the FDA was that there would be a long-term uncontrolled follow-up of a significant number of people who had had VNS to see whether or not there was a sustained benefit which would be extremely unusual in treatment-resistant depressed patients, as I showed you from the data this morning. So that was the plan. And the reason for it was simple, it's direct, has minimal patient exposure, long-term safety could be established. It's the most efficient path to approval, and it just made a lot of sense. And we had already done the D01 open trial to indicate that in fact there was a reasonable chance of a reasonably good benefit in the 10-week time period.

So as you know, the results, the primary outcome failed the positive finding on the secondary outcome. Then the question obviously is raised, gee, was the sample size too small? Well, I guess if it had been made larger we might have achieved a difference in the primary outcome. And what about a longer duration, a subject you already raised.

So as Dr. Rudolph said, at the end of that trial period, we couldn't then go back. The trial had already shut down. So, next slide. We then considered a variety of next-step options. One is to just simply conduct a longer term acute trial. Let's go out nine to twelve months. You have now maximized the duration, and if you don't change the medications, you can attribute with absolute certainty cause to VNS. It's a terrific design except it's not feasible, it's not ethical, and it's not safe. And you couldn't have any patients sign up for it. Because these are treatment-resistant depressed patients. The number of medication changes that occur over even a six-month period just to keep the patients intact, safe, and alive, is significant. So we would have lost all the patients probably by even four to five months, or at least a large proportion. So that then, the long term here mitigates -- prevents us, really, from attributing absolute 100 percent certainty cause to VNS, because medicines will change. And I'll come back to that in a second.

One of the difficulties is if VNS works in one group, and the other group does not receive real VNS, there's a differential management with medications. One group will have medicines changed at a different rate or time or so on. We'll come back to that.

Another possibility that we thought of is simply go after electroconvulsive therapy versus VNS acute, in an acute trial modality perhaps. A couple, two, three, four weeks -- I'm sorry, four weeks to, say, eight to twelve weeks. The problem is ECT is a terrific acute treatment, but can't be given over the long run in most patients. And VNS is not an acute treatment and has to be given over a long run. So it just wouldn't make any sense. And secondly, even as I showed you and Dr. Rudolph showed you, the patients entering the VNS trial, 40 to 50 percent had already had ECT and had failed. Thirty-some percent, 38 percent in the current episode. So we'd have to change the patient population.

Another possibility we discussed at length, and really brought this around, was the idea of taking the patients who had received VNS who had benefited from the D01 trial or the D02 trial and simply randomizing them to a discontinuation. Just turn off the device. Two difficulties with that. One is, of course, there's risking of un-blinding. But more important than that, we would probably need an even larger sample pool than could be generated. The problem is we went to patients. We asked them, we surveyed the patients directly. I spoke to a number of patients, other investigators did. To a person, and you heard it also from a patient today, the patient said with this emotion you are not going to turn this device off. I do not want this device to turn off. So we would have to go after people with really minimal benefit who might have been more willing to have the device turned off, but that's not the population that would be appropriate for discontinuation trial. Next slide, please.

DR. ELLENBERG: If you don't mind, can I interrupt?

DR. RUSH: Sure. Yes, sir. Please go back.

DR. ELLENBERG: In the first bullet.

DR. RUSH: Yes.

DR. ELLENBERG: If I remember correctly, there were a significant number of meds changes in D02.

DR. RUSH: That's correct.

DR. ELLENBERG: So --

DR. RUSH: And a significant number in D04, which I think --

DR. ELLENBERG: Understood.

DR. RUSH: Take D04 as evidence that you couldn't do this with medications not changing.

DR. ELLENBERG: Well, that's what I want to follow up on.

DR. RUSH: Okay.

DR. ELLENBERG: So, the protocol as I recall, once the acute phase was finished allowed but discouraged medication changes in D02. And with that allowance you got, I think it was 60 percent medication changes. So in real life, that probably is going to be the way VNS will be given, with the allowance for medication changes.

So why then -- I understand your point based on the assumption that the cleanest way to do this is with no medication changes. But why couldn't a randomized trial be done where medication changes were discouraged but allowed, since effectively for the long-term results in D02 that's what happened? So why couldn't that trial be redone with the medication change discouraged but allowed?

DR. RUSH: Oh, I think at this moment it could. I think that the data that we now have in terms of long-term safety, which we didn't have at the end of the D02 acute. We still didn't have long-term safety data. We had a good feeling from epilepsy, and some modest sample size from the D01, but we really didn't know and couldn't tell the patients what the longer term safety risks were. So we, for example, have noticed the induction of mania, or hypomania.

DR. RUDOLPH: I think Dr. Ellenberg, you've essentially described the D02/D04 paradigm, except as a randomized study.

DR. RUSH: Exactly.

DR. RUDOLPH: I think that's what you're suggesting.

DR. RUSH: Let me just go through two more, because actually the last one deals just with what you're asking.

We thought of another possibility, which is we would prescribe -- we would control the medications in a reasonable standard of care. So we'd have less deviation across doctors. So we wouldn't have too much exotic, potentially dangerous pharmacotherapy. And both -- then one group would get algorithm only, and the other group would get algorithm plus VNS. That's terrific, except when you look at the level of resistance in these patients. Nobody knows how to write an algorithm for any of them, because they're at level 6 and the algorithms -- that is that half are beyond ECT. So we have no evidence with which to write the algorithms.

And secondly, because of the huge number of treatments that have been tried. I mean, some of these patients -- remember, if it's two to three ATHF treatments, it's exposures to 12 different treatments, not counting combinations that were used. And then when you go to four to five, and remember the average here is four, you're looking at 16 clinical treatments, not counting all the potential combinations that you layer on one drug after the other after the other.

So in order to do that experiment, which would be very interesting, the only suitable population is one that's much less resistant, one that's maybe had one or two steps. Then we could take three more steps in an algorithm developed by consensus. That could be done. That's not the population that's going to be the primary target for this treatment. And so the relevance of the research is limited due to generalizability issues.

And then the last point is the point that you were just raising in the question. Could we do such a trial, that is randomize what we've done now? Yes. Now that we know long-term safety, I think it's possible to do such a trial. You will still have interaction with medication changes that are likely to differ between the two groups, if there's an effect of VNS, of course. And the other risk which is always true with this level of depression is both groups will still be exposed to exotic, unstudied combinations. And finally, at the time that this was done -- we could do it now -- we had no idea of effect sizes. We wouldn't know how large to make the sample to figure out whether we have a difference.

So, last slide please.

DR. RUDOLPH: Can I -- before you go on just comment on that?

DR. RUSH: Yes.

DR. RUDOLPH: So while that's feasible, our assessment was there wasn't much gain to do that over the strategy we followed. Because you would still be faced with trying to establish the comparability of medications, as you still had that major issue. Yes, the randomization would give you a greater level of certainty with regard to baseline covariates. You'd still have, even in a randomized control -- we've heard a lot of talk about unmeasured covariates. Even in randomized control, that's still potential problem. So you're still left with that.

And frankly, with VNS therapy, you always have an issue of blinding. Blinding is never perfect with VNS because a good number of patients do perceive when they're getting stimulation. So yes, maybe there's a slight gain to that particular paradigm, but it's only a slight gain I would say.

DR. RUSH: Let me just -- I know we're getting near lunch period, so let me just finish if I could, this slide and the very last one. So, this is in brief of course what we have. And put the next one up for me to see. No, no, the other one. Back, please. Okay.

So what we have done is exactly everything short of the randomization, comparing -- trying to find out what is the long-term outcome of TRD at this level given standard care. Remember when we started this was never defined. This was not known. So we had no idea whether these patients are getting better, getting worse. Obviously D04 shows really not much benefit. So that's the first study that's available to comment on this. That's D04, not yet published.

D02, then as you know, simply was extended with doctor's choice. Again, the first study of long-term adjunctive VNS establishes safety. So now, we now have the effect size and safety and expected outcome. So clearly at this moment we could do such a randomization, but up until this point we could not do that.

And as you know, the results, we've reviewed them, so I'm not going to detail that other than to indicate that most of the responders at three months were still there at a year, and some of the non-responders at baseline -- I'm sorry, at three months, actually responded by a year.

The one comment is this design actually provides a better control of longer term VNS than was originally agreed to with the FDA. Because it has a comparison. Not randomized, but a comparable group, if you will, by which to gauge the most salient portion of this treatment, which is the longer term outcome. And then if I have just the final slide.

So you are asking, in a sense, how do we deal with a non-randomized evidence base, and is it really safe and okay to support this treatment. So, this is sort of the way I put things together. The question is is VNS the cause of better outcomes in 02 versus 04. How certain are we. And Dr. Rudolph went through a number of explanations, and illuminated a placebo not likely in TRD, not a sustained effect, typically occurs early. Medication changes, yes there were more in 04 than 02. That would be consistent with more efficacy in 02 due to other causes. The EC differences were equivalent in terms of percentage use and unrelated largely to outcome.

Sample differences, we went through that at baseline. Unmeasured sample differences. Here we could look at anxiety. Anxiety subscales. We anticipate, based on the Hamilton, the anxiety subscale highly correlates with the total on the Hamilton, the two totals are the same and we'll probably have that number for you this afternoon.

The other issue is access to personality disorders. Personality disorders are very common in TRD. And in fact, many patients with TRD, once the TRD is fixed, actually have a resolution of these personality disorders. The incidence is on the order of 75 to 85 percent. And in fact, they have not been found to affect outcomes except for borderline personality disorder in Dr. Sackheim's ECT trial, and Tracy Shea back in 1990 reported the NIMH collaborative psychotherapy trial. Personality disorders did not affect symptomatic outcome. It did affect social functioning, but that was it.

The final thing is maybe there are differences that we don't know that we didn't measure. So I call them unknown sample differences. Now if they're present, they would have to explain the 12-month outcomes -- that's possible -- the fact that they're sustained largely -- that's very unusual in this group -- and the fact that we have an increasing number of beneficiaries in the VNS group over time. Whatever that thing is, we would have to create it out of whole cloth. So I feel, personally and scientifically, very persuaded by this level of data, which I must say is a bit short of the one-year, randomized control trial that we can now do based on all the evidence that we have. And the question becomes how necessary is it to establish with that level of certainty for this population in the context of this device for epilepsy already widely used.

DR. ELLENBERG: I think the last bullet is the key bullet on this. On that slide. I have a nave question and then I'll stand down. In terms of the masking of the IDS-SR or the HAM-D, looking on the Web, it seems that while that is a self-report, both of them are self-reports by the patient, there is someone that's working with them to complete the questionnaire. Someone asking the questions. Is that not correct?

DR. RUSH: No, sir. If you're asking about the IDS-SR?

DR. ELLENBERG: Yes.

DR. RUSH: No. Typically that's given to the patient and they're asked to fill it out in the waiting room.

DR. ELLENBERG: So they're doing that all alone. Alright, then that's the primary outcome.

DR. RUSH: Some of these patients are very depressed. And so we may have to explain some of the words. But that happens at just pretty much the baseline. Most of the patients are able to fill this out throughout on their own. They may ask a query, but generally IDS-SR is totally a self-report. This population needed a little bit of help.

The Hamilton is -- the interviewer you saw, Dr. Husain, doing that on one of the tapes. You know, that's a clinician interview.

DR. ELLENBERG: They were prompting, and helping them.

DR. RUSH: Exactly. And that was independently rated.

DR. ELLENBERG: Then going to just the IDS, the subjects, both the delayed onset and the immediate onset subjects in D02 were being measured by the IDS with let's assume it's full self-report. Those patients all knew that they were turned on.

DR. RUSH: Well, you're asking the degree of blindness with regard to the subjects.

DR. ELLENBERG: Correct.

DR. RUSH: And we purposely did not ask the patients whether they thought they were receiving active treatment or not.

DR. ELLENBERG: In the long-term, after the three-month period?

DR. RUSH: Well, you heard --

DR. ELLENBERG: Did the patients know that they were going on?

DR. RUSH: Not necessarily. Not necessarily.

DR. RUDOLPH: After three months.

DR. RUSH: Not after the acute they didn't. They knew there was three months and three months, no? I'm sorry. Let me --

DR. ELLENBERG: My impression was --

DR. RUDOLPH: After three months considered the study to be un-blinded.

DR. ELLENBERG: Yes. Okay. So the outcomes we're looking at, in addition to the repeated measures, which included the three-month period, but most of the data was coming from a point in time where all patients knew that they were on treatment in the D02. In the D04, all of the patients knew that they were not on VNS.

DR. RUSH: Yes. Correct.

DR. ELLENBERG: Could you assess for us what impact that might have had in terms of one group knowing that they weren't on treatment and the other group knowing that they were?

DR. RUSH: It is true the D04 patients knew they were not getting VNS, and the majority actually weren't offered VNS. So some, as pointed out by the report, were eligible for both and might have had a discussion with both. Most of them when there was a choice were offering D02. So the knowledge that they didn't get VNS may not even be in the heads of people getting D04, but they know that was just routine care.

They are having their medications changed. They're under still expert care. And so the medication changes, they would have an anticipation, I would think, as any patient would, of making a change that might work out for the better, but they're often very discouraged because many other medication changes have not worked. If the fact that the patients receiving VNS, they say this is terrific, I can feel it, I know, I'm un-blinded, hooray, I'm in the right ballpark because I have nothing else left, you would have to say that that is a -- if that's the placebo response rather than the VNS, right, because I'm aware of it. It would have to last for another nine months and it would have to be characterized by a sustained benefit. Okay? We don't know of any placebo that does that.

DR. ELLENBERG: I'm not sure I understand why it's a placebo effect. If the group for nine months of the year in which they're followed knew that they were on the VNS.

DR. RUSH: Right. You're asking couldn't that have been cause for the difference between the groups.

DR. ELLENBERG: Have caused their response -- yes.

DR. RUSH: Right. And I'm saying it would either be the active VNS itself, or you'd have to say it's non-specific effects. And I'm saying I don't think it's non-specific because of the magnitude and the growth over time. So I'd have to attribute it to the active VNS itself.

DR. RUDOLPH: Additionally, you have clinician ratings being done at the same time, and the clinician ratings were videotaped and sent out to a third party blinded rater which established the legitimacy, if you will, of the clinician ratings.

DR. ELLENBERG: That was done both with the 04?

DR. RUDOLPH: That was done for D02.

DR. ELLENBERG: D02.

DR. RUSH: Just one other comment on the IDS and I'll get out of here. That would avoid more questions. The IDS is a self-report. And while it is highly related and has been published in a number of articles, the psychometric properties, it's highly related to the Hamilton and to the clinician-rated IDS. It was accepted as an outcome measure off-the-shelf, ready-made, at least the IDSC was, the clinician rating at an NIMH conference held a couple of years ago, with the IDS-SR. If a self-report were acceptable to the FDA, that would be an acceptable metric.

One comment, though, that we know, especially in chronic, longstanding depression, self-report tends to be a little more sluggish to change than clinician ratings. And you heard from the patients, a couple this morning. My family members saw me change before I knew I was getting better. That is very common, especially when you're looking at multifunction impairment over a long time. We know we're better when we can do things we love to do. And we can't do those right away. The symptom changes by the clinician rating will happen a little quicker. So let me get out of here.

DR. SACKHEIM: Hi, I'm Harold Sackheim, a professor of psychiatry and radiology at Columbia. Just two comments. I believe that part of what you may be getting at, Dr. Ellenberg, is the idea that patient expectancies can have an impact on outcomes. In this population, I know of only -- or a population that's relatively similar -- only one attempt to look at the relationships between expectancies and outcome. And that's what we've done over the years in terms of asking ECT patients whether they expect to get better, whether they expect to have cognitive effects, and so on. In that population, there is no association between patients' beliefs about whether the treatment is going to work for them or not and the final scores.

And I think that would be pretty much comparable here given the history of these patients in terms of repeated failures with other treatments. In fact, the correlations tend to be negative. Those patients who are most pessimistic tend to do better with the treatment. So that's one point.

Then the second point regarding the validity or the bias that could've entered into the ratings by the raters at the individual sites. It was at Columbia that we did the blinded ratings of these tapes. And it was time blind as well. So the people that were rating the tapes didn't know when in the course of treatment -- could I have the slide please -- they were rating. And these are the ICCs for the ratings by site. And as you can see, the ICCs are not bad at all, particularly since there's variability in the ends at each site.

Overall in this study, the reliability of the Hamilton ratings -- we're talking about an assessment of 400 different interviews -- was on the order of 0.93 was the overall ICC. Can I have the next slide, please?

To give you a sense of how the tight the ratings are, but even more importantly, the absence of a bias in the ratings. You see that the intercept for this regression is essentially zero. That the blinded rater did not see these patients as either more or less well on average than did the ratings at the site. And that there's a very tight association. These are the individual points for each of the interviews that were rated.

DR. ELLENBERG: So this is the clinical taped evaluation?

DR. SACKHEIM: On the Y axis.

DR. ELLENBERG: Versus the HAM-D or the IDS?

DR. SACKHEIM: No, this is the HAM-D where it's a blinded rater at Columbia rating it versus the original rating at the site. And these are each of the individual ratings, where the rater was time blind, not knowing when in the course of treatment, whether this was baseline, or after a year, or what have you.

And then just to echo the remarks of Dr. Rush. What at least to me is very unusual, having spent many, many years working with treatment-resistant patients, is the fact that not only is there a group of patients who showed benefit late, but the people who benefited on average by my estimate, we're talking about 70 percent of them, holding it for at least a year or two years. Now, when we contrast that with the relapse rates that we see with other treatments, particularly in treatment-resistant patients. And so can I have this slide, please.

These are the percent of patients in an analysis that we've completed of D01 and D02 who were responders at three months, who would continue to be classified as responders at one year. And you can see that it's 72 percent in D01 and 63 percent in D02. If we look at the --

DR. RUDOLPH: And let me just interrupt. The reason why these percentages are a little different than I showed this morning is because it's a different set of criteria.

DR. SACKHEIM: This is, in fact, a more conservative set of criteria. We're not allowing in the 25 to 49 percent, but looking at people who were 50 percent and above at three months. And we allowed them wiggle room down to 40 percent. So how many of those who were at least 50 percent and above had at least 40 percent improvement at one year. And you can see those figures there. And obviously we do have the data now on the longer term. And it's very, very promising.

CHAIRPERSON BECKER: I think with that we'll hold any further questions or comments till after lunch. So if everybody could return in an hour, let's say at 1:45, we'll pick up with the deliberations.

(Whereupon, the foregoing matter went off the record at 12:48 p.m. and went back on the record at 1:48 p.m.)

CHAIRPERSON BECKER: It's now 1:50 and I'd like to call the meeting back to order.

And I'd like to remind the public again that while the meeting is open for public observation, public attendees may not participate except at the specific request of the panel.

We'll now begin the panel deliberations. Two voting members of this panel will open this part of the meeting with their remarks.

Dr. Philip Wang will give his remarks. Dr. Wang will give his remarks on the clinical information.

And Dr. Ellenberg was going to address the statistical analysis but I understand he's in agreement with the FDA analysis and will not be making any specific comments.

After this, the panel will have a general discussion at which the panel will focus their deliberations on the FDA questions.

There will then be a second public opening hearing and FDA and sponsor summations.

Then the panel will conclude the deliberations and vote on the recommendation concerning this PMA.

I want to remind the panel that they can

ask the sponsor or the FDA questions at any time.

So at this point, I'd like to ask Dr. Wang to take the microphone and open this part of the panel's deliberations.

MEMBER WANG: That's perfect, thanks. Great. I'm going to be brief because everything I have to say you've already heard.

Again, just to recap, this is a pre-marketing supplement application for a new indication for VNS.

It was approved in `97, again for -- as an adjunctive therapy to reduce the frequency of seizures in patients who are refractory to anti-epileptics.

Now it's being considered for approval for a new indication as an adjunct long-term treatment for chronic or recurrent major depression that hasn't adequately responded to two or more treatments.

Again, it's similar to the previously approved device. It has an implantable pulse generator, a lead to the left vagus, an external programming wand, programming software and a compatible computer to run the software.

The VNS clinical studies include these two which are perhaps the most relevant to our discussions today.

There's the D-02 Pivotal Study which had two phases, a 12-week acute phase in which 235 patients with chronic or recurrent treatment-resistant major depressive episodes were implanted and then randomized to either receive the stimulation or sham-control treatment.

This was then followed for both arms by a 12-month long-term phase in which all patients were then given VNS.

Also of relevance, perhaps most relevant to our discussion today is this D-04 Observational Study in which 138 patients with chronic or recurrent treatment-resistant major depressive episodes were given usual care and then followed for a year.

Other studies we've heard about, the D-01 Pilot Study that originally got things going, a D-03 European Post-Marketing Study, D-05 Videotape Study in which there was an examination of the inter-rater reliability of depression assessments in the D-02 study, and the D-06 Pilot Study of not necessarily depressive episodes but rapid-cycling bipolar disorder.

In terms of the D-02 acute phase results, this now is again the first 10-week portion of the D-02 study. The difference on the primary end point, as defined by a 50 percent reduction in HAM-D scores, again was -- there was a tendency but that wasn't statistically significant to favor VNS.

The sponsor has brought up that their full VNS effect may take longer than the ten weeks in the acute phase portion of the trial, especially considering that the first two weeks of that ten-week period was for adjustment of the VNS device.

So this was the rationale for the D-02/D-04 12-month comparison, which we heard about, in which outcomes during the long-term phase, in which everyone in D-02 was receiving VNS was compared to the 12-month outcomes in the D-04 study.

On this primary endpoint analysis, which again it was the change in monthly IDS-SR scores from a repeated measures linear regression model, on this primary endpoint it was statistically significant in favor of the VNS therapy.

There were also secondary endpoints in this D-02/D-04 comparison and there was also statistically significant differences in favor of VNS. This included an analysis of responders, as defined by 50 percent reduction in IDS scores, complete remission as defined by having less than a 14 on the IDS. That's actually not greater, it should be less than.

Also in terms of an analysis of responders as measured through 50 percent reduction in HAM-D scores and also an analysis of complete remission as defined by -- that's less than nine on the HAM-D.

Issues, however, were raised in the FDA review in terms of this D-02/D-04 comparison. Principal among these are the patient and disease characteristics may have differed between subjects who were in the D-02 and the D-04 study due to the fact that it was not a randomized trial.

There was a propensity score adjustment analysis that was done in terms of the primary endpoint analysis to reduce potential bias. However, as was brought up in the FDA review, there's still confounding possible in this primary endpoint analysis due to the possibility of unmeasured variables. And I added there also poorly measured variables.

Also brought up was the fact that the simple comparisons are proportions as was done in the secondary analyses. Secondary endpoint analyses were completely unadjusted. There were no potential confounders controlled for.

The other issues brought up by the FDA include placebo effects, which may have been greater in D-02 than in D-04 due to the higher expectations. D-02 was known to be an intervention study while D-04 was billed and known to be only a control observational study.

There were also allowed changes in concomitant therapies including antidepressant drugs and ECT, which may have potentially altered the apparent efficacy of VNS.

And there was an analysis presented in which patients were censored if they added or increased treatments. And this reduced the observed effects of VNS not only in terms of the statistical significance but as you can see also, in terms of the observed effect size.

In terms of safety issues of VNS, there are issues that make it difficult to assess the safety of VNS for depression. The primary reason for this is that the safety data were not systematically collected in D-04. So there's no comparison group for the D-02 long-term phase data.

Another difficult when just looking at the acute phase data of D-02 is that as was brought up, the treated end-sham groups both received a lead that was attached -- they received the implantation and a lead was attached to their vagus nerve so the adverse event rates do reflect that.

Just looking at the incidence of adverse events in D-02 is also difficult to interpret due to the fact that there are non-negligible background rates of many of the adverse events examined. And I just put down there reproduced the cardiovascular events as were seen in the D-02 study by phase.

There were three deaths in D-02. One was a sudden death and considered to be possibly related. It was the only one of the three that was considered to be possibly related to VNS but unfortunately no autopsy was done so it's hard to conclude much about that.

As was raised, the safety data from epilepsy studies may be informative here. What I've reproduced here is just the treatment emergent adverse events in a trial that was done of VNS in epilepsy patients, E-05. And what's shown here is the adverse events that occurred in greater than ten percent of subjects and were statistically significantly different between the baseline and the treatment phases of the study.

I think what's potentially reassuring here is that 99 percent of the side effects were rated as mild or moderate.

In terms of the long-term safety of VNS, there is some data. And that, again, comes from this E-05 trial of VNS in epilepsy patients in which after five years, 31 of 51 patients were still -- still had their VNS therapy system implanted and were presumably receiving stimulation.

And as you can see here, these are -- here's the profile of adverse events that were found. And, again, providing some reassurance is the fact that this profile of long-term side effects is generally similar to the adverse events that were reported during the E-05 trial.

I raise here, just sort of in closing, a few things. One relates -- there are some issues related to training that are worth considering. One is who should implant this device? Two is who should be programming this device? Should it be psychiatrists? Neurologists? Any physician?

Should there be mandatory training courses for these -- whoever is going to be doing the implanting and programming? And also what kind of guidance can be given on titrating the output current? And based on what data?

And finally there would be some implications of an approved device, which I've put up here. As was mentioned, there are currently very few options for treatment-resistant depression. Some options currently used in typical practice have very little evidence to support their efficacy or safety.

But on the other hand, would there potentially be unrealistic expectations. I remind everyone that in absolute terms, the apparent effect sizes seen for VNS were relatively small. Would there also be pressure to forego effective but stigmatized modalities such as ECT? So these are all some issues to ponder.

CHAIRPERSON BECKER: Thank you, Dr. Wang. Does anybody have any questions for Dr. Wang?

(No response.)

CHAIRPERSON BECKER: I guess if not then we'll move on to the general discussion portion of the panel's deliberations. At this time, the panel may ask the sponsor or the FDA any questions that they might have.

And I think in order to kind of start with a fairness to everybody in the panel, we'll just go around the panel and solicit questions and comments. And we'll start with Dr. Ellenberg at the end.

MEMBER ELLENBERG: I don't have any additional questions. Thank you.

MEMBER JAYAM-TROUTH: Hi. I had a couple of questions. One is I'm looking at the D-01 study and, you know, there were 25 out of 55 responders at one year and 18 of 42 responders at two years. My question is were the stimulation parameters equivalent, you know, to the D-02 study?

And was it a drop out of these patients, 55 patients to 42 patients? You know, why did the patients drop out if VNS was that effective?

MR. TARVER: I'm Brent Tarver. I'm a Senior Director in Clinical and Regulatory -- in Medical Affairs at Cyberonics.

About the stimulation settings, the settings were similar in D-01 and D-02. Out over the long term, the D-01 settings tended to be maybe a quarter of a milliamp higher but which would be very similar.

And the second part of the question?

MEMBER JAYAM-TROUTH: Why was there such a drop out from 55 to 42 in the responders? In, I mean, the groups, 25 to 55, they were responders. And then by two years, there were only 42 people still having VNS.

MR. TARVER: Right.

MEMBER JAYAM-TROUTH: You know, who dropped out?

MR. TARVER: Okay. The drop is in -- what you're looking at is the rating scores. There were a number of patients, about ten, that did not have the rating at two years but were still receiving stimulation in the study. So they're only included in the last observation carried forward analysis.

MEMBER JAYAM-TROUTH: So it wasn't a real actual drop?

MR. TARVER: Well, to some extent it was, and Dr. Rush, who was -- there were only four investigators in D-01. Dr. Rush can comment at least on the experience at his site.

DR. RUSH: Actually, I can do it for the whole study. I'm sorry, I should have brought the paper that had to be submitted.

As I recall, there were, I think, four or five people that had either had it turned off or were explanted. And that was because of lack of efficacy, not due to side effects, starting with the original 59.

There were probably, as mentioned, eight or ten patients we just couldn't get back in that particular time period. One of the interesting things is those individuals, the young man I presented this morning, the graduate student, is out in California. He doesn't like to deal with doctors any more because he's undepressed. So some of those individuals are not coming back because they're well. So that accounts for the large shift in -- the drop in the number.

But the vast majority, I think it would be fair to say 90 percent still had the device implanted, maybe two had it turned off.

MEMBER JAYAM-TROUTH: I see.

DR. RUSH: So it's a pretty high retention at two years.

MEMBER JAYAM-TROUTH: You know, along the same lines, you know, Dr. Rush, along the same lines my question is if I was a responder, you know, I would want to keep my, you know, device on --

DR. RUSH: Yes.

MEMBER JAYAM-TROUTH: -- you know, and therefore I should see a bias in greater numbers, you know, out of 42, I should have seen more numbers, you know, who were still continuing to be responders, not a drop from 25 to 18, you know?

DR. RUSH: Yes, but in fact what actually happens is it's not biased in that favor because we can talk to the patients on the phone. So some individuals, of course, are coming back hoping that we can readjust the parameters or their doctor just changed the medication and now with the parameters changes, maybe help them.

So there is a tendency to come back actually if you're a little more ill than if not. It's not just one way or the other.

MEMBER JAYAM-TROUTH: Is there a greater drop between the first year and the second year? I know this is a one-year study. And there was a steady improvement, you know, with the VNS. But was there a difference between the first year and the second year?

DR. RUDOLPH: Dr. Sackheim was one of the other four investigators in that first study.

DR. SACKHEIM: I think what would help would be to show you the data on the number of patients that sustained improvement from three months to one year and from three months to two years. So can we have that slide?

You'll see this separately for D-01 and for D-02. I'll just comment on it while we try to find the slide. This is uncharted territory because what percent of patients should maintain a response to say that a treatment has a persistent on-going benefit? Nobody has ever put a benchmark on that. Can we show this?

And what you see here is the D-01 where we actually are breaking down the patients into whether they had a 50 to 60 percent improvement, 60 to 80 percent, or 80 to 100 percent, or simply the total group in D-01.

And in this slide, we are looking at the percent that are showing sustained benefit over time. And we're doing it -- we should have had the markings on the bottom. But to the left, we're looking at the group that were responders at three months, and maintained it for one year.

You'll notice that D-04 is listed there as well. Of the patients who had a -- who were responders at three months in D-04, none were responders at one year. That was a small number of patients.

There was only seven patients that -- because D-04 actually obviously had poorer results. But none maintained it while we have a 70 percent rate of maintaining it in D-01 basically. And we're around 63 percent in D-02.

Now the next set of bars presents --

DR. WITTEN: Excuse me, can I just ask -- is this information that's in our file?

DR. RUDOLPH: Yes, I was going to say at the end of the comments for disclosure, it's not in your file. These are not even the sponsor --

DR. WITTEN: Okay, and we haven't --

DR. RUDOLPH: -- these are not even the sponsors' analyses.

DR. WITTEN: -- received these yet, okay.

DR. RUDOLPH: These are analyses that were done by Dr. Sackheim.

DR. SACKHEIM: The next set of analyses looks at patients who were responders at three months, had at least a 50 percent reduction in their Hamilton scores in this case, and then we're looking out at two years. Were they still responders?

And here, again, we're defining response as at least a 40 percent improvement so that we wouldn't punish the people who went from let's say 52 percent to 48 percent.

And again what you see is that we're, certainly with D-01 and D-02, in D-02 where over 70 percent of the people who were responders at three months held it at two years. And this to me was quite remarkable, unexpected really.

Then finally we have data on a very interesting group of people, the people who were not responders at three months but who at one year became responders, the late emerge. And those I would have thought would have been a very difficult group because they didn't benefit initially but benefitted later on.

Did they hold it? And so we then go from the end of the first year to the end of the second year. And what you can see is we're certainly above 50 percent for both the D-01 and the D-02 there as well.

So these, I thought, were very compelling data that the most treatment-resistant group of patients I think anyone has really ever studied, where we expect them if they benefit at all to lose it very rapidly, we're not seeing that rapid loss. Actually over a two-year period of time, we're seeing it sustained.

DR. RUDOLPH: So again, just to make it clear, these are Dr. Sackheim's own analyses of our data set. The sponsor hasn't seen them and they certainly have not been submitted to the Agency.

MEMBER JAYAM-TROUTH: Can I ask one more question?

CHAIRPERSON BECKER: Sure.

MEMBER JAYAM-TROUTH: The -- I understand that you kind of stimulated the -- in the D-02 group, not the sham, but the ones who had the stimulation on over a two-week period. Now did these patients come in every single day and you kept increasing their milliamps? The parameters?

You know, how did you do that? I mean was it rapid stimulation that you did? And then you held it at one spot? Or you continued to change the parameters during the acute phase and during the long-term phase?

DR. RUDOLPH: During the acute phase for those patients who were assigned or randomized to the active treatment group, their adjustments were all done in the first two weeks after the device was turned on.

MEMBER JAYAM-TROUTH: Every single day they would come and you would increase it?

DR. RUDOLPH: The protocol gave wide latitude to the investigators. I think -- and some of the investigators can correct me if I'm wrong. I think, in general, however, the patients were only seen at weekly intervals because that is what one is used to doing when they're conducting a clinical trial.

So although they had latitude to see the patients more frequently, I don't think practically that actually happened.

Then during the long-term extension phase, anybody that had entered that phase could have stimulation parameters adjusted at the investigators' discretion.

MEMBER JAYAM-TROUTH: Into the long-term phase?

DR. RUDOLPH: All throughout the long-term phase --

MEMBER JAYAM-TROUTH: A throughout --

DR. RUDOLPH: -- all throughout it.

MEMBER JAYAM-TROUTH: -- so there were variations in what parameters they had?

DR. RUDOLPH: That's correct.

MEMBER JAYAM-TROUTH: Was there a difference in the non-responders and the responders?

DR. RUDOLPH: No, there was not.

MEMBER JAYAM-TROUTH: So they also had stretched to the limit of tolerability --

DR. RUDOLPH: Yes.

MEMBER JAYAM-TROUTH: -- and they didn't respond.

DR. RUDOLPH: In fact, as you know, as you are probably aware, what happens in trial where an investigator is given sort of free range, if you will, to make adjustments, the non-responders are often those that are receiving the higher doses because they're not getting better and they're constantly pushed up in dose.

But there was no distinction between the responders and the non-responders in terms of their settings.

DR. RUSH: I just wanted to make it clear that in both the D-01 and D-02, there was only a two-week period during which the parameters could be adjusted. And after that, they were fixed for the entire duration of the trial.

MEMBER JAYAM-TROUTH: Okay.

DR. RUSH: So it's a very truncated opportunity to make the adjustments. And in that context, with weekly visits, while we could send the patient away for a few hours and come back to see if we can adjust further up in terms of current, it's a truncated time period for adaptation and a truncated opportunity.

So we may have, in a sense, injured ourselves in terms of what might really have been more effective.

MEMBER JAYAM-TROUTH: So for the whole year, they didn't change after that?

DR. RUSH: No, no, no. I'm talking about just the acute trial.

MEMBER JAYAM-TROUTH: Acute phase.

DR. RUSH: Just the acute phase, yes. Then after the acute phase, then adjustments can be made.

MEMBER JAYAM-TROUTH: Then adjustments -- and for the sham people, they were also adjusted over two weeks? And then they --

DR. RUDOLPH: Once the sham people crossed over --

MEMBER JAYAM-TROUTH: Yes.

DR. RUDOLPH: -- into an acute phase, if you will, following the acute study, they went through the same procedures as the original VNS group did.

MEMBER JAYAM-TROUTH: Okay.

DR. RUDOLPH: And then --

MEMBER JAYAM-TROUTH: And then they could be adjusted?

DR. RUDOLPH: Well, they could be adjusted over two weeks first. And then they had to have another ten -- another eight weeks at a fixed dose range increment. And then beyond that, then they were also available or they were allowed to have unlimited changes at the investigators' discretion.

MEMBER JAYAM-TROUTH: Okay.

CHAIRPERSON BECKER: And actually before Dr. Fochtmann asks her questions or makes comments, I just want to follow up on this particular issue.

Do you have data on how many adjustments were made in each patient? And I guess in particular, I'm interested in how the different number of adjustments would correspond to the different changes in their medications as well. Is that data available?

DR. RUDOLPH: I don't believe so. No, no it's not.

CHAIRPERSON BECKER: And on Slide 56 of the sponsor's presentation, it gives the time course of response in D-02 and D-04. And it looks like the biggest benefit is actually very acute on that non-adjusted graph. And I was wondering if you could comment on that.

Since the efficacy is really at the long term, why is it that the benefit in this analysis appears to actually happen straight away?

DR. RUDOLPH: It's true that a good deal of the benefit occurs early as is the typical pattern in drug trials over a shorter period of time. And just as a reference point, could we have E-151? So that's typical of trials in general.

It's also true in this case, and I'll show that in a minute. It's also true that the separation between the D-02 and D-04 groups continued to grow after that initial phase. Slide on please.

So this is actually taken from a real drug trial. We washed it of the identifiers as to what drug it was. But this is a very typical pattern that you see in an antidepressant drug trial. So most of the drop does occur early.

And then where you really start to get statistically significant separation is usually at the later part when the placebo group will start to flatten out and the drug continues to improve.

I think what's -- as we've been -- as we're probably sounding like a broken record but what we've tried to underscore throughout the whole morning is -- what's probably really important here is the sustained nature of the response, which is what you don't -- wouldn't expect in this group without an effective treatment.

CHAIRPERSON BECKER: Well, I guess that one could make the argument that they're getting some kind of augmentation on the placebo response because they're coming in for frequent readjustments of their stimulation.

Having said that, we'll as Dr. Fochtmann if she has any questions.

MEMBER FOCHTMANN: I have a number of questions.

First of all, to follow up on the questions about the stimulus settings, you said that there was no difference between the responders and non-responders in the stimulus settings. Was that just for the acute trial? Or was that also analyzed for the open phase, open label phase of the followup trial?

DR. RUDOLPH: Let me check with the clinical team. Doctor? You want to come up?

DR. BRANNAN: Hi, I'm Steve Brannan. I'm one of the Medical Directors at Cyberonics. The -- let's see, both for the acute and for the long term, the parameter settings did not differ between the responders and the non-responders.

When you have, just at the highest settings, there was a slight preponderance of people who were non-responders who had some of the higher settings.

And again, without having your kind of trial set up to be looking at a fixed dose, then that ends up the people who are necessarily doing very well. And so they're increased on their parameters.

So you actually see that better in the long term than you do in the acute phase where their parameters, again, were set after the first two weeks and so no adjustment could be made after those first two weeks of stimulation.

MEMBER FOCHTMANN: Was there any dose response relationship to the observation of adverse effects?

DR. BRANNAN: The only adverse event I am aware of would be the voice alteration that you have.

MEMBER FOCHTMANN: Okay.

The next question I had related to the proposed indication and there were two aspects of it that I wanted to inquire about.

You specifically defined chronic or recurrent depression as a current major depressive episode that is of at least two years in duration or a current major depressive episode in a patient with a history of multiple prior episodes of depression.

While those could be overlapping groups of individuals, they do seem to be also in some instances very discreet groups of individuals. And I wondered whether you had analyzed the data to determine whether you had comparable efficacy in individuals who just had a chronic episode lasting more than two years versus individuals who had shorter lasting episodes but who had had multiple recurring episodes?

DR. RUDOLPH: Yes, well, we did. I presented a slide this morning where we did it in the form of an exploratory analysis. And on the slide, if we can pull it up from my presentation, it focused on the chronic group only.

And in the chronic group, which was -- and I'm talking about the D-02/D-04 comparison, in the chronic group, which was again two-thirds -- next slide -- no, not that one. The next slide. Yes, there we go.

In the chronic group, which was two-thirds of the patient population for D-02 and D-04, and these again were patients defined as having an episode lasting at least two years of duration, and see the overall response rate in D-02 is 29 percent, 10 percent for D-04, and if you'll recall, and this is from the Hamilton, if you'll recall for the entire sample set, it was 30 versus 13 percent.

So by extension, you can figure that the recurrent patients had approximately the same rates.

MEMBER FOCHTMANN: Okay.

DR. RUDOLPH: So the bottom line is --

MEMBER FOCHTMANN: But did you analyze it specifically for the recurrent --

DR. RUDOLPH: No, we didn't.

MEMBER FOCHTMANN: Okay.

DR. RUDOLPH: But, you know, I think you can probably make that -- you can surmise that by subtracting out those patients. It's not going to differ that much. And so I think we can conclude that the overall rates of response, and particularly the difference between D-02 and D-04, are similar whether we're looking just at the recurrent patients or the chronic patients.

MEMBER FOCHTMANN: Okay. I was curious in terms of -- from a statistical standpoint whether you had been able to demonstrate efficacy in both of those subgroups of patients that you were including in the indication statement.

DR. RUDOLPH: We did not do a statistical test on this because it was an exploratory analysis.

MEMBER FOCHTMANN: Okay.

The second question that I had about the other definition in the indication statement was related to the definition of failed adequate treatment which mentioned a failure to respond to electroconvulsive therapy or an established antidepressant drug administered at an adequate dose for an adequate duration.

And my question there was whether electroconvulsive therapy adequacy was assessed either in terms of stimulus titration methodologies or in terms of electrode placement or in terms of numbers of treatments or any of the usual ways of assessing ECT adequacy.

DR. RUDOLPH: I'll ask Dr. Sackheim to answer that because the standard scale that we used for assessing adequacy was the antidepressant treatment history form.

DR. SACKHEIM: Right. On the -- I'm Harold Sackheim. On the antidepressant treatment history form, ECT is one of the treatments that is assessed in terms of whether or not the patient has had an adequate trial.

There is a slight bias on that form. One can have a higher rating for bilateral than for unilateral. We don't believe that we can determine usually retrospectively whether a stimulus test titration was used or what the dosage settings were or the adequate seizures.

We try and get that information when it's possible but it is based primarily on the number of treatments with the threshold number of treatments to be considered adequate and they differ for unilateral and bilateral.

MEMBER FOCHTMANN: Okay. Was -- I guess the follow up to that question would be is there a reason that the indication did not include failure to respond to an adequate trial of electroconvulsive therapy since adequacy was mentioned in terms of the antidepressant treatments.

DR. RUDOLPH: The reason the proposed indication was written as it was is because we were trying to parallel as close as possible the inclusion and exclusion criteria in the D-02 protocol.

MEMBER FOCHTMANN: Okay. The next question that I had was related to the other treatments that were used.

You censored the data based on a change in antidepressant treatment and yet as has been pointed out, the individuals in these trials are on multiple medications in addition to the antidepressants.

And one of the reasons for not changing the medications was presumably that any of these medications that they were on would have some potential to be of help in managing their depressive disorder.

And so even though these weren't specific antidepressants per se, I wondered whether you had looked at the effects of changes in other medications that people might be on be they antipsychotic medications, benzodiazepines, stimulant medications.

Some of the characteristics that at least appeared to perhaps be different in the frequency of use between the D-02 and the D-04 subgroups.

DR. RUDOLPH: I think and I'll ask for some confirmation. I think that the censoring actually encompassed more than just traditional antidepressants. It was -- we encompassed -- tried to encompass a whole group of medications that would be used to treat mood disorders.

DR. BRANNAN: It is certainly true that most of these patients were on concomitant medications. It is also true that when we looked at this, and we looked at it very carefully in a number of ways that the data do not show that they had an effect on outcome.

Can I have the slide with Responders/Non-Responders? Slide up please. Thank you. This one.

All right. When looking at this, and I think something like this was already shown when Dr. Rudolph initially had his presentation, one would think if you were having additional medications, you know, just did that benefit who was getting well?

And the answer, and I'll show you two ways, this is the first. It is clearly not. When you just concentrate on the two columns on the left and in the middle, the responders actually had fewer changes in the ARR than did the non-responders.

And again, similar to the parameter settings that we were just talking about, part of this is when people are not doing well over the course of year, then they're going to have more changes than those people who are actually doing better.

Can I see the next slide please? Another way of looking at that is to look at the patients who had ARR scores who increased or showed no increase in their ARR score.

And what you see clearly is is the people with no increase in their ARR score actually had a 51 percent response. So quite a bit better than the group as a whole. And so those with increases you actually see did less well overall.

But in addition to this, there are other parameters that are kind of hard to quantify. How do you know exactly? Do we have the slide of the medications that they were using during the one-year?

MEMBER FOCHTMANN: When you classified antidepressant medication changes, are you including only antidepressants per se?

DR. BRANNAN: No. I think that --

MEMBER FOCHTMANN: Are you including benzodiazepines, antipsychotics, stimulant medications, any of the medications that they were on that would be a psychotropic medication are included in the censoring?

DR. RUDOLPH: I apologize. And I think that was the core of your original question.

MEMBER FOCHTMANN: Right.

DR. RUDOLPH: And we are encompassing in there more medications than just standard antidepressant. It does include atypical antipsychotics, stimulants, benzodiazepines. They're all captured through the use of the ATHF form.

DR. SACKHEIM: Just to clarify what some of the difficulty may be, a benzodiazepine could never, as a treatment alone, be considered adequate in the treatment of depression.

MEMBER FOCHTMANN: Right.

DR. SACKHEIM: And so it would not, ATHF criteria for what constitutes adequacy are quite different than the information it captures on all the psychotropics. So we would capture the atypicals and all the others that were mentioned. But there is a real distinction between what can reach the level of adequacy.

CHAIRPERSON BECKER: Okay. We're going to continue to go around that table. Dr. Wang, do you have any questions?

DR. BRANNAN: We actually did have a slide on the atypical antipsychotics --

CHAIRPERSON BECKER: Okay.

DR. BRANNAN: -- if that would be helpful. E-164.

As I'm sure some of the panel members are aware, atypical antipsychotics in combination with other antidepressants are thought potentially to be something that might be very useful for patients who are not responding so as augmenting agents.

Slide on please. We do see, and I had a slide looking at all sorts of medications but this will kind of show you for the antipsychotics. There is a difference in the percentage of patients in D-02 who had atypical antipsychotics, 47 percent in D-02 and 32 percent in D-04.

What you see in the D-02 patients, so those 47 percent, they had a response rate of 26 percent as opposed to 30 percent as measured by the Hamilton Depression Rating Scale.

So there was no benefit from the -- in fact a slight decrease for patients in D-02 who were on an atypical antipsychotic.

Of the 32 percent of the patients in D-04, 18 percent had a response, which actually is slightly better than the 13 percent that they had. So the response is somewhat higher than the overall response.

So in terms of does this create a little bit of a confound, the answer is yes. But it makes the difference between the groups appear more robust in that this actually gives more advantage to D-04 and less advantage to D-02. So it should hide the difference.

DR. RUDOLPH: And the antipsychotics, I might add, were the only category of medication where there was really a substantial difference in usage between the D-02 and D-04 patients.

CHAIRPERSON BECKER: Dr. Wang?

MEMBER WANG: Yes, I'd like to go back to the FDA's major concern about residual confounding, you know, in the D-02/D-04 comparison that used propensity score adjustment.

Dr. Rudolph, you mentioned that you didn't think that the distribution of unmeasured variables might be equally distributed because of the large size of D-02 and D-04? How is that possible since that principle only really applies if you are randomizing?

DR. RUDOLPH: Randomization certainly gives you that assurance but also we made that evaluation because the sample sizes were so large here and the measured covariates were so equally distributed that I think it's a reasonable assumption, not -- obviously not proven, but a reasonable assumption that for unmeasured covariates, it would probably be equally distributed as well.

Obviously there is an assumption there but we -- Dr. Davis has some information that I think bears on this question that you'll find interesting.

DR. DAVIS: Hi. I'm Sonia Davis, Director of Biostatistics at Clean Tiles.

First, as Dr. Rudolph had described before, we evaluated through the propensity score the differences for the covariates that we had measured at baseline.

Slide up please. And this is a list of all the 17. This happens to be the input that each of the 17 had on the propensity score but of all of these 17, only four of them were significantly different between the groups at baseline. Two of them were quite related to each other, ECT and lifetime, or in the current episode. So we used these parameters to incorporate the propensity score.

Next slide please. As Dr. Rudolph already presented, when we used the propensity score to adjust for our primary efficacy model, on the top row, you can see, in the top right box, did not have a significant impact on the changes over time in the IDS scores. So this tells us that there were very minimal differences between the groups even when we combined them all together.

Even down -- looking at the very bottom row of this table, if we did not adjust for propensity scores at all, we still get a very strongly significant linear study effecting the primary parameter.

Now you might say well the propensity score maybe didn't catch all these parameters that were different at baseline. So is you could look at the next slide -- slide up -- we did an additional exploratory analysis that actually put each of these covariates -- there was a strong influence into the logistic regression and also for the propensity score and also were different at baseline to see are these covariates having an effect on our primary analysis with our linear study effect.

And in the first column we can see that after putting each of these predictors that were the strongest ones in there, we see there is essentially no differences to our linear study effect. We always had very consistent and very strong answer results.

So this led us certainly to conclude that everything that we measured, although we saw very small differences, had no impact in the difference between D-02 and D-04.

Switching now to what about the covariate that possibly had not been measured. Slide up please. In order to have a strong impact into the response rates that we saw over time, the unmeasured covariates would have to be correlated with response.

They would have to be in balance between the studies, otherwise we wouldn't have an impact.

They would have to be recurring in a reasonable number of patients in order to have an impact on response. If they were very small number of patients, it wouldn't have a meaningful impact on the overall group.

And they would need to be uncorrelated with all the other covariates that we have adjusted for.

And we feel that these four things are quite hard that we would reasonably expect that it's very, very unlikely there would be some unmeasured covariates that we did not measure.

And I'd like to turn it over to Dr. Brannan now if I could to talk about possible --

DR. BRANNAN: Sure. Just as an example, one of the unmeasured covariates talked about by the FDA had to do with thyroid disease. We did not measure that specifically but we did have an estimate for it. Slide up please. And that was thyroid medication use.

Now again, this is not going to tell us exactly whether people had thyroid or not because a fair number of these patients may be having this as thyroid augmentation which is one of the strategies one has for treatment-resistant depression. So again there is some appropriate caveats.

But what it is reassuring is when you look at the difference in thyroid medications either lifetime or in the current episode, you see it's just almost equally distributed between the two groups.

Kind of bearing up again those four measures are pretty hard to find something that's going to be -- have that imbalance between the groups, have a strong correlation with outcome, and even treatment doesn't have that strong a correlation with outcome in depression. And also not even be correlated at all with any of the covariates that are already there and measured.

DR. RUDOLPH: This might be an opportune time to correct something that you either heard from the FDA this morning or you saw in their review and that was the statement that there is no covariate adjustment in the secondary analysis. That was not true. There was covariate adjustment in the secondary analyses.

And let me just ask Dr. Davis to comment on that because I think it's a very important point. It's come up a couple times.

DR. DAVIS: Yes, this is Sonia Davis. It's true. All of our analyses for D-02 versus D-04 adjusted for the propensity score, for the baseline value whether it be Hamilton or the IDS, depending on what measures we were looking at for the outcome, and also for the pooled site.

So all analyses, not matter what we did, adjusted for those parameters.

MEMBER WANG: Did you model in the categorical analyses?

DR. DAVIS: That is correct. The categorical analyses were modeled with logistic regression. And I would have to point out that we did follow that up with an exact logistic regression due to relatively small sample sizes of our events.

If you looked at that exact logistic regression, because of the sample sizes, we could only adjust for one covariant. We adjusted for the propensity score in every case, our exact logistic regression adjusting for the propensity score gave very similar results to the logistic regression that adjusted for all the covariates.

And also our results were quite similar with the Fisher's Exact Test that Dr. Lao performed during his review.

MEMBER WANG: Actually, Dr. Davis, if you could -- probably I'm going to be asking this to you. I mean this issue of unmeasured confounders is, you know, we're probably not going to come to any sort of clear resolution because it would probably be sort of speculating.

But in terms of estimating whether there is residual confounding, one thing to wonder about is also how you categorized -- how you dealt with your propensity score. And I see you used -- you left it as quintals.

To sort of begin estimating whether there is residual confounding, did you try more categories? You know, even through it in as a continuous variable? And if so, did it change the results? Because I might --

DR. DAVIS: Yes, we did. And it did not. Slide T-50, T-50 please. Slide up. The middle row here shows the results of the primary model if we treated the propensity score as a continuous measure rather than the five levels.

The five levels that we used were what we already specified in our analysis plan but this row is an exploratory analysis, supportive analysis to confirm that results are not different.

MEMBER WANG: My last set of questions have to do with this concomitant treatment issue. In the D-02 analysis, I mean in the -- when people who had either adjustments to their medication or ECT added or dropped, the apparent benefit, you know, it not only became less significant but the actual effect size went down which -- there's a couple ways to interpret that.

You know, one potential way to interpret that is some of the apparent efficacy of, you know, VNS, may actually be attributable to the superior efficacy of the rescue treatment.

I'm wondering if you have some either data or some way to sort of reassure us that that is not happening.

DR. RUDOLPH: While Dr. Brannan is coming back up, I will remind you that the censoring analyses were done as sensitivity analyses. They're very -- we had available to us a number of methodologies. The methodology I presented this morning was actually the most conservative of the methodologies that we could envision.

And even with that methodology, as I presented this morning, although statistical significance wasn't achieved, it came awfully close at .052. The confidence -- the 95 percent confidence interval for that range from -.37 to zero. So it never crossed zero.

And one should bear in mind that it did truncate the VNS effect to about seven months. So the patients in that censored D-02 group did not have even the full benefit of VNS assuming that there is an accruing benefit over time.

DR. BRANNAN: Let me just ask again -- the point of your question again was?

MEMBER WANG: It's not really the significance issue.

DR. BRANNAN: Okay.

MEMBER WANG: It's the effect size.

DR. BRANNAN: Okay.

MEMBER WANG: You know when you censor --

DR. BRANNAN: Yes.

MEMBER WANG: -- the effect size reduces suggesting that, you know, maybe the rescue treatments, you know, ECT, maybe that's what has this appeared efficacy. Anyway, just sort of -- you partially answered it. Actually Dr. Rudolph partially answered the potential reasons why but --

DR. BRANNAN: Okay, as he mentioned, this was part of a number of sensitivity analyses. We actually looked at several things. Can I have the slide up a second?

So there's actually a lot of different ways to kind of do the censoring. The original is on the top. If you actually censor both, which is not necessarily very helpful to answer the question, there's still a huge difference.

One of the things that was attempted was not to censor D-04 until after the first three months. As you will recall, in the first three months, people were not supposed to have medications added in D-02. So that was done. And that actually does reduce what you saw from the original one some but it's still fairly large.

Then an attempt was done just to do a censoring without anything -- without carrying forward anything but treating everything as missing value. And that again had such a large value.

So you are correct. When we actually go to the extent of going and censoring at the first ARR change and then carrying that forward, LOCF fashion, which again truncates both the benefit in terms of whatever interaction or VNS effect that you're getting as well as shortening effectively how long they are in the trial, so on average, they're about seven months instead of twelve months, at that point then you do see this decrease.

And I think it's a good point that since this is a sensitivity analysis, what are the P values really talking about. But you still see actually a fairly substantial -- you do cut it down by about half. But it's still fairly different.

DR. WITTEN: Can I just request of the sponsor that you -- when you present these slides, just clarify which analyses were in the submission and which weren't -- for each -- just in general.

DR. RUDOLPH: Oh, I'm sorry.

DR. WITTEN: For each slide.

DR. RUDOLPH: Okay, for this one, slide back on, for this one, the original and then the bottom were done in the original submission. And then I believe only the second from the top was also submitted during some of the questions back and forth. Yes?

DR. DAVIS: Hi, this is Sonia Davis, I just want to add a summary about this to bring home this point. The analysis that we did with the LOCF censoring, which is the last bottom line here that was submitted, is an exceedingly conservative post hoc analysis that we did.

We expected it full well to be very conservative. We didn't present some of the others to the FDA because they are quite non-conservative. So the idea was for us to say under this very unusual situation where almost 50 percent of the time that people were on D-02 was taken away from the analysis.

Is there a difference between that and a full 12-months of the D-04 analysis? And we said yes, we still see a signal even in the most exceedingly conservative situation.

CHAIRPERSON BECKER: Dr. Jensen?

DR. JENSEN: My questions primarily have to do with safety. So if you have a safety person, want to stick him up?

First I noticed in the study that the wound infection rate was eight percent for the VNS group and two percent for the sham control. And you've mentioned only a 1.4 percent infection rate in your epilepsy study that required explantation. So were these patients, did they require explantation? Or were they treated only with antibiotics and they recovered from that?

And if not, if they required explantation, then why do you see such a substantial difference in infection rates?

DR. WINGARD: My name is Peggy Wingard. I'm one of the Medical Directors at Cyberonics.

In answer to your question about the difference in the rates of the infection, eight percent versus two percent -- is that the information that you are referring to?

DR. JENSEN: Well, that's in the depression study but you compared that to the infection rate in your epilepsy group, which I believe was 1.4 percent required explantation due to infection.

So it's not clear to me whether this eight and two percent required just explantation -- required explantation? Or if they were treated with antibiotics?

DR. WINGARD: I see. Okay. In answer to your question, there was only one patient that required explantation due to infection. And the rest of them were treated with antibiotics.

DR. JENSEN: And that is in the depression group?

DR. WINGARD: Yes, ma'am.

DR. JENSEN: Which means it is equivalent to what you saw in the epilepsy group? Is that true? In your -- because you were comparing the safety data between those patients that have had the device for epilepsy. And I just want to make sure that between the two groups, you're not seeing a difference in --

DR. RUDOLPH: Well, it actually results in a lower rate because it is one out of 235 implants.

DR. JENSEN: Okay.

In terms of the patients who do not respond and have a permanent implant, what is the company's position on how to cancel those patients as to what to do with that explant -- with that implant? Explant it or leave it in?

DR. RUDOLPH: You have two options. You can have it explanted. You can leave it in. If the device is left in, the general precautions that are communicated are those that are in our label regarding the risks of having this implant, permanent implant in, which have to do with MRI risk, full-body MRI, not head MRI but full-body MRI. And the risk of receiving diathermy.

DR. JENSEN: And the defibrillation risk? Anything? If they require defibrillation, that doesn't do anything?

DR. RUDOLPH: No.

DR. JENSEN: Okay. All right. So you've got a group of people that are relatively young now but they have an implant in. And chances are at some point in their lifetime, they will require an MR.

And furthermore, they are depression patients so they will probably at some point in time get a brain MR, possibly a high field strength MR, so is it -- to me it looks like you may be precluding other forms of either treatment or at least evaluation of patients if the implant remains in since they now cannot have an MR with that.

DR. RUDOLPH: You can still have head MR.

DR. JENSEN: Even high field strength? When we go to three Ts?

DR. RUDOLPH: We're going to -- we'll get an imagine expert up here to answer this. This is --

PROFESSOR GEORGE: Hello, I'm Professor Mark George. I'm a professor of psychiatry, radiology, and neurology at the Medical University of South Carolina. And I did a research imaging fellowship here at the NIH and I'm in Charleston now.

And we've done extensive FMRI studies in these VNS patients. So it's not the case that they can't have the MRI scans. It's just that they can't have certain types. And those are whole body, where the gradient actually are the large whole body gradients and they cause heating of the electrodes.

So if you have a send/receive head coil, even at three tesla, then VNS is fine. So it's not the case that having the device in would preclude diagnostic MRI.

DR. JENSEN: Of the head.

PROFESSOR GEORGE: Of the head.

DR. JENSEN: But if they ever at some point down the road need a body MR, that could be problematic?

PROFESSOR GEORGE: Correct. And of the cervical spine, that could be a problem as well, yes.

DR. JENSEN: Okay.

So along those lines, there are other issues, too, just with operating in the carotid sheath, which is once you have fibrosis that's set up after surgery, you now make other operations, you know, redoes difficult, for example, the complication rate with carotid endarterectomies goes from 3 to 16 percent when you're looking at patients with carotid redoes.

So again you have a relatively young group of patients right now who at some point in time might need to have carotid surgery and you may be precluding them from being able to have it if the device has caused some sort of fibrosis.

So do you have any long-term data on the patients who have had the device implanted for seizures in terms of battery corrosion, wire corrosion, vascular perforation, operative problems in the carotid sheath afterwards?

DR. RUDOLPH: I'm going to ask one of our engineers to come up and answer that technical question for you.

MR. ARMSTRONG: Hello, I'm Scott Armstrong, Director of Electrical Engineering, Cyberonics.

And as far as the corrosion, no we don't have any problems with the leads or the device or the battery. They're all -- the can it titanium, which is very stable. And all the materials are the same that are used in pacemakers and defibrillators so there is an extensive history of this same type material and have not seen any type of corrosion.

DR. JENSEN: Okay. But in terms of surgery, we don't really have any data on repeat carotid surgery? Got a surgeon in the group?

(Laughter.)

DR. RUDOLPH: Unfortunately we don't have a surgeon but one of my staff members is a clinical engineer and he spends a lot of time in the OR so --

MR. PARNIS: My name is Steve Parnis. I'm the Senior Manager of Clinical Engineering, Cyberonics.

No, we don't have specific data about the number of other surgeries that have been done but we do know of a lot of surgeries that have been done to remove the lead, to replace the lead in cases of lead breaks, or patients who want the device removed.

As you know, any operation, whether it is carotid endarterectomy, whether it's a patch graft that has to be put in the neck, surgeons do go back in. We do know of the complications. There's no difference in the complications between our device being in and a patient who has had previous carotid surgery.

DR. JENSEN: I'm sorry. Say that last line again.

MR. PARNIS: We've looked at the complication rates of patients who had carotid endarterectomies, especially redo carotid endarterectomies, and the complication rates for a replacement of our device has been no different than the replacement or a redo for a carotid endarterectomy.

DR. JENSEN: Okay. So you're not -- what you're saying is you're not seeing, for example, a higher recurrent laryngeal nerve palsy occur in removing your device versus those that occur with carotid endarterectomy redo? In other words, those are similar, the cranial nerve palsy is similar.

I would just submit though that if a patient has this device in has to have a carotid endarterectomy, they're already in the higher risk group, than that whose a virgin carotid. That's just my point.

MR. PARNIS: That's correct. And the redo rates for redo carotid endarterectomies do run up to 28 percent. And we have looked at those numbers. And there hasn't been -- we don't have that high of a complication rate as we have seen in the literature for redo surgeries with our device. We haven't looked at other surgeries associated with vagus nerve stimulation.

DR. JENSEN: Okay.

My next question has to do with --

MEMBER JAYAM-TROUTH: Okay, can I follow through on that?

DR. JENSEN: Oh, yes, go ahead.

MEMBER JAYAM-TROUTH: I don't think that you answered it correctly. I mean I don't think it is still to the point. I think the question that was raised was not that the morbidity for your operation is, you know, as good as a carotid endarterectomy. The question was is the morbidity greater for a carotid endarterectomy after you've done your operation.

DR. JENSEN: Well, you're putting the patient in the higher risk category as opposed to the three percent group.

MEMBER JAYAM-TROUTH: Would you put a patient -- suppose in the future this patient needs a carotid endarterectomy then would this patient be in greater jeopardy doing a carotid endarterectomy after you have done this operation?

DR. RUDOLPH: We don't have data to bear on that but presumably yes.

DR. JENSEN: In terms of your training requirements, how are you choosing which surgeons are allowed to place the implant? Which physicians are allowed to program the implant?

And do you have any sort of program of proctoring for those new physicians who are programming the implant to be sort of overseen by a member of the -- a trained member of the company to make sure that the initial patients are done properly?

DR. RUDOLPH: Mr. Parnis has been working on developing our training program so he's best suited to answer that as soon as he comes -- we lost him -- as soon as he comes back up here.

MR. PARNIS: Hi, Steve Parnis again. If I understand your question correctly, it's -- first of all what surgeons would be implanting our device?

DR. JENSEN: Yes, how do you pick your surgeons, yes. For example, in the NASDA trial, surgeons had to have a five percent or lower complication rate in order to be in the trial. So are you just saying anybody who is a neurosurgeon or vascular surgeon or do you have to have some criteria?

MR. PARNIS: Okay, slide up please. In our labeling, we do recommend that surgeons be experienced in surgery within the carotid sheath. That is in our current labeling and that will be -- and it's in our depression labeling, in the draft labeling as it sits today.

Sixty-four percent of our surgeons are neurosurgeons today. We do have other surgeons. There's ENTs, general surgeons, vascular surgeons could do the procedures. So we do recommend that surgeons be experienced in working within the carotid sheath.

DR. JENSEN: But you're not going to have a minimum number of carotid procedures performed per year? It's just all that they can say is I'm experienced?

MR. PARNIS: No, no.

DR. JENSEN: Okay. All right. What about who can program it?

MR. PARNIS: T-21 please. As far as programming, we do have -- for psychiatrists, we will have a training program in place. The training will consist of the device overview as well as product labeling. Going over the experience that we do have in VNS and epilepsy as well as going over the training itself as far as the programming, diagnostics, and the experience that we have in the D-02 studies.

A member of Cyberonics, our therapeutic consultants, clinical engineers, clinical technical services do perform the training for psychiatrists.

DR. JENSEN: Okay. Is there any plan to have any sort of proctor program where you send somebody to the site for the first however many --

DR. RUDOLPH: We do recommend that physicians do consult with an experienced VNS user before prescribing VNS. That's in our current labeling today as well as in the labeling for depression.

In addition to that, we do offer proctoring.

DR. JENSEN: Is there any plan for any sort of registry to follow up the first however many number of patients?

DR. RUDOLPH: Yes, definitely. We've been actively talking about and planning a depression -- a treatment-resistant depression registry.

DR. JENSEN: And do you have an idea of how many -- how many numbers of patients you would enroll in the registry and how long you would follow them for?

DR. RUDOLPH: We -- as you may or may not know, we have an epilepsy registry. And we're somewhat modeling it on that although we're trying to make it a new and improved version.

I'll ask Dr. Wingard to come up because she has the primary responsibility for developing the registry so she can give you the specific numbers.

DR. WINGARD: This is a draft form of the registry. But we actually do have a protocol for a TRD registry for the United States. And we're initially going to have a TRD registry so that patients who have VNS therapy as well as those who do not have VNS therapy will be allowed to come into the registry.

We're initially going to have it at 20 sites. These are our sites that did our investigative studies in D-02 and then this will expand to about approximately 60 sites in the United States.

And when they are enrolled in the registry, we will be asking all kinds of demographic information, patient history as well as their psychiatric history, medical history.

And then we will be following them at least on a quarterly basis for approximately three years.

DR. RUDOLPH: How many total patients?

DR. WINGARD: In the end, we're planning to have about 9,000 patients in the TRD registry that we will be following.

DR. JENSEN: Thank you.

CHAIRPERSON BECKER: Dr. Ortiz?

MEMBER ORTIZ: As a follow up to that, I have another question. So -- and the purpose for the non -- the patients that are not going to receive the device, what will be the purpose of those people in the registry?

DR. RUDOLPH: To better understand the course of treatment-resistant depression. You know, as you probably appreciate this more, and actually the published literature is fairly scant on longer-term outcomes in treatment-resistant depression patients.

MEMBER ORTIZ: Okay. My questions are more clinical.

What can you tell us about the co-morbidity of the patients in your studies? I'm interested -- it seems like a number were on atypical antipsychotics. And I'm wondering if they had depression with psychotic features or they had concomitant psychotic problems?

I'm also interested in a little bit more about the personality issues, cognitive issues. Can you elaborate a little on that?

DR. RUDOLPH: Yes, first of all, the use of the atypicals was as a treatment for treatment-resistant depression didn't represent a psychotic depression, there were some specific exclusions of patients in the D-02 protocol.

And those included patients with psychotic depression, patients with a drug or alcohol abuse, and patients with a schizoaffective disorder. And if I'm forgetting any, perhaps Dr. Rush could chime in.

DR. RUSH: Well, I just want to --

DR. RUDOLPH: And in terms of the Axis 2 diagnosis, they were not specifically excluded.

DR. RUSH: And I just want to emphasize there were no patients with a current or lifetime history of psychotic depression in the trials. They, of course, could have bipolar 1 disorder in the depressed phase, not mixed, not manic, and not rapid cycling.

DR. RUDOLPH: Rapid cycling was the other category excluded. So I think I covered all of them. Schizoaffective, rapid cycling, psychotic depression, and drug and alcohol abuse.

MEMBER ORTIZ: Schizoaffective was included or was not?

DR. RUDOLPH: Was excluded.

MEMBER ORTIZ: Excluded. Okay.

The other question I have is from the FDA presentation. On page 15, and again you seemed to address a couple of the variables. They have a list of variables that they were --they have a question mark about the -- if there were any notation about it. But I guess this is more for the sponsor.

I'm interested if you did have -- I know you had thyroid up there that you showed us. Then you showed us some data on ethnicity. But premorbid personality, family history, other losses, was there any data on those areas as well?

DR. RUDOLPH: I'm sorry. We thought you were asking the FDA.

MEMBER ORTIZ: No, no.

DR. RUDOLPH: We were hoping to leave the table for a little while.

MEMBER ORTIZ: I was asking you. I wanted a follow up to the FDA question is what I want.

DR. RUSH: Sorry, no, no, we misunderstood.

Let me take them one at a time. With regard to personality disorders, there was no formal structured interview for personality disorders. And, therefore, we did not diagnose them. I would put in context that the consent form is a multi-page, single-spaced document which says something like you have received an implantable device, the safety and efficacy of which is unknown for your condition.

This tends to take certain personality disorders and move them to the side. But it's not a -- it's just a practical screener. So we really do not know the types of personality disorders that were included and certainly none were excluded.

Any social personality disorders tend not to sign up for this sort of thing. Borderline personality disorders, especially with the multiple baseline requirement and the complex consent tend to shy away from it. I'm just speaking clinically.

But we have no data for you. As you know, in TRD, the incidence, the prevalence of personality disorder is very high. It's high in chronic depression, over 60 percent from the Keller Study done a few years ago. Probably on the order of 75 to 80 percent in treatment-resistant depression.

Some of that is actually due to the chronic nature of the depression. And as I mentioned this morning, when you treat the depression, you often -- it has been documented in trials, that some of the people with so-called personality disorders even diagnosed by structured interview, studied 12 to 14 weeks later when not depressed, no longer meet those criteria for that personality disorder.

So we know from the work of Akiskal and others that the personality disorder range is very high but also fluid, highly dependent on the state of depression.

And finally from the work of Klerman and Hirshfield back in the 80s and replicated by others since, I'm sure that you recall that the reliability of personality diagnoses in the midst of significant depressive symptomatology that lasts for a time is really not very high.

So that was the reason to elect not to attempt to diagnosis these personality disorders. So that's one.

Family history of mood disorders, if I recall, we reported and I want to say it's around 50 percent in first degree relative, something like that. It's significant, as you would expect, in this kind of condition.

And the others?

MEMBER ORTIZ: They talked about -- they asked about losses and substance abuse.

DR. RUSH: No, we had no codification of losses. Our assumption here was that the chronicity and/or recurrent nature required. The two years or greater in the current episode or the four more episodes in the lifetime would identify people with serious depression that would be long lasting and, therefore, require an implantable kind of intervention.

And that people that were suffering depression from losses alone would have recovered from the losses in the typical time, at six months, as you know, with DSM, or they would have -- the loss would have triggered a major depressive episode that now would meet those criteria, that is two years or more, or they would have multiple episodes, typically not all triggered by losses, as you know.

So we didn't codify losses and we don't have that.

What was the last one?

MEMBER ORTIZ: Substance abuse was the other one.

DR. RUDOLPH: That was excluded --

DR. RUSH: Drug and alcohol was excluded.

DR. RUDOLPH: -- by the protocol.

MEMBER ORTIZ: Okay.

DR. RUSH: Six months, I think, six or twelve months, one year. Thank you. Twelve months exclusion.

MEMBER ORTIZ: Okay. Thank you.

DR. RUDOLPH: It might be worth mentioning that these typically aren't measured in antidepressant drug trials either.

DR. RUSH: No, but often drug and alcohol is excluded.

DR. RUDOLPH: Yes, that's true. Usually I would say.

CHAIRPERSON BECKER: Dr. Malone?

MEMBER MALONE: I don't know how many questions I have but I have some general comments on design. But I don't know if it is appropriate to make them now.

CHAIRPERSON BECKER: Sure, go ahead.

MEMBER MALONE: I come from the psychopharm advisory committee, so we're used to looking at drug trials.

And in many ways, I think the same criteria should be used for judging this data because it is a treatment for a psychiatric disorder. So it's the kind of disorder we're usually used to looking at.

And most psychiatric -- and depression, like most psychiatric disorders, has spontaneous remissions, variable treatment responses, variable placebo responses in each study.

You know for instance, if you read Kahn's reviews of the FDA data, he shows that you get all these differences every time you do a study. And so he recommends that you use double blind placebo-controlled studies with randomization.

And I really think that that's the kind of standard you generally need for most psychiatric disorders to show that a treatment works.

DR. RUDOLPH: Could I respond to that? Or --

MEMBER MALONE: You could but I'm not finished yet.

DR. RUDOLPH: Okay, sorry.

MEMBER MALONE: I also read, you know, the articles that you provided us and this is where -- I mean I had these ideas from other sources but these ideas are also in the articles you gave us.

And actually Thase, I think, starts talking about treatment-resistant depression. And even though he gives these rates of zero to ten percent, if you read further in the article, he starts talking about adjunctive treatments for treatment-resistant depression.

And what he does is he starts criticizing studies that don't have randomization and parallel controls. Now I guess -- I mean I wouldn't say that you don't need placebo controls. I think you do.

But even if you wanted to argue that you don't need placebo controls, I think he says that you need -- and I believe you need randomization in parallel groups so that, you know, both groups have to be studied out of one study with randomization.

And I think this is generally true in psychiatry because of the many unknowns in psychiatric disorders with regard to outcome and treatment response. And I think they dictate that you need a certain type of study in order to show clear evidence of efficacy.

I don't know what the tradition in devices but I think that those sorts of standards should be used in looking at studies that are assessing a treatment in a psychiatric disorder.

CHAIRPERSON BECKER: Does the sponsor want to make a response?

DR. RUDOLPH: Would you like comment?

Most of the literature you cited pertains to more common type depression. And it doesn't necessarily apply to treatment-resistant depression.

We'll ask Dr. Rush to comment. One of the citations you gave, the Thase one, Dr. Rush happens to be the second author on that. So he might be particularly appropriate to comment on that.

DR. RUSH: That position is known as the senior author in academia.

(Laugher.)

DR. RUSH: I had to say that.

First of all, I agree with your general contention that randomized controlled trials are essential when they can be conducted in a safe and ethical and feasible manner. And when you know that the outcome of the disorder is not uniformly terminal. That is the preferred gold standard.

And if I could design a study today, as opposed to what we were working with several years ago, as I mentioned this morning, one can now, given that we have long-term safety established in this population of very resistant patients with VNS, and that we know effect sizes, we are now in a position that that could be done.

The question, I believe, is whether a randomized controlled trial is necessary given the data we have in this condition at this time.

So let me expand on that just a little bit. If I could have I think it's 059. It's one of the first slides I had this morning. I just want to go back to that because we discussed an option in there that was very close to I think what is going to be feasible in this population.

We -- if the outcome of interest here is long term, which I think it must be given the nature of the treatment that we're dealing with, then we cannot fail to change the treatments in an ongoing way in these patients and hold them on a constant medication regimen for a year. It's not feasible. It's not ethical.

I think we'd have -- doctors wouldn't sign up. IRBs wouldn't let us do it. I wouldn't want to do it. I wouldn't want to be in it. I don't think it should be done.

So the only way you can do that is to then allow the treatments to vary individually, being individually managed. And one group that gets VNS and one group that does not get VNS over a long period of time. That is now. It was not feasible then because we didn't have long-term safety. It is feasible now.

Even under those conditions, you're going to have an interaction between the treatment, VNS if it is effective and medications and the medication management.

In other words, if VNS is being helpful to the medications, the need to change medications will shift. So you'll have fewer medication changes over time in the group that receives VNS as compared to the group that's not receiving VNS.

That's actually what we found in the non-randomized comparison, the control, okay?

So if you are allowing a concomitant treatment to change while you're giving a fixed treatment to one group and not to the other group, you are going to have trouble being absolutely positively certain of the level of certainty that we had hoped to get out of the D-01 trial -- sorry, D-02 acute trial, that if the patients differ in outcome, it's all due to VNS because the concomitant treatments are changing.

And when you change the concomitant treatments in one group differently than in another group, you have now two confounds. One got VNS, one did not. One had these kinds of changes, one has those kinds of changes.

So even there, while you have what I would think would be very strong evidence, you don't have, you know, totally convincing evidence of the type you get with acute ten-week trials of the D-02 acute that we tried to set up.

So one final comment. This was actually brought up and debated with -- I think the sponsorship was under the Manic Depressive Association, EGIS, we're talking about studies in bipolar disorders, especially long-term studies.

And there was a consensus there that whey you're dealing with long-term studies in a chronic and recurrent illness, that the feasibility and safety of doing a controlled trial for a long period of time with placebo and fixing all the other regimens is really -- it's just not possible to do that.

So because of the extremely depressed, treatment resistant, disabling, lethal nature of the conditions that we're dealing with here, we're dealing in the range of a lymphoma. We're going to lose a certain number of people to this condition in the course of a one-year trial.

I mean please go back to the 1,000 people per month with treatment-resistant depression that actually kill themselves. So we're dealing with a really different group. These people are totally ineligible for any pharmaceutical company-sponsored or, by the way, NIH-sponsored trial I've ever been involved in.

I am running a sequence treatment alternative to relieve depression trial. That starts with people who are not treatment resistant. They begin with -- citolapram is the first treatment. They're randomized to four different switch with three different augment treatments as Level 2.

If that fails, then they're randomized -- still randomized comparison to active switch treatments in Level 3 to active augments in Level 4 to switch treatments.

So randomization is possible. But you'd have to start with patients where it is safe, feasible, and ethical. This group has been through really everything. So half have received ECT. If we say well you are in the algorithm that requires ECT, then it would be unethical to give them ECT having already failed on it.

So I just -- I need to help you appreciate the nature of the condition which I do think changes the requirements of the trial. Not to come up with any less science than we otherwise can feasibly, we really want to do the best science.

But as I walked through this morning, when we were designing the studies and what we knew about the long-term safety of the treatment, in fact the short-term safety to say nothing of the efficacy, these were the very best trials that we could do.

This is the first trial in the D-04, the first time any of these patients at this level of severity have ever been studied at all much less for a year.

We didn't know if they would get better, they would get worse. We didn't know how many would kill themselves. We had no idea. No one has ever reported this.

And please remember also the adjunctive VNS on to the standard of care, the D-02, you know, post random -- post acute, the long-term D-02. It's the first time, again, anybody had ever studied VNS beyond three months in these kinds of patients in significant numbers.

So we were wrestling with a -- really a totally new territory, a terribly difficult illness with a very high risk of disability death, you saw a hospitalization, we had a patient suicide who was a physician and so on.

So I think that the question I'm sure for the panel and certainly one I ask myself or I wouldn't be here is are the data sufficient given what we have to deal with and what we have acquired?

Because really any other explanation for a growing long-term benefit, which you don't see following ECT, and you don't see with maintenance medication, you don't see with any long-term treatment, there seems to be at least a sustained, if not growing, long-term benefit in patients who have received VNS at a level of severity and disability so bad that half of the patients receiving ECT in the New York metropolitan area would not qualify for this study. That's the issue, I think.

And what more -- what degree of certainty is required given the nature of the condition and the long-term outcome, which grows in benefit rather than wanes, which is true for all other treatments.

CHAIRPERSON BECKER: Dr. Ellenberg has a comment.

MEMBER ELLENBERG: Can I comment?

CHAIRPERSON BECKER: Sure.

MEMBER ELLENBERG: I mean if I could follow up on Dr. Malone's question.

CHAIRPERSON BECKER: Actually I think Dr. Malone wanted to follow up on his question.

MEMBER ELLENBERG: Oh, I'm sorry.

MEMBER MALONE: I think it would be ethical to do another study. First of all, you already did a placebo-controlled study with a sham and it didn't work. So I don't know why it couldn't be done again.

And when you did you D-02/D-04 comparisons, the response was very quick. So the explanation that you needed more time in the D-02, I mean I don't understand it.

The other thing is if you have people going in the D-04 study for one year on treatment as usual, I don't know why part of them couldn't ethically be randomized to an adjunctive treatment. And that you couldn't use some sort of controls in the at least blind assessment of the patients.

And I still am not convinced that you can't do those things. And I, myself, do studies in serious psychiatric disorders, too. Currently we're doing studies in autism. And there's no way we could do drug treatment studies without placebo controls.

Now people might argue that autistic children are not going to change much over a short period of time. And they've argued this for OCD and various other psychiatric disorders.

But the truth of the matter is that when you do placebo-controlled studies, because of the variability and diversity of response in course for these disorders, you always get responses in all the arms. And you get various responses.

And every time you do a study, you get different rates of response. So when you don't have a concurrent control with randomization, I'm not sure what the data means.

DR. RUSH: Could I try --

CHAIRPERSON BECKER: We should maybe have Dr. Ellenberg's comment and then I'll let you respond.

DR. RUSH: Sure, thank you.

MEMBER ELLENBERG: Well, Dr. Malone fairly well covered it. But I think in terms of the issue of change of treatments, I just don't see why that is a problem to allow the change of treatments and yet have as an adjunctive therapy randomized controlled clinical trial an assessment of VNS in the field, randomly assigned to those subjects that are given standard of care is totally analyzable and will give you solid results.

So that's just a further comment on what Dr. Malone's point is.

CHAIRPERSON BECKER: And actually, before you responded, I come from the cerebral vascular, cardiovascular disease world where that's the norm. You would evaluate if a statin prevents heart disease. These patients are on lots of other therapies. And this is the way the trials are done.

It's an add on to all of the other adjunctive therapies that they get so that, again, I think it speaks to the fact that it's not valid to say you can't do the study where medications are changing in the background.

DR. RUSH: Well, let me clarify. First of all, that was not my position. So let me be clear. What I was saying was at the time that we started, where we were going we had no evidence of long-term safety or efficacy of VNS. Could we do a long term -- could I have the slide up?

Could we do a long-term trial now in which patients are treated with doctors' best choice, severely depressed, and half get VNS and half do not? Yes. Because now we know the long-term safety and have some clue about efficacy of VNS that we didn't know earlier.

So I'm saying in the course of development, which I went through this morning when you and I were talking, that was not an option. We couldn't do a long-term. We didn't even have an idea of safety in the short run. No idea that there was a signal at all for antidepressant effect.

So I'm not saying that it can't be cone. Or it's not ethical or feasible given our current state of knowledge, which has taken six years to acquire.

But way back when we started, we didn't have that knowledge base and couldn't, in my view, make that judgment with any evidence, okay?

The issue of variable outcome, I want to have two slides, one is the slide from the ECT followup. Do you know what I'm talking about? The non-responders and responders from the Harold Study. And the other is just the -- I think you have the D-04 IDS or something like that. But Harold's is the most important.

The fact of the matter is that it is true that patients who enter efficacy trials for depression drug development have a very wide variation in outcome. Placebo responses all over the board. Studies have been done to show that more than half the time the drug doesn't even separate from placebo.

What kinds of individuals enter those studies? I've done them for 30 years so I can tell you. And many of you know. You sit on the panels and so on and done the studies. These are individuals who are symptomatic volunteers, taking no other medication, who are willing to go through a drug washout, who are not acutely suicidal, have minimal co-morbidity, psychiatric and/or general medical, are capped at two years in the current episode.

Most of the trials in the last ten years have done that. You cannot be in the episode more than two years. You cannot have failed on more than one medication in the current episode. And they are acute eight- to ten-week trials. And you look for a signal. And they have to accept a placebo randomization, of course.

Now that is entirely feasible with symptomatic volunteers. The reason that we did not have a placebo in Star D is because when you move into real patients, and Star D only allows real patients, no symptomatic volunteers, the first thing that you are struck by is massive comorbidity, many of these patients would not be allowed, because of the co-morbid illnesses into the standard efficacy trials with symptomatic volunteers.

Second is their length of illness, the length of illness in these patients is on average 20 years, the current episode is 20 months. This is a sample drawn out of primary care and specialty care practice. These are real patients, not symptomatic volunteers, okay?

More than half the patients in that trial are not eligible for efficacy trials run right now for the purposes of developing drugs for regulatory approval. That means they're really not representative.

We throw out the co-morbid, the general medical conditions, 60 percent have a concomitant general medical condition. May of them are not eligible for placebo-controlled efficacy trials because they don't know the safety of the drug that they're using, they don't want to give it to people at risk, which is very reasonable.

So I really -- I must tell you these patients are totally, completely different from patients that go through depression trials. I'm not saying you can't do a randomized trial. I want to be very clear about that.

What we know now with this treatment over the long run in terms of safety, you can do a long-term trial. The one thing, though, that you will have naturally is you are going to have to let treatment change over time. You cannot take these people off of all their drugs and make them go onto placebo in my judgment.

No one agree to it. I frankly wouldn't take them off. Many of these patients are in and out of the hospital, barely holding on, with multiple medications to keep them as outpatients. I mean I would -- my IRB would not allow a pure placebo control. I would not do a placebo control.

And I don't know a patient, short of psychotic depression, that would take one. But you could do an active treatment, and an active treatment plus VNS. When you -- can I have that slide up?

CHAIRPERSON BECKER: And actually we're going to have to kind of curtail your comments a little.

DR. RUSH: I'm going to finish -- I'll finish in one minute. Just one slide that says it all. Can I just have that one up that's here?

This is the issue of probability of spontaneous improvement in resistant depression. This is a group that received ECT here, okay, and they either did not hit a remission or they did hit remission. This is from a Sackheim study, a Prudic's study, I'm sorry, of ETC beneficiaries in New York.

The people that benefitted from ECT as a group started to lose it, as I showed you this morning, so they worsened. The group that did not benefit or they benefitted but didn't hit remission, they got improvement with ECT but didn't hit remission. Hamilton is now only 20 so they're eligible for studies.

Notice their course over the subsequent 24 weeks. These patients are not spontaneously improving. Treatment-resistant depression does not spontaneously improve over time as a group. And, therefore, you may not need that kind of control.

MEMBER MALONE: Well, I mean it sounds to me like you're saying you're ready to do a pivotal study now you know the parameters. And I still think you need to randomize treatments and have concurrent control.

It sounds like you're saying you're ready to do it.

CHAIRPERSON BECKER: And I think we're just going to continue to move along and see if Ms. Wells and Mr. Balo have any specific questions that they need -- they would like to ask.

Ms. Wells?

MEMBER WELLS: I just have a couple.

CHAIRPERSON BECKER: And if I could ask the sponsor to limit their responses to directly answering the question.

MEMBER WELLS: The D-03 study is still ongoing?

DR. RUDOLPH: That's correct. It's still enrolling patients.

MEMBER WELLS: Do you have any intermediate results on that?

DR. RUDOLPH: Are you asking about effectiveness? Outcomes? Or --

MEMBER WELLS: Yes, effectiveness.

DR. RUDOLPH: Yes, we do. The D-02 is an open study so I put that qualification out -- D-03, I'm sorry, is an open study. Their interim results, the results so far are similar to the D-010 in terms of response rates.

MEMBER WELLS: Okay. During the D-02, was it ever suspended by any IRb during the study course for SAEs or AES?

DR. RUDOLPH: No, it was not.

MEMBER WELLS: Okay.

MEMBER BALO: I just have one question. I'm going to give the sponsor a little rest because I'm going to ask the FDA this question.

In light of Dr. Davis's information that she put up, there seemed to be a lot of questioning about the propensity analysis and also the covariant analysis. With the data that she put up, I'm wondering if this basically answered some of the concerns that Dr. Lao had, relevant to the propensity analysis.

DR. LAO: This is Chang Lao. I see the propensity score repeated measured linear regression analysis which was done reasonably well. But for the comparison of the two response rates, I reviewed it statistic plane everywhere they did talk about logistic regression and covariance.

But for some reason, there are many, many different volumes of submissions. To compare two proportions of response, how come I didn't see this logistic regression and covariance and it means compare two proportions to adjust for covariant. So I would like to know which volume the analysis was there. And that's my concern.

MEMBER BALO: But she also provided that they did do adjustment of all the covariance, which was one of your concerns in your presentation. And so I would imagine that that would sort of resolve at least one of the issues you had with your statistics.

DR. LAO: Well, propensity score here in the repeated measure and integration includes 17 covariates.

MEMBER BALO: Yes.

DR. LAO: There's three covariates in terms of percentage of the ECT use during the current episode, current use or lifetime use were very highly significant before an adjustment.

But after the adjustment, they become non-significantly different between D-02 and D-04. So an adjustment procedure works for the second covariant before an adjustment.

But the reason the propensity score was non-significant in the repeated measured regression, it means after an adjustment they reclassified each individual patient into one of the five subgroups based on the propensity score probability.

Like if each individual patient has a predicted probability after D-02 assignment, like you can roughly classify each patient into probability into zero to .2, .2 to .4 and up to .8 and 1.0, rank and order. Then rank and order, then reclassify.

So I think that the propensity score did a good job here --

MEMBER BALO: Okay.

DR. LAO: -- in the repeated measure linear equation.

MEMBER BALO: Okay. Thank you.

DR. LAO: Thank you.

MEMBER BALO: Dr. Pena, can I ask you a question?

I'm just wondering with D-04, you know, dealing with the sponsor why there was never discussion about safety data.

DR. PENA: In D-04?

MEMBER BALO: D-04.

DR. PENA: Okay.

MEMBER BALO: I was wondering in your discussions with the sponsor when they were submitting D-04, did the FDA ever request them to have safety data?

DR. PENA: The D-04 was conducted local IRB jurisdiction so it didn't require any FDA approval. In addition, when they submitted the revised statistical plan back on September 3, 2002, FDA responded with a correspondence letter saying that we had serious concerns with this comparison.

We additionally had conference calls that further underscored those concerns. So I think we had a lot of concerns about that comparison and use of that open label controlled study, observational controlled study.

MEMBER BALO: Okay, thank you.

CHAIRPERSON BECKER: Dr. Witten, do you have any comments or questions?

DR. WITTEN: No.

CHAIRPERSON BECKER: And I think that before we move on, Dr. Fochtmann has three questions that she would like to address to the person responsible for safety issues.

MEMBER FOCHTMANN: The first question that I have -- actually the first two questions, I believe on the exclusion criteria for the study, was mentioned patients with carotid stenosis as shown by ultrasound and the other group of patients that was mentioned were patients with a diagnosis of obstructive sleep apnea because of the chance of increasing apneic episodes with the stimulation.

My question relates to whether there is a need for either initial specific screening for those disorders in people who would be using this clinically and/or ongoing assessments? Certainly, I know, obstructive sleep apnea is often undiagnosed in community samples and in a group such is this which, as your data show, have an increased body mass index, there might also be further increases in sleep apnea.

So is that something that needs to be taken into consideration from a safety standpoint in terms of future use in the general population?

DR. RUDOLPH: There is already a warning in our labeling with regard to obstructive sleep apnea. It doesn't require screening of the patient, however. And we have the epilepsy safety experience which shows that that warning -- it would suggest that warning has been sufficient to protect the safety of the populous.

MEMBER FOCHTMANN: Okay. But warning specifically relates to diagnosed sleep apnea. And my concern is about people who may have it that are just not diagnosed.

DR. RUDOLPH: No, I understand. And that's how the warning is currently written in the label.

MEMBER FOCHTMANN: Okay. The other question that I had related to the issue of patients who are not adherent with treatment. And it was specifically mentioned both in the presentation and the graph labeling information that this might be a particular treatment that could be considered in patients who are non-adherent.

My concern with that relates to the issue that the patient brochure and some of the other information in the volumes we received mentions the need for individuals to continuously carry the magnet with them in the event that the needed to turn off the stimulation.

In a patient who we would see clinically that we would think may not be totally reliable and, therefore, non-adherent, would we have reason to be concerned about their reliability in not always carrying the magnet with them from a safety standpoint?

DR. RUDOLPH: The magnet is mostly a convenience for temporarily turning off stimulation for minor side effects like a common situation where it's used in turning off stimulation to stop voice alteration in a patient who might sing in a choir or who has to do public speaking. So it's there more for these nuisance side effects. And it doesn't, you know, the absence of carrying the magnet wouldn't impose any undue major safety risk on the patient, which, I think, is a short answer to your question.

DR. RUSH: There's just -- there's a little convenience factor, in order for the magnet to stop the stimulation, it has to be held over the device.

So you'd have to intentionally tape the magnet over the device and walk around with that taped on 24/7 in order to stop the device. So the likelihood in our clinical experience is that really is not likely at all.

MEMBER FOCHTMANN: Yes. My concern was more that there would be an adverse event. That the patient would have left the magnet at home. And they wouldn't be able to turn it off.

DR. RUSH: Oh, we've had some patients ask -- yes, we've actually given patients several magnets. One for the car, one for the office, and one for home. Several of our patients actually like that.

MEMBER FOCHTMANN: Okay.

CHAIRPERSON BECKER: Another question here.

MEMBER JAYAM-TROUTH: I had actually a couple of questions.

Does it effect it when you go through the airport screening process you know?

DR. RUDOLPH: No.

MEMBER JAYAM-TROUTH: There's no magnetic interference then?

DR. RUDOLPH: No.

MEMBER JAYAM-TROUTH: So you don't have to readjust it or reset it?

DR. RUDOLPH: No.

MEMBER JAYAM-TROUTH: What about this imaging? You know I was just looking at those and I was a little puzzled. And apparently all of the stimulation on the PET scan appears to go to the left brain? You know is that then true that the major connection is to the left brain except the single place where it crosses over and goes to the right brain?

DR. HENRY: Yes, Thomas Henry, Associate Professor of Neurology, Emory University. I would like to disclose that I did imaging studies. And Emory was reimbursed by Cyberonics as well as my participation in the epilepsy E-05 study. And my transportation to this meeting was paid.

If I could have the PET slide that is in question here. I think this is --

MEMBER JAYAM-TROUTH: 38.

DR. HENRY: Well, this one.

MEMBER JAYAM-TROUTH: Slide 38.

DR. HENRY: Oh, slide 38, okay.

I'm not sure that this is the slide you're looking for. This is one that address your question showing that acutely during vagus nerve stimulation there are significant blood flow increases bilaterally as well as some significant blood flow decreases.

This is a group of five patients in the epilepsy E-05 study who were scanned within the first 24 hours after VNS was turned on for the first time. So this is an acute stimulation effect. PET scans were compared during vagus nerve stimulation versus without stimulation within subjects and then co-registered to MRI here across five subjects.

So with one centimeter spacing on these axial images, subject left on image right, the usual convention, we were able to discern areas of significant blood flow increase in the dorsal rostrum. The medulla -- here are other brain stem regions along known pathways projecting up to autonomic and limbic centers in the hypothalamus, the thalamus bilaterally.

And then bilateral orbital frontal cortex, insular cortex, and other relevant areas of the limbic system. Or posteriorly, however, in the singlet and hippocampus decreases were seen, the main area of significant asymmetry is in the subjects who felt left cervical paresthesias during stimulation.

Only the right sensory strip, precentral gyrus was really activated. And you can see a specificity there for this somatasensory distribution here just on one side.

But most of the other stimulations may have been a little asymmetric but overall were bilateral.

I hope that addressed your question.

MEMBER JAYAM-TROUTH: Well, yes, but your PET Scan No. 38, you know, in your slides was almost lateralized to the left side.

DR. BRANNAN: This is the image you were asking about, is that correct?

MEMBER JAYA-TROUTH: Yes.

DR. BRANNAN: Okay. Also get ready for the next slide, 39 -- not this one but the next in the sequence.

In this particular -- when you're looking here, you actually see a lot of midline activity in the singlet but you do see in these particular slices activity on the left. But when you're looking at one slice, you're not looking at the whole brain.

And so similar to what Dr. Henry was saying, 052 please, 052, slide up, here, I think, this is one-year scan data that is available from University of Minnesota. And let me just draw your attention -- let see -- right down here, so you see very nicely there's bilateral, almost mirrorlike activation or deactivation patterns here.

So there's really bilateral activation. It doesn't mean that there aren't some areas that are asymmetric but you're not seeing left-sided activation in the PET studies or FMRI studies either.

CHAIRPERSON BECKER: Thank you.

MEMBER JAYAM-TROUTH: Thank you.

CHAIRPERSON BECKER: In the interest of time, we're going to move on to the FDA questions. And do you want to put the FDA questions up for us?

I think the first question that the FDA has is one that we've spent a lot of time discussing already, and that's the limitation of the long-term D-02/D-04 comparative analysis.

And that the comparisons are not from a randomized data set but rather comparison of outcomes from an investigational device study and observational control study.

And while the sponsors did do a propensity adjustment strategy, there are still potential biases that exist.

And so the FDA would like the panel to discuss the impact of a comparative analysis of non-randomized subject data, comparison of outcomes from an investigational study and the observational study and the unmeasured patient variables upon efficacy outcomes in this PMA submission.

And I think we'll just go around the table and get comments from the different panel members. And we'll start with Dr. Ellenberg.

MEMBER ELLENBERG: This is a non-randomized comparative controlled trial with a single blind on the primary outcome measure. And in my view, in spite of the extraordinary analyses presented by the sponsor, attempting to demonstrate that the baseline observed differences and other characteristics that might effect the nature of the patients that were entered into the two arms, that this type of analysis by showing that there were no difference -- there were differences seen, either clinically or statistically, does not replace the concept for randomization.

And it does not specifically address the issue of all of those variables that we cannot measure, did not think about, and come into play when you compare two arms as has been done here.

And so my sense is that we need to stick to the standard of a randomized controlled trial in order to evaluate the VNS. And that's a set standard. It appears from the discussions we've had with sponsor that such a trial could be done today although it couldn't be done perhaps at the time that the original D-02 was done.

So my conclusion is that there could be a major impact on these results that we cannot see, we cannot measure. And we can guess all we want. We can speculate. But I don't find this at the acceptable level of a randomized clinical trial.

MEMBER JAYAM-TROUTH: While I agree that there is a problem there, and that we do have, you know, no definite randomized trial here, there's no comparison, but I do see the point that, you know, at the time that this was taking place, such a randomization could not have occurred.

You know I also feel that when I look at the two groups of people and I look at the D-04 and the D-02, that the D-04 certainly had, perhaps, you know, patients who were better. You know from the slides that we can see, they did not need as much ECT. And they, you know, had not been into as many multiple trials, et cetera, you know, as the patients who were put into the D-02 studies.

So I think that even though these are not really truly comparable, I think that, you know, having used it as data for comparison, even though it doesn't fit into randomization, I feel that, you know, at the time that this was a study that we could kind of look at and we could say, okay, you know, there is a comparison there that can be made.

If at all, it's skewed towards worse patients in the D-02 study.

CHAIRPERSON BECKER: Dr. Fochtmann?

MEMBER FOCHTMANN: I would certainly concur with both of those impressions. And I would also really emphasize the point that has been made by the sponsor that this compared to studies of depressed patients in other studies is a very, very unique group of individuals.

And one of the groups of patients that we as clinicians, even those of us who have expertise in treatment such as electroconvulsive therapy, are always confronted with how to assist these individuals with these obviously devastating illnesses.

And so although I would agree that in an ideal world it would be nice to be able to do, at this point knowing what we know now, further study, I'm concerned about the potential burden to patients who might not be able to receive a viable treatment for this very severe illness.

And so I would want to seriously weigh both sides of the issue.

CHAIRPERSON BECKER: I would just add that I think the data look very promising and do suggest that there's a benefit there although it's really difficult to be sure given the difficulty in comparing the two groups.

And this seems like to me the right time to do the pivotal control trial.

Dr. Wang?

MEMBER WANG: Yes, just sort of echoing what I said earlier, in terms of the fact that you didn't see differences after, you know, before versus after your propensity score adjustment, there's several ways to interpret that. One is, you know, maybe you didn't have a very good propensity score. You know you didn't either have the unmeasured variables that you needed.

There are other ways that you can also have a poorly performing score, you know, how did you categorize your variables? What did you do with your missing information? You know did you bury it into the extremes? That sort of thing.

But what I do find promising is your acute phase data which is randomized. And maybe this is a point for later discussion but I'm just curious why there wasn't sort of a push to do more -- not as Dr. Rush was saying long term but acute phase randomized. Why not acquire more of that data? Because it looked like you were about on the threshold of seeing a significant result.

DR. JENSEN: I sympathize with your situation. As an interventional nerve radiologist, I deal with a lot of groups of patients who have no other viable alternatives except what is being offered. I liken this particular situation with ours concerning percutaneous vertebroplasty, which is a treatment of patients with osteopartic compression fractures who have failed all medical therapies.

And what we found was that there was a very high response. But when we started out with this, we didn't do a randomized control trial. We did best medical therapy versus vertebroplasty with using patients as their own internal controls.

When we then went back and tried to do a randomized controlled trial to show the data, it was impossible because vertebroplasty was now too widespread. It was available everywhere. And patients would not consent to being randomized.

So for me one of the issues is of timing. One of the differences between this particular study and vertebroplasty is we had consistently across different sites 80 to 90 percent response. And yours is 30 percent.

So for me one of the issues is timing. This may be the only time to actually get the data that you need to prove without some of the doubts that have been raised here that this is truly efficacious.

MEMBER ORTIZ: I agree with what's been previously said that it's unfortunate this wasn't designed differently but I think it's understandable given the nature of this group, that the blinding that was built into this study.

And my impression is that both the anecdotal reports as well as the long-term symptom reports and the comparison with the K-04 group suggests that this would provide a significant alternative treatment to what's available.

CHAIRPERSON BECKER: Dr. Malone?

MEMBER MALONE: I guess I already said that I don't think that you can use that sort of control. I think that the sponsor did demonstrate they could randomize to a sham treatment and carry out such a study. You know I think that's what's needed.

It is possible that this is a viable treatment. But it's also possible that it's not a treatment. And there are ethical issues on both sides of the fence here.

So I'm not sure that it's quite ethical to give a treatment for which there is not, I don't think, substantial treatment. You may just be providing people with more side effects and no increased efficacy.

And, you know, there's something I don't think the PA analyses can ever get out, when we do our studies, we screen people for studies. And once they find out it's a drug study, there are people, and we never can predict from any demographics, who say no, I don't want my child on a drug.

I only have to think that the group of patients who will consent to have this procedure done, because I don't think it's -- it's not getting your tooth pulled, is different in some way that we can't really find in these PA analyses.

And so I think, you know I think that's the failure of those analyses, that you may be pulling different groups of patients because of the interventions. Some people will agree to some interventions and some won't.

And there's no way from the data that I've seen that you can tell who would or wouldn't agree.

CHAIRPERSON BECKER: Ms. Wells?

MEMBER WELLS: I agree with Dr. Ortiz.

CHAIRPERSON BECKER: Mr. Balo?

MEMBER BALO: I think we've heard a lot about the design of the study, whether it's randomized or non-randomized. And I think the company really -- I think Dr. Rush really explained it pretty explicitly about they really didn't know what they had when they started the study. It was never really a long-term test for this population that was so unique that we didn't know how the VNS was going to operate.

I think from a randomization perspective, I think in devices, sometimes randomization studies are not done and devices get approved. Obviously the optimal would be do randomization.

But my feeling is that the company actually went out, dealt with the FDA, looked at the data after three months, saw that they needed to get some long-term results because they -- and I think I also agree with Dr. Wang that, you know, the acute data did have some promise to it.

And I do feel that maybe if they would have continued with this study a little bit longer, it would have given them a little bit better data. And we wouldn't be in such a controversy right now.

But I do also feel that the analysis that was done by Dr. Rush and by the sponsor did try to show that there was some potential benefit. And I do feel that there is some potential benefit to the device.

CHAIRPERSON BECKER: So it sounds like the panel thinks that the sponsor did a really good job in dealing with the data that they had. But the data that they had was not the optimal data. And that there are limitations in comparing the two groups that exist.

DR. WITTEN: Thank you.

CHAIRPERSON BECKER: Next we'll move on to question 2 which is the sponsor believes that D-02 long-term outcomes are not due to a placebo effect. The data provided in the PMA includes a placebo effect rate, 20 percent, in sham treatment controlled subjects at acute phase exit as defined by HAM-D score less than 18.

Patient expectation of participating in an investigational study for new therapies, such as the D-02 study, may have also been greater than the expectation of participating in an observational control study.

Please discuss the placebo effect and impact upon clinical outcomes presented in the PMA.

And I think, Mr. Balo, we're going to start at your end of the table and come around this way.

MR. BALO: I have no comment about that.

CHAIRPERSON BECKER: Ms. Wells?

MEMBER WELLS: I have no comment either.

CHAIRPERSON BECKER: Dr. Malone?

MEMBER MALONE: As I said before, when Khan reviewed all the FDA data and I know that Dr. Rush doesn't think it applies but I think some of it has to apply. It's the best data that we have.

The placebo response rates are different in every study. And so I think that it's hard to really know what the -- what placebo response there is in this study.

The other thing is when Khan examined all of the FDA data -- well, I don't know if it was all, it was a ten-year period of recent antidepressant trials, and there have been a lot of them recently, what he found was that the HAM-D scores decreased for every group. So they decreased for the drug treatment group. They decreased for the drug comparator group. And they decreased for the placebo group.

So when you have a long-term study and you get a decrease in scores, it's really hard to know what that means. One would actually expect scores to decrease in the long term. So that, for instance, when the scores decrease across time in D-02, it's hard to know why that happens without what I would think would be an adequate comparator.

CHAIRPERSON BECKER: Dr. Ortiz?

MEMBER ORTIZ: I guess my only comment is that psychiatric studies placebo responses are often high. And I think this particular population is so complicated and probably does have a very high incidence of Axis 2. It's hard to interpret the placebo response.

CHAIRPERSON BECKER: Dr. Jensen?

DR. JENSEN: I agree with Dr. Ortiz. I do appreciate the sponsors pointing out though that placebo response is normally short and not long term, which is certainly what we saw with vertebroplasty, too. Patients would get better immediately and then go back to having chronic pain.

I think another big confounding factor for me is is that it's very difficult to blind this study because I think a lot of patients probably knew whether or not they actually had the device turned on. And so for me that confounds what the placebo effect might have been.

MEMBER WANG: Yes, I think this is another one of those issues where it's probably -- there is something probably still there despite the very sort of rigorous reassurances, including the fact that in the acute phase, there was, you know, 11 percent of people responded to the sham treatment.

The -- I'm still curious, though, earlier I raised this issue about sort of the IDS, the difference in the outcomes when you look at the IDS versus the HAM-D which the HAM-D is, we think, is the gold standard. But we see that the responses were somewhat, you know, more robust at the IDS.

And I'm wondering is it that the IDS is more prone to -- because it's a self report and not a clinician-administered instrument, is it more prone to placebo effects or, you know, other kinds of information biases? Because it may be a more relevant measure for depression than the maybe antiquated HAM-D.

CHAIRPERSON BECKER: I suspect that many of the benefits seen of vigorous stimulation in the study were related to the placebo effect but not all. And part of me wants to say well, so what if it was a placebo effect? This is a very treatment resistant group of patients.

And if this placebo effect works for them and others didn't, that should be fine. But I think there's enough safety concerns with the device, especially as brought up by Dr. Jensen with the young patients who are being implanted now are going to have these devices in for a very long period of time, that we really do need to be sure that the effect is more than placebo.

And I think only a true randomized controlled trial is going to answer that for us.

MEMBER FOCHTMANN: I certainly would concur with the concern about the difficulty in interpreting any sort of placebo effect. I believe one of the previous presenters emphasized the difference between a placebo effect and an actual response as measured by rigorous definitions of the term response. And also persistent response.

And I think that those are three very different parameters that should be considered independently.

I'm also concerned about the short term, the blinding in the short-term study as well. But I'm not sure, given the nature of the treatment, how one could adequately prevent people from knowing or prevent the investigators from knowing based on the fact that the side effects seem to be at least in some instances dose related, related to the stimulant's intensity.

I'm not sure how you could design a study that would totally blind those effects.

MEMBER JAYAM-TROUTH: I agree that yes, I mean there's really nothing in the acute phase that separated the two groups, you know the sham as well as the D-02 groups.

But my own feeling is that this is an invasive procedure. You know people are looking for something to happen. And then you're coming there and stimulating them almost every four hours, every day. They don't know that they're getting stimulated.

And I think that in itself probably set off, you know, neuro epinephrines and every other agent inside the brain and I think, you know, that type of an invasive process possibly is responsible, you know, that term, that 12-week term probably was not enough, you know?

And possibly if that sham period had continued a little longer, we might have seen a difference. But since the study was not set up to show that, we do definitely see a difference in the long-term study. And it seems like it is a consistent, it is a sustained difference.

And even though I agree that this was definitely the sham in the acute phase in the D-02 did not show, you know, any significance, I think the long-term studies kind of outweigh that.

MEMBER ELLENBERG: I concur with Dr. Becker.

CHAIRPERSON BECKER: So Dr. Witten, it sounds like the panel, in general, feels that without a randomized study, it's very difficult to know what to make of the placebo response and how much of the response of VNS stimulation is due to the placebo response.

Although there seems to be some general belief that there probably is an effect that isn't completely placebo related, we just don't know how to measure that at this point.

So the third question that the FDA has posed has to do with concomitant medications in ECT use, which were not standardized in either the D-02 long-term study or the D-04 observational controlled study.

So please discuss the impact of concomitant medications in ECT use on interpretation of the efficacy of VNS therapy for treatment-resistant depression.

And we'll start with Dr. Ellenberg this time.

MEMBER ELLENBERG: Well, this certainly proved to be a very interesting issue. And again I think the sponsor did an extremely nice job in trying to tease out the impact of concomitant meds.

I would agree or I am sensitive to the comment that Dr. Wang made that it's difficult to sort of speculate when there is a change in medications and you start dealing with less observation carried forward or dropping medications or other forms of censoring, it's very difficult to speculate as to what that means in terms of the outcome.

I would find it difficult to argue that because the average time to change the medications for the DOT group, the combined DOT group with the immediate and delayed start of VNS, that that group was disadvantaged in the sense that they only had seven months of treatment rather than the full year. It's not clear to me that one can speculate on that. Some additional things that I would like to see, which I couldn't find in the volumes, would be the distribution of the change of medication times for those on the DOT component rather than just the average. And that might help us to better understand the impact on the analysis.

The second point, again I think this came out of Dr. Wang's questioning, but when the chart was put up for the slope coefficients, looking at, I believe, five different types of censoring, it seems to me that there were dramatic changes in the slopes presented with the different types of censoring.

And if you disregard the issues of statistical significance, that sensitivity analysis, to me, was screaming that this whole process is not robust to changes in the definition of how you censor or how you treat the censoring in the analysis.

So I think this is a question that needs further study. And it's certainly very interesting.

CHAIRPERSON BECKER: Thank you.

MEMBER JAYAM-TROUTH: The way I see it, you know, even if you did have, you know, ECT interfering and people could change ECT anytime they wanted, they could change their medications anytime they wanted, I mean there was really no randomization, there was no algorithm.

And I guess it's the nature of the treatment that's the nature of the disease. But if it was skewed, it was skewed towards, you know, the D-02 actually having worse patients, you know, and patients who had had to seek more ECT, you know, as compared to the D-04s.

And I think that the fact that they needed much less medication adjustment, you know, I think does go along with, you know, that there was some effect in there. So to me I think that even though there was no definite data that you could compare and there was a lot of alterations being made, if at all, it went skewed towards the D-02 study.

CHAIRPERSON BECKER: Dr. Fochtmann?

MEMBER FOCHTMANN: The issue of the concomitant medications is one that I continue to have questions about, the issues that I raised earlier, which were answered, but also because of the opposite side, and that is could concomitant medications be influencing the efficacy of VNS, specifically the medications with anticonvulsant properties, given what we know about ECT efficacy being impaired, in some instances, by a medication such as benzodiazepines.

And so I think that without some attempt at standardizing some of the concomitant medications, it's difficult to know how to interpret one way or another what impact that the medications might have on the VNS efficacy.

The other issue is just in terms of the wide variety and the number of medications that people were taking concomitantly, which makes it difficult to know how to interpret. You could argue that because there was -- that was present in both groups that it should wash out across the groups but, again, it's hard to know.

But at the same time, hard to make a standardization given the number of failed trials that these individuals had already experienced.

CHAIRPERSON BECKER: I have nothing to add.

MEMBER WANG: I think this issue, you know, allowing changes in the concomitant treatment makes this data we're actually looking at not the efficacy of a device but we're now looking at the efficacy of sort of strategies, you know, and it really is hard to sort out because, well, for that reason.

And, again, as has been sort of raised again, this issue of the reduction in the magnitude of the effects estimate after you censor people who made changes or, you know, added ECT or that sort of thing, suggests that the rescue treatment may have been more robust, you know, a good rescue treatment. And maybe that is partially explaining the efficacy.

But on the other hand, what makes you analyses that you showed us conservative is the whole issue of ceiling effects. I wonder to the extent to which, you know, the fact that you allowed everyone to be on concomitant treatments, did they max out and are you not able to see sort of efficacy because everyone is on, you know, good regimens potentially.

DR. JENSEN: I agree with Dr. Wang.

MEMBER ORTIZ: My comment about this would be that it would be helpful, I would think, to get further information both about the types of ECT, at what point it was used, the specifics of antidepressants.

As Dr. Fochtmann was saying, some of the antipsychotic medicines actually lower the seizure threshold as well as does buproprion. And again those kinds of issues I think will be very important for clinicians to understand better because I think though the request is only for the VNS, the reality is clinicians will be combining it.

And the more information they have the better.

CHAIRPERSON BECKER: Dr. Malone?

MEMBER MALONE: I agree that it's difficult to know the effect of the concomitant medicines in ECT. I don't know if there's any way a round having this. It could maybe be more standardized in a protocol.

But I think obviously it would have some effect on the outcomes.

CHAIRPERSON BECKER: Ms. Wells?

MEMBER WELLS: I have no comment.

CHAIRPERSON BECKER: Mr. Balo:

MR. BALO: I just think, you know, like everybody said, it's going to be pretty difficult. It seems like this is a very difficult patient population and the amount of ECTs or the amount of different medications they take would be very difficult to sort out.

And I think the sponsor has really done -- at least like Dr. Ellenberg said, teased out as much as they could from the study that they did.

CHAIRPERSON BECKER: So in summary, it sounds like the panel believes that because this wasn't the randomized trial, it's hard to know what to make of the concomitant medications, especially in light of the fact that there's no standardized approach to medically treating these patients.

The sponsor did a good job in trying to sort it out but I think we're still left at the end of the day without really knowing what to do with concomitant medications.

CHAIRPERSON BECKER: Next we move on to questions of safety and efficacy. The FDA regulations, specifically 21 CFR 860.7(d)(1) states that there must be a reasonable assurance that a device is safe when it can be determined that the probable benefits to health from use of the device for its intended uses when accompanied by adequate instructions for use and warnings against unsafe use outweigh any probable risks.

And so the question for the panel is do the clinical data in the PMA provide reasonable assurance that the device is safe?

I'll start with Mr. Balo.

MR. BALO: I'm not a medical doctor. Basically I'm an industry representative. You know in dealing with these studies and putting these studies together, industry basically works closely with the physicians, with the medical community, and with the FDA to put forward a study that they feel will be safe and will be efficacious.

I think the sponsor -- and to my opinion, from the data they showed, I believe there's a lot of points that are made by Dr. Jensen, by Dr. Fochtmann, if I said your name correctly, about the safety of the device. There are some concerns with young patients and the future effects that the device may have.

But looking at and listening to some of the patients speaking today about, I guess, their new lives that they gained back, I would think that from a safety perspective, I think it's a balancing act for me.

I would really have to look at the patient. I would look at the condition. But I do think that the data they did show today, at least to me, showed that it was a device that would be safe.

CHAIRPERSON BECKER: Ms. Wells?

MEMBER WELLS: Again, I think the options are so limited for this particular disease process that we have to consider especially the patients that came forward this morning and spoke to us about their device experiences.

So I think this is something that we really need to consider as a panel.

CHAIRPERSON BECKER: Dr. Malone?

MEMBER MALONE: I would consider safety against efficacy or, you know, the cost benefit ratio and since I'm not sure that they've shown benefit, I think there are safety concerns. So I think the safety outweighs the benefit.

CHAIRPERSON BECKER: Dr. Ortiz?

MEMBER ORTIZ: Yes, I believe that the safety is documented by the data presented on the depression studies as well as the seven years with the use in epilepsy.

CHAIRPERSON BECKER: Dr. Jensen?

DR. JENSEN: I think you've met the burden of saying that this is a safe device when compared to the patients that have epilepsy. I didn't see any increased incidents of problems in this particular group so I don't think the disease process, having the device with this disease process makes a big difference.

Again, my big issue is just the 70 percent of patients that have an implantable device that does not work that they now have forever and the long-term implications that go along with that, particularly in further imaging and/or potential surgeries.

Having said that, I still feel that the device is safe but I think the company should certainly look at some way of addressing those patients who have a device that does not show any improvement in their condition. And how, if they so desire, would like it removed, have that done.

MEMBER WANG: I have nothing to add beyond what's been said.

CHAIRPERSON BECKER: It appears to me that the device is safe. It has some annoying side effects but in general it appears quite safe.

MEMBER FOCHTMANN: My impression is also based on the data presented, that the device shows adequate safety, particularly when weighed against the risks of continuing, persistent, treatment-resistant depression.

The -- I believe that the registry plan that was outlined earlier would be extremely helpful in providing further information about the long-term effects of the treatment. And I don't know whether it's possible as part of that to also look at specific issues of safety. For example if individuals need future ECT, safety issues at the time of the ECT with having this device in place, issues along those sorts of lines.

So I think that with the evidence that has been presented with the registry follow up plan that I would be comfortable with the safety.

MEMBER JAYAM-TROUTH: I agree with Dr. Jensen and I think that maybe, maybe you could evaluate your data a little bit more closely and see why are some people responders and why are some people not responders. Then maybe you don't need to implant it into everybody in the first place.

You know you might be able to glean some extra data and see if you need to put it into those 70 percent of people who are "non-responders." You know, and as far as the safety in epilepsy now I think it's been established. It's been there for a long time.

And there are only a few problems there. But I do not know of long-term studies, you know, on infants. You know I know they have put some of these in infants with Lennox Gastro Syndrome and infantile myoclonic spasms. And these are growing infants. And I do not know if they have any safety data, you know, on whether this was okay, you know, in those situations.

I think that too should be considered because they are among the epilepsy studies.

MEMBER ELLENBERG: My sense is that the safety profile has been adequately defined for the age population being considered but the cost benefit ratio issue I agree totally with Dr. Malone. That we don't have, the cost benefit ratio at hand on which to base the safety profile.

CHAIRPERSON BECKER: So in summary, Dr. Witten, it sounds like the panel believes that the device is generally safe but based on what is questionable efficacy, it's unclear whether the safety benefit ratio rises to the point that make it something that we should achieve to use.

So the final question has to do with efficacy, we're leading right into it then. And this is based on the FDA requirement 21 CFR 860.7(e)(1) which states that there should be a reasonable assurance that a device is effective when it can be determined, based on valid scientific evidence, that in a significant portion of the target population, the use of the device for its intended uses and conditions of use, when accompanied by adequate directions for use and warnings against unsafe use will produce clinically-significant results.

Considering your response to questions 1, 2, and 3, do the clinical data in the PMA provide reasonable assurance that the device is effective.

So Dr. Ellenberg, would you refresh the microphone?

MEMBER ELLENBERG: I don't believe that we have seen adequate evidence of efficacy from the data presented albeit the data has been presented in an excellent way.

And I believe that a randomized clinical trial will be the way that we have to see the efficacy determined.

MEMBER JAYAM-TROUTH: I agree that it appears that the device is effective.

MEMBER FOCHTMANN: I think we have seen some evidence of efficacy. Whether that meets the rigorous standard required in this question is not totally clear to me. Obviously a more rigorously designed study would help in answering that.

CHAIRPERSON BECKER: I think there are certainly hints to efficacy. I think it's not been proved in the way that we're used to seeing other treatments proved in medical trials.

MEMBER WANG: I basically think the D-02/D-04 data are essentially not really contributory. But again, I'll just emphasize, I think the acute phase data are extremely positive. In my mind, you know, you had a tendency on the HAM-D and you had a significant finding on the IDS-SR. So I do think there's some evidence, albeit weak for efficacy.

But let me just say there's really two questions. One is is it effective? And then second, is it as effective as other modalities such as ECT?

And from a public health perspective, that second question is also relevant since you don't want to necessarily divert people from, you know, other potential modalities that might help them.

DR. JENSEN: I think I'm struggling with the same issues as the rest of the panel. It appears to be efficacious in certain patients. And I'm also sensitive to the fact this is a very difficult patient population. Again, I see similar patient populations.

Part of me says yes, I'd love to see randomized controlled trials but in my heart I know it would be very difficult to do that with this particular patient populations.

I also don't was to see what happened in the Pro Act II Study, which is where we had data of efficacy for intraarterial thrombolysis only to be told we then needed to have another study and the company then decided not to pursue that. And it was never made available to the population.

MEMBER ORTIZ: I agree with the comment Dr. Becker made.

CHAIRPERSON BECKER: Dr. Malone?

MEMBER MALONE: I think in order to show that a treatment is effective in a psychiatric disorder, you need a randomized controlled trial, which is positive. And we don't have one.

So I don't think it shows efficacy. I do think that it is possible to do these studies because you did one. It just was a failed study.

CHAIRPERSON BECKER: Ms. Wells?

MEMBER WELLS: I agree with Dr. Jensen. I think her remarks are right on.

CHAIRPERSON BECKER: Mr. Balo?

MR. BALO: I sort of agree with Dr. Jensen and Ms. Wells but I also think, you know, we're sort of looking at this with a drug perspective and when you look at it from a company perspective, they're running this as a device study.

And from what I see what the company had did and the long-term effects, there are a group of people that have this disease that could benefit from this device.

And, again, balancing that act, I still would say that they have shown that there are patients who could benefit. And this would be effective for those patients.

CHAIRPERSON BECKER: So Dr. Witten, it appears we have a little consensus on this question. It seems that some of the panel members believe that the device has been shown to be effective. Others think more data is needed. And still others think that the device hasn't been shown effective for all patients but at least the hints of efficacy in this very treatment-resistant depression group might signal that it should be okayed for use.

So I think with the end of the FDA question, we'll move on to the second open public hearing on the Cyberonics Vagal Nerve Stimulation System, PMA 97003, Supplement 50.

Is there anybody from the audience who would like to address the panel now? If so, raise your hand and come toward the podium.

(No response.)

CHAIRPERSON BECKER: Okay if that's not the case, I think what we'll do is take a ten-minute break. So if everybody could return at 4:25 and we'll vote on the PMA.

(Whereupon, the foregoing matter went off the record at 4:16 p.m. and went back on the record at 4:30 p.m.)


CHAIRPERSON BECKER: It's 4:30, and we'll get started and try to finish this meeting up.

I think that we gave the sponsor a bit of a scare forgetting to mention that we will have summations now, and we'll start with the FDA summation if there is one, Dr. Witten.

DR. WITTEN: There is none.

CHAIRPERSON BECKER: So we'll move on to Mr. Totah and the sponsor's summation.

MR. TOTAH: At this point ‑‑ this is Alan Totah, Vice President of Regulatory Affairs ‑‑ at this point, I'm going to defer to Dr. Rudolph, but I will join in in a moment. Thank you.

DR. RUDOLPH: What we decided to do is we'd like to have several of us address the panel, and I'm going to start. Mr. Totah's going to contribute. Dr. Rush and Dr. Sackheim are both going to contribute as well.

The VNS safety data that we presented today, I think the panel agreed with us that although there may be some specific safety issues in genera, the safety is well established, both in the depression trials and in actual clinical use for epilepsy. Side effects do occur, they're generally mild, stimulation related, tend to diminish over time and rarely cause the patient to discontinue. We didn't find any indication for any specific safety concerns for this specific indication.

We'd like Dr. Sackheim to sort of make ‑‑ we have several topics we want to address, and we're going to ask Dr. Sackheim to talk about clinical benefit in this very ill patient population.

DR. SACKHEIM: Yes, thank you, and I understand that this is an important and difficult issue for many of us.

When we think about the niceties of research and the purity of designs, we also have to think about the population in which they're going to be applied. One of the things that I think certainly deserves emphasis here is that the types of individuals that are being considered for this treatment are individuals in whom the likelihood of a placebo response, even the consideration of a placebo response are quite small. These are not children, these are not individuals who haven't had many, many opportunities to demonstrate placebo responses before.

What reminds me in the less severe population in our work with electric convulsive therapy at Columbia we have forms of ECT where we have 17 percent of the patients responding after full course, depending on where in the brain we stimulate and with what type of electricity. That's acute, and what that means is that there's very little in the way of placebo response in this severe population. That's been demonstrated in studies of oncolic patients and the psychotically depressed patients.

But what's really unusual and what really actually stirred me in looking at the findings with VNS, because for a long time I've been quite critical, was the identification of the long-term benefit in these people, that I simply don't know in any treatment that we can point to that has as much promise in terms of sustaining a benefit if you get there. It's not that a lot of people get there, but if they get there, it looks like they hold and they hold it for a long time.

I spent a career working with patients with treatment-resistant depression. I worked clearly in the area of ECT where our expectations now are that in treatment-resistant patients if we get them better, if they remit, that they become virtually asymptomatic. Seventy percent of patients will lose that benefit within six months. That's pretty much the standard view. This is a context where we have a treatment where it looks like 70 percent will maintain it maybe for two years.

So I think there is tremendous promise here, and what we're debating hinges on the importance of one word: randomization. I'm the first to say that hard core clinical work is certainly to be valued, but I also think as you think through this little bit that control over concomitant treatments, the strength of inference in the randomized design in some way become comprised and lack feasibility in this population.

And I say that for the following reasons, I'll just give you one quick vignette. Where do you go with standards of care with these patients who have had 20 years in some cases of being in the same episode in Hamilton at 40 and have been treated by some of the best people in the country? Where we've gone has been to placed in pharmacology that put these people at risk, that's really on the outside of pharmacology, because they're hanging on by a threat. So standard of care is often very dangerous, unacceptable in many ways but have to become acceptable to these individuals.

We are going to have a lot of problems with concomitant medications, because you can't keep people for a long-term study in a narrow bind, particularly with these disorders. I would submit that randomization also is going to be a problem because of the selection bias that that would involve. It's that we are offering the same of nothing versus being randomized to something that might be helpful. There may be many patients who would reject that type of compromise and would go then on study.

In any case, to summate, the benefit we've seen so far is something that I haven't seen with any intervention for treatment-resistant depression. It's quite unusual. And it echoes, of course, what has been suggested for epilepsy. I think that very benefit and its nature indicates an effect that can't be accounted for by a fluke of randomization, a fluke of the assignments to different studies and is very, very unusual in the context of treatment-resistant depression. Thank you.

DR. RUDOLPH: I want to pick up the randomization theme a little bit, because what I picked up in listening to the panel deliberate is that was the most troubling aspect of the program that we presented to you today.

So, first, I would like to talk a little bit about the D-04 as a control. It was obviously a non-randomized control, but it should be thought of as, I think, something more than just a haphazard control. It had many of the elements that would give you a high degree of confidence in its ability to determine effectiveness. It did come from a prospectively designed study, there were overlapping sites, same exact principle enrollment criteria, and it was conducted over a similar time period. So this by itself should have ensured a lot of comparability. And in fact, as I showed you from the data, it did.

And in terms of considering, okay, if that's not good enough, you want a randomized trial, we did talk a lot today about what alternative trial designs might be, and I think for the most part the panel understood the limitations of many of the possibilities, particularly an extended placebo control trial wouldn't be viable in this population. An active treatment control with a single therapy wouldn't work in a population that's already churned through so many different treatments. And, again, if I understood the panel deliberations correctly, I think what most people gravitated to was essentially the D-02, D-04 comparison that we did but do it in a randomized control fashion.

I'd ask you to consider a few things. Even randomized control trials, while they are our gold standard, they do not necessarily guarantee that these baseline covariates are equally distributed between groups. And the other issue that was raised was the concomitant medication issue, but even in your deliberations, the way I understood them, you still came back to allowing the possibility of pretty much access to a variety of medications, as we did in D-04.

So, ultimately, what I took away from the discussion was that you would prefer a randomized trial, and the one thing that would do that the D-04, D-02 comparison did not do was it would provide a higher level of confidence that the patient populations did not differ in any significant way. So I think, ultimately, what we're asking the panel to consider is, is that by itself or the greater confidence that you would gain from a randomized control trial, is that by itself enough to delay approval of this therapy; that is, would you gain that much more confidence from randomization which essentially wouldn't address the medication issue any better than the paradigm we used, it would only perhaps, in theory, give you some greater level of confidence that baseline covariates were equally distributed.

And as you're considering that question, consider some of the analyses that you saw during the day, particularly, I would say, not only those that show that the patient groups were very well matched and that the propensity adjustment added some further confidence that the patient groups or that baseline covariates were not the explanation for the difference, but also consider some of the analyses that you may have forgotten about that Dr. Davis presented where we did look at what would be the effect of a single covariate, and we used all the covariates that did differ significantly, the measured covariates, and we used those as examples of if you adjust for that, what is the impact on the effect size, the linear effect, that is, or the p-value and confidence limit. And you saw that any one of those didn't contribute in any meaningful way to the overall statistical significance of the study.

So, again, I guess to kind of shorten it, the bottom line would be I would ask you to consider would randomization, which would essentially, if I understand correctly, mainly benefit only in terms of giving some greater theoretical confidence that the patient groups would be comparable than we've already shown, is that worth delaying approval of this product for?

One other issue before we have Dr. Rush close, that we weren't sure that the panel, particularly the people with more of the psychopharm background fully appreciated was the standards for approval of a device, so Mr. Totah will address that, and then Dr. Rush will close.

MR. TOTAH: Thank you, Richard. Again, I'm Alan Totah, Vice President of Regulatory Affairs. When the FDA quoted 21 CFR Part 860.70, which has to do with scientific evidence, and charged the panel with the questions that you went through today, what they didn't give you is the full context of that regulation, and I'm going to read to you because we had a question from Dr. Malone that didn't get answered. I tried to get up here but we ran long, and so now's my chance to answer his question.

I'll quote out of 21 CFR 860.70. device regulations. "Balanced scientific evidence is evidence from well-controlled investigations, partially controlled studies, studies and objective trials without matched controls, well-documented case histories conducted by qualified expert and reports of significant human experience with a marketed device."

Now, I think important to keep in mind after hearing that regulation for those of you that come from the drug side or pharma side but you may not be familiar with this part of the regulation is keep that mind. Active controls obviously fall within the scope of this regulation.

Now, what else I want to tell you is in my earlier speech but just a few more details: History of approved PMAs on the medical device side. Fifty-five percent of all approved PMAs were supported by non-randomized clinical trials. This is for all time. Forty-eight percent do not include randomized control trials. Patients as their own control, or non-randomized active control, fall into that group. Seven percent include no controls whatsoever, and 45 percent ‑‑ only 45 percent ‑‑ include randomized control trials.

Now, the basis for what I'm giving you is a CDRH Staff College report on least burdensome provisions of the FDA Modernization Act of 1997, and I'm giving you information from a March 19, 2000 presentation by the CDRH Staff College. I think you need to keep that in mind, or at least I respectfully request that you do that, because that is the difference, one of the differences between the drug side and the device side. Thank you.

DR. RUSH: Just briefly, I want to add a brief comment to the issue of efficacy. There are some things in medicine when you see them are pathognomic. You don't see them often but when you see them, it really means a lot, like they really have the illness. So what is pathognomic here about efficacy? I'll just put three on the table.

One is the induction of bipolar disorder in 22 percent of patients. We see that in effective antidepressants. I know it's uncontrolled. You have patients who have lost the battery, a battery shutdown, their depression came back. The battery was replaced, the depression went away. That's pathognomic of activity.

And, thirdly, you have a predictable course of treatment-resistant depression, unlike other kinds of depression. The follow-up from the ECT patients, the non-responders to ECT that I showed you in the graph from Dr. Sackheim, continue to be in a terrible state, no better for a year. For the D-04, unchanged as a group for a year. From the Texas Medication Algorithm Project, single-digit sustained response rates, 14 percent, in the best case with algorithm done twice as well in using that as a benchmark, and they're not TRD.

So if you have an improvement that grows over time, which appears to be true, looking at it in an uncontrolled fashion with D-02 long term that's pathognomic of activity of the course of illness, is either the same or worsening.

Finally, I just want to point out, and I'm sure you are aware because of the patients' testimony and your own knowledge, that this treatment-resistant depression for which we have no other available effective treatments at the moment, is highly lethal and during the time it will take to do another randomized control trial, we'll lose another 1,000 a patients a month, 36,000 if it takes three years. There's a desperate need out there for this treatment, and I understand that if you look at it and you look at the pathognomic evidence of efficacy as well as the randomized trial evidence, that I think you would persuaded to ‑‑ safety having been established ‑‑ to approve this device at this time. Thank you.

CHAIRPERSON BECKER: Thank you. Ms. Scudiero will now three possible panel recommendation options for pre-market approval applications.

DR. MALONE: The biggest threat of regulation that has all these different standards of evidence or something, I can't imagine that you can just pick whichever one you want but that you would be trying to pick the level of evidence that was appropriate to the device; is that right?

CHAIRPERSON BECKER: I'll ask Dr. Witten to comment on that.

DR. WITTEN: Yes. That's what I was going to say. Those all are acceptable forms of evidence for us, all the ones that he listed. And then for each specific case, as in this case, we're asking the panel to evaluate whether based on what they provided, whether reasonable assurance of safety and effectiveness has been provided. But that means that everything could be accepted if it provides reasonable assurance of safety and effectiveness.

DR. MALONE: But each sort of device could demand a different level; is that true?

DR. WITTEN: Yes. I mean, in part, that's part of why we're here is we're asking for your recommendations on this data set for this device.

CHAIRPERSON BECKER: Ms. Scudiero?

MS. SCUDIERO: Okay. These are on the back of the meeting handouts, the fourth page. The medical device amendments to the Federal Food, Drug and Cosmetic Act, as defined by the Safe Medical Devices Act of 1990, allows the Food and Drug Administration to obtain a recommendation from an expert advisory panel on designated medical device pre-market approval applications, PMAs, that are filed with the agency. The PMA must stand on its own merits, and your recommendation must be supported by the safety and effectiveness data in the application or by the applicable publicly available information.

Safety is defined in the Act as reasonable assurance based on valid scientific evidence that the probable benefits to health under the conditions of intended use outweigh any probable risks. Effectiveness is defined as reasonable assurance that in a significant portion of the population the use of the device for its intended uses and conditions of use when labeled will provide clinically significant results.

Your recommendation for the vote are as follows: One, approvable if there are no conditions attached; two, approvable with conditions. The panel may recommend that the PMA be found approvable subject to specified conditions such as physician or patient labeling education, labeling changes or further analysis of existing data. Prior to voting, all the conditions of approval should be discussed by the panel. Three, not approvable. The panel might recommend that the PMA is not approvable if the data do not provide reasonable assurance that the device is safe or if a reasonable assurance has not been given that the device is effective under the conditions of use prescribed, recommended or suggested in the proposed labeling.

Following the voting, the Chair will ask each panel member to present a brief statement outlining the reason for his or her vote.

CHAIRPERSON BECKER: Is there a motion from the panel? Dr. Wang?

DR. WANG: Approvable with conditions.

CHAIRPERSON BECKER: Is there a second for the motion?

PARTICIPANT: Second.

CHAIRPERSON BECKER: I hear a second. So at this point, I guess I will entertain an amendment to the main motion for the first condition of approvability. Is there a motion for a condition of approvability?

DR. WANG: Yes. The condition, I wonder if it wouldn't be helpful to have a condition for both scientific and also public health reasons that there be a failure of more than two or more trials, to maybe something like four or five, and here's my reasoning. The scientific reason is I think we may be going beyond the generalized ability of the data. You showed us data suggesting that these people have ‑‑ they had nearly four, on average, fail trials just in this episode, and on average I think it was 12 or so failed trials. So to extend these results to a population that may have only failed two trials may be going beyond the limits of this data.

The second is a public health reason, and that is given the, let's say, less than robust data right now on efficacy, I think there's a concern, public health concern, which I alluded to earlier, that patients who have only failed two trials, and you can get there pretty fast, you just have to fail two medication trials in the span of a few weeks and you'd be eligible for this, you might forego modalities which have a much stronger evidence for them. And by that I mean ECT, lithium augmentation, maybe dual modalities, psychotherapy, plus medications that haven't been tried.

So I think if you raise the bar to four or more failed trials or five or more failed trials, something like that, you at least would ensure that patients have had a chance to go through some of the modalities that have stronger evidence bases than I think currently exist for VNS.

CHAIRPERSON BECKER: So, Dr. Wang, would I be correct in saying that your motion would be that patients need to fail at least four trials of approved medical therapy?

DR. WANG: I would take guidance here from sort of the other ‑‑ the clinicians in the room how many sort of modalities do we think have at least as much evidence suggesting their efficacy. My guess is four or five, something like that.

CHAIRPERSON BECKER: Is there anybody who seconds that motion?

DR. FOCHTMANN: I would second that motion.

CHAIRPERSON BECKER: Do you want to add on a little bit?

DR. JAYAM-TROUTH: Can I kind of add on a little bit.

CHAIRPERSON BECKER: Sure.

DR. JAYAM-TROUTH: Thank you. I think at this point that indication statement where you have indicated in your write-up that VNS therapy indicated for use as an adjunctive long-term treatment of chronic or recurrent depression for patients over the age of 18 who are experiencing a major depressive episode but has not had an adequate response to two or more adequate antidepressant treatments needs to be definitely modified. I agree with Dr. Wang, and I also think that it should be for treatment-resistant depression that should be considered.

CHAIRPERSON BECKER: Would anybody like to discuss this motion for approval ‑‑ this condition of approval, I mean? No further comments? So if that's the case, then I guess we're ready to vote on the first condition of approval. We'll do each one individually, so we'll go on the first one.

All in favor of the first condition of approval, which is that patients must fail at least four or more trials or somewhere thereabouts of medical therapy prior to implantation with a VNS, please raise their hand.

So Dr. Jayam-Trouth, Dr. Fochtmann, Dr. Jensen, Dr. Wang, so that's four.

All opposed to the first condition of approval, please raise your hands. Dr. Ortiz.

And all abstaining from voting on this condition of approval. Dr. Ellenberg and Dr. Malone.

DR. MALONE: I wouldn't vote for approval, so I don't know how to vote on this condition.

CHAIRPERSON BECKER: So you're abstaining then.

Does anybody have a second condition that they would like to move for approval?

DR. JENSEN: I have a couple, actually. The first has to do with MD education once the device is approved and anybody can use it, and it's not going to be in the 20 centers where the best of the best are doing procedures. So I think we have to look at the lowest common denominator. I think surgeons should be identified by the number of nectosections they do a year, and there should be a minimum number of nectosections they do a year to show that they are actually capable of implanting the device.

I think there needs to be identification of the psychiatrists and their ability to show appropriate use of the device in patients, not just in the lab and ‑‑

CHAIRPERSON BECKER: We need to go one by one.

DR. JENSEN: One by one. Okay.

CHAIRPERSON BECKER: Does anybody ‑‑ would anybody like to second Dr. Jensen's condition 2 that the surgeons need to be identified for the number of nectosections they do and their ability to perform those nectosections? A second for that motion?

DR. FOCHTMANN: Second.

CHAIRPERSON BECKER: So we have a second from Dr. Fochtmann. So at this point, we need to vote on that second condition. All in ‑‑ or any discussion before we move on this motion? Anybody want to ‑‑

DR. JAYAM-TROUTH: Yes. I have a point, and that is that many people will be starting new and fresh, and you can't tell them, "How many have you done," when they've none at all. I think there should be teaching for them so that they are familiar with it and they can do it. But then to stipulate that you have to have ten when they have to start, I think it's not feasible.

CHAIRPERSON BECKER: I guess I would say that most vascular surgeons or neurovascular surgeons will have done nectosections, so that shouldn't be an issue. Putting the leads on may be new for them and that's probably not as big a concern, but I think the surgeons should at least know how to get into the neck safely, into the carotid sheath safely.

All right. So let's take a vote for the second condition of approval, which is that the surgeons need to be identified for their ability to operate in the carotid sheath.

All in favor of the second condition of approval raise your hands. And that would be Dr. Ortiz, Dr. Jensen, Dr. Wang, Dr. Fochtmann and Dr. Jayam-Trouth.

All against this condition of approval? All opposed?

And all abstaining? That would be Dr. Malone ‑‑ and Dr. Ellenberg, sorry.

Any motions for a third condition for approval? Dr. Jensen?

DR. JENSEN: Again, in terms of MD education, I similarly like to see the psychiatrists identified as their ability to show appropriate use of the device in patients, not just necessarily in the lab, but they should be required to take a course and then have their first three, four patients checked in some manner to make sure the programming is appropriate.

CHAIRPERSON BECKER: So Dr. Jensen would like psychiatric training for programming the VNS device. Anybody second that motion?

DR. JAYAM-TROUTH: Second.

CHAIRPERSON BECKER: Dr. Jayam-Trouth. So anybody like to discuss that point that psychiatric education should be included in the condition for approval?

DR. ORTIZ: I wonder if it should be expanded, because it's not necessarily just psychiatrists who might program that. I mean we may be talking about behavioral neurologists or other people, so it's more of an engineering kind of thing.

CHAIRPERSON BECKER: So essentially any of the clinicians who will be taking care of these patients and need to have training prior to being able to have a patient implanted.

DR. FOCHTMANN: Would this be ‑‑ would you perceive this as requiring as some sort of specific certification or just showing that you've gone to a continuing education course?

DR. JENSEN: Well, I think that, clearly, you have to go to a course that should be run by the company on how to use the device, but for me the issue always comes down to when you're doing it in the lab by yourself for the first time, did you do it right? And so I think there needs to be a mechanism to make sure that you've said you've set this thing to a certain standard and that's correct.

Now, I don't use the device, maybe it's very simple and all it would take is somebody from the company coming behind and saying, "Check," or somebody else who already uses the device in the hospital checking or whatever, but I just want to make sure that when people are getting an implantable device, that it's being programmed correctly, and that that's somehow documented.

CHAIRPERSON BECKER: Okay. So I think it's time to vote on this condition for approval, which is that the clinicians who are caring for these patients who have devices implantable need to show some sort of documentation that they are able to understand how to program and change the parameters of simulation on the device.

All in favor of this motion raise their hands. So Dr. Ortiz, Dr. Jensen, Dr. Wang, Dr. Fochtmann and Dr. Jayam-Trouth.

All opposed to this condition for approval raise your hand?

And everyone abstaining from this vote. Dr. Ellenberg and Dr. Malone.

Are there motions for any other conditions for approval? Dr. Jensen?

DR. JENSEN: For patient education, I think that it needs to be clearly stated, and the patients know that the device implant may affect their ability to have diagnostic or therapeutic procedures in the future and that they do have the option of having the device removed if need be, if they are a non-responder or whatever and they want to have it out and that they receive some sort of identification bracelet, card, et cetera, that identifies them as having this implant and what the implications are for MR use or other sorts of imaging surgeries.

CHAIRPERSON BECKER: So it sounds like this condition for approval has to do, in part, with labeling, and, obviously, as part of, I think, any delivery of medical care, we'd want to inform our patients completely of the risks and benefits involved in having the device implanted, which include the risks of having a limited ability to obtain diagnostic radiographic tests, and so that needs to be very clearly spelled out to the patients and perhaps have some sort of identification that they carry with them like someone who has a pacemaker and carry an identification with them.

Is there anybody who seconds this motion?

DR. JAYAM-TROUTH: Second.

CHAIRPERSON BECKER: There's a second. So everybody in favor of condition 4, which is patient education and some sort of identification for the patients that they have this VNS device implanted, please raise your hands. I'm sorry, before we raise our hands, discussion points, I'm sorry. Seems pretty straightforward. Anybody want to discuss that point? No?

All right. So now everybody in favor of Condition 4 raise your hands. Again, it seems like our usual group: Dr. Jensen, Dr. Wang, Dr. Fochtmann and Dr. Jayam-Trough and Dr. Ortiz. Did I see your hand there or not?

DR. ORTIZ: Yes, you saw it.

CHAIRPERSON BECKER: Okay. All opposed to this condition for approval raise your hands.

And all abstaining from voting? Dr. Malone and Dr. Ellenberg.

Any motions for further conditions for approval? We've exhausted you, Dr. Jensen?

DR. JENSEN: Well, I do have one question about the registry. Can we ask that certain data be culled from the registry? They're planning on having a registry, but one of the things that I would like to see is that the parameters that are looked at in the registry are evaluated for patients who are non-responders looking for particular group types that are non-responders, and if we find one, that this is a group of patients who do not respond to this device, unequivocally, that that could then end up being a contraindication for use.

CHAIRPERSON BECKER: So let me ask, actually, Dr. Witten, is that something that we can request that the sponsor do to collect certain data in the post-marketing registry to the FDA?

DR. WITTEN: Yes, especially given that Dr. Jensen has stated a specific purpose for this.

CHAIRPERSON BECKER: So with that motion for identifying specific clinical data be collected in the patient registry, is there anybody who'd like to second that motion?

DR. JAYAM-TROUTH: Second.

CHAIRPERSON BECKER: Dr. Jayam-Trouth seconds that motion. And anybody want to discuss this point any further?

DR. FOCHTMANN: I'd like to discuss this. I think that it's a reasonable idea to collect the data. I am a little bit concerned about identifying specific subgroups of individuals who would be then designated as being contraindicated to receive this device, because I think, as has already been quite well described, this is a treatment that people go to when they really don't have other viable options, and I don't think that any statistical analysis that shows that one subgroup was less likely to respond is going to be an absolutist sort of subgroup, and so I would be very concerned on the basis of subsequent analysis thereby denying a potentially effective treatment to individuals in need.

DR. JENSEN: Yes. I guess I should have clarified when you asked the question about what FDA is allowed. If we got information that showed that there were certain prognostic factors which gave you a better idea of when a patient would respond, that would typically be used in a labeling update. Typically, contraindications are when there is a safety problem that's been identified with the device. So I was just responding to the question about whether you could ask for this data in the registry, not about the other part of the recommendation, about contraindication.

DR. FOCHTMANN: I would just like that information in the labeling.

DR. JENSEN: Right.

DR. JAYAM-TROUTH: One other thing: Would it interfere with the HIPAA and all that, the regulations, with the demographic, et cetera, that we collected and put out there in the registry, anybody could get access to it? Wouldn't that come in the way of patient confidentiality?

DR. WITTEN: If the Sponsor's planning a registry, then I'm assuming that they've looked at how they were going to do that in such a way that wouldn't be in contradistinction to what they're required to do under HIPAA.

DR. FOCHTMANN: Would that same issue of HIPAA also apply to manufacture or checking device parameters and operation, that they would check on that as well? I'm not used to a situation where manufacturers are watching me or in the room with me when I'm taking care of patients.

DR. WITTEN: Let me say, I'm not an expert on HIPAA, but a lot of it would depend on how it was done and what the patient was informed of, I think. But what we're looking for from you, the Panel, is your recommendations about the kinds of ‑‑ if you're recommending things in a registry or further information to be collected, we're looking to you for recommendations about the kinds of information you'd be interested in and how you would see this information being used to further public health and not the specifics of exactly how something would get done. That's something we would discuss later if we decided to implement those conditions. That's the kind of thing we would discuss specifics with the sponsor. So I don't think you need to be concerned about those questions. We'd like to hear what it is you think is needed or you'd like to recommend.

DR. JAYAM-TROUTH: Can I kind of voice another concern? I don't know if it's a recommendation but my concern is that this is still only a short time for this device. We're talking about just a few years, and I'm not sure down the road maybe some of the side effects will come. Can we stipulate now that at the moment that if some extra side effects are seen or something else happens, that it's brought to the attention of the FDA or is that normal?

DR. WITTEN: There is a process by which sponsors are required to report to us on a periodic basis and report to the MDR system also new safety information about the device. That doesn't need its own condition, unless there's some specific thing you're asking us to look for.

CHAIRPERSON BECKER: Dr. Witten, would you like the Panel now to make some recommendations about the data to be collected or is this something that could be worked out after the meeting between the FDA and the Sponsor?

DR. WITTEN: Well, so far what I've heard is a registry to try to identify prognostic factors to determine who might best benefit from this device. Is that right?

DR. JENSEN: Correct.

DR. WITTEN: Okay. And so if you have any ‑‑ it's up to the Panel. The Panel can either stop there or if the Panel has any specific suggestions about the kind of data that they think would be useful to collect in an effort these prognostic factors, that would be useful too. So it depends on whether you want to add anything to that.

CHAIRPERSON BECKER: Does anybody on the Panel have any thoughts about what other pieces of information should be collected of what else we should look for in collecting information for the registry? What pieces of information do we want to get out of it, what do we want to learn?

DR. FOCHTMANN: I would think, obviously, information about the stimulus settings and not just including pulse widths and the other aspects of the stimulus parameters, information, obviously, about efficacy in terms of patient perceptions and in terms of clinician perceptions, obviously information about safety, adverse effects. Those would be the key elements in addition to the other patient characteristics, concomitant medications, things like that.

CHAIRPERSON BECKER: Any other thoughts or discussion?

DR. ORTIZ: I guess I would want to follow what you were suggesting. It seems like, at least from the Pharmacology Committee, that a lot of that is better worked out between the FDA and the Sponsor directly.

CHAIRPERSON BECKER: So with that, I think it's the recommendation of the Panel that the pre-market approval application, P970003 ‑‑ there's more conditions? I'm sorry. So I think there is a motion for another condition.

DR. JENSEN: Well, have you voted on the registry?

CHAIRPERSON BECKER: I think we voted on it before we talked about the specific information to be collected.

DR. FOCHTMANN: The motions that I had related to specific aspects of the wording on the labeling claim. Is that something that is appropriate to comment on?

CHAIRPERSON BECKER: Sure.

DR. FOCHTMANN: The first comment that I would have is that on Points 2, 3 and 4, I think it should specifically state 12-month open label follow-up of the randomized control trial so that it doesn't give the impression that there was a 12-month randomized control trial to the person who is not totally familiar with these studies.

CHAIRPERSON BECKER: Is there a second to the motion in changing that labeling information?

DR. WANG: Second.

CHAIRPERSON BECKER: Dr. Wang seconds it. Any discussion on that point?

Everybody in favor of changing the labeling to reflect the fact that it was not a 12-month randomized control trial raise their hands. Dr. Ortiz, Dr. Jensen, Dr. Wang, Dr. Fochtmann and Dr. Jayam-Trouth.

Everybody opposed to that motion raise their hands.

And everybody abstaining? Drs. Malone and Ellenberg. Thank you.

DR. FOCHTMANN: The second wording issue that I would have would be in Point 4. Since there was a degree of variability in the results of the trials depending on what outcome measure was used, I think that the phrase, "highly statistical significant p less than 0.0001," should be changed to, "showed a significant," and there's a word missing there, "effect for treatment."

CHAIRPERSON BECKER: So the motion is to change the wording from, "highly statistically significant effect," to just, "a significant effect." Is there a second for that motion?

DR. FOCHTMANN: And to delete the, "p less than 0.0001."

CHAIRPERSON BECKER: And delete the p value. A second for that motion?

DR. WANG: Second.

CHAIRPERSON BECKER: Dr. wang. Any discussion on that motion?

Everybody in favor of changing that labeling information raise their hands. Dr. Ortiz, Dr. Jensen, Dr. Wang. Dr. Fochtmann, Dr. Jayam-Trouth.

Everybody opposed?

Everybody abstaining? Drs. Malone and Ellenberg.

DR. FOCHTMANN: The next wording point that I would have is in Point Number 6 where is says, "VNS therapy should be considered." I would like to suggest that that be changed to, "VNS therapy may be considered," since I don't believe that it's fair to say that any treatment absolutely has to be considered for every patient.

DR. JAYAM-TROUTH: Second.

CHAIRPERSON BECKER: Second. Thank you. Any discussion on that point?

Everybody in favor of changing the labeling to, "VNS therapy may be considered," as opposed to, "should be considered," raise their hands. Dr. Ortiz, Dr. Jensen, Dr. Wang, Dr. Fochtmann, Dr. Jayam-Trouth.

Everybody opposed?

And everybody abstaining? Drs. Ellenberg and Malone. Thank you.

DR. FOCHTMANN: And my final suggested amendment is Point 12, which states, "Brain imaging studies have demonstrated that VNS modulates blood flow and/or metabolism in many areas of the brain that are affected to mood disorders." I would suggest that the data that's presented, although very interesting, is in small groups of individuals and would be considered, I believe, to be most people to be preliminary data, and I would suggest that this point be deleted entirely.

DR. JAYAM-TROUTH: Second.

CHAIRPERSON BECKER: Thank you. And is there any discussion on this point?

Everybody in favor of deleting information on blood flow changes with VNS stimulation may I see your hands? Dr. Ortiz, Dr. Jensen, Dr. Wang, Dr. Fochtmann and Dr. Jayam-Trouth.

Everybody opposed to deletion of this point?

And everybody abstaining? Dr. Malone, Dr. Ellenberg.