UNITED STATES OF AMERICA
FOOD AND DRUG ADMINISTRATION
CENTER FOR DEVICES AND RADIOLOGICAL HEALTH
MEDICAL DEVICES ADVISORY COMMITTEE
* * * * *
CIRCULATORY SYSTEM DEVICES PANEL
* * * * *
* * * * *
JUNE 22, 2005
* * * * *
The Panel met at 8:00 a.m., in Salons A, B and C of the Gaithersburg Hilton, 620 Perry Parkway, Gaithersburg, Maryland, Dr. William H. Maisel, Chairman, presiding.
WILLIAM H. MAISEL, M.D., M.P.H. CHAIRPERSON
SHARON-LISE NORMAND, PH.D. MEMBER
RICHARD L. PAGE, M.D. MEMBER
JOHN C. SOMBERG, M.D. MEMBER
CHRISTOPHER J. WHITE, M.D. MEMBER
CLYDE YANCY, M.D. MEMBER
EUGENE H. BLACKSTONE, M.D. CONSULTANT
JEFFREY BORER, M.D. CONSULTANT
ROBERT M. CALIFF, M.D. CONSULTANT
THOMAS B. FERGUSON, M.D. CONSULTANT
NORMAN S. KATO, M.D. CONSULTANT
CYNTHIA M. TRACY, M.D. CONSULTANT
THOMAS A. VASSILIADES, JR., M.D. CONSULTANT
GEORGE W. VETROVEC, M.D. CONSULTANT
JUDAH Z. WEINBERGER, M.D. CONSULTANT
MICHAEL C. MORTON INDUSTRY REPRESENTATIVE
LINDA MOTTLE, MSM-HSA, RN, CCRP, CONSUMER REPRESENTATIVE
GERETTA WOOD EXECUTIVE SECRETARY
C O N T E N T S
Conflict of Interest Statement ................. 4
Introductions .................................. 6
Voting Status Statement ........................ 8
George Hawkins .......................... 11
Dr. Spencer Kubo .................... 14, 70
Dr. Douglas Mann ........................ 18
Dr. Mariell Jessup ...................... 39
Dr. Michael Acker ....................... 76
Dr. Michael Berman ..................... 114
Dr. Illeana Pina ....................... 140
Dr. Julie Swain ........................ 161
Dr. Brock Hefflin ...................... 169
Dr. John C. Somberg .................... 196
Dr. Clyde Yancy ........................ 204
Panel Discussion ............................. 233
FDA Questions ................................ 321
P R O C E E D I N G S
CHAIRPERSON MAISEL: Good morning. My name is William Maisel. I'd like to call to order this meeting of the Circulatory System Devices Panel.
Today's topic is discussion of a premarket application for the Acorn cardiovascular CorCap, CSDP040049.
Geretta, would you please read the conflict of interest statement?
MS. WOOD: The following announcement addresses conflict of interest issues associated with this meeting and is made a part of the record to preclude even the appearance of an impropriety. To determine if any conflict existed, the agency reviewed the submitted agenda 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 interests.
However, the agency has determined that 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 have been granted for Drs. Eugene Blackstone, Robert Califf, Judah Weinberger, and Christopher White for their employers' interest in the sponsor's study. The waivers involve a grant to their institution for which they had no involvement and have no knowledge of the total funding.
The waivers allow these individuals 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.
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 him or herself from 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 would also like to note for the record that Cynthia Tracy was unable to attend the meeting today.
CHAIRPERSON MAISEL: Thank you, Geretta.
At this point I'd like to have the panel members introduce themselves.
I'm William Maisel, a cardiologist at Brigham and Women's Hospital, and why don't we start with our industry rep., Michael?
MR. MORTON: I'm Michael Morton. I'm the industry rep., and I'm employed by Medtronic.
DR. KATO: Norman Kato, cardiothoracic surgery, Los Angeles California.
DR. NETROVEC: George Vetrovec, Chief of Cardiology, Virginia Commonwealth University, Richmond.
DR. BLACKSTONE: Eugene Blackstone, Director of Clinical Research in the Department of Thoracic-Cardiovascular Surgery, Cleveland Clinic.
DR. WHITE: Chris White, cardiologist, New Orleans, Louisiana.
DR. NORMAND: Sharon-Lise Normand. I'm Professor of Health Care Policy and Biostatistics at Harvard Medical School and Harvard School of Public Health.
DR. FERGUSON: Tom Ferguson, cardiothoracic surgery, Washington University School of Medicine, St. Louis.
DR. YANCY: Clyde Yancy, heart failure and heart transplantation, UT Southwestern Medical Center in Dallas.
MS. WOOD: Geretta Wood, Executive Secretary.
DR. SOMBERG: John Somberg, Rush University, Chicago.
DR. CALIFF: Rob Califf, Duke University.
DR. BORER: I'm Jeff Borer from Wile Medical College, Cornell University.
MS. MOTTLE: Linda Mottle, Gateway Community College, Phoenix.
DR. VASSILIADES: I'm Tom Vassiliades, cardiovascular surgery at Emory University in Atlanta.
DR. ZUCKERMAN: Bram Zuckerman, Director, FDA, Division of Cardiovascular Devices.
CHAIRPERSON MAISEL: Thank you.
Geretta, would you please read the voting status statement?
MS. WOOD: Pursuant to the authority granted under the Medical Devices Advisory Committee charter dated October 27th, 1990, and as amended August 18th, 1999, I appoint the following individuals as voting members of the Circulatory System Devices Panel for this meeting on June 22nd, 2005:
Eugene Herbert Blackstone, M.D.
Thomas T. Ferguson, M.D.
Norman S. Kato, M.D.
Thomas A. Vassiliades, Jr., M.D.
George W. Vetrovec, M.D.
Judah Z. Weinberger, M.D., Ph.D.
For the record, these individuals are special government employees and are consultants to this panel under the Medical Devices Advisory Committee. They have undergone the customary conflict of interest review and have reviewed the material to be considered at this meeting.
The agency also would like to note that Dr. William Maisel has consented to serve as Chair for the duration of this meeting.
Please strike that last statement. Dr. Maisel is our permanent Chair.
And that's signed by Daniel G. Schultz, M.D., Director of Center for Devices and Radiological Health.
I also have a separate temporary voting status. Pursuant to the authority granted under the Medical Devices Advisory Committee charter for the Center for Devices and Radiological Health, dated October 27th, 1990, and as amended August 18th, 1999, I appoint Dr. Califf and Dr. Borer as voting members of the Circulatory System Devices Panel for the June 22nd, 2005, session of the meeting.
For the record, Dr. Borer is consultant to the Cardiovascular and Renal Devices Advisory Committee of the Center for Drug Research and Development.
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.
And this is signed by Sheila Derryberry Walcoff, Esquire, Associate Commissioner for External Relations, and dated June 13th, 2005.
CHAIRPERSON MAISEL: Thank you.
At this point I'd like to begin the open public hearing session of the meeting. Both the Food and Drug Administration and the public believe in a transparent process for information gathering and decision making. To insure such transparency at the open public hearing session of the Advisory Committee meeting, FDA believes that it is important to understand the context of an individual's presentation. For this reason, FDA encourages you, the open public hearing speaker, at the beginning of your written or oral statement to advise the committee of any financial relationship that you may have with the sponsor, its product, and if know, 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 relationships.
If you choose not to address this issue of financial relationships at the beginning of your statement, it will not preclude you from speaking.
At this point I'd like to invite Mr. George Hawkins to address the panel.
At the podium, please.
MR. HAWKINS: Good morning. I'm George Hawkins, and I am a congestive heart failure survivor and would like to thank the Advisory Panel for the opportunity to speak about my experience with the Acorn device.
Before taking a few minutes to share information about myself, my heart condition, and my recovery, I would like to assure the panel that I am not paid by Acorn or anyone else. They did not pay $155 for me to stay at the Hampton Inn last night, and I have not spoken to anyone at the Acorn company regarding my statements here today.
I'm a 49 year old native Washingtonian and received congestive heart failure notice in '97, probably due to family history. Both my father and brother have heart murmurs.
I enjoyed a satisfying professional career in human resources training and traveled a great deal. Unfortunately, I had to retire in 2000 due to congestive heart failure.
My physical activities have included jogging, walking, tennis, weight training, and during the early summer of 2001, my physical condition worsened so much that I was unable to climb the steps in my house without resting to catch my breath.
My heart surgery was successfully completed at the Washington Hospital Center by Dr. Mercedes Dullums. In July 2002, to repair a leaky valve and to insert the Acorn device, the medical team informed me at that time that my heart was one of the largest that the doctor has operated on.
The success of my recovery can be attributed to the Washington Hospital Center staff, which includes Dr. Carlos Ross Cooke and Janice Richey, who is here with me. My recovery involved physical therapies, diligent medical follow-up, effective medications, and assistance from special people in my life.
Fortunately, I have not experienced any returns to the hospital for heart related issues at all. One year after surgery I was walking at least several miles a week and allowed a low impact tennis and weight training. Also, several months ago, in 2004, I worked part time as an HR consultant. Almost three years after surgery, my heart has not gotten any larger, and as a matter of fact, it has gotten a little smaller.
On my pre-op exercise bike test in 2001, I scored 8.6. In January 2005, I scored 15.4. So to sum up, my cardiologist, my surgeon, Dr. Dullums and the Washington Hospital heart center team took a big risk with my surgery because I am surviving two other chronic illnesses, and I'm just hopeful that my experience will help to underscore that the Acorn device is a major factor in enhancing the quality of my continued life.
Thank you very much.
CHAIRPERSON MAISEL: Thank you for your comments, Mr. Hawkins.
Are there any other members of the public who wish to address the panel this morning?
CHAIRPERSON MAISEL: Seeing none, we will close the open public hearing at this point.
At this point I would like to invite the sponsor to give their presentation. I will remind each of the speakers to introduce themselves and state their conflict of interest statements.
DR. KUBO: Good morning. My name is Spencer Kubo. I'm the Senior Vice President, Global Medical Director, and a full-time employee of Acorn Cardiovascular.
We very much appreciate this opportunity to discuss the cardiac support device with you this morning, a new technology for patients with dilated cardiomyopathy and heart failure.
The work summarized today includes extensive animal testing in three different animal models that prove the concept that the cardiac support device would, in fact, work. This animal work, as well, defined the mechanisms, histologic, biochemical, molecular as to why it works.
There is also extensive patient testing that's culminating in one of the largest randomized, prospective, controlled trials ever conducted for a permanent device implant that requires cardiac surgery, and we all feel that this work is potentially important because it fills an unmet need for patients with heart failure.
We're pleased today to have three outstanding speakers who will be sharing the data with you and discussing their results. All three were part of a five-person steering committee who were critical in the design, execution and recording of this trial, and they include Dr. Douglas Mann who recently accepted the position as Chief of Cardiology of
Baylor College of Medicine; Mary L. Jessup, who is Professor of Medicine and Director of the Heart Transplant Program at the University of Pennsylvania; and Dr. Michael Acker, who is Chief of Cardiac Surgery, also at the University of Pennsylvania.
I also want to acknowledge that the Steering Committee represents an extraordinary group of investigators, cardiologists, surgeons, and study coordinators from the 23 centers who participated in this trial. This trial today reflects their dedication and commitment to patient care, and we are delighted that many of them could take time out of their very busy schedules to attend this important meeting as our invited guests.
And I would ask the investigators and study coordinators to stand at this time and be recognized for their extraordinary contributions.
Thank you very much.
Our presentation today is divided into four parts. After a few introductory comments from myself, Dr. Mann will discuss the core concept, the preclinical studies, and the trial design, followed by Dr. Jessup, who will present the results of the trial to you, and I will come back for some final summary comments.
Our presentation today is meant to support an intended use statement which we had proposed and summarized here in this slide that the CorCap cardiac support device provides beneficial changes in cardiac structure associated with a reverse remodeling effect as defined by a reduction of left ventricular size, an increase in left ventricular ejection fraction, and a change to a more elliptical shape.
The CorCap cardiac support device also provides a decrease in the need for additional major cardiac procedures that are associated with the progression of heart failure and in an improvement in quality of life.
Our bases for this intended use statement comes from the demonstration of safety and efficacy, which is based on the four following points:
First, that the randomized trial performed achieved its primary endpoint at a P level of 0.024;
Two, that a number of secondary endpoints, including those that measure cardiac structure, such as the LV end diastolic volume, the end systolic volume, and the sphericity index, as well as secondary endpoints that deal with patient functional status, such as the Minnesota Living with Heart Failure questionnaire and the SF-36. All demonstrated significant clinical benefit of the CorCap cardiac support device.
Third, there were no safety issues identified, indicating that the device was safe.
And, four, based on all of this information, that the CorCap provides an effective therapy for patients with LV dilation and heart failure.
With that as a short background, I'd like to introduce Dr. Douglas Mann, who will introduce the CorCap concept and the preclinical studies.
DR. MANN: Good morning. My name is Doug Mann. I'm a paid consultant for Acorn Cardiovascular. I have no financial interest in the company.
My charge this morning is to briefly review the following three areas. We're going to focus on left ventricular dilation and the importance of that to the syndrome of heart failure. We will briefly mention that there's currently an unmet clinical need for patients who have large hearts and who have progressive symptoms despite optimum medical therapy, and then lastly, we'll review the scientific foundation for the CorCap, including three proof o concept studies which will briefly touch on the cellular and molecular mechanisms. We'll review some of the safety studies, and then we'll present the basics for the clinical trial, which my colleague, Dr. Jessup, will show to you shortly.
Progressive left ventricular dilatation produces a number of adverse consequences for the ventricle which are reviewed on this slide. First of all, as the ventricle begins to dilate and the walls begin to thin, there's an increase in wall stress. This, in turn, directly translates into an increase in after load for the ventricle, which can, in turn, lead to increased oxygen consumption and episodic subendocardial ischemia.
Furthermore, the progressive increase in left ventricular size can leave to stretch activation of a variety of maladaptive genes which are sufficient to activate the fecal gene program.
And finally, there's increasing evidence now that this progressive left ventricular dilatation can pull the papillary muscles apart and lead to progressive mitral regurgitation, which leads to a sustained volume overload on the ventricle.
It has been recognized now for a number of years that progressive LV dilatation heralds a worse prognosis for patients with heart failure. Shown on this slide are two studies, one by Hammermeister in Circulation in 1979, and the second by White and colleagues in Circulation in 1997.
As shown on the left-hand panel on this slide, adverse outcomes following an acute infarction were directly related to changes in left ventricular end diastolic volume and changes in end systolic volume.
Very similar findings were reported by White in Circulation, and as shown here, the relative risk of dying after an infarct is directly related to the end systolic volume of the patient following the infarct.
In addition to changes in left ventricular size, we now recognize the changes in left ventricular shape are also important in terms of determining patient outcomes.
Shown on the left-hand portion of the slide is the normal prolate ellipse shape of the ventricle, and you can see here that we break wall stress down into a circumferential wall stress, which is dependent on the length of the ventricle, and a meridional wall stress which is dependent on the diameter of the ventricle.
One of the things that we recognize now is as the ventricle remodels, the heart undergoes a transition from a prolate ellipse to a more spherical ventricle, and as it does this, there's an increase in the diameter of the ventricle such that meridional wall stress directly increases.
The reason why this is important is most ventricular shortening occurs in the short axis dimension. Very little shortening of the ventricle occurs in the long axis such that the increasing wall stress, meridional wall stress here directly impacts the amount of fractional shortening of the ventricle and can directly create a mechanical burden for the ventricle that didn't exist before.
The concept that ventricular size and shape is important is also borne out by the study by Douglas, et al., shown in the left-hand portion of this slide. They looked at left ventricular dimensions. As shown here the patients who have the larger hearts have the worst outcomes. Seven out of seven patients died who had ventricles greater than 7.6 centimeters, and then, again, in terms of the shape of the ventricle, you can see here the people who had the more spherically shaped ventricles, who had an increase in the ratio of the diameter to the length, also had the worst outcomes.
So both shape and size matter in terms of patient outcomes.
So what I've tried to show you over the last several slides is that patients with left ventricular dilation and progressive symptoms are at a high risk for limitations in the quality of life. They have frequent hospitalizations. They often need transplant and left ventricular assist devices, and as I've shown you there, an increased risk for high mortality.
Unfortunately, we have limited treatment options for this subset of patients. We know that cardiac resynchronization therapy is effective and will induce reverse remodeling, but it's effective really for only 20 to 30 percent of the patients.
We know that both mitral valve repair or replacement is effective, and that coronary bypass surgery is effective, but it's important to emphasize that neither of these two modalities have ever been tested or proven in clinical trials.
And lastly, we know that left ventricular assist devices and transplants are the last option for patients with advanced heart failure. So, in summary, we have limited treatment options for patients with progressive symptoms and large ventricles.
The CorCap cardiac support device is a fabric mesh device that's surgically implanted around the ventricle. It's intended to provide end diastolic ventricular support to reduce left ventricular wall stress and, hence, myocardial stress. It reduces the stimulus for ventricular modeling, and as we'll show you in preclinical studies, it also induces reverse modeling.
It is intended to improve cardiac structure and patient functional status in patients with moderate to advanced heart failure.
The CorCap cardiac support device looks like a very simple device, and yet it's a very complex device that has a number of key features which I'd like to review for you.
First of all, it's a multi-filament yarn, a knit fabric. It has four key design features. It has optimal compliance. It stretches enough so that it doesn't compress the ventricle, and yet it doesn't stretch too much so that it doesn't provide end diastolic support.
It has bidirectional properties, that is, it stretches more in the longitudinal direction than it does in the interior/posterior direction, and this tends to urge the ventricle back into a more elliptical shape.
It has a 31 microfiber construction so that it allows a smooth fit or a conformal fit under the surface of the heart, and last but not least, it has long-term biocompatibility. The polyester material that has been used has been used in other implantable devices.
How does the CorCap cardiac support device work? Most people in heart failure believe that the syndrome begins after some initial index event or injury to the heart that produces a decline in the pumping capacity of the heart. This decline in pumping capacity can lead to an increase in left ventricular wall stress and increase in myocardial stretch. Both of these components are then thought to lead to ventricular remodeling.
As I articulated on the previous slides, ventricular remodeling is sufficient to beget worsening cardiac functioning and worsening remodeling so that you end up with a vicious downward spiral.
The CorCap cardiac support device is intended to prevent the increase in wall stress and prevent the increase in dilatation, thereby preventing further cardiac remodeling, which we believe leads to an improvement in heart failure symptoms and better outcomes with patients with heart failure.
What I'd like to do now is to review a number of preclinical studies that have been compiled, and this is really an extensive preclinical database that shows the safety and efficacy of this device in experimental models, and it will provide some basis for examining the biochemical, cellular, and molecular mechanisms that underlie this unique device.
This slide is from a study by Tony Sabbah, and what they did was to use his microsphere injection model of heart failure. This, in my opinion, is the best model for studying heart failure. What they do is to progressively embolize the coronary artery with small microspheres. This, in turn, leads to microinfarcts and the injury which I mentioned previously. This, in turn, leads to progressive ventricular remodeling, and that's shown here in the control slides. There's a progressive increase in end diastolic volume, and these dogs will undergo the microsphere injection method.
Three months after implantation of the cardiac support device, you can see that there's a decrease in ventricular volume. If the device was just constraining the ventricle, the volumes would be unchanged, but what we see here is actually reverse remodeling.
This, in turn, translates to an improvement in overall pump performance for the ventricle, particularly in comparison to the control hearts, where there's a progressive decline in ejection fraction.
This slides shows the histologic findings of the CorCap cardiac support device. As shown here, it elicits a mild fibrotic response that covers the device. Importantly, there's no invasion of this fibrous tissue into the myocardium, and that's shown in the upper panel here. You can see here's the cardiac support device shown here. This green material is actually fibrous tissue, and you can see that there's really no invasion of the myocardium.
Furthermore, there's no compression of the arteries of the veins. This is the cardiac support device shown here, and you can see there's no compression of the artery and the vein. So it's really safe in preclinical models.
What are the components of reverse remodeling? This is, again, a study by Dr. Sabbah, and what he's done here is to look at a number of key signal transduction molecules that are involved in cardiac growth beginning with the p21ras, which is linked into endocrine signaling. You can see that there's actually up regulation of the amount of protein in heart failure. This is down regulated with the CSD device.
P21ras can activate a variety of signal transduction pathways shown here as the p38 pathway which has been linked into hypertrophic growth and signaling. You can see that the protein amount is increase in heart failure and then downregulated with the CorCap CSD.
And lastly, c-fos is a transcription factor that has been implicated in cardiac hypertrophic growth. Again, the amount of protein is increased in heart failure and then down regulated with the CorCap CSD.
So a variety of signal transduction pathways that we think are important for cardiac growth are up regulated in heart failure and are down regulated by reducing wall stress.
This not surprisingly translates into a decrease in myocyte size. Shown here are normal cardiac myocytes from the canine model. These are canine myocytes from a heart failure model showing an increase in width and length of the cells, and then three months following implantation of the CorCap CSD you can see that the myocyte size, both the length and the width, are both decreased.
In addition to the changes in myocyte size there are also changes in myocyte function, and that's illustrated on this slide. These are cell shortening curves as shown here. This is the cell at rest. This is the cell at the end of shortening. The amount of shortening is shown by the length of this line.
In heart failure we know that there's a decrease in the amount of shortening of the myocyte, and as you can see here, implantation of the CorCap CSD partially returns myocyte function towards a more normal shortening.
What I've done now is to provide the preclinical basis for the human safety studies which I'll show you on the next several slides.
This is a slide from one of the early safety studies done in Charite Hospital, and it has really two important features which we've found to be consistent in the large clinical trial, which my colleague, Dr. Jessup, will show you.
First, you can see that there's a progressive decrease in left ventricular end diastolic volume in these patients who had the CorCap CSD implanted. Furthermore, this change in end diastolic volume is durable.
Secondly, there's an improvement in ejection performance of the ventricle, and again, this improvement in the ejection performance is durable over time.
This slide shows pressure volume loops from a single patient that was enrolled in the Charite safety study, and it has several important features which I'd like to direct your attention to.
Shown on the vertical panel here is left ventricular pressure and on the horizontal panel is left ventricular volume. These are pressure volume loops of the ventricle and for the patient before the CorCap CSD was implanted. If there was cardiac compression, one would expect that the pressure volume, of course, would have been shifted upward and to the left. That doesn't occur with the CorCap CSD.
What we see instead is a reverse remodeling, a true reverse remodeling with a decrease in the pressure volume curve and the ventricles operating on a much more favorable pressure volume curve here.
Also note that the area of the pressure volume loop increases, which implies that there's an increase in cardiac work. So the ventricle is operating more efficiently. There's more work at less pressure.
In addition to reductions in the volumes in the ventricles and the pressures in the ventricles. there's a reverse remodeling in terms of cardiac mass. Shown here is a decrease in cardiac mass with the CorCap only, and a decrease in cardiac mass with the CorCap on top of mitral valve repair.
So in summary, what I've tried to show you over the last series of slides is that left ventricular dilatation is directly related to adverse patient outcomes. We've shown you briefly a series of animal studies that demonstrate proof of concept of reduction of wall stress leads to reverse remodeling of the cellular and molecular level.
And lastly, we've provided some safety studies that confirm the findings of the animal studies. The final step, of course, is the proof in a randomized trial.
What I want to do now briefly is review the trial design for the CorCap CSD. This slide shows the inclusion and exclusion criteria for the trial. We enrolled men and women age 18 to 80 years. They could be New York Heart Class III or IV heart failure of ischemic or nonischemic etiology. They had to have had a left ventricular ejection fraction of less than 35 percent and a large ventricle with a left ventricular end diastolic dimension of greater than 60 millimeters.
The two exceptions to these previous statements are that patients who are enrolled in the mitral valve stratum could have New York Heart Class II and/or an ejection fraction of less than 45 percent was allowed. The patients had to be functionally limited. They had to have had a six minute walk test of less than 450 meters, and they had to be on stable optimal medical therapy defined as ACE inhibitors and beta blockers plus or minus an aldosterone antagonist.
The exclusion criteria shown below, the patients could not have had a CABG, nor could they be on an active transplant list.
This slide shows the randomized trial design. We enrolled 300 patients who, as I said, were on optimal medical management. If, depending on the site investigator, the patient required mitral surgery, they were entered into a mitral surgery stratum and then randomized in a one-to-one fashion to either a control arm, which consisted of mitral surgery alone, or mitral surgery plus the CorCap, which we just referred to the treatment arm.
If, on the other hand, the site investigator deemed that they did not require mitral surgery, they were randomized in a one-to-one fashion to the control arm, which was optimal medical therapy, no surgery here, or optimal medical therapy plus the CSD.
The trial was designed according to an intention to treat analysis. It wa powered for 300 patients. The data analysis plan prespecified pooling of both strata and reporting as one cohort, and we felt that that was justified because the inclusion criteria in both strata were virtually identical, and the endpoints for both strata were identical.
This slide shows the primary endpoint at the trial, the clinical composite. It's important to emphasize that each component was clinically relevant and was detectable by the patient. The three components that comprised the worsening category could account for every clinical outcome for a patient with heart failure. For example, patients who were considered worsened could either have died during the study, could have had a major cardiac procedure that was adjudicated by a blinded committee to be because of worsening heart failure, or could have had worsening New York Heart Association class as assessed by a blinded New York Heart assessor.
If the patient was improved, they had to have had an improvement in New York Heart Association as assessed by a blinded assessor, and they couldn't have had anything that would have categorized them as worsening during the trial.
We underwent a number of careful measurements to assure safety of the device, which my colleague, Dr. Jessup, will review for you. I just briefly want to touch on them. As mentioned, we looked at cardiac mortality. We looked at major cardiac procedures that we felt were indicative of worsening heart failure. We catalogued a variety of serious adverse events, and then finally we looked at the combination of serious adverse events or death.
So we've undergone extensive analysis to prove safety in the device.
This slide summarizes the secondary endpoints for the trial, including cardiac structure and function and changes in patient functional status. So we examined left ventricular end diastolic volume and systolic volume, ejection fraction, sphericity index as a measurement of left ventricular shape. We looked at left ventricular mass and the amount of micro regurgitation severity.
We also looked at patient functional status in terms of the Minnesota Living with Heart Failure questionnaire, SF-36, as the generic functional status measurement, New York Heart Association class, all cause hospitalization, peak VO2, and finally six minute walk.
We recognized going into this trial that it was a device trial, and as such was unblinded. So we went through a number of careful steps to try to reduce study bias in the trial to maximize the scientific integrity of the trial.
First of all, the design of the primary endpoint included what most people would consider as a hard endpoint, mortality. We also designed the three components that went into worsening to be interdependent. So that, for example, if one didn't undergo cardiac transplantation, they would show up a worsening heart failure. So there's really no way to hide with the way that we designed the primary endpoint.
All core labs were blinded to a treatment allocation, and these were the core labs that made the important measurements of both the primary and secondary endpoints, and the sponsor and the investigators were kept blinded to the aggregate data.
A second implementation that was made was the development of a clinical events review committee that was blinded to the patient treatment allocation with respect to a number of important outcomes.
So shown here, patients who underwent mitral valve surgery, tricuspid valve surgery, biventricular pacing, the CERC committee had to adjudicate whether these were done because of worsening heart failure, and they were blinded to treatment allocation, both VADs and cardiac transplants, the CERC was not blinded as to outcome. We considered that these were indicative of worsening heart failure.
And the third final element that was really implemented at the behest of the FDA was the development of a blinded New York Heart Association core laboratory assessment. This was implemented to reduce a potential bias. It utilized a questionnaire that was administered to the patient by the blinded site clinician. The questionnaire was validated prior to implementation. The questionnaire was then sent to a cardiologist who was blinded to treatment allocation, who then assigned a New York Heart Association class.
The core New York Heart Association class was used in all of the analysis of the primary endpoint. Unfortunately this was implemented as the trial was rolling forward. So they were missing baseline core values that were -- they were missing patients because the analysis was implemented as the trial rolled on.
It's important to emphasize that there are really two types of data in this trial because it can be a little confusing, and I wanted to walk you through these briefly. First of all, there are data that are driven by the common closing date, and this includes deaths, all adverse events, and major cardiac procedures. So all of thee events were captured within the trial.
There were also data that were collected at follow-up visits, including three, six, 12 and every six months thereafter, and this included the echocardiographic assessment of LV structure and function, the New York Heart Association class, the quality of life data, and finally the exercise testing data.
What I'd like to do now is to introduce my colleague, Dr. Mariell Jessup, who will review the main trial results with you. Dr. Jessup is the head of heart failure transplant at the University of Pennsylvania.
DR. JESSUP: My name is Mariell Jessup. I'm a member of the steering committee for the Acorn CorCap randomized trial and was also a co-principal investigator at our clinical site at the University of Pennsylvania. I have no financial interest in the company.
I'm very pleased to present the results of this randomized trial. As reviewed by Dr. Mann, there were 300 patients who had already undergone optimal medical management with standard heart failure therapy. One hundred and ninety-three patients were placed in the mitral surgery stratum, patients in whom the site investigators determined that mitral surgery was required.
These 193 patients were then randomized in a permuted block design for each stratum, into mitral valve repair replacement alone in 102 patients and mitral valve surgery plus the CorCap cardiac support device, CSD, in 91 patients.
The remaining 107 patients were in the no mitral surgery stratum and were randomized to continue on optimal medical therapy as the control group in 50 patients or the optimized medical therapy with the CSD in 57 patients.
Data was collected from the beginning of the study in June of 2000 until the common closing date, July 4th, 2004, so that each patient contributed different amounts of follow-up data by the end of the study.
Specifically, there was a minimum plan follow-up of one year, but there were only 37 percent of patients who were followed for this minimum time of 12 months. Twenty-one percent were followed for 18 months; 23 percent were followed for 24 months; and 19 percent were followed for 30 months or greater. Therefore, the median follow-up was 23 months.
In general, these patients were similar to multiple other low EF heart failure trials with a few notable exceptions. The patients enrolled were slightly younger, with a mean age of 52.5 years. There was a higher percentage of females enrolled. There was a higher number of non-white patients in this study compared to most other trials, and the most common heart failure etiology was idiopathic as compared to ischemic in other trials.
This study does, indeed, however, represent a population of chronic heart failure patients, since the mean duration of heart failure in this group was at least five years.
This slide shows the baseline structural and functional characteristics of our study population. The mean left ventricular end difolic (phonetic) diameter was enlarged at 69.8 millimeters. Peak V dot O2 in this patient population was 15 mLs per kg per minute. The mean left ventricular ejection fraction was 23 percent. The Minnesota Living with Heart Failure score was elevated at 59.3, and the six minute walk distance achieved was only 340 meters.
A small group of patients were designated as NYH Class II by the site investigators.
You will remember that the study design allowed these patients to be entered if they were going to undergo mitral valve surgery. The majority of the patients, however, were in NYH Class III.
The patients' baseline medications were to include optimal medical management. The investigators of the study adhered to these instructions, and I think the high percentage of concomitant medical therapy in this trial should be taken into account as I present the results to you.
Fully 90 percent of all patients were either on ACE inhibitors or angiotensin receptor blocker, or ARBs. Eighty-five percent of all patients were on a stable beta blocker dose for at least three months. Almost all patients were on a diuretic and almost half of the patients were on aldosterone antagonists.
This represents a concomitant medical therapy or optimal medical management that really is noteworthy in contrast to many other earlier heart failure trials.
Randomization in the study yielded comparable groups between treatment and control, except for three baseline covariates. These included gender. As more women were randomized to the treatment arm, the core lab peak V dot O2 as the treatment arm in this study had a lower value for V dot O2, and diastolic blood pressure, especially in the MVR stratum. There was no identifiable cause for this imbalance, and as specified in the data analysis plan, therefore, covariate adjustment of the primary endpoint was necessary for these variables.
This table depicts the primary composite endpoint results. Patients were placed into three categories: improved, same, or worsened. In this trial the treatment group had more patients improve and less patients worsen than in the control group. The proportional odds ratio was 1.73, indicating that the treatment patients with the CSD compared to control had a 73 percent greater likelihood of being in a better outcome category. This was statistically significant at a P value of .024.
Thus, the primary objective of this trial was met.
The primary composite endpoint allowed for interdependence of the components that made up the composite. Thus, the three components in this trial accounted for every clinical outcome for each patient.
For example, heart failure patients can worsen in three ways, either deaf, a major cardiac procedure indicative of worsening heart failure or worsened NYHA classification. Each possible event is accounted for in the primary endpoint, and when it occurred, the patient was considered worsened.
I, therefore, would like to discuss each of the components individually. First let's look at survival.
This slide depicts the survival of the entire 300 patients over the 24-month period of follow-up. This is a Kaplan Meier curve showing no significant difference in survival between the control and the treatment arm. There was likewise no difference in the mode of death between groups.
As requested by the FDA, all patients have been followed through a secondary closing date of April 15th, 2005. During the additional follow-up, there was no accrued discordance observed.
Survival is depicted in an alternative manner in this slide. This table shows all deaths reported of the extended follow-up of April 15th, 2005 organized by the time period for each stratum. In the MVR stratum at the top, there were three deaths within 30 days of randomization out of 183 operations, for an operative rate of 1.6 percent
In the no MVR strata, there were four operative deaths within 30 days of randomization. One patient died prior to the surgical procedure our of 51 operations for a 7.8 percent mortality at surgery rate.
After the 30 days usually considered to be the perioperative time frame, mortality in both strata was exceeded in the control population as compared to the treatment group.
Clearly, the operative mortality rate in the no MVR stratum was of concern. This table details the four patients who died in that early postoperative period. All four patients were significantly compromised with respect to heart failure, as reflected by either a very depressed peak V dot O2, a markedly depressed ejection faction, or significantly enlarged LV volume.
However, they all did fit into the study entry criteria and, therefore, were not excluded. Two of the patients were initially operated on without heart-lung bypass and were subsequently placed on bypass pump due to hemodynamic instability.
After the third death and as a result of these observations, recommendations were made to our investigating surgeons and discussed extensively at an investigator's meeting so that patients with far advanced disease did have surgery with concomitant cardiopulmonary bypass. An entrerk (phonetic) balloon pump was used if there was any hemodynamic instability.
Subsequently, as compared to the first year of enrollment where there were two deaths out of 12 implants in the no-MVR stratum, in the following year there were two deaths in 20 implants. In the final year, out of 19 implants there were no deaths. This does represent a learning curve phenomena that has been demonstrated in other surgical trials.
Now, I will turn to the second component of the composite endpoint, freedom from major cardiac procedures. At every point in the trial the treatment arm had fewer major cardiac procedures than the control arm. This was statistically significant at a level of .01.
This table shows the major cardiac procedures indicative of worsening heart failure included in the trial. It depicts in the first column the number of patients in the treatment arm and in the second column the number of patients in the control arm.
It is important to remember that some patients had more than one event. As can be seen, the number of patients experiencing events in the treatment arm was significantly less than the control group. Treatment patients have significantly less major cardiac procedures, specifically a marked reduction in the need for cardiac transplantation and/or ventricular assist devices.
There were fewer repeat mitral valve surgeries and fewer biventricular pacemakers implanted. In this component of the composite, we analyzed the results excluding biventricular pacing with no effect on the overall outcome of the statistical significance.
As our third component of the composite, this table shows the change in the core lab NYHA classification. This summarizes the core lab NYHA classification from baseline to the last follow-up visit. For the purposes of this analysis, patients who had VADs, transplants, and other major cardiac procedures that were adjudicated as worsening heart failure and, therefore, were considered NYH Class IV.
Please note that this scoring excluded patient depths. This analysis demonstrates that more treatment patients improved and fewer treatment patients worsened with respect to NYHA functional classification. The proportional odds ratio was 1.74, similar to the odds ratio, the primary endpoint which was statistically significant.
There are 38 percent more patients improving by at least one NYHA class in the treatment group compared to control.
In summary then, after review of the three composites of the primary endpoint, this slide serves to remind you that the primary endpoint was achieved. The three components comprehensively accounted for every clinical outcome of each patient, either death, a major cardiac procedure, or a change in the NYHA classification.
It's important to underscore that the favorable treatment effect observed in this trial could not be attributed to referral bias for the major cardiac procedures because the primary endpoint would have accounted for worsening in other ways, either through death or a change in the NYHA classification.
This clinical composite endpoint is also important because each endpoint is relevant and detectable by the patient.
I would now like to turn to the safety profile of the CSD. This table details the serious adverse events that occurred in greater than a five percent incidence over the extended follow-up of April 15th, 2005.
Overall, 78 percent of patients in the control group and 81 percent of patients in the treatment group had a serious adverse effect. There was a higher percent of patients experiencing hemodynamic or renal compromise in the CorCap treatment.
Recognize, however, that all these patients had surgery, whereas many patients in the control group did not undergo any surgery. There were, however, no adverse events or complaints reported related to sizing or fitting of the CorCap.
There was a theoretical concern about the possibility that the CorCap CSD would inflict a constrictive physiology on the hearts of our patients. Therefore, a comprehensive echo surveillance program was initiated every six months on each patient. These echoes were reviewed at the Mayo Clinic in our core echocardiographic lab and were blinded as to treatment strata. The standard protocol was designed for the early detection of any echo abnormality.
There were 59 total patients with an abnormal echo possibly suggestive of constrictive physiology. However, 80 percent of these reports were isolated, with no repeat abnormality demonstrated on follow-up echo, and there was no association of the echo cardiographic finding with morbidity and adverse event or mortality.
Specifically, there were no patients with clinical symptomatology that was associated with the possible constrictive physiology seen on echo. In summary, we could find no evidence that the CorCap caused clinically significant constriction.
If one were to look at the entire patient cohort using the Kaplan Meier analysis, examining freedom from death or a serious adverse event, there was no statistical difference between the control and treatment arm, despite the up front cost surgery in the treatment arm patients.
This same analysis, freedom from death or serious adverse event, is depicted for the mitral valve surgery stratum alone showing no difference in the curves for the control and treatment arm. Please remember that in this strain both the control and treatment patients underwent mitral valve surgery.
Indeed, if one were to look at the same Kaplan Meier curve, freedom from death or serious adverse event in the no MVR stratum, the impact of surgery for this group of patients only half of whom underwent open heart surgery becomes apparent. Nevertheless, there was a narrowing of this difference by the end of the extended follow-up of April 15th, 2005.
Thus, we would submit that an adverse event summary contains three important points. One, there was an increased adverse event risk related to surgery.
Two, there was no significant overall difference between treatment and control with respect to serious adverse events.
And, three, there was no evidence of adverse clinical outcomes related to the theoretic possibility of constriction.
I would now like to move on to the secondary endpoints of this trial. This figure illustrates a longitudinal regression analysis for the change in left ventricular end diastolic volume from baseline to 18 months. Both the control and treatment groups demonstrated a reduction in a left ventricular end diastolic volume, indicating a decrease in left ventricular size.
I would suggest to you that in the control group both the optimal background medical therapy and the mitral valve surgery could account for the observed reduction in left ventricular volumes. The average reduction in LVED volume over the 18-month follow-up period was greater, however, in the treatment compared to the control group, with an overall treatment difference of 17.9 milliliters, highly significant at a p value of .008.
Note that there was a progressive reduction in left ventricular size over the follow-up period rather than an abrupt decrease immediately after surgery. These results can be considered consistent with reverse remodeling, not an acute girdling effect of the device.
This slide illustrates a longitudinal regression analysis for the change in left ventricular end systolic volume from baseline to 18 months. Both the control and treatment groups demonstrate a reduction in end systolic volume, again, indicating a decrease in left ventricular size. The average reduction in end systolic volume over the 18-month follow-up period was greater in the treatment compared to the control group with an overall treatment difference of 15.2 milliliters, which was statistically significant.
This decrease in left ventricular end systolic volume followed the same pattern as left ventricular end diastolic volume.
This figure illustrates the longitudinal regression analysis for changes in left ventricular ejection fraction from baseline to 18 months' follow-up for the entire study population. Both the control and treatment groups demonstrated a slight increase in ejection fraction throughout follow-up. The treatment group demonstrated a slightly greater increase in EF than the control group with a difference of .83 units, but this difference was not statistically significant.
Pre/post comparison, however, within the treatment group was significantly different so that the treatment group showed a significant improvement ejection fraction compared to baseline, which the control group did not show.
Cardiac sphericity index is calculated as the ratio of left ventricular length to left ventricular width, both measured in diastole. A normal cardiac sphericity index is approximately 1.58. As the heart enlarges and changes shape from an American football to a soccer ball, the length/width ratio approaches one, which is a perfect sphere.
Therefore, any intervention which increases the length to width cardiac sphericity index to greater than one would be returning the heart to a more normal shape.
This figure demonstrates an increase in sphericity index for both groups, indicating a beneficial change in shape. There was, however, a larger increase in sphericity index for the treatment group. This overall difference was statistically significant at a p of .031, and the treatment difference was approximately .042.
The change in shape is likely related to a design feature of the CorCap CSD in which the compliance or stretchiness is greater in the longitudinal, the base to apex direction, as compared to the transfers or circumferential direction. No only is the heart smaller, but it has a more normal shape which provides mechanical and bioenergetic advantages.
These results support the secondary study objectives and the intended use statement.
Turning to the clinical response to this device, this figure illustrates the longitudinal regression analysis for changes in Minnesota Living with Heart Failure for all patients. A lower score translates to a better quality of life.
Both the control and treatment groups demonstrate a reduction in Minnesota Living with Heart Failure score. However, the treatment group had a significantly greater reduction with an average difference of 4.47 units. This was significant at a p of .04.
The improvements in clinical score were evident in three months and were sustained over 24 months.
Likewise if one examines the change in the physical function domain of the SF-36, a higher score in this clinical tool indicates a better quality of life. The treatment patients had a greater score with a treatment difference of 5.41, indicating a better quality of life in the physical function domain compared to control patients. This was likewise statistically significant.
Again, the improvements were evident at three months and were sustained over 24 months.
The rehospitalization rate is tabulated in this slide. Baseline hospitalizations were excluded because all of the control patients in the no mitral surgery group did not undergo surgery and were accounted for in the hospital mortality and adverse events assessment.
As this table shows, there were no difference in the total number of rehospitalizations between the treatment and control. There was a difference in the total number of rehospitalization days, the total number of ICU days, the mean and median length of stay between treatment and control, all favoring the treatment arm, but was not at a statistically significant level.
This then was a conservative study design because hospitalizations were not adjudicated for heart failure relatedness. This parameter, all cause hospitalizations, has been similarly shown to be unchanged with other highly effective heart failure therapies as seen in Miracle ICD and the VALHEFT trial.
Exercise response in this trial was examining by assessing both peak exercise capacity as measured by maximum oxygen capacity of peak V dot O2 and by the six minute walk test.
Unfortunately, there was an excessive amount of missing data primarily because the patients who did not perform the tests were sicker or in the hospital. More tests were missing in the control population.
For these reasons, the exercise responses were analyzed by rank analysis to appropriately deal with the missing data. Patients were assigned into one of six categories ranging from the best to worst response. At 12 months the odds ratio for a six minute walk was 1.27 favoring the treatment group. At 12 months the odds radio for a peak V dot of two was 1.37, again, favoring the treatment group.
Before I proceed to analysis of the individual stratum in this trial, I'll summarize the secondary endpoints of the trial as a whole, examining the treatment effects observed.
There was statistically significant results favoring the CorCap CSD treatment in changes in left ventricular end diastolic volume, left ventricular end systolic volume, left ventricular sphericity, Minnesota Living with Heart Failure, and the physical function domain of the SF-36. There were trends favoring treatment and rehospitalization and exercise testing.
The data analysis plan called for analyzing each of the two stratums separately. Although the same statistical analysis was performed, it was always presumed that it would be unlikely that analysis of each individual strata would reach statistical significance even with similar treatment effect because of a reduced power.
First, let me turn to the no mitral surgery arm. This is a fundamental analysis, as it probably represents the truest measure of the efficacy of the CorCap CSD alone. This figure illustrates the Kaplan Meier curve for survival in the no MVR stratum. These curves are different from the all patient analysis because there was an early risk for the treatment group compared to control, which was expected due to the initial risk of surgery in the treatment group.
Remember that the control group in this stratum did not undergo surgery.
Survival of the control group does catch up to the treatment group by 12 months so that at 24 months there were ten deaths in the treatment group and eight deaths in the control group. There was no statistically significant difference between the two treatment arms in overall survival.
As requested, this graph shows the survival follow-up through April 15th, 2005 for the no mitral surgery arm. This table summarizes the time period analysis of mortality showing all deaths as of April 15th, 2005, reported by time period after surgery. In the no MVR stratum, there were four operative deaths within 30 days of surgery, one patient dying before surgery, out of 51 operations for a 7.8 percent 30-day mortality rate.
Again, subsequent recommendations to the operating surgeons about the use of heart-lung bypass and/or balloon pump in unstable patients resulted in an improvement in the operative risk as the trial proceeded.
I mentioned before that in the final year of the trial, there were no operative deaths in this stratum.
This table examines the serious adverse events that occurred in the no MVR stratum both for the period within 30 days of surgery and the follow-up exceeding 30 days. The excessive adverse events seen in the treatment arm occurred in the early postoperative period, in contrast to the control group that does not undergo surgery.
There is a balanced adverse event rate after the initial 30 days in both arms.
This table shows a primary composite endpoint for the no MVR stratum. This does provide the opportunity to answer the fundamentally important question does the CorCap CSD by itself provide benefit compared to patients treated with an optimal medical regimen.
In this arm of 107 patients, there are no confounding effects of concomitant micro valve surgery. The CorCap CSD treatment group had a greater frequency of improvement and a lower frequency of worsening compared to a controlled group.
The proportional odds analysis of this distribution revealed an odds ratio of 2.57. This was statistically significant at a p value of .032, illustrating that the treatment group had over two and a half times better odds at being in an improved category compared to the control group. The fact that the primary composite endpoint was statistically significant in the no MVR stratum was not expected because of the small sample size or reduced power.
That it occurred at all is because the effective treatment was very large. The odds ratio was 2.57 versus 1.73 in the all patient cohort. This shows the major cardiac procedures in the no mitral surgery stratum, again. There is a significantly improved chance of being free from the need for additional major cardiac procedures as a result of the CSD.
This table shows that in the no mitral valve surgery stratum the treatment patients had significantly less transplants, ventricular assist devices of VAD devices, as well as biventricular pacing. Again, this reproduces the results of the overall study.
This table illustrates the change in the core lab assessment of NYHA functional classification in the no MVR stratum. Again, the treatment arm had a higher chance of being improved in this functional classification as measured by the core lab. Thus, 84 percent more of the treatment patients improved by at least one NYHA class compared to control.
The next few slides summarize the results of the 193 patients randomized to the mitral valve surgery arm of the study. All patients in this stratum underwent mitral valve repair or replacement. Half of the group underwent CorCap CSD implantation as well.
There was no significant difference in survival between the treatment and control groups in the MVR stratum similar to the study as a whole.
Looking at the survival curve in data accrued through April 15th, 2005, there is a trend towards a better survival in the treatment group compared to patients who had mitral valve surgery alone.
This table shows the serious adverse events in the mitral valve surgery stratum depicting no adverse events in the treatment group compared to control either in the first 30 days or greater than 30 days after surgery.
As we look at the primary composite endpoint in the mitral valve surgery stratum, it does provide the opportunity to answer an additionally important question in this trial. In the group of patients undergoing mitral valve surgery, does the CorCap CSD provide incremental benefits when added to mitral valve surgery?
This question has significant practical implications since our over 50,000 valve procedures are performed in the United States each year. Any adjunctive therapy that increases the efficacy of these valve procedures would be valuable.
The CorCap CSD treatment group had a greater frequency of improvement and a lower frequency of worsening when compared to the control group. The proportional odds analysis of this distribution revealed an odds ratio of 1.51. Although this odds ratio was not statistically significant, the effect size was similar to the overall patient cohort.
The freedom from major cardiac procedures likewise showed a trend toward fewer procedures in the treatment group versus control, although because of reduced power was not statistically significant. Nevertheless, in the group of patients who have mitral valve surgery in CorCap, there were fewer cardiac transplants, fewer VADs, fewer mitral valve surgeries, and fewer biventricular pacers implanted.
An analysis of the change in the core NYHA functional classification in the MVR cohort, again, reflects a benefit of the CorCap CSD in improving the odds of being in a more favorable functional class.
As mentioned earlier, the data analysis plan called for analyzing each of the two strata separately. Although the same statistical analysis was performed, it was always presumed that it would be unlikely that analysis of each individual strata would reach statistical significance, even with similar treatment effect because of the reduced power of the individual stratum.
This slide summarizes the secondary endpoints of left ventricular end diastolic volume, left ventricular end systolic volume, cardiac sphericity and Minnesota Living with Heart Failure questionnaire for the entire study population in the first column; the no MVR stratum in the second column, and the MVR stratum in the final column.
The results of all the secondary endpoints in both strata are similar in magnitude to the secondary endpoints in the main trial showing a consistency of effect both in reducing left ventricular size and improving patient symptoms.
To summarize no MVR stratum, therefore, with respect to safety, the operative risk was 7.8 percent. There were increased adverse events due to the surgical implant compared to the medical treatment arm, and this risk was mitigated through training and labeling.
With respect to efficacy, there was a significant effect in this arm in the primary endpoint observed. There were nearly twice as many patients in the treatment arm improving by one or more NYHA class. There were fewer cardiac procedures and fewer transplants. Hearts were smaller in the treatment arm, and the patients experienced an improved quality of life.
In the MVR stratum, the operative risk was a surprisingly low 1.7 percent, and there was no increased risk of adverse events in the treatment arm of this stratum.
With respect to efficacy, the magnitude of the treatment in this arm was similar to that seen in the study as a whole. There were fewer cardiac procedures, fewer transplants; left ventricular size was reduced, and there was an improved quality of life seen in the treatment patients.
Finally, I'd like to address what these trial results mean to an individual patient. Remember that these are patients who had been maximized on standard medical therapy and continued to be symptomatic.
In addition, many of them had significant mitral regurgitation.
Patients similar to our study population are told that they are at risk for continued deterioration of their symptoms or cardiac size or function and might even need a transplant if they're candidates. Realize that the options available to such a patient besides transplant are limited to select patients. They do include biventricular pacing, coronary bypass surgery or mitral valve surgery.
The use of the CorCap resulted in significant benefit to comparable patients. Their functional class improved at least one in rate class in 38 percent of patients. Quality of life improved by 4.5 units as measured by the Minnesota Living with Heart Failure Questionnaire.
The need for transplant of that was decreased by 55 percent, and the heart size decreased in volume in a meaningful amount.
I'd like to turn the concluding remarks over to Spencer Kubo.
DR. KUBO: Thank you, again, Dr. Jessup.
My name is Spencer Kubo. I'm a full-time employee of Acorn Cardiovascular.
We've heard quite a bit of information today summarized. I'd like to provide just a few concluding comments to place it in the appropriate context.
In this slide, we've looked at the changes in LV end diastolic volume that reported in our trial and related to other trials involving both drugs on the left-hand side and device therapy with cardiac resynchronization therapy in the middle panel.
In this study in this slide, you'll see that there are controlled groups in the gray bars and the treatment group is in the red. With drugs such as Enalapril as demonstrated first in the salt trial (phonetic), we know that the control group will get progressively larger. This is the progressive enlargement of remodeling, and that process can be attenuated with the use of a very effective drug, an ACE inhibitor.
The ANC trials was one of the first demonstrations that beta blocking therapy can actually reverse this process so that the ventricle becomes smaller. This is a very exciting observation, but the effect size is rather modest.
With the advent of device therapy such as biventricular pacemakers, as demonstrated in both Miracle trials, we see a somewhat larger reduction in LV size, a somewhat larger effect on reverse remodeling, although the effect size between treatment control was larger in the first Miracle study.
In our trial, we demonstrate the treatment control difference first in the no MVR stratum and then in the MVR stratum. So this bar would be the control group getting just medical therapy, showing that effective medical therapy with beta blockers and ACE inhibitors will, in fact, lead to reverse remodeling, but that effect can be augmented with placement of the CorCap.
In the MVR stratum, we see the effects of the control medical therapy plus the effect of the mitral valve surgery leading to a rather significant effect on LV size, but that effect also being augmented with the addition of the CorCap.
We see here that the effect, the treatment control difference is the same in both strata, but the starting point is different because the control therapies are different in both strata.
Similarly, if you looked at the changes in Minnesota Living with Heart Failure questionnaire, here a reduction in the score indicates an improvement in quality life. We have data that we can compare our results to, three different or four different CRT trials, contact, the two Miracle trials, and rhythm ICD, all reasonably showing a consistent reduction in Minnesota Living with Heart Failure questionnaire score indicating an improvement in quality of life.
We see the same step-wise function or a similar step-wise function in the Acorn trial. Continued medical therapy will improve quality of life. That effect can be augmented with the addition of the CorCap.
In the MVR stratum, the mitral valve therapy, in addition to medical therapy, will lead to a large improvement in quality of life. That effect can be augmented with the addition of a CorCap, so again showing that medical therapy, a surgical therapy, and then perhaps two surgical therapies combined.
Based on all of this information, we would make the following summary concluding statements.
First, on safety, there was no difference identified between treatment and control in terms of mortality and overall serious adverse events.
However, the CorCap showed a significant reduction in the major cardiac procedures that are associated with progressive heart failure compared to control.
Third, that the risk of the CorCap and the implantation surgery can be mitigated through training and labeling.
On efficacy, the CorCap trial achieved its primary endpoint at a level of p equals 0.024. It also demonstrated that a number of secondary endpoints, including cardiac structure, as well as patient functional status as listed there, also demonstrated significant clinical benefit of the CorCap and corroborated the findings of the primary endpoint.
Third, that the CorCap showed a significant reduction in major cardiac procedures.
And, fourth, all of these data in the randomized trial are consistent with the preclinical and safety study results that have been reported previously.
Based on these findings, we would propose the following intended use statement for the CorCap. The CorCap cardiac support device provides beneficial changes in cardiac structure associated with a reverse remodeling effect as defined by a reduction in left ventricular size, an increase in left ventricular ejection fraction, and a change to a more elliptical shape.
The CorCap cardiac support device also provides a decrease in the need for additional major cardiac procedures associated with the progression of heart failure and an improvement in overall quality of life.
Our indications for use, as we propose them in your panel pack, are listed here. First, it is indicated for patients diagnosed with dilated cardiomyopathy, patients who are symptomatic despite treatment with optimal heart failure medical management; third, patients with a dilated heart, as demonstrated by an increase in the LV end diastolic dimension greater than 60 millimeters or an indexed LV end diastolic dimension greater than 30 millimeters per meter squared; and finally, in patients with a left ventricular ejection fraction less than or equal to 35 percent or less than or equal to 45 percent if planned mitral valve repair or replacement surgery.
The agency has provided us with a summary list of important questions that they have indicated after their review of the panel pack. We'd like to take this time to respond to just a few of the important questions.
The first question is listed here regarding the evaluation of device safety. The question that's posed is: does placement of the CorCap cardiac support device and the resulting increased difficulty for follow-on surgery, especially coronary bypass operations, compromise patient safety during a subsequent operation?
And I'd like to ask Dr. Acker to address this question.
DR. ACKER: Hello. My name is Michael Acker. I'm Chief of Cardiothoracic Surgery at the University of Pennsylvania, principal investigator in this study. I have no financial interest in the Acorn company.
The increased operative difficulty the previous question refers to is due to the presence of adhesions that are encountered during redo operations after CorCap placement. It has been suggested by some that the adhesions encountered during transplantation in CorCap patients would result in bias, specifically, that transplantation operations otherwise indicated would be withheld secondary to the fear of poor outcomes due to the difficult operation.
The data indicates, however, that safety of redo transplant operations was not compromised. Dense adhesions will be encountered after CorCap procedures, but are often encountered after other cardiac operations. Adhesions of similar density and severity are often seen by transplant surgeons after patients who have had multiple bypass operations or patients who have LVAG or BIVADS placed for a significant period of time.
Redo cardiac procedures, specifically transplantations, were performed safely and with good outcomes. And an expert panel on reoperations made significant recommendations on patient management which was incorporated in labeling and in training.
Significant adhesions are reported in 100 percent of the patients transplanted greater than 30 days following initial CorCap surgery in contrast to 70 percent of patients with previous mitral valve surgery alone. Despite these adhesions often being dense, the cardiopulmonary bypass times, which usually reflects the overall difficulty of an operation was increased only by 15 minutes, or eight percent over control patients.
This slide provides the actual mortality and morbidity after transplantation in both groups. The number of transplants in CorCap group was seven in contrast to 16 in the control patients. There were no deaths in patients receiving transplantation after CorCap in contrast to two in the control patients.
The total number of adverse events per patient in the CorCap group was 1.7 in contrast to 1.9 in the control group. The number of adverse events within 30 days of transplant, which one would expect to reflect the overall safety of the operation, was four in the CorCap group in contrast to ten in the control patients.
Despite the dense adhesions encountered, there was no bleeding complications in the CorCap group in contrast to one patient returned to the operating room for bleeding in the control group.
Finally, one would expect that the postoperative length of stay to reflect how safely the transplant operation was performed, as well as the number and severity of adverse events arising from the transplant operation, postoperative length of stay in the CorCap group was 12.3 days in contrast to 23.9 days for the control group.
A reoperation advisory panel was held by the investigating surgeons and a consulting pathologist to discuss the reoperation after CorCap. Findings were reviewed in depth with all of the principal investigators both in October of 2003 and in February 2004 and were widely disseminated. The panel members are listed in that slide.
Conclusions based on the above data are the following.
Number one, adhesions, sometimes dense, are encountered in redo operations after CorCap, making dissection difficult at times.
Two, coronary bypass surgery would be extremely difficult after CorCap implant in its current form, and I would not recommend it at this time.
And finally, because of these expected adhesions, we recommend one to two hours are allocated for dissection and cannulation prior to the donor heart returning to the operating room to minimize ischemic time and that dissection would be facilitated often by the initiation of cardiopulmonary bypass.
With these expectations and following these recommendations, transplantation after CorCap can be done successfully, safely, and without added complication.
DR. KUBO: Thank you, Dr. Acker.
This is Dr. Kubo again.
The second question that we'd like to address from the FDA memo is involving the device effectiveness. Question No. 4: does the imputation of NYHA class compromise the analysis of the composite primary effectiveness endpoint which includes NYHA class as one element?
Our response is that, no, the imputation of NYHA class does not compromise the analysis of the primary endpoint, and we make that statement based on the following three points.
First, a multiple imputation is a well established and frequently utilized method to account for missing data and, in fact, was recommended by the FDA. Three different imputation models were conducted and all yielded similar results.
And finally, analyzing the primary endpoint without imputed data provides the same information, and that information is shown on the next two slides.
In this slide, we compute the primary endpoint as demonstrated exactly in the same format that Dr. Jessup presented, but restricted this analysis to the 121 patients who had correlate NYHA available at baseline and during the last follow-up visit. So for these patients there is no imputation of baseline data as done for the primary endpoint.
In this subgroup of patients of 121, there were, again, more patients improved, fewer patients worsened in the treatment group. The odds ratio of this distribution was 1.75, very similar to the odds ratio in the overall analysis.
The p value for this distribution is only .12. So it's not statistically significant, but it is less so because of the reduced power going from 300 to 121 patients, the odds indicating however, that the effect size was quite similar.
Similarly, in this slide, we look at the primary endpoint now reflected or represented as the status at the end of the trial. All patients in the trial had an NYHA status completed at the last follow-up visit, and so this analysis does not require the use of any imputed data which we're missing at baseline.
In this distribution we look at patients in Class I, II, III, or IV. We combine in Class IV those patients who underwent a major cardiac procedure, and the fifth category would be death.
Again, we looked at the distribution of these mutually exclusive categories in the treatment group compared to the control group, and we see here that there are more patients in the treatment group who had a better category compared to the control patients. The proportional odds analysis of this analysis or the proportional odds ratio for this analysis was 1.57 and was statistically significant at the 0.42 level.
Therefore, calculating the primary endpoint data without the use of imputation yielded the same results as with imputation.
The next question that we'd like to address is Question No. 6. Did physician treatment bias affect the outcome of the primary effectiveness endpoint? And if so, to what degree?
Our response to this question is that, no, patient bias for reoperations did not affect the outcome of the primary endpoint, and our basis for that statement is on the following three points.
First, we agree and know that bias can always occur in a trial, especially in an unblinded trial, but that we employed and implemented several interventions to reduce bias in this trial.
Second, the structure of the primary endpoint and the analysis of the data were designed to prevent bias from creating a false signal of efficacy, and that is related to the interdependence of the composite components.
If treatment patients were clearly worsening and there was a reluctance to refer for a major cardiac procedure, by the specific design of the composite endpoint we would expect to see more deaths or more patients deteriorating to NYHA Class IV and that simply was not seen.
And that is shown in this slide or in the next slide. This refers to the interdependence of the primary composite endpoint so that if patients were clearly worsening and they were not getting referred for a major cardiac procedure, we could pick them up as either more deaths or more worsens NYHA.
That is shown in this slide in which we look at the status at the end of the trial and the treatment group in the first column and the control group in the second column. We do see that the treatment patients had fewer major cardiac procedures than the control patients.
However, there was no increase in the number of deaths and no increase in the number of patients who were considered NYHA Class IV. The reason that there were a reduced number of major cardiac procedures is that there are greater numbers of patients int he better NYHA classes in the treatment group.
The last question that we'd like to address in this session is Question No. 11 regarding labeling. The question reads: can the results of this study be extrapolated to an ischemic cardiomyopathic population?
Our answer to that question is that there is no evidence that safety and efficacy are any different in the ischemic patient subset. This involves only 30 patients as outlined by Dr. Jessup. We make that statement based on the following three points.
First, there was no difference in mortality, adverse events or major cardiac procedures.
Secondly, there was no difference in the primary endpoint.
And, third, there was no difference in the secondary endpoints.
On the safety side for these 30 patients with ischemic heart disease, for mortality there were two control and two treatment patients who died. In terms of adverse events, the rate of serious adverse events were similar between treatment and control patients and between ischemic patients and the all patient cohort.
For the ischemic group treatment had 81 percent versus control, 86 percent. In the all patient cohort it was 81 and 78 percent, respectively.
For morbidity, there was one treatment patient who had a transplant. There was one control patient who had a VAD, and for bi-V pacers, there were two control and two treatment patients, showing no significant difference.
In terms of efficacy, we have the primary endpoint which showed no statistical evidence of a difference between ischemic and non-ischemic etiologies. This was investigated through the use of an interaction term.
We do note, however, that with only 30 patients, the detection or the power to detect a difference was very small.
In terms of secondary endpoints in the table below we look at a number of the important secondary endpoints, including end diastolic volume, end systolic volume, ejection fraction, sphericity index, Minnesota Living with Heart Failure score, and the SF-36.
In this we look at the overall treatment difference, that is, the treatment versus control difference in the ischemic patients in the first column and the non-ischemic patients in the second column. Although there are some small mathematical differences, none of these differences were statistically significant, and there was no interaction term, indicating that in general, the effect size on the secondary endpoints was similar in the ischemic patients and the non-ischemic patients.
In conclusion then, we would make the following points regarding safety, that there was no difference between treatment and control in mortality and serious adverse events; that the CorCap showed a significant reduction in major cardiac procedures compared to control; and that the risk of the CorCap and implantation surgery can be mitigated through training and labeling.
For efficacy we make the following four conclusions:
One, that the CorCap study achieved its primary endpoint at a p equals 0.024 level.
It also achieved statistical significance on a number of clinically important secondary endpoints, including cardiac structure, as well as patient functional status.
Third, that the CorCap showed a significant reduction in major cardiac procedures.
And, fourth, that all of these data as represented in a randomized trial are consistent with the large amounts of preclinical and safety study data already presented.
Thank you very much for your attention.
CHAIRPERSON MAISEL: Thank you for a very thorough presentation.
At this point I'd like to invite the panel members to question the sponsor. I'll remind the panel that we will have ample opportunity this afternoon and that each panel member will be able to individually question the sponsor and the FDA later. So I'd ask you to limit your comments to just very important points of clarification or burning questions.
Why don't we start with Dr. Borer?
DR. BORER: First of all, Spencer, Doug and Mariell, I thought that was really a wonderful presentation.
I have some questions really that deal with clarification of your data. First of all, in one of your slides -- I'm sorry I'm not open to it now -- you noted that approximately 11 percent had heart failure secondary to valvular disease. I wonder if you can tell us exactly what kind of valvular disease if it wasn't mitral regurgitation and to what extent are you convinced or what evidence do you have that this was primary rather than secondary valve disease?
I ask the question for two reasons. Let me tell you why so that you can think about this because it may come up later.
First of all, if you were treating patients who had primary valvular disease, the myocardial processes that are involved in generation of heart failure may have been slightly different than those involved in the other patients, and that may have altered outcome to some extent, although I can't certainly suggest that that's true, but it may have.
But more importantly, with regard to the safety issue, the issue of bias that was raised and to which you alluded in response to one of the questions was couched in terms of avoiding a second operation for patients who had adhesions, were known to have adhesions after the CorCap plus mitral valve procedure.
I would wonder if, in fact, most of these people had secondary mitral regurgitation, whether the CorCap might not have actually been beneficial in preventing secondary operations since, as the surgeons know better than I, secondary mitral regurgitation generally is due to abnormalities in the support structures, and if you actually provided more support somehow with the CorCap that might make things better.
And your data do show fewer re-ops for mitral valve disease.
So the question then is: how did you determine that the heart failure was secondary to valve disease? What kind of valve disease was it? And what evidence do you have to clarify that for us?
DR. KUBO: Thank you.
This is Dr. Kubo.
Thank you, Dr. Borer, for that question. I'd like to ask Dr. Acker to come up to describe this information, but I can tell you for the most part the valvular disease that was indicated here was related to mitral valve disease, but also a second point is very important that most of these patients did, in fact, have secondary valvular dysfunction related to the ventricular enlargement. So in my part the valve structures were considered normal without having any anatomic abnormalities, and the regurgitation was a functional, quote, unquote, functional due to that because only a small percentage of patients had mitral valve repair or replacement.
DR. ACKER. Yes. May I have Slide 39 on just to refresh everyone's memory on what Dr. Borer is referring to?
Here we see that idiopathic is caused by 61 percent and valvular is 11.3 percent. In no cases did this refer to aortic valve or pulmonary valve or tricuspid valve as the primary ideology, and the vast majority was a functional MR, functional leak secondary to dilatation, most predominantly from a dilated ventricle, ten percent from an idiopathic etiology of the dilatation, and 60 percent from ischemic etiology in ten, and 60 percent in idiopathic.
When the operating surgeon would look at the valve, there was an occasional rheumatic valve that was identified or significant calcification, and in this case the patients were classified that the dilatation was perhaps secondary to a primary valve problem rather than the MR being secondary to a primary dilatation. In both cases we had a dilated ventricle.
Does that answer your question, sir?
DR. BORER: Probably as close are we're going to get. Thank you.
CHAIRPERSON MAISEL: Dr. Normand.
DR. NORMAND: Yes, I was wondering whether somebody could clarify whether or not in the main presentation when you presented p values, for example -- I think it's on Slide 43 -- are those p values for the composite endpoint based on the multiple imputation?
DR. KUBO: Yes.
DR. NORMAND: And that's using three data sets?
DR. KUBO: Right.
DR. NORMAND: Three imputed data sets.
DR. KUBO: Correct.
DR. NORMAND: Okay. Second question, and I really have only three short --
DR. KUBO: Excuse me. Dr. Brown would like to add to that.
DR. BROWN: My name is Scott Brown. I'm a statistical consultant to Acorn. I have no financial interest in the company.
The only clarification, the reference to three analyses that you saw in one of those slides made reference to an original imputation and two validations, two entirely separate validation imputations. There were 100 imputed data sets in the imputation analysis, which is far beyond what the literature deems necessary, but we went to 100 just because of the fact that there was a good amount of missing data.
DR. NORMAND: So just to clarify, your p values based on Slide 43 use 100 imputed data sets.
DR. BROWN: That's correct.
DR. NORMAND: Okay. How much missing data?
DR. BROWN: There was -- can I get -- pardon me a moment.
While they're getting the slide up, there were 300 patients in the trial as a whole, 126 of whom -- slide on, please -- here's the imputation slide I was referring to. NYHA core lab assessment at baseline was missing for 174 out of the 300 patients.
DR. NORMAND: So that's more than 50 percent.
MR. BROWN: More than 50 percent. That is mitigated by a couple of factors. One of them is in talking about the primary endpoint, if you'll note the note at the bottom, of those 174 patients, 70 of them were classified as adverse for reasons other than NYHA, for instance, death or NCP.
So what that means is that for the purpose of computing a primary endpoint, only 104 patients required imputation of core lab NYHA to actually get the answer in the primary.
DR. NORMAND: In the primary, but for the remaining analyses, I think that's about 60 percent missing.
DR. BROWN: Right.
DR. NORMAND: Which is really not ignorable.
DR. BROWN: For the core lab NYHA, if you look at our core lab NYHA analyses, for example, which also use imputation, that's going to require imputation of all 174 patients.
DR. NORMAND: And you did that. So for primary analysis every time we see p value that involves that, it actually involves 100 imputed data sets.
DR. BROWN: Yes, that's true.
DR. NORMAND: Okay, and then the last question of clarification relates to -- could somebody explain to me what a Minnesota -- it's just a question -- a Minnesota Living with Heart Failure decrease of four points is? What does that mean? Can I walk upstairs now? What does that actually translate to clinically?
DR. KUBO: Yes. This is Dr. Kubo again.
The Minnesota Living with Heart Failure questionnaire is a quality of life instrument that is specific for patients with heart failure. So it's not usable in patients with angina or any other conditions. It is specific for the disabilities experienced by patients with heart failure.
It is a scale of zero to 105. So 105 is the worst possible quality of life. The best possible quality of life without any limitations would be a score of zero.
In general, a patient with New York Heart Class I might be ten to 20. A patient with Class II might be 20 to 30 or 20 to 40, 40 to 60, and over 60 we get Class IV patient. Our mean score was about 57.
In many other trials, a score of about five is considered clinically significant. That is detectable by a number of different interventions, and we've also done some other studies in which we've asked patients, for example, "Would you accept an increase in mortality if you could improve your quality of life by five points?" and many of them did accept that.
So we think that that is a clinically valid, clinically relevant and sustainable improvement.
DR. NORMAND: So just to clarify, I think what I heard you say that a change of about five points or greater is something that would be clinically meaningful.
DR. KUBO: A difference of four to five points, correct.
DR. NORMAND: Five or four? I know I'm pushing here, but I just wanted to -- the studies have said five, greater than five, equal to five points.
DR. KUBO: Yeah, it's not as --
DR. NORMAND: Cut and dried?
DR. KUBO: -- cut and dried as that. I think in the patient scores anything between four and five.
DR. NORMAND: Okay. Thank you.
CHAIRPERSON MAISEL: Dr. Somberg.
DR. SOMBERG: A couple of fast questions if I may. One is it's said repeatedly throughout the presentation and the packet of information that these are severely ill patients on maximum medical therapy. I see that they're on a number of different classes, but I don't see the dose or the duration.
Can you have follow-up material on that? And while you think about that for a second, I would also say that I just heard today that the patients had five years of congestive heart failure diagnosis. That's the means duration. Can you say what class they were?
That seems, you know, just parenthetically, a very stable population, you know. Most patients have a shorter.
And last but not least is on your New York Heart Association classification, I understand that there was 51 percent of the data was not available in, if you will, the committee that was evaluating it. So the site specific had much more data, and that you pointed out was not significant.
When that is factored into the overall analysis without taking into the worsening, better, and all of that other classification which may or may not be accepted, but just to take the three components of mortality, which is not significant, then New York Heart Association would not be significant if one took that on a site specific basis and you just have a major medical procedure. Would that compute out to being a significant difference?
DR. KUBO: Okay. If I could -- this is Spencer Kubo again -- you've asked three different questions. I think I've gotten two of them and I'll just go through them if I could.
I'll take the last one first, and that relates to the use of a site NYHA as an indicator of efficacy. That, of course, the site NYHA is open to all of the biases associated with the physician, and in multiple discussions with the agency, that was felt to be the most biased assessment that was possible and would not be acceptable as an endpoint, and so all of the analysis that we presented here relate to the core lab NYHA, which has at least reduced one form of bias. That is of the investigator.
So the site NYHA was reported, but not really discussed at any length.
DR. SOMBERG: But you can't give me a final answer in terms of using that for the composite endpoint? Because that did have much more data present, and you did mention that it wasn't significant when you looked at it. So I just wondered when you factored into the overall.
DR. KUBO: Yes. If we could have slide on, please, this is looking at the change in the site assessed, NYHA, just as a reflection of -- I think, the types of questions that you're looking at on the left-hand side is the no MVR stratum. On the right-hand side is the MVR stratum. The treatment group is in the solid line. The control group is in the interrupted line.
Here for the no MVR stratum you see a marked reduction in the NYHA class as assessed by the site, that difference being statistically significantly greater than the control group at 0.04, indicating an improvement in NYHA class.
That effect is less marked in the MVR stratum, but again, we're not discussing this in any large detail because of the importance of bias in implementing or affecting the site NYHA status.
The second question, you asked about doses of medications. We don't report doses here. We can provide them to you, of course. The doses of medication and a question that comes up to us many times is what happens over long term. Are patients taking more or less drugs?
And as many of you ar aware, the doses of medications can change. The types of drugs can change from core egg (phonetic) to metoperol, and the doses might be different there. So in the absence of having a clear-cut guidance on what equivalent doses are, it's very difficult to know when someone goes from 25 of metoperol to 6.25 of core egg whether that's any different or not, but we can provide you that.
The third question I think you asked is is this really a chronic heart failure population because of the NYHA or the years of heart failure diagnosis being over five years. We don't have accurate records on their NYHA status during that five-year period of time. As you probably are aware, in many clinic situations not every physician is indicating what their NYHA status is at each visit.
Having said that, we do have a very firm estimate of what their level was at the time of enrollment and all the data are quite consistent with an advanced heart failure population. Importantly though, they were clinically stable because they were about to undergo cardiac surgery. So that was an important point of making sure that these patients were stable, and that, I think, was reflected in the medication requirements, that they had to be on a beta blocker for three months, and the doses of the ACE and beta blocker had to be stable for at least one month prior to entry into the trial.
Did that answer your questions?
DR. SOMBERG: We'll come back to it later.
CHAIRPERSON MAISEL: Dr. Califf.
DR. CALIFF: Thanks for a very detailed presentation, and there's some really fascinating dilemmas here. So I have a couple of questions just to get your thoughts before we get into our own discussion.
Could you describe in a little more detail the data monitoring committee and how often they met and what data they had access to?
DR. KUBO: Certainly. Could I have the slide on the Data Safety Monitoring Board?
I'd ask Dr. Mann to come up and describe that. Actually we don't have a slide on that.
DR. MANN: We don't have a slide on that.
So the Data Monitoring Safety Board met periodically, and they were charged with the ability to stop the study for safety. They did not have the ability to stop the study for efficacy, and they met periodically.
It was headed by Dr. Gary Francis.
DR. CALIFF: Could I ask a follow-up question there? I'm trying to understand. Maybe the most important issue I'm trying to understand is how to position -- you say stop for safety. Obviously you have an imbalance and early death, not a complete surprise when you do an operation. We see that frequently, but it does play into the issue ultimately of how you judge the balance of risk and benefit of any kind of therapy.
Were there any instructions or was it just an open ended look at safety and if you think it's bad stop the trial?
DR. KUBO: Yes. If I could add, this is Spencer Kubo.
The SMB met every six months with an interim conference call at every three month interval. The guidelines that they have were specifically outlined in their charter, which we can provide you or have provided for you in the panel pack, but I don't recall if they had specific guidelines with respect to stopping rules, but based more on an open ended thing.
The other members besides Dr. Gary Francis is the chair was Dr. Chris O'Connor from Duke, Jim Neaton (phonetic), David Holmens, a cardiologist from Minnesota, and Bill Curtis, a surgeon in Seattle.
PARTICIPANT: Now I'm really worried with O'Connor on that committee.
PARTICIPANT: That's an off-the-record comment.
DR. MANN: If I could just add one, this is Doug Mann again.
If you look at the kinds of deaths that happened early, two of them were arrhythmic deaths and wouldn't have happened in the post SCUDHEFT (phonetic) era, so I think that the individuals on the Data Monitoring Safety Board, although I can't speak to what they were thinking, if you look at the kinds of deaths that occurred, they're not atypical for heart failure populations who are undergoing complex surgical procedures.
DR. CALIFF: Thanks. I have just a couple more questions. I think they're relatively discrete.
I was interested, you know. This breakdown of etiologies it not typical for the United States. Was that planned? Did you think you were going to have mostly idiopathic?
DR. KUBO: That's a very good question. I can comment on this. This is Spencer Kubo again.
The reason for this was that we did not allow bypass surgery, and the reason for not allowing bypass surgery is then we would have had in addition to four substrata or two strata, four groups, we'd have six or maybe eight, no MVR, MVR, CABG and perhaps even two more, CABG plus MVR, and that would have led to a situation in which there were so many subgroups that interpretation would be very, very challenging.
And so we excluded bypass surgery as a concomitant intervention, and because we excluded concomitant bypass surgery, our evidence of patient or the incidence of patients with ischemic heart disease was markedly reduced.
DR. CALIFF: So you did expect it would come out that way?
DR. KUBO: Yes, exactly.
DR. CALIFF: Okay. Two more questions, and maybe you want to refer to your statisticians here. I'm a rabid advocate of imputation for missing data, but I've never been involved in imputing more than half the data for a key endpoint, and I suspect we'll have a lot more discussion about this, but I'm interested during your part of the presentation to just get a view of how much of a deterioration in confidence one should have about the accuracy of an estimate of a p value as a function of the amount of data that is imputed.
You know, we used to have a saying in the South that you can't take chicken salad out of chicken whatever, and there's obviously some point if you had 100 percent missing data you'd have no confidence. Here for at least one of the key endpoints you've got half the data missing.
DR. KUBO: Yes, a very important point. Dr. Brown would like to address to address that.
DR. BROWN: Scott Brown.
Just refer again to the amount of missing data. As it was noted before, there are 174 patients for whom the core lab NYHA is missing at baseline, and we did use a multiple imputation scheme to account for this missing data built in consultation with Dr. Kinley Lawrence.
Next slide, please.
A brief background on MI. For those of you who aren't as familiar as the questioner, multiple imputations have been around for about 20 years. It's a statistical technique used to account for missing data. It has been applied in a number of settings, including clinical trials.
And the key thing about multiple imputation is that each missing value in the data says replaced by a distribution of plausible values modeled and other predictors collected from the trial.
What does that mean? It means that multiple imputation preserves to the best of our ability valid statistical inference by accounting for the error in imputation process. When you multiply impute data, you pay a penalty for the fact that the data is being imputed for missingness.
Imputed data is not as good as data that you actually collected during a trial. So the question of how much missing data is too much is partly self-correcting. If you perform a multiple imputation and you have a great deal of missing data and a poor ability to model the missing data, the p values will get worse and worse until the point where significance couldn't be obtained in a trial like this.
Next slide, please.
Now, I just want to remind everyone of one thing. We do have a good amount of missing data on New York Heart Association at the baseline. I just remind everyone the primary endpoint has three components only one of which had missing data. The other two are mortality and incidence of major cardiac procedures which are present in their entirety. So we have one of the three endpoints.
What's more, the primary endpoint relies upon a change in the New York Heart Association classification. So we compared the baseline value to the final follow-up value.
The final follow-up values were available. What was missing was baseline. So you've got half of the change score available to you. You're missing the other half. If you were going to choose which half you wanted to have missing, if I said to you you're going to have to do a change score and you've got to have one-half of the change score missing, which one is it going to be, you would choose the baseline in a randomized trial.
The reason is that in a randomized trial we expect the groups to be comparable at baseline, which makes it a lot easier to handle the missing data. That is the chunk of the data that's missing in this trial. So it's the less difficult half to model.
Having said all of that, there is a statistical measure you can apply to try to measure the impact of an imputation on a final estimate of a number like this. In this case it's the odds ratio for the primary endpoint.
Rubin's 1987 work which introduced this conduct defined a quantity called the rate of missing information, which it's a statistical evaluation of the fraction of the variance in this estimator which is due to the imputation process. It's a number between zero and one, and what it is meant to do is measure the impact of the imputation process on the value that you're estimating.
And for this Acorn trial, that rate of missing information is .09 relative to the primary endpoint. That is actually a fairly low number. If you review the literature imputation, you will see rates of missing information of .2, .3, .4 without necessarily causing severe difficulties with confidence and with inference. So .09 is actually quite low for the primary endpoint as a whole.
DR. NORMAND: I think it's fair to say that that argument holds when it's missing at random. We have more than 50 percent missing data. Most statisticians would assume it's nonignorable missing data, and I don't think those arguments apply.
DR. BROWN: I think it's a fair point that with a large amount of missing data, the missing at random assumption is particularly relevant.
Now, the data in this trial is not missing completely at random. That is, if you think of the data as a matrix of patients and fields, it's not as if we plucked out missing data by throwing darts as a dart board. That's not the form of the missing data.
This data is missing for a structural reason, the fact that the New York Heart Association classification system from the core lab was not available at baseline, but that's not required for imputation. For an imputation model to be valid, what we need to have happen is that the values that are missing need to be predictable from the values that are present, and although it is a lot of missing data, we don't have any evidence -- this is hard to assess. I agree with Dr. Normand.
But we don't have any evidence that the missing at random assumption in terms of predictability is violated in this case.
CHAIRPERSON MAISEL: Tom, you had one question.
DR. VASSILIADES: Yes.
CHAIRPERSON MAISEL: We'll take that as the last question before our break.
DR. VASSILIADES: My question relates to the decision to perform mitral valve surgery. In looking at the data on the design of the trial, it appears that the investigators are allowed to determine the need for mitral valve surgery and then they were randomized. Is that true or did the design of the study have any criteria that specified what constituted the decision to perform mitral valve surgery?
DR. KUBO: This is Spencer Kubo again. The design is exactly as you pointed out. It was at the discretion of the site investigative team. So they made the determination where or not mitral valve surgery was indicated, and that was determined whether the patient would go into the mitral valve surgery stratum. There was no attempt to standardize the recommendations for mitral valve surgery or the indications for that.
CHAIRPERSON MAISEL: At this point, why don't we take a 15-minute break? I have about ten o'clock. We'll reconvene at 10:15.
(Whereupon, the foregoing matter went off the record at 10:01 a.m. and went back on the record at 10:19 a.m.)
CHAIRPERSON MAISEL: Why don't we get started?
At this point I'd like to invite the FDA to give their presentation, please.
DR. BERMAN: Good morning. My name is Michael Berman. I am the lead reviewer for this file. I am a full-time employee of the Food and Drug Administration.
For the record, this panel is convened to consider a premarket application, P040049, for the Acorn CorCap cardiac support device. These are the key members of the FDA review team. I am the lead reviewer. Clinical review was done by Dr. Illeana Pina and Dr. Julie Swain, both of whom are consultants to the FDA.
The statistical review was done by Dr. Laura Thompson, an FDA biostatistician. Preclinical review of different aspects of the device system were done by Eric Chen, Keith Foy, Sharon Lappalalinen, and Bill Reimenchneider, and an assessment of the proposed post market study was done by Dr. Brock Hefflin, a member of the Office of Surveillance and Biometrics.
The order of the FDA presentation will be this brief introduction by me, followed by the statistical review, which will be presented by Dr. Thompson; the clinical review, which will be presented first by Dr. Pina and then by Dr. Swain. Dr. Pina is a heart failure cardiologist in practice. Dr. Swain is a cardiothoracic surgeon. And then Dr. Hefflin will discuss the post market survey or registry.
The device is the Acorn CorCap CSD, as you can see, it is a proprietary polyester mesh which is fit around the heart. It covers both of the ventricles as attached. It is sewn to the heart, and it is sized. The device comes in several sizes. There are accessories to the device system which allows the surgeon to size the heart and choose the proper size CorCap device, and the CorCap device can be customized right at the time of placement.
This is the proposed indication for use as provided by the sponsor. This is in your panel pack. It's indicated for use in adult patients with dilated cardiomyopathy, symptomatic despite treatment with optimal heart failure meds. Appropriate patients are those with a dilated heart with an LVEDD of greater than 60 millimeter or an index greater than 30 millimeters per meter square and an ejection fraction of less than 35 percent unless the patient is indicated for mitral valve repair or replacement, in which case the LDF can be as high as 45 percent.
Let me remind you please that the FDA is operating under applicable law and regulation. So we need to determine whether the data provided by the sponsor provides us with a reasonable assurance of safety and effectiveness. It doesn't have to be perfect. It has to be reasonable, but both safety and effectiveness must be established. Safety and effectiveness are determined as defined by law for us. We need to look at the patients who will be using the device, the conditions of use that are prescribed or recommended, and the probable versus the probable injury.
We perform a preclinical evaluation for all devices that come to us for market clearance, and the things we look at in the preclinical phase are the manufacturing, sterilization, packaging, shelf life, shipping of the device, biocompatibility, mechanical safety and animal studies, and in particular, we look at the animal studies only as an indicator of whether it is reasonable to move forward to human studies. The animal studies are not intended to prove anything other than to suggest that the device is reasonably safe, safe enough to move forward to human clinical trial.
We determined that the preclinical items that we look at were all satisfactory. We have no concerns, and I remind you that this device has no electronics, no software, and so none of that is an issue.
We do, however, have some remaining concerns, some clinical, some statistical. Our clinical concerns revolve around the effectiveness of the device. Some patients receive major cardiac procedure for worsening heart failure and whether or not those procedures were received as a result of worsening heart failure were adjudicated by a clinical events committee, and we don't agree with some of that adjudication based on the records that we've seen.
Our concern is that major cardiac procedure was an element of the composite endpoint, and so should patients have been adjudicated differently, it may have affected the endpoint outcome.
We are concerned that there appears to be a differential effectiveness for the CorCap used in combination with MVR versus the CorCap used alone, and we are concerned with some of the measures for reverse remodeling (a) because the data set is rather limited and (b) because there is a lack of agreement between the agency and the sponsor as to what is an acceptable definition of reverse remodeling and what are acceptable surrogates to demonstrate reverse remodeling.
With regard to safety, we are concerned with the difficulty of reoperation in patients who have a CorCap device, and all of these issues will be addressed by our clinicians.
With regard to our remaining concerns about the statistical analysis of the data set, you heard some discussion that data, particularly data regarding baseline New York Heart class, was imputed. More than 50 percent of the patients had their baseline class imputed by a model, mathematical model. We're concerned about that because the change in New York Heart class was an element of the primary effectiveness endpoint, and if the baseline New York Heart class is problematic, it may make interpretation of the primary effectiveness outcome problematic.
I will remind you that the reason there were more than 50 percent of the patients who needed to have that data imputed is because the sponsor chose to go forward with enrollment into the trial prior to reaching agreement with the agency as to whether or not it was necessary to have a blinded core lab assess New York Heart versus site assessment.
So by the time that agreement was reached, they had enrolled quite a few patients, thereby requiring imputation. However, the agency did not tell them to impute data using any particular model. It was merely a case of saying, "You have a problem. We're aware of it. You need to do something about it. One possible way is to use a model to impute the data," but we did not require them to do it.
We also are concerned statistically that at least one assumption made for the analysis of the primary effectiveness endpoint may be problematic, and in particular, as Dr. Thompson will explain, the assumption of proportional odds may be problematic, and she notes as well that the Type 1 error rate was not controlled for most of the secondary endpoints, which means conclusions drawn from such secondary endpoints are not sufficient to support a labeling claim such as the idea of the reverse remodeling.
So having given that introduction, I will now ask Dr. Thompson to please come and discuss the statistical aspects of this trial.
DR. THOMPSON: Thank you.
I'm Laura Thompson, statistician for FDA, and I'll be doing the statistical review of Acorn's CorCap cardiac support device.
First I'd like to --
MS. WOOD: Excuse me, Dr. Thompson. Please pull the microphone a bit closer.
DR. THOMPSON: Is this better?
MS. WOOD: Yes.
DR. THOMPSON: Okay. I'll discuss the sponsor's study design, their primary endpoint analysis, some concerns that FDA has brought up with respect to the analysis, separate analyses of the components of the primary composite endpoint. I will also discuss analyses of secondary endpoints. I'll conclude with analyses of the primary endpoint by MVR strata, and then present a summary.
For reference here, I give the study design, listing the four groups resulting from the stratification below. The primary analysis compared CorCap to control, pooling across MVR strata. The justification for pooling included the results from a test of interaction between MVR and treatment group which was not found to be significant and also the prespecified notion that any treatment effect would be the same within each strata.
For reference I give the primary composite endpoint definition. I list the three components here, which included all cause mortality, change in core lab NYHA class assessment from baseline, and number of additional major cardiac procedures indicative of worsening heart failure, and I'd like to remind the panel that LVAD and transplant were automatically counted as indicative of worsening heart failure and were not subject to adjudication by the Clinical Events Review Committee.
I also give the ordinal scoring which contained categories improved, same, or worsened. The patient was denoted as improved if they improved on NYHA class and did not die and did not receive major cardiac procedure. They were denoted as worsened if they died or received major cardiac procedure for worsening heart failure or worsened on NYHA class and were denoted as same otherwise.
I also reiterate the differences found in baseline covariates. The four lowest p values are given in the table. We would expect based on the number of covariates examined that about two or three would be significant at the five percent level just by chance alone.
And the sponsor modeled the first three covariates listed in the table in their primary analysis model.
So in addition to treatment group, these were the explanatory variables used in the primary endpoint analysis: MVR stratum, site size. A small site had less than 11 patients. A medium site had between 11 and 16 patients. A large site had greater than 16 patients.
The length of follow-up, early enrollees were enrolled prior to the implementation of the blinded NYHA assessment, and they had greater than 18 months follow-up and late enrollees were the complement of that group, and also the three baseline covariates mentioned in the previous slide.
The primary endpoint was analyzed using a proportional odds model, which models the cumulative probability of an ordinal response. The response has three order categories in the descending order: improved, same, and worsened.
On the next three slides I'd like to give you a background or description of this proportional odds model.
There are two possible binary logistic regression models that could be fit to the responds variable. For the first, we can call a patient a success if they're improved on the composite and call the patient a failure if they were the same or worsened on the composite.
Alternatively we can call the patient a success if they were improved or same on the composite and a failure if they were worsened on the composite. So there are two ways to dichotomize the three categories into success or failure.
The proportional odds model is analogous to fitting both of the above binary logistic regression models simultaneously, but with a common treatment effect or a common odds ratio. Thus a single odds ratio summarizes the treatment effect over all possible cut points in the ordinal response.
For illustration, you can see in this hypothetical picture that the treatment effect called mu, which is the difference in log odds across treatments one and two, is the same regardless of how we cut the ordinal response to dichotomize it, and in the hypothetical illustration it's equal to one.
Because of this constant log odds difference, there is a proportionality property in that the odds of any higher category for treatment one are lambda times the odds for treatment two, regardless of where we decide to dichotomize the ordinal response.
If there is no treatment effect, then the difference in log odds would be zero, and the proportionality constant would be one. Thus, the no hypothesis of no treatment effect is given here.
Now, this is contrasted with the situation where the difference in log odds changes depending on where we select the cut point for success or failure. In this picture here, there is a greater treatment difference when we discriminate between improved versus same or worsened than when we discriminate between improved or same versus worsened.
And we asked that the panel keep these descriptions in mind during the discussion of the primary endpoint analysis.
There were missing data in the composite endpoint for NYHA. An assessment of NYHA class by a site position is available for most patients at both the baseline and common closing data. However, this assessment is unblinded.
An assessment of NYHA class by a core lab based on a questionnaire administered by the site physician and sent to a blinded lab cardiologist was missing at baseline for more than half of the patients.
And as has been mentioned previously, the reason for missing baseline assessments was because the instrument for measuring blinded NYHA was implemented part way through the trial. Thus, only 42 percent of patients have baseline core lab NYHA assessments.
Furthermore, the sponsor has shown a low concordance between the site assessed and core lab NYHA. So we shouldn't substitute the site assess NYHA at baseline in place of core lab assessed NYHA at baseline to solve the missing data problem.
Thus, 58 percent of baseline core lab NYHA assessments were filled in or imputed using an imputation model.
The imputation models considered by the sponsor used observed variables to predict core lab baseline NYHA, and those variables are given in this slide.
The sponsor fit at least two different imputation models. The first model was a linear regression model that predicted baseline NYHA from observed variables. The predicted value was continuous and was rounded to get a predicted NYHA class.
The second model used the ordinal nature of NYHA class to predict baseline NYHA. Both models were used within multiple imputation to adequately address between imputation variability. Fifty-nine percent of CorCap and 57 percent of control baseline NYHA values were imputed.
With multiple imputation we assume missingness at random, that is, baseline NYHA is missing due only to observe variables, in particular, that it is due only to enrollment time. Anyone enrolling before the implementation of the blinded NYHA measurement has a missing baseline core lab NYHA assessment.
We also assume that the baseline NYHA measurements across the two enrollment time periods are the same barring random variation.
Missing not at random would imply that baseline NYHA for early enrollees is distributed differently than for later enrollees, and here early enrollees were those enrolling before the implementation of the blinded NYHA assessment.
In an unblinded trial, there was a concern of selection bias in choosing patients who enter the trial. In this trial, a concern is that later enrollees may be less sick than earlier enrollees. However, we note that if a selection bias existed it might be expected to affect CorCap and control roughly equally, and this is because similar percentages of CorCap and control were imputed. That's 59 and 57 percent, respectively, as was seen in the previous slide.
Nonetheless, we can check whether other baseline variables related to NYHA differ across patients within each enrollment time period to give an indication whether baseline NYHA might differ as well.
In this slide, I compare selected baseline means for early and late enrollees. I've highlighted in yellow the group with mean the worse of the two across early and late enrollees. Note that the worst means mostly apply to the first time period, the early enrollees. However, when comparing means across groups using ordinary T tests, none of the tests are significant at a five percent level.
Now I'd like to turn to the analysis of the primary endpoint. This table gives results from that analysis. I compare results across the two methods of imputation and also with no imputation using only available data, that is, patients with non-missing values on the primary endpoint. The latter analysis was done by FDA.
First, the analysis with no imputation gives an estimated odds ratio of 1.57 which numerically favors the CorCap group, estimating an average of 57 percent better odds of being in a better category on the composite, but the sample size here being about two-thirds of the enrollment size does not provide high power to detect a significant effect.
The two imputation methods, the linear regression imputation and ordinal regression imputation, gives similar results showing on average about 70 percent better odds of being in a better category for CorCap over control, and both the imputation methods show significant treatment effects.
FDA recommended that the sponsor consider imputation as one solution to the problem of such a large amount of missing data. Nonetheless, there are concerns about imputation in this context. More than half of the patients are missing core lab assessment of NYHA. With such a large amount of missing data, result from the primary analysis may be sensitive to the violation of the missingness at random assumption. We ask that the panel please discuss the reliability of analyses that use imputation.
There is also a potential concern about the primary endpoint analysis model. The model used assumes a common odds ration across category cut points. However, if we fit a binary logistic regression model using each of the two possible dichotomies, then the resulting estimated odds ratios show a difference in magnitude.
There appears to be a greater treatment difference in favor of CorCap in improved versus same or worsened than in improved or same versus worsened. In the former, the odds of improving are an average estimated to be two times more likely for control than for -- I'm sorry -- the odds of improving are on average two times more likely for CorCap than for control, and in the latter, the odds of not worsening are on average 1.45 times more likely for CorCap than for control.
We ask that the panel please discuss the appropriateness of the proportional odds assumption here, whether a violation is conservative with respect to rejecting the null or not.
Although the primary endpoint analysis was only appropriately powered to detect a significant treatment effect in the composite, the elements of the composite endpoint were looked at individually in order to determine which components might contribute relatively more.
Toward this end, I will present separate analyses of each of the components. Note that because the family-wise error rate was not controlled a priori for these component analyses, the p values that are presented are not interpreted in the same way as they are for the primary endpoint. The significance level with which to compare the p values is not known. However, a bond for any correction which treats the components as independent of one another would imply a significance level of .017.
In a separate analysis of mortality, a log ranked test of the difference in Kaplan Meier survival curves up to the common closing data gave a p value of .85. There were 25 deaths in each group, and this analysis was presented by the sponsor.
In addition, a Cox proportional hazards model incorporating covariates gave similar results with respect to a possible treatment effect on mortality.
A separate analysis of change in the NYHA component of the primary endpoint does not use patients who had an FCP or died because these patients apparently do not have recorded NYHA at the common closing date.
A proportional odds model was used that used the same categories as for the primary composite, but only using change of NYHA class.
The first row in the table uses only those patients with both a baseline and a final NYHA class and shows an estimated odds ratio of 1.75, p value .12.
The second row uses imputation model number one, and gave an estimated odds ratio of 1.64 in favor of CorCap, p value .18, and the other imputation model gives similar results.
But these two analyses don't use patients who had major cardiac procedures. Instead, the sponsor has assumed that patients who got an MCP for worsening heart failure would be classified as four on the NYHA scale at the common closing date.
If we use that classification, then we get an estimated odds ratio of 1.74 and a p value of .049, but note that this p value is probably too high to override multiple testing concerns.
Also, note that assuming Class IV for NYHA, the worst class, is anti-conservative because there were more controls with major cardiac procedures.
This table gives the percentages of major cardiac procedures by treatment group. There were 22 percent of control patients with major cardiac procedures and 12.8 percent of CorCap patients. The sponsor reported that a Cocker Mantle Hansel test comparing the two treatment groups and also controlling for site size and VR stratum and length of follow-up found at the control had significantly more MCPs.
So it appears that the number of patients who received additional major cardiac procedures for worsening heart failure contributes a great deal to the statistical significance of the composite results.
There were several prespecified secondary endpoints. Five of these were originally denoted as major endpoints, reduced to four by the sponsor during an IDE supplement part way through the trial. The fifth secondary major endpoint used to be six minute walk. These four endpoints were subjected to tests of significance controlling for multiplicity using a Hochberg criterion.
In the following slides I present the individual p values adjusted for multiplicity. Because they are adjusted, these p values can be compared to a .05 criterion.
First, I would like to present the following reminder regarding testing multiple secondary endpoints to support regulatory claims. If and only if the primary endpoint is met, then prespecified multiple secondary endpoints can be tested as a family at an additional overall significance level.
For any secondary endpoints for which multiple testing issues were not considered a priori, statistical significance cannot be interpreted. This is because it is not clear how to adjust p values for the fact that you are using the same data set to test many different hypotheses. The chance could be too high that the randomization to treatment groups resulted in an artificial significant difference on a few of many secondary endpoints just by the way the randomization happened to turn out.
So here are the Hochberg adjusted p values for these four secondary endpoints. The sponsor already presented the p values unadjusted for multiplicity. As you can see from the table, only the p value from left ventricular end diastolic volume is less than a .05 criterion.
Other secondary endpoint tests were not controlled for multiple testing issues. So p values are not interpretable with respect to significance. We ask that the panel think about the use of tests of other secondary endpoints in making statements about intended use.
The sponsor has presented data in the panel pack correlating cardiac structural changes to functional endpoints. The magnitudes of all correlations presented by the sponsor are relatively low, in the range of .1 to .35.
To illustrate a correlation within this range, FDA has plotted in a figure below, the change in Minnesota Living with Heart Failure score from baseline to 12 months, the change in left ventricular end diastolic volume from baseline to 12 months for those patients with values on both of the changes.
The calculated correlation is .22, but the plot below shows no evidence of an association despite the p value being very low at .003.
Statistical significance here would imply that the correlation is significantly different from the absence of any correlation whatsoever, and a confidence interval would imply that we are highly confident in the low correlation. Thus, the p value itself does not indicate a degree of concordance. Rather, the magnitude of the correlation reflects the degree of concordance, and here there is a low degree of concordance.
In fact, the magnitudes of all such correlations presented by the sponsor reflect a low degree of concordance between cardiac structural changes and functional endpoints.
Finally, I would like to present results of the primary analysis with an MVR strata. Before I do that, I would like to give a reminder for stratum specific analyses.
First, we would power the study to detect a stratum by treatment interaction at a prespecified significance level. If that interaction test is significant, we would perform tests within each stratum. Then a within stratum analysis with a significant result can claim a treatment effect.
However, a sample size is not large enough for the interaction test. Then tests within strata can be made for exploratory purposes. The sponsor prospectively intended to examine a treatment effect within each MVR stratus, although the proposed sample size was not sufficient for 80 percent power within each stratum.
A post hoc test for an interaction requested by FDA between MVR stratum and treatment group that controlled for the covariates in the primary analysis was not found to be significant. Thus, no claim can be made for any significant results within strata.
Analyses within strata are done here for exploratory purposes only. In the table, within the MVR stratum with 193 patients the estimated cumulative odds ratio from the primary endpoint was 1.51. Within the no MVR stratum with 107 patients the estimated odds ratio was about 70 percent higher, at 2.57. Thus, the magnitude of estimated treatment effect is higher in the no MVR stratum.
The component that showed the greatest difference across MVR strata was major cardiac procedures. A much larger reduction in major cardiac procedures for the CorCap group was found in the no MVR stratum, a stratum where the control group received no operation.
The estimated odds ratio in the no MVR stratum is more than double that in the MVR stratus.
There was somewhat less of an observed difference in cumulative odds ratios for change in NYHA alone and almost no difference in the comparison of mortality goods.
In summary, for within stratum analyses of the primary endpoint, the MVR by treatment interaction was not found to be statistically significant, although the study was not powered to detect a significant interaction.
Examination within strata showed a larger observed treatment difference in the stratum with the smaller sample size, the new MVR stratum.
Finally, the observed treatment difference across strata might be worth examining further, and we ask that the panel please comment.
In summary, the sponsor met the composite primary endpoint using imputed data at a five percent significance level. However, the large amount of missing data may make inference uncertain. Examination of the separate components of the composite shows a strong influence and reduction in major cardiac procedures.
There were a similar number of deaths in each group.
Results from major second analyses were mixed with respect to finding a significant CorCap benefit. Measures of cardiac structure do not show an association with functional status, and Dr. Pina will discuss this further.
And finally, treatment effect across MVR strata may not be consistent.
DR. PINA: Good morning, ladies and gentlemen of the panel. My name is Illeana Pina, Professor of Medicine at Case and Director of Heart Failure at University Hospitals. Dr. Julie Swain and myself -- Dr. Julie Swain is a cardiovascular surgeon -- have been the primary clinical reviewers for this PMA and are consultants to the FDA and the CDRH Branch.
You have heard a lot of information to day. I am not going to reiterate what you have already heard, but I think there are some clinical points that do need to be clarified.
The intended use of the CorCap, as you see in the slide, is to provide beneficial changes in cardiac structure associated with reverse remodeling. This is an important terminology, as defined by a reduction in LV size, a change in EF, and a change to a more elliptical shape, and that the device also provides a decrease in the need for additional cardiac procedures associated with the progression of heart failure and an overall improvement in quality of life.
This is a prospective, randomized, controlled, two-arm trial of heart failure patients either with mitral insufficiency requiring a mitral valve procedure as determined by the site, or without mitral insufficiency, and they're stratified by mitral valve repair or replacement.
The hypothesis was that the CorCap would improve patient functional status a measured by a clinical composite consisting of mortality, major cardiac procedures which you will see as MCP in this presentation, and change in New York Heart class.
The primary objective you have already heard several times and very nicely stated by the sponsor. The secondary objectives are listed here to determine the rate of death and other adverse events experienced by patients who were randomized to the implant and to compare this rate for patients assigned to control and then to compare the patients' functional status and structural changes between the two groups.
The primary composite efficacy endpoint, again, you have heard it several times, is composed of these three items that you see.
There were multiple secondary efficacy endpoints that are listed in your panel pack. I will be discussing briefly the changes in brain naturetic peptide or BNP.
However, the baseline characteristics do need a bit of clarification. There were multiple exclusion criteria, actually 21 exclusion criteria. One of those excluded patients who were felt to be a high operative risk, and the high operative risks were defined as four of any of several factors, which included a peak VO2 of less than 13, left ventricular diameter of greater than 80 millimeters, heart failure duration of greater than eight years, a lower six-minute walk of less than or equal to 350 meters, previous cardiac surgery and signs of renal dysfunction with a BUN of greater than 100.
Therefore, a higher level of sickness was excluded. This population, as you can see -- and, again, the sponsor has stated this -- is primarily a non-ischemic population, and just to remind the panel that ischemic heart failure is the number one cause of heart failure in the United States, and in most data sets, it's 50-50 in most of our heart failure programs, with the rest of these being nonischemic, including a series of valvular, as determined by the site investigators. These were dilated patients. The peak VO2 looking at this age group, if I compare this to our current NIH-HF action trial where we have Class II and III, our peak VO2 is about 14.8.
And I will speak further to the Minnesota Living with Heart Failure questionnaire. You've heard Dr. Berman tell us one of the reasons for the lack of New York Heart class core lab data where the sponsor continued enrolling patients prior to reaching an agreement with the agency on the core lab assessment.
But it's also worth mentioning that there were missing tests in about 47 percent of the patients, and in this bar graph you can see multiple of these tests, more missing in the control group, the peak VO2 and the six-minute walk could be due to the sickness of these patients.
The primary composite endpoint which we have seen before reached statistical significance using the imputation models, the New York Heart class that have already been reviewed.
The mortality was kindly updated by the sponsor for us as of April 15, 2005. There were seven patients who died within the first 30 days of surgery in the treatment group and one in the control group. One patient in the treatment group died prior to surgery, but is analyzed as an intent to treat analysis on the evening of randomization, and so the true perioperative mortality is 4.3.
However, the sponsor has shown you this bar graph in a different format, that the 30-day operative mortality dropped with the institution of interaortic balloon pump and earlier cardiopulmonary bypass.
Another possible explanation for this is that the earlier patients were, in fact, sicker, and therefore, had a higher 30-day operative mortality.
I just want to review very briefly. I have taken this table and adapted it from our colleague Lyn Stevenson's presentation and the rematch panel showing where this trial fits with other heart failure trials.
Systolic blood pressures very often tell us the level of sickness. The escape trail, which has been more recent which was a pulmonary artery catheter trial, the systolic blood pressure wa 106. The mean in this trial for CorCap was 111. Left ventricular ejection fraction is similar to that of the escape. We do not have any data, have not seen any data on serum sodium.
Here's the six month mortality, and if you want to equate it or compare it to the other trials, the VMAC trial is neserotide (phonetic). This is optimal with milrinone, first with flolan, the rematch group of optimal medical therapy, and the escape trial.
Similarly, to look at trials that have been done in -- clinical trials -- in heart failure patients, namely, Solvd, Dr. Kubo has mentioned Solvd briefly. The consensus trial, these two are ACE inhibitor trials, Copernicus, a beta blocker trials, and the well known Rales spironal lactone trial, showing the difference between control and treatment. For one-year mortality, these are percentage, and this is where the CorCap trial sits.
There are some remaining concerns regarding the points that I will make at each of my following slides. There are still some disagreements with adjudication by the endpoints committee of several of the major cardiac procedures.
A question about bias against re-op of patients with CorCap. Dr. Swain will follow my presentation with a presentation on this topic.
I will discuss briefly the Status 2 transplant patients in this data set. The issues about reverse remodeling, the significance of BNP, and Sharon-Lise, the clinical relevance of the Minnesota Living with Heart Failure differences.
The major cardiac procedures, they were defined as surgical interventions for worsening heart failure, and the procedures that were adjudicated are shown here, bypass, mitral valve repair or replacement, tricuspid valve repair or replacement, and BiV pacing.
Transplants and LVADs were not adjudicated.
Progression of heart failure was determined by any one of these clinical parameters, including the history, physical exam, lack of clinical response to conservative therapy, or numbers on the right heart CAS (phonetic). So these are fairly standard processes to determine worsening heart failure.
In this slide we see the differences between the treatment and the control group for the procedures, and we have reviewed the surgical op. reports of most of these patients, and I would just like to point out several of these. If you look at the mitral valve repair or replacement as a second procedure, two of those patients had significant mitral stenosis without gradients. One patient with tricuspid valve returned to the operating room because of tricuspid valve endocarditis. Another patient in the MVR control group had a transplant, and it was felt that the transplant was due to worsening heart failure. The surgical report states that the mitral valve had tethered and that this led to a worsening clinical condition and, therefore, transplant.
So there are still questions about some of these procedures.
This is the no MVR group showing, indeed, that there were more cardiotransplants in the control group than in the treatment group, and you have seen all of these slides before.
BiV pacing became available during this trial, and has been received by many heart failure teams with great enthusiasm. I just want you to note here the proportionality of BiV pacing, which of course is a closed procedure as opposed to an open procedure in the no MVR treatment group, and again, the issue of bias comes up once again.
Cardiotransplantation, there were patients who were, in fact, listed prior to enrollment under a previous version of the protocol. An amendment that occurred later excluded patients who were listed for cardiac transplantation.
Of the 19 of the 23 patients who were transplanted, the date of listing is, in fact, after the randomization. Four of the patients that were listed for transplant prior to randomization, three of these were Status 2s, and one had been placed in the inactive list; therefore, was Status 7.
Just to review, Status 2s are patients who are not inotrope dependent, could be at home, are waiting for their transplant.
One (b) and 1(a) implies inotrope dependence, and 1(a), a sicker individual with perhaps two inotropes or a higher level of milrenone and some kind of hemodynamic catheter.
Patients who were transplanted as Status 2 would have been automatically counted as New York Heart Class IV, but in fact, these Status 2 patients may have been at home and may not have been a Class IV.
Other points to be made. In the LVAD group, and remember that the first procedure was what was counted as an MCP. If it was an LVAD and then followed by the transplant, the LVAD was what was counted.
None of the 11 patients who ultimately received an LVAD had been listed prior to enrollment. Six of those patients were listed, and the LVAD was used as a bridge to transplant. Three patients were not on the list, and two were listed after the LVAD was placed.
There were three patients who received LVADs that were never listed for transplant. One of these is the patient who expired prior to any other surgery, but still had been randomized to the treatment group, and the other two were patients who did not do well during their initial surgery, and so acutely clinically worsened and received an LVAD.
I want to address now the functional measures to remind everyone that the placebo effect is possible in the less effective measures such as quality of life and New York Heart class site assessed, and that the placebo effect is less likely in more objective measures. So functions such as cardiopulmonary testing for peak VO2 and a six-minute walk.
You have seen, again, some of these data in different formats. These are the observed New York Heart class by core lab without any imputation. There were more patients in the MVR treatment group that showed improvement. This is percent of patients improved who had core lab New York Heart class assessed than in the MVR control group.
And a question had been raised about the site assessed, and in that same group, the site assessed New York Heart class is exactly the opposite. The MVR control group, a higher percent of patients showed improvement in the MVR control group than in the MVR treatment group.
The quality of life assessment was done by the Minnesota Living with Heart Failure questionnaire. There are questions in this questionnaire that relate to the inability to have employment, and so I often wonder how well this questionnaire should be used for a sick population.
But nonetheless, in the no MVR stratum group, and these are patients who had both baseline and 12-month data, both groups improved. The recent AHEFT trial, which has been published in the New England Journal, used a change of five or greater as showing clinical improvement.
The difference between these two groups is not clinically significant even though it may be statistically significant. In the MVR stratum both groups decreased quite significantly, and this difference is even smaller than in the no MVR group.
The six-minute walk is shown in this slide. The percent of patients who actually had baseline and 12-month data. A group of experts from the sponsor determined that greater than 65 meters in the six-minute walk would be a clinically significant change, and you can see that the groups really cluster. Here's the MVR control group and the MVR treatment group being quite similar in this improvement.
In a similar fashion, the cardiopulmonary exercise test, the sponsor with a group of experts determined that an improvement of greater than 0.7, and these are mLs per minute per kilogram of peak VO2, was considered clinically significant, and there, of course, is a large amount of missing data. The MVR treatment group tends to have a greater number of patients in the 0.7 group.
So in functional summary, the placebo effects are most likely in subjective testing, such as quality of life for New York Heart class. You have seen from our statistician, Dr. Thompson, neither the six-minute walk nor the CPX test show clinically significant improvements in the CorCap.
With a large amount of missing data, it is very difficult to correlate functional changes with changes in ventricular dimensions.
The structural endpoints bear on some discussion as well. You have heard the terms "reverse remodeling" being used several times. Reverse remodeling is more than simply a smaller ventricle. It does include changes at the molecular level which, of course, we don't have that data here, but it also includes improvements in LV mass, and you have seen the improvements in left ventricular end diastolic volume and in systolic volume, although with a very, very modest improvement in ejection fraction, a sphericity index that's tending in the right direction, but notice that the LV mass, which is calculated by echo -- by the way, probably the gold standard for calculating LV mass is the MRI -- shows already a reduction in the control group and in the treatment group.
If we stratify this now according to MVR or no MVR, the major amount of LV mass reduction has occurred at the time of the MVR with little added in the treatment group. The amount of improvement in LVEDV or end diastolic volume remains higher in the treatment group than in the control group.
I want to just spend a few minutes going over these data. The sponsor has shown you earlier in their presentation data from Charite. Charite is a hospital in Berlin. The data set that you have been shown is data set that's derived from 29 patients in a single center, non-randomized study, of which 12 patients received the CorCap alone and 17 had other valve surgery that had been predetermined.
There are very large baseline imbalances in these two groups, including sickness severity, such as duration of heart failure of beta blocker use and number of hospitalizations, and there were four in-hospital deaths.
I have reviewed some of these data and show you here that in the small data set it is very difficult to tell differences between the patients who simply had a valve procedure and the patients who had CorCap, both for LV and diastolic dimension, in this case not volume, or New York Heart class.
You have seen a form of this graph, also shown to you by the sponsor. These were six of those 29 patients who had baseline incomplete data at each of these endpoints, showing the reduction in left ventricular end diastolic diameter to shown that these changes happen early and then tend to plateau.
So a very important question with this device is this really reverse remodeling. In the treatment group at six months, 30 percent of patients did not have data available, and in 12 months, 34 percent of patients had no data.
In the control group at six months, 37 percent of patients had no data, and at 12 months, 42 percent of patients had no data.
Notice that the changes in LV volume have already occurred by three months, and of course, we don't have data prior to three months, and then both curves tend in the same direction with the same magnitude.
The change in sphericity very similarly improves early and then tends to drop as does the control group.
So in the questions of reverse remodeling, we have data that are missing, which we now realize may not be at random, with more data missing in the control group. Remodeling, however, is a time related process, and according to Constam Conan and Ender Anant, should be linked to favorable outcomes if you're going to, in fact, use it as a surrogate.
And there are trials that have shown this. Dr. Kubo talked about the carvedilol trials and he talked about the SOFT trial showing changes that led to favorable outcomes, such as improvements in mortality, improvements in hospitalization and even improvements in sudden death.
In this trial there are no differences in mortality, nor are there any differences in hospitalization. Most of the changes occur early, as you would expect true reverse remodeling to continue through time. I believe this is less likely to be true reverse remodeling, and in fact, most of the LV mass decrease is accounted for in the MVR group.
The sponsor collected data on BNP. BNP is now being used widely as a determination of worsening heart failure even though strict values are very, very difficult to assess, and there were data collected at baseline and at six months, and you can see in this slide the number of patients in each group, and in fact, the treatment group had an increase in BNP as opposed to a decrease. I would expect true reverse remodeling to decrease BNP, not to increase it.
So in summary, for the structural parameters, the structural changes reflect the benefit in LV mass reduction due to mitral valve repair or replacement and not to the CorCap added to the mitral valve. The structural changes mostly have happened by three months, suggestive of an early mechanical effect and not to reverse remodeling which should occur and improve with time.
The BNP measures do not support an improvement in filling pressures in the treatment arm. Correlations then between structural and functional changes are difficult to interpret due to the large amounts of missing data which you have heard before.
I just want to briefly talk about the adverse events related to hemodynamic compromise when a mechanical constraint is placed on the ventricle. It does not allow one of the compensatory mechanisms of heart failure patients, which is, in fact, using the Frank Startling mechanism to increase end diastolic volume, and this would be impossible to do with a mechanical constraint.
We all know that mitral valve replacement or repair in the early postoperative period is a tough patient to manage. So you would expect to see this in early hemodynamic compromise, perhaps requiring more inotropes or more vasodilators early. We do not have the data on the early administration of these drugs, but there is a significant difference in hemodynamic compromise in the treatment group early on after surgery.
Constrictive physiology has been brought up several times. We have reviewed all the data that the sponsor has given us. There were 252 patients who had echo data, and the sponsor did, in fact, have an echo core lab.
There were 18 patients in the treatment arm and 30 in the control arm that did not have a follow-up echo, and there were 33 percent of patients in the treatment group and 13 in the control group that at least had one echo that was suggestive of possible constriction.
No patient had any action taken, nor were there any adverse events related to constriction. So at least in the data that we have been given, there are no signs related to constricted physiology. However, constriction can occur across time, and we have no data beyond the 18 months, but at least within the 18 months there do not appear to be any concerns.
So, in summary, the sponsor has met the primary endpoint. The only component of the composite primary endpoint that is significant is the MCP, and you have heard the statistician presentation on this. There are no differences in mortality or rehospitalization.
There are large amounts of missing data which may not be at random, including the baseline core lab New York Heart class. This is an unblinded trial with potential problems of known and unknown treatment and assessment bias. There is an up front mortality cost to the device and the surgery. Only ten percent of the patients tested had an ischemic etiology of heart failure, which is the most common cause of heart failure in the United States.
And I thank you.
DR. SWAIN: Thank you.
As you have heard, one of the concerns we have is that the surgical adhesions might have affected components in this trial, and there are problems with dense surgical adhesions. To give you an idea of the scope of the problem, seven patients had open operations, meaning MVR transplant, after the CorCap procedure. So we're dealing with a small n in these considerations, and 23 patients had control operations. Six of those were in the no MVR virgin chest group, and 17 after MVR.
So today we have the privilege of having four cardiac surgeons on the panel, and you will be able to use your experience to judge this and the medical group on the panel, we've included all of the op. reports, including the control ones.
So it may be instructive to look at a couple of things about adhesions. Now, one of the things about adhesions in that all of the most prominent databases used by surgeons to predict mortality and complications contain redo operations as a component of the ROC (phonetic) curve creating that database. So it is a powerful predictor of problems, meaning mortality or complications after surgery.
So when we refer anyone for surgery, it's always a risk-benefit analysis. If you have a previous operation, you're already up to more of a risk. So that may well change the need to look at benefit. So, therefore, result in a different proportion being referred for surgery.
Well, as Dr. Acker said, not all adhesions are created equal. You know, some of the redos are chip shots. If they've had a mitral valve procedure perhaps ten years ago, that's easy.
We know that there are certain other types of reoperations that are very difficult. Right ventricular outflow reconstruction with prosthetic material and getting back into that chest is sometimes very difficult after infections, things like that. So we're looking at a chest, going back in a chest with a large amount of foreign body present, and what are the potential complications? Why do we care about adhesions?
And I have to say I really admire the sponsor for creating the panel on adhesion or redo operations because it indicated that a challenge, a severe challenge was recognized, and they created a panel to try to determine changes in operative technique that might help change some of the complications.
Well, future coronary bypass operations. None were performed in this study, and the question is: could you ever perform a coronary bypass? Dr. Acker, I think, has answered that.
Myocardial injury. You look at the operative reports, which you all have, and we'll discuss a couple of those in a few minutes.
Phrenic nerve injury. You can't find complications if you don't look for complications, and in order to determine phrenic nerve injury, you very often need a sniff test under fluoroscopy to look at diaphragmatic motion. Very often we say, "Well, it's a piece of the left plural effusion and" da-da-da-da, but we just don't have information on phrenic nerve injury.
Many of these operative reports mention the care that was taken to preserve the phrenic nerves. Other structural injury, such as mammary artery injury, which again speaks to the ability for revascularization, and there was a note on one of the operative reports of injury.
Increased operative time. We try to dig out the heart in non-redo coronary bypass operations off pump, and most of these op. reports you see most of the dissections were done off pump, and we don't have operative time data. We have cardiopulmonary bypass data, which may not indicate the operative time, but by the recommendation of the committee that before you get a heart transplant, a heart back in the room, you ought to put aside one to two extra hours for the dissection.
And we also don't have information on blood and blood product use on these patients, and we did see some data on postoperative stay, but when you're looking at an n of seven compared to an n of 23, and you're looking at mean, and I noticed one of the outliers was a 46-day hospitalization, it really makes it difficult for me to interpret the data.
When we look at efficacy, again, you're looking at a risk-benefit analysis for referral for operation. So there could be a possible higher bar for referral, knowing that one would have the probability of encountering dense surgical adhesions and what that implies.
Now, we're not allowed to talk about where operations are done or who the surgeon was. So that if you all will look at your, the panel members, the extra sheets that we gave you of operative reports, and maybe first of all go to page 11, and you'll just notice it's a standard operative report, and we have the demographics of operative reports, where the operation was done, who did the operation, things of that sort, and that's common throughout all of these operative reports. So we'll get to page 11.
And you look at certain comments. The heart and great vessels were encased in some of the most dense mediastinal adhesions I have encountered, and that's the one I think everybody is pretty much at, page 11, Dr. White. He's getting there.
So, you know, that's a fairly interesting statement, and I must say I have not seen that in operative reports that I've either dictated or read over the last 20 or 25 years.
Then if you look at page 2, it's a heart transplant removal of an LVAD as opposed to mitral valve repair, Acorn BIVAD. It's a multiple operation, and we received an MVR, medical derive report, regarding this, and you do not have that medical device report in your pack, but the MVR stated that there were extremely dense pericardial adhesions obliterating the plains between the heart, pericardium and surrounding tissue, and those adhesions were extremely dense adhesions and almost made transplant impossible.
The patient received over 20 units of blood, blood products to control post operative bleeding. Again, we don't know the amount of blood products given to these cohorts.
Page 23 was a heart transplant, and again, when you do a heart transplant, it doesn't matter getting the previous heart out whether you injure the myocardium. That would matter to other operations, such as valve replacements, coronary bypass, things like that.
So on this they were a very intense and difficult dissection for the period of approximately two hours, severe dense adhesions throughout the mediastinum, and the procedure for freeing the heart was extremely tedious and long. So that gives you a little better idea of what's going on in that case.
And if you look on page 28, another one that's a heart transplant after mitral valve repair in Acorn, and we made a subepicardial dissection to peel the epicardial layer off the myocardial fibers. So that speaks to having to essentially shell out the superficial layer of the heart to get the heart out and leave the device in place.
And finally, page 19, you can see talking about adhesions under the complexity part that the adhesions in the chest were extremely dense, adherent, and exuberant, and developing a plate around the heart was impossible.
The recipient cardiectomy was performed with great difficulty. The left side of the pericardium was inadvertently detached from the diaphragm and the inferious aspect of the diaphragm had to be reconstructed.
So that indicates, you know, damage to surrounding structures. So I think that what we have, and again, to be complete and fair, you have all of the control operative reports, and you have to judge the amount of statements of difficulty percentage-wise in the seven patients in the device group versus the ones in the control group, and the surgeons use their experience.
So, in summary, there are questions that remain about the effect of adhesion formation after CorCap on both the safety and the efficacy analysis of this device.
DR. HEFFLIN: Good morning. My name is Brock Hefflin.
CHAIRPERSON MAISEL: Can you please fix the microphone.
DR. HEFFLIN: Good morning. My name is Brock Hefflin. I work as a medical epidemiologist in the Office of Surveillance and Biometrics, and this presentation provides a summary and assessment of the CorCap cardiac support device condition of approval study.
First, some general principles. The objective of conditional approval studies is to surveil or evaluate over an extended period after premarket establishment of reasonable device safety and effectiveness, device performance, and potential device related problems.
The purpose does not include evaluation of unresolved issues from the premarket phase that are important to the initial establishment of device safety and effectiveness. In other words, the conditional approval study should not be used as a substitute for the premarket study.
The sponsor's proposed condition of approval study is five-year surveillance of up to 348 patients with the CorCap CSD. As in the pivotal clinical trial, principal outcome variables include NYHA class, mortality, adverse events, and echo measurements.
Inclusion criteria are similar to those of the pivotal clinical trial. The condition of approval study has three components: a group of treated patients extended from the pivotal clinical trial plus two new groups.
Data from these components are to be combined, stratified, analyzed by study variables, for example, demographics, MVR versus non-MVR outcomes, and submitted to the FDA in an annual report.
The stated objective of the condition of approval study is to evaluate the long-term performance of the CorCap CSD in the general population. The proposed surveillance provides appropriate variables to meet this objective. However, to obtain the most meaningful analytical results, the following items should be considered for the study.
The study should include an appropriate comparison group, and this is the subject of Question 14 to the panel. Such a group would facilitate the interpretation of device safety and effectiveness of data results.
Continued five-year follow-up of the control group from the pivotal clinical trial would be appropriate. Alternatively, historical controls from the literature that have had several years of follow-up may also be suitable.
If a comparison group is utilized, then clinical outcomes and meaningful differences are needed to make the comparison. Rates of mortality, for example, might be compared between the two groups, a meaningful difference in mortality than would need to be proposed and accompanied by a rationale to support it.
Finally, there is no indication that the proposed surveillance will attempt to reflect the distribution of prevalent heart disease etiology in the general population, notably ischemic disease.
Heart failure etiology may impact device effectiveness. Therefore, the accurate representation of prevalent heart disease etiology in the general population may be wanted for the study, and this is the subject of Question 15 to the panel.
We believe these additional elements are needed to make this study one that will provide results that can be interpreted with greater objectivity, and that can be applied to the general population.
CHAIRPERSON MAISEL: Thank you very much.
At this point I'd like to open the floor to questions from the panel for the FDA.
DR. FLEISCHER: Just for the record, I'm Dina Fleischer. I'm the Branch Chief of Circulatory Support for Prosthetic Devices.
And since our team is so large, I think I'm going to try to field the questions and then defer appropriately when possible.
CHAIRPERSON MAISEL: Rob.
DR. CALIFF: I'm confused about the second guessing of the events committee. It might be worthwhile to hear a little bit about the process issues there. Typically in a prospective clinical trial, one appoints an events committee that has a systematic, blinded as much as possible approach to adjudicating events.
Typically, if one does a re-review of cases, you'll find disagreement with itself ten or 15 percent of the time I would say would be a sort of standard, but you're talking about clinical adjudication here. So I'm just unclear as to why there was felt to be a need to go back and second guess an independent committee.
Was there a concern about the process? If so, it would be useful to hear about it.
DR. FLEISCHER: Well, in the course of the review, yes, I mean, looking at the patient reports, but actually I'll let Dr. Pina actually address because that was her actual comment.
DR. PINA: Rob, obviously, I agree with you that that's the purpose for endpoints committees. We were given the op. reports for the re-op. patients, and we read all of the re-op. reports, and at least in two of those, as I've stated before, it was very clear that the re-op. was due to, for example, metrostenosis and not to worsening heart failure from deterioration of LV function, which is the implication of putting a mechanical constraint to prevent further remodeling.
So, yes, I agree that it is a bit unusual to go back and do that, but we were given the reports, and the reports were read.
DR. CALIFF: So I'm still not completely understanding this because finding a few discrepancies would be routine, but are you saying that these were such great discrepancies that we can't trust the overall work of the events committee?
DR. PINA: No, we're not saying that at all. As a matter of fact, I just pointed out to the few discrepancies that I found and the rest have not been questioned.
DR. BORER: May I just follow that up, please?
CHAIRPERSON MAISEL: Dr. Borer.
DR. BORER: Illeana, when you did that re-review and raised the questions, was it on the basis only of the information in the op. report or did you go back and retrieve charts and look at notes of the attending cardiologists, et cetera?
DR. PINA: You know, the sponsor has been very good at providing clinical summaries when they have been requested, and in at least two of those patients, the clinical summary was given back to us, which is exactly what the CRC would have seen.
And after reviewing that I personally still disagree with the adjudication, but, no, they have provided everything that we have asked for.
CHAIRPERSON MAISEL: Dr. Somberg.
DR. SOMBERG: I just wanted to clarify my understanding of the statistical evaluation, and to that extent, when one looked at the primary endpoint, the composite, and one took into account the subanalysis the FDA did with the group regarding New York Heart Association class and looking at when the data was imputed, et cetera, it was a nonsignificant difference.
When that was all taken together, was it still the composite endpoint significant taken all together? And I understand that the major surgical interventions is the major moving force of that, but when one takes away there's no positive immortality; there's a real question with heart failure and one makes that nonsignificant, would that then be the composite endpoint, still be significant?
I didn't see that data. I asked the sponsor. They may want to answer that, but I would like to see what the FDA has to say.
DR. FLEISCHER: Do you actually have that analysis, don't you?
DR. THOMPSON: Dr. Somberg, I'm not quite sure I understand your question. Are you asking whether we disregard major cardiac procedures entirely and then -- okay. Please ask.
DR. SOMBERG: No, no. To be specific is if you take at face value the major cardiac procedures as the way it was, but now with your reanalysis of the New York Heart Association classification, which I saw was not significant in that subset, we didn't have to impute large amounts of data. I think there was no data imputation. Was the composite endpoint still significant in your reanalysis?
DR. THOMPSON: I'm sorry. I still quite don't understand. We have --
DR. SOMBERG: New York Heart Association, not significant by itself.
DR. THOMPSON: Right. When I looked at just --
DR. SOMBERG: Mortality, not significant by itself.
DR. THOMPSON: Right.
DR. SOMBERG: And then you look at the composite endpoint. Is the composite endpoint still significant?
DR. THOMPSON: Well, the composite endpoint is only looked at when you have all of the data in the composite, and so there's no --
DR. SOMBERG: No, I mean put it all together then. So you have nonsignificant --
DR. THOMPSON: I don't understand what you mean "put it all together." That's what I'm not quite understanding.
DR. SOMBERG: Well, initially there are three determinative positive endpoints.
DR. THOMPSON: That's right.
DR. SOMBERG: One of them is the mortality. That stays the same. New York Heart Association is no longer different in terms of being significant if one takes out the imputed data and then one just has the major cardiac procedures.
And when one takes all of that data into account, what is the final p value?
DR. THOMPSON: When you only use the available data on the composite endpoint?
DR. SOMBERG: That's right.
DR. THOMPSON: I did present a summary of that analysis, but I'll show you some more detail in the back-up.
DR. CALIFF: Is that summary on Slide 15, Dr. Thompson?
DR. THOMPSON: It may, in fact, be on Slide 15.
Yes, I believe so, and it would be the first row. Let me pull up some details regarding the percentages in each of the composite categories.
Sorry. You get to see all of this.
I apologize. This is more detail regarding the first row of the table, and this includes available data, which means all patients who had a measurement on the primary endpoint, and the odds ratio, the estimated odds ratio, is still in the direction in favor of CorCap, and actually the percentages are more or less analogous to those that were presented by the sponsor when you use the imputed data.
I believe there is a little bit of increase in worsening. The percentages for these two are somewhat higher in both groups.
DR. SOMBERG: But that's only in the roughly 93-95 patients where all data is complete.
DR. THOMPSON: That's right.
DR. SOMBERG: Okay. So if somebody has all data in the major surgical procedures but they only have incomplete data with New York Heart, they're dropped from the study?
DR. THOMPSON: No.
DR. SOMBERG: I mean they're dropped from this analysis.
DR. THOMPSON: In this particular analysis, no, no. If they had a major cardiac procedure, then they were counted as worsening according to the composite, and they were not removed from this analysis.
DR. SOMBERG: Okay. I see. Thank you.
CHAIRPERSON MAISEL: Dr. Borer.
DR. BORER: Just for clarification purposes really, Illeana, you mentioned this and I wanted to ask it of the sponsor before. The issue of the BiV pacing, the popularity of that procedure has dramatically increased recently with the publication of a paper that suggested improved mortality rate, reduces mortality rate, but clearly it wasn't applied to all of these patients. How many patients would have been eligible for BiV pacing given the current criteria with a QRS greater than or equal to 0.13? Do we know, you know, how many or what percentage of the population would have been eligible and what percentage of those that were eligible actually had that procedure done so that we can put this into some context concerning the added benefit that might have been inferred?
DR. PINA: It's a terrific question, Jeff. I haven't seen every single piece of data, but I believe that the sponsor does have the QRS, and from what I was shown of the QRS, it was an inappropriate indication for the QRS lengthening, and I didn't see any QRS lengthening that did not get the BiV pacer. So I had very few concerns about that.
DR. BORER: So does that mean that there was a disproportionality because more people in the control group got biventricular pacing? Does that mean that there was a disproportionality in the QRS duration distribution in the population?
DR. PINA: The people that got it did appropriately have that. Now, having said that, it was interesting to note the rate at which the BiVs went in was very correlated to the enthusiasm. Now, the clinical endpoints committee who adjudicated every BiV pacer decided that a group of them had not been put in for worsening heart failure, and so they did not fall into the major cardiac procedures, and they were taken appropriately out.
DR. BORER: Okay, okay. Thanks.
CHAIRPERSON MAISEL: Michael.
MR. MORTON: I'd like to thank the agency for a thorough review. I'd also like to recognize that OSB made a presentation here regarding the post approval study. I think that's very important, but I have a question. Obviously, one of the issues that we have on the table today is this NYHA classification by core lab, and in looking at the chronology of the history of this review, it looks like the IDE was approved in, say, June of '01 and then the protocol evolves.
Could we understand how that issue came up about core lab NYHA?
DR. FLEISCHER: Can we discuss that at the forum?
Well, I can also actually do just a sort of hypothetical situation. Also I want to make it clear that there are cases where issues regarding protocol that are not safety related do happen to be discussed in studies that can be ongoing, and that it is a risk that sponsors will take while agreements are being made or discussions are being taken on whether or not agreements have been made that the studies will be ongoing, that that -- gosh, what am I saying? -- that they will agree to be going on their study while we are still in discussion on that particular topic.
But we did look back to see what the actual chronology was on that particular topic, and Dr. Pina can tell you.
DR. PINA: Yeah, I wasn't present at every single one of those discussions, but New York Heart class, even though it's a very imperfect system because it's based on a lot of subjective sense from the clinician of what the patient is telling them and the patient's subjective sense of what they can or what they can't do. It's still what we use very commonly to reflect in our mind what these patients look like.
When patients are undergoing an unblinded trial where there's an intervention, there's always that sense of the placebo effect of the intervention and perhaps a clinician not being totally objective about the New York Heart class. And so the discussions appropriately came up about a core lab, but these discussions -- I was only present at one of them by phone -- took time to go back and forth and discuss, and the sponsor was making a comparison between site assess and core lab assess that perhaps Dr. Kubo, who has published now on this, could discuss this, but it took some time.
But during that time that the discussions were going on, patients kept getting enrolled, and the sponsor could have had the choice to stop and wait until the finalized agreement with the agency had occurred, but they kept enrolling. So it is a time dependent phenomenon with multiple discussions.
CHAIRPERSON MAISEL: Well, just as a point of clarification, I believe you stated that was the IDE approved by FDA prior to FDA expressing a concern about the core lab?
DR. FLEISCHER: You can answer that. I believe it was.
DR. PINA: As far as we're concerned, yes, it was.
CHAIRPERSON MAISEL: Okay. Thank you.
DR. ZUCKERMAN: Dr. Maisel, I would state that a little bit differently. Frequently we approve IDEs conditionally, meaning that we don't have any preclinical safety concerns. There may be ongoing questions with the sponsor, but the sponsor can utilize the conditional approval letter for two purposes.
One is to start talking with IRBs to get the trial ramped up.
Two, if the sponsor disagrees with the agency about the particular significance of an agency question, the sponsor can proceed at their risk.
CHAIRPERSON MAISEL: Ms. Fleischer, do you want to?
DR. FLEISCHER: Yes, that's correct. Also I wanted to also say that we did approve a feasibility study for the child and then we also were working through lots of issues with the pivotal trial.
CHAIRPERSON MAISEL: Clyde.
DR. YANCY: Thank you.
I have two discrete questions that I'd like to have addressed. In Slide 10 of Dr. Thompson's presentation, there is a statement that the sponsor showed a low concordance between the site assess and core lab NYHA. My specific question is whether or not you could look at that degree of discordance and determine if the bias was in favor of treatment or in favor of MVR.
DR. THOMPSON: Actually, I believe the sponsor may be better equipped to answer this question. I did find a -- just when I looked at the data I found a low correlation, but I don't remember looking at what the sponsor had presented. I don't believe there was any indication of any sort of bias one way or another, but they may wish to correct me on that.
DR. YANCY: And while you're at the mic, let me just ask you one other question, please, about statistical integrity, for lack of a better word. We've heard quite a bit about the missing data vis-a-vis NYHA class, which may have been based on design issues, but other statements were made about missing data regarding objective measures of exercise, missing data regarding structural measures of ventricular function and size, and there are statements in briefing documents regarding missing data on BNP assessments.
What is your statistical gestalt when you see this kind of a profile of missing data? At what point do you question the integrity of the database?
DR. THOMPSON: Well, for the so-called objective measures of functional status of peak VO2 and exercise data, the sponsor has said that the missing data, which was a fairly substantial percentage, was missing not at random. It was missing for the sicker patients.
And normally under a circumstance like that, we would expect that if it's missing for the sicker patients, then the missing value itself would be different from the data that are not missing, and so we would not like to then analyze the available data because we could get a biased result.
Regarding any other missing data in the structural measures, I don't believe there was a high amount in any of their secondary endpoints besides just those two, with a relatively low amount, you know, like five or so percent. Really it probably doesn't make that much of a difference, but for the two particular secondary endpoints, peak VO2 and six-minute walk, I would somewhat question those results.
CHAIRPERSON MAISEL: Judah.
DR. WEINBERGER: It's clear that one of the big problems here is that we don't have baseline NYHA data that was adjudicated, but we do have 100 percent total NYHA data on patients who survived to the end of the study who didn't fall into the other events.
So this is probably statistically invalid, but if one were to look at the distribution of NYHA classes, would there be any way of determining an ordinal endpoint off of that, off of that 100 percent evaluated data?
DR. THOMPSON: This time I know where the slide is. Unfortunately this slide was not put in, but I can recite the analysis that I did.
I looked at the distribution of NYHA class at the end of the trial, comparing treatment and control, and I did this in terms of a proportional odds model where the categories are ordered, Class I, Class II, Class III, Class IV, and finally debt. I excluded patients who had major cardiac procedures, and this is because, for one thing, I didn't know where to put that category. I didn't know necessarily to put it, you know, after debt before debt, between Class III or IV or whatever. So I don't include those patients, but it does include everyone else, and the total patients in the treatment group would be 131, and the total number in the control group is 118.
The proportional odds, the estimated odds ratio is 1.50. It is in favor of control group. I'm sorry. It is in favor of CorCap group. So by interpretation that would mean the CorCap group has on average a 50 percent better chance of being in a better category.
Now, the p value is .093. Across the five categories the distribution or the percentages of treatment and control are actually very similar for a Class I, NYHA Class I. The treatment group had 1.5 percent, the control 3.4 percent.
In Class II, treatment group, 28.2 percent, control, 23.7 percent.
Class III, the treatment group had 32.1 percent, control 28 percent.
Class IV, treatment group 19.1 percent, and control 23.7 percent.
And the death, 19.1 percent of the treatment group, and 21.2 percent in the control group.
I don't know if you got all of those, but I apologize for not having this slide available.
CHAIRPERSON MAISEL: Thank you.
DR. NETROVEC: Is there any evidence or any data that suggests that the etiology, most specific ischemic, though very small percentage, did it disproportionately drive any of the endpoints?
DR. THOMPSON: Well, from what I understand, there was only a very small percent in one of the particular etiology groups. So my answer, just my guess would be no. I didn't specifically look at that.
DR. BLACKSTONE: You mentioned the problem of proportional odds. Could you tell us whether you looked at a nonproportional odds model?
DR. THOMPSON: Yes. Well, I did fit a nonproportional odds model, but before I present those results, if they're even there, I do want to say a couple of things regarding fitting an ordinal response.
The problem with fitting a nonproportional odds model, if we just take the particular model that I'm talking about here and just make the two treatment effects different, you have to obey a particular ordering. In other words, the probability of being, let's see, same or better has to be at least as large as being improved. So there's a cumulative ordering that has to be obeyed.
And when you do different treatment effects, that may not necessarily hold. Nonetheless, if you want to take a look at what the results are, I can show you, but for the most part, they're similar to when I fit the separate models, and in fact, let me just show those because then I don't have to flip through all of these slides because it's basically the same numerical result, and I can remind you of the slide number. This is in the slides you have. I believe it's 17.
Okay. My concern was that these two estimated odds ratios were not the same. We did request that the sponsor justify the proportional odds assumption, and they did present a score test that showed a nonsignificant result, meaning that the proportional odds assumption held.
However, that test looks at all of the covariates or all explanatory variables together, and I was particularly interested in just the treatment effect or -- I'm sorry -- just the treatment group, and this is what I got.
If you notice they are both at least in the same direction, you know, in favor of CorCap.
DR. NORMAND: Isn't it true that the lower interval is contained in the upper? I mean those intervals overlap.
DR. THOMPSON: I'm sorry?
DR. NORMAND: Well, the 1.45 is contained in the -- those intervals overlap. So --
DR. THOMPSON: Right, right, and as I have mentioned to the sponsor before, this concern has been somewhat abated since it has been included in the panel pack. So as far as I'm concerned, I'm not as concerned about it as I was previously.
CHAIRPERSON MAISEL: Dr. Borer.
DR. BORER: I have two, again, clarification issues. My recollection, number one, is that there were about eight patients, I think it was, maybe it was seven, who were randomized and refused surgery, who randomized to surgery and refused surgery. I don't know that we can infer very much from their outcome, but I'd sort of like to know what it is, whether it's more like the control group or more like someone else, one of the other groups. That's one thing.
And then in sort of a follow-up to George's question although I could easily understand how we can't learn much by trying to analyze the group of patients with prior known coronary disease alone in such a small study, there were about 15 patients with coronary disease who were operated on, and some of them had had bypass grafting procedures before, and even though this is not in the data set that was presented, there have now been several years since they were originally operated on.
Has any of them required or been considered for bypass grafting procedures subsequent to their study procedure, and if so, do we know what the degree of difficulty was if such an operation ever was performed?
CHAIRPERSON MAISEL: Those sound like more appropriate questions for the sponsor. So maybe we can ask the FDA to step back and give the sponsor a chance to answer that, if that's okay.
DR. ZUCKERMAN: Well, those are excellent questions that would be very appropriate for this afternoon's session. We're just trying to wrap up any remaining FDA questions on points of clarification now.
CHAIRPERSON MAISEL: Any other questions for the FDA?
CHAIRPERSON MAISEL: Lunchtime. Reconvene at 1:00 p.m.
(Whereupon, at 12:00 noon, the meeting was recessed for lunch, to reconvene at 1:00 p.m.)
CHAIRPERSON MAISEL: Good afternoon. Why don't we begin our afternoon session?
And we'll start by having our lead reviewers ask questions of the sponsor and make some comments. We'll start with Dr. Somberg.
DR. SOMBERG: Well, good afternoon. When I first saw this PMA application, I said this is very, very interesting, and I thought it was a very novel idea and the sponsor should be congratulated for pursuing this. The concept of essentially a low tech intervention that might benefit congestive heart failure sounds very promising.
The review of the material that I made though raised in my mind several disturbing points, and I'd like to focus in on what's been discussed, I think already extensive and will be more extensively discussed.
One of them is the primary composite endpoint, and while it looks very reasonable, and I understand the sponsor's rationale for trying to capture all possibilities, I think in reality it turned out that it raises more questions than it answers, and my observation is, as well as everyone else's, is that there's no difference in mortality between the intervention and the control group.
There is a problem with the New York Heart Association class, and I can empathize with essentially the changing standard that occurred, but with that said, I would have gone back and said, "Well, you know, if we're missing about half of the data set, and it's usually an initial evaluation which is kind of critical in that regard, what we should do is go back and find something that could help us correct that, as well as try other techniques to impute data."
Well, I think that was well intentioned, the imputation. With this much missing data, it really raises a great deal of questions in my mind, and thus, I would have gone for the site selected or designated criteria and with that on the sponsor's packet there on page 52, it turns out to be nonsignificant. So two of the three of the composite endpoints don't seem to show any difference, and the one that does is the major cardiac procedures.
And when one looks into that, the question that the FDA reviewers raise was is there a bias because of the potential for surgical difficulties, complications and the arduousness of the task which I think everyone realizes of the reoperation.
And what strikes me as brought out in the sponsor's and several other places, too, but on page 36 of the table, looking at the different procedures, and if one goes across that, one sees that there's really no difference between CRT, and obviously in an unblinded study, CRT is the only intervention that the proponents of the intervention we know is not going to require surgical operation on patients who may pose a great hurdle.
So I think by having that equally pretty much distributed between the treatment and the nontreatment group, it sort of confirms to me that there was a bias in the investigator's minds, and that could be conscious or unconscious, and I'm not impugning anyone's integrity here, but there is a bias favoring less procedures in the cap treatment group.
Now, supposedly, okay, you have less procedures in the cap treatment group. The New York Heart Association class should shoot up because you didn't intervene appropriately in the control population, and they're going to get sicker and they're going to die more.
Well, that's only true if they're really on the cusp and if they didn't have these interventions, it would be a life or death situation, and maybe that's not the case. Maybe the trigger was just too early pulled in the control population and it wouldn't be picked up by mortality and we're not going to know it by the New York Heart Association class because of the problems with the data collection or the imperfectness of New York Heart Association classification to pick up subtle differences in any event.
So whatever we say with these things, really the primary endpoint when one dissects it is wait, and what's at the secondary endpoints and it turns out the ones that are harder like six-minute walk, MVO2, a lot of missing data there again, seem to not show a difference and the ones that are weaker are the ones that do show the difference.
Now, there is a difference that we see in terms of heart size, and I think that was the primary or the driving hypothesis that if one reduces the heart size, one sees an improvement in this patient population. So it's nice to know the heart size did get smaller, but I did not see a very tight linkage with the demonstration that if one reduces the heart size, one sees a difference in outcome here.
And in fact, most of the studies where we're approved leaving the general regulatory community has approved drugs and devices has been to see an improvement in mortality, and it's not impossible to show that. I mean, the very crude ACE inhibitor studies early on showed marked reductions in either sicker or less sick populations or reduction in hospitalization, and neither of those two endpoints moved at all with this particular intervention, which raises some very serious questions in my mind.
Then I looked at the other major concern, and there are a lot of more minor concerns, but I'm going to raise in a short period of time the other major area of concern, and that was the population we're serving.
The population we're serving is patients with severe end stage congestive heart failure, usually etiology ischemic heart disease, and I thought it was, you know, just the vagaries of life and the way the study was sort of designed to exclude, but it looks like there was a conscious decision that ended up pretty much excluding 90 percent of the patients who -- well, that may be an exaggeration -- but a large percentage of the patients who have the etiology we're most likely to deal with, and that's ischemic heart disease.
So this study had only ten percent. Patients with ischemic etiology where we're usually 50, 60 percent, the general patient population. So this is very worrisome, and I would say that if the study was very strongly positive and the primary endpoints were all consistently moving in the right direction, I would still say that we would have to limit the use of this device to that subset of patients with idiopathic heart disease because we would really need to do further studies on patients with ischemic disease, and as we've heard today, you know, surveillance post marketing studies are really not for that purpose, but really it should be done pre-market approval.
So I'm very concerned of the safety, if you will, because of the limitation of the population that was studied, and why is it a safety issue? Well, we hear that it's virtually impossible to operate on patients after this device has been implanted for a period of time.
So at the least one would have to very clearly define the patients and make sure that if this device was implanted we would have excluded ischemic heart disease as a further exacebater of the condition and have to deal with all ischemic problems up front because later on our most useful modality of intervention which is still coronary artery bypass surgery would be taken away for the most part from this population once the cap has been implanted.
So I think we have not studied the appropriate population, and we have not demonstrated the efficacy of the procedure. I think there are other concerns that come to mind, that is, that there is a cost paid for this device, and therefore, it has to be balanced and, of course, be up front operative mortality.
Alibi, possibly could be mitigated with appropriate learning by the surgeons, et cetera, and liberal use of on pump bypass to place the device, which of course makes it even more invasive.
But still, I think we have a serious adverse event up front profile with this device, and that has to be balanced against a very significant benefit, and we really don't see that signal from the primary or the secondary endpoint.
And finally, I think it was to the sponsor's credit. This is data I didn't see in the initial pack. So they presented it today, this morning in the review where they try to place in context the cap in terms of other therapies and how much of a benefit one sees by it.
But I was startled, to use the word, by the dramatic benefit of the control population here compared to all the others, and that seems to tell me either it was miraculous intervention in this particular population and they were so good at manipulating medical therapy that the controls were better or that this was a very interesting subset of patients that behaves differently than one of mostly the case in the other control populations, and I think it's the ladder.
Therefore, I think it's a nonrepresentative patient population where I do not see a distinct marked clinical benefit demonstrated, and thus I have considerable concerns about the approval of the cap device.
CHAIRPERSON MAISEL: Thank you, Dr. Somberg.
DR. YANCY: I'd like to provide my comments in the form of several questions. I think we've all heard summaries of the protocols. So I'd rather not do that.
Let me, first of all, comment that I believe that today's presentation was probably the most fluid and comprehensive I've heard in two years on the panel. So all of you should be commended for that.
Secondly, as a practitioner and one who deals with this patient population, I have no angst whatsoever about the inclusion primarily of patients with non-ischemic etiologies. We've not had an appropriate surgical intervention for that patient population and having something that extends medical therapy and device therapy for that cohort, I think, is a reasonable objective.
I also think the investigators should be acknowledged and commended for doing a more than credible, but an excellent job of meeting evidence based treatment strategies because that really should be the standard against which we should judge interventions, and I think to a certain extent it makes the ability to demonstrate efficacy a bit more challenging.
Let me start with the primary composite endpoint and just raise questions, and I'll leave it to the chair to decide if we should chronicle these or answer these as we go.
My first question is a question of inclusion. Within the part of the primary composite that demonstrated the most directional change, the major cardiac procedures, BiV pacing is incorporated. Certainly times have changed since the study was initially started and so now we have an intervention that carries an indication independent of some of the measurements that were used in the study, but I'm just curious as to what the original thought process was because obviously it's not the same as transplant or LVAD or cardiac surgery.
So a comment from the sponsor about the rationale behind including BiV would be helpful.
CHAIRPERSON MAISEL: Do you want to do one at a time or --
DR. YANCY: It's your choice.
CHAIRPERSON MAISEL: Your choice.
DR. YANCY: Well, let me do the inclusion and exclusion and then I'll let you comment on that.
The thing that was notably excluded from the primary composite was hospitalization, and I think all of you spoke very eloquently to the issue of having a primary composite that matters to patients, and I find it odd that hospitalization wasn't in that composite because clearly that does matter, and it is a standard that we've previously held other device trials to to incorporate hospitalization even if the intervention itself required a hospitalization.
Because part of what's important is to understand if the downstream hospitalizations are impacted vis-a-vis heart failure or are still in toto the same because there are other complications that occur. So let me see if I can stop here and have you address the issue of inclusion with the BiV and then the issue of exclusion with regard to hospitalizations.
CHAIRPERSON MAISEL: Could you actually stand at the podium? We like to leave that table empty.
DR. KUBO: Spencer Kubo again.
Thank you very much for your comments.
I'll respond to Dr. Yancy's questions first. The first question is about the inclusion of BiV pacers as part of our major card procedures. This, of course, was a problem for us because it was approved as an intervention while the trial was being conducted, and so we had to handle the issue of biventricular pacers.
And we met with the entire investigative group and had a discussion about our two options, which were to not allow them into the trial and, therefore exclude the noise factor that could occur or to allow them for patients who might need them,b ut to adjudicate them but to only include them in cases of worsening heart failure.
The discussion was quite extensive, as you might imagine. We also decided, by the way, that if patients were considered a candidate for BiV pacers, they were to get the BiV pacer first to see how they worked, and then if they were only sick and met entry criteria for a minimum of three months, then they were to get into the trial.
So we tried as best as we can to minimize the impact of an additional therapy on our endpoint.
DR. YANCY: Because if I can follow what you just said, the concern, of course, would be that the BiV pacing per se would invoke a reverse remodeling effect in some substrates.
DR. KUBO: Correct, correct. So that's why we wanted to have it done beforehand. You know, of course, that there are patients who don't respond to BiV pacing, and then that would be a very good candidate for the trial, and if they did respond and were better, then they wouldn't have met the entry criteria and they wouldn't have gotten in to the trials. So that was a good way for us to minimize the impact.
The decision analysis that went into this discussion was to say we should allow BiV pacers. That would be very difficult to continue enrollment in a trial in which an effective and approved therapy was withheld from patients.
And so the vote of all of the investigators was to allow it to occur. Now, once we allow BiV pacers to come into the trial, we want to sort of limit their access or only use it in cases where it is clearly indicated, and that's where the teaching about only doing this for cases when there's worsening heart failure.
Now, there was clearly an element or potential for bias in that the criteria for entering a patient into getting a BiV pacer was smaller than what you might see with a transplant, and that was precisely the reason why we established the CERC committee so that they would review these events and only count those events that were clearly associated with worsening heart failure counting towards the endpoint.
So That was the intent. That was how we dealt with the advent of a newly approved therapy, and I think we did quite well.
Having said that, one concern about BiV pacers is that they might have affected the major cardiac procedures and they may have affected the primary endpoint. We have done an analysis of the primary endpoints in which we excluded all of those patients as worsened who received BiV pacers.
And if I could have the slide on, please, this is the same primary endpoint with excluding biventricular pacers. All BiV pacers are out and not counted as worse, and we followed patients to their logical endpoint. We find that the odds ratio is .168, very similar to what we saw in the primary analysis, and that is still significant at the 0.03 level.
So we think we've been able to allow BiV pacers and allow an ethical and rational treatment, but to minimize its impact and noise effect on the outcome of the trial.
DR. YANCY: I have done the arithmetic as well and thought that it looked a bit more in favor of treatment. Tell me about hospitalizations.
DR. KUBO: Okay. The question that now you point is: why did we exclude hospitalizations from a primary endpoint, and of course, hospitalizations is being managed in many different trials. The difference in a heart failure trial as opposed to an intervention that requires surgery is there are many hospitalization that occur after surgery that are related to the surgical process, the post-op afibs, the pleural effusions. You get a little bit of wound infection, and so forth., and we were concerned that that noise level, if you will might lessen the efficacy signal that we might expect to see with reduced hospitalization due to the effectiveness of a device. And so we were concerned about counting hospitalizations in that.
We did not have an adjudication process for the causes of hospitalizations. That might have been a way around this issue so that would get at the heart failure related causes.
That, of course, has been done in other trials. It is something perhaps in retrospect we might have done. We didn't do it because we really were concerned about the extra noise, if you will, about hospitalizations.
DR. YANCY: Because even though it's noise, arguably it's something that's important to the patient.
DR. KUBO: Absolutely, absolutely.
DR. YANCY: Let me go to a separate line of questioning. Thank you, Spencer.
Let me focus on the secondary endpoints, if we can, for a bit, the structural endpoints, in particular. It looks to me as if the magnitude of change specifically in the LVF was less than we would have hoped it would have been, even though we had favorable changes on sphericity, diastolic and systolic volumes.
Just in terms of helping me understand what your perspectives are as principal investigators, how do you feel about the observations on LVF? I think the slide that Dr. Jessup showed demonstrated a one unit change that was in favor of treatment, but none of us would think that's clinically meaningful, and especially since the phraseology "reverse remodeling" is incorporated in the overall indication statement.
The community physician identifies a change in LVF as the metric for reverse remodeling. So if one of you can address that and amplify that a bit that would be terrific.
DR. MANN: Doug Mann.
These are both good questions, Clyde, and I want to also address partially Dr. Somberg's question, too.
I think the first thing to emphasize is that we were comparing ourselves to what we believe to be an active therapy group in that two-thirds of the cohort was undergoing mitral surgery, which as you k now reduces the amount of volume loading that the ventricle has and will lead to reverse remodeling in and of itself.
So this is a trial where all of the odds were stacked against the CorCap and we still came out and showed benefit in that we had an active therapy arm.
Can I have the slide on, please?
So if you look at the ejection fraction changes, there's about a three -- from baseline -- there's about a three percent change in the control group, or 2.5, and about a 4.8 in the treatment group. As shown, the changes -- so there's an improvement in the ejection fraction, both groups, which we attribute to, in part, the mitral surgery and the good medical therapy, but as you can see, the change within the control group was not significant over time, whereas it was highly significant within the treatment group.
The problem with echo measurements of VF is there's a lot of inherent noise, and so if one compares treatment to control, at the end of 18 months there's really not a significant change between groups. Yet within groups there is a significant change within the treatment group. So that's where my comfort level comes from, and again, I think that the difference in comparing treatment to control has to do with the fact that we're looking at an active therapy group.
We know beta blockers improve therapy. We know the ejection fraction. We know biventricular pacers will, and we have some evidence now that maybe mitral surgery will.
Have I answered your question?
DR. YANCY: Well, that slide, actually is very helpful. So thank you for showing that.
DR. CALIFF: What's so helpful about that slide?
CHAIRPERSON MAISEL: Why don't we let Clyde finish first?
DR. YANCY: No, I'll yield.
CHAIRPERSON MAISEL: Each member will get a chance to comment. Why don't you continue, Clyde?
DR. YANCY: I was just curious. I'm sorry. What are the units on this?
DR. KATO: These are EF units.
DR. YANCY: EF units. So a percentage ejection fraction?
DR. KATO: Yes, absolute unit changes.
DR. MANN: Rob, the only reason that I said that is helpful means that it looks like within the treatment group baseline to post intervention there's a signal that the EF is going in a direction that you would hope it would.
MS. WOOD: Just to clarify, we're dealing with the primary reviewer's reviews right now. If you have questions or comments regarding the information the sponsor may show in relation to the questions Dr. Yancy is asking, can you please hold them until the general discussion following this review?
DR. YANCY: Let me move to another question, please, and that's on BNP. I think that this is a troublesome area because, as I've looked at this several times and saw yet again today, within the treatment group there was an increase in BNP. Now, there may not yet be uniformity on this concept, but many people believe that BNP, more so than being a marker of filling pressure, is, in fact, a marker of left ventricular stress and/or maybe even a marker of remodeling, and so, again, along the lines of addressing this important question of whether or not reverse remodeling is impacted, I find it counterintuitive to see the BNP increase in the treatment group.
So additional explanation or amplification of that observation would be helpful.
DR. KUBO: Thank you very much.
We have also been interested in BNP and following the efficacy of using BNP in identifying patients and also tracking patients, and as you pointed out, we do not have the types of anticipated efficacy that we were anticipating with the BNP in that the treatment group had a higher BNP than the control group.
Some of this is related to the acute effects of surgery, which have now been published to raise BNP, and in the no MVR stratum, as you know, the treatment group gets surgery where the control group does not, and that is in part related to it.
The other point, I think, which made the BNP analysis for our study difficult is shown on this slide, if I may.
This looks at the baseline BNP levels among the patients who are in the trial, and this was performed through a core lab, John Burnett at the Mayo Clinic, and you see the number of patients listed there, as well as their baseline level of BNP.
And fully 55 percent of patients had a BNP level less than 100 picagrams per deciliter, which is in the normal range. This was a finding that was at first a surprise to us because we expected to see much higher levels of BNP, but, in fact, other publications recently from Gary Francis at the Cleveland Clinic has indicated that stable patients, despite having advanced heart failure, can have a normal BNP level.
And I think our finding here is that this also reflects, this normal BNP level also reflects the fact that we had stable patients, very well treated with background medical therapy including ACE inhibitors and beta blockers, and that that clinical stability is what led to a normal BNP level.
So lowering a BNP level from a normal level is very difficult and, I think, reduced our ability to find an effect here.
DR. MANN: Doug Mann again.
I want to just quote Yogi Berra here. There's a quote that "what gets you in trouble is not what you don't know but what you think you know but ain't necessarily so." And I would suggest that our knowledge of BNP today is incomplete. I mean, we started out with this as a biomarker for heart failure. We're now finding that it's elevated in a variety of conditions for which we have no ready explanation.
My interpretation of the data set here would be that it's in response to cardiac manipulation. I would remind the committee that we think BNP is a good protein. It's a vasodilator. It's antihypertrophic. It has a variety of biologically good effects.
I wouldn't necessarily try to extrapolate beyond what we know about BNP today. So I think the BNP data set were a wash. I don't view them as being of concern, and I would argue that we don't really understand the full significance of BNP, particularly in the context of cardiac surgery.
DR. YANCY: Let me just pose another perspective here because what I'm looking for are signals of consistency, consistency about the patient population and consistency of other treatment effect.
And when I see a mean LV of -- I'm sorry ‑- a mean VO2 that's 15 and I see a six-minute walk test that is in the upper 300s, a characterization of 2.93 or so for the NYHA class, but then measures being p levels that are lower, assuming that we have a sufficient sample size, it's inconsistent because you are correct. The database is still in evolution, but several of the things that we all accept is that at least groups of patients tend to have BNPs that are roughly parallel their symptom classification, and this does not meet that same standard.
And so it raises a question in my mind about consistency because we do see treated patients with low numbers, but that's not the norm. People tend to have a BNP that goes with their symptomatology.
Let me do one more question if the Chair will let me, and I hope I've interpreted this slide correctly. This is Slide 57 of the sponsor's presentation, and it's in the MVR stratum, and it's freedom from death or significant adverse event.
And if I understand this correctly, the two lines, treatment and control, identify the cohort that receive mitral valve surgery, but the difference is that the treatment group received mitral valve surgery plus CorCap, and I see no signal of benefit with regards to the important issues of mortality adverse events.
Would this tend to make one pause and think that in this setting by these measurements that CorCap adds nothing beyond them mitral valve procedure? Have I misinterpreted this?
DR. KUBO: Yes, thank you.
Could you put Slide 57 on? I'd like Dr. Acker to respond to that.
DR. ACKER: Yes, this slide is really not meant to show a difference in efficacy, but rather to show that there is no safety concern. Importantly, they both had open heart surgery. So it's really the best test to say is there adverse events associated with the CorCap alone, and this slide would suggest there is not.
There's no change in difference in mortality, but more importantly, which we already know, there's no difference in adverse events. If you look at the MVR stratum, then we can't differentiate the CorCap from heart surgery, from a sternotomy, from ventilation.
This one does control for that, and it reassures me that there is no safety concern for the CorCap alone.
DR. YANCY: But using that same logic, when I look at this and I consider trying to identify, again, areas where the data are consistent in terms of the benefit, even if it wasn't statistically significant, I would like to have seen something that demonstrated that the treatment group fared better than the control group.
Maybe this wasn't the right metric to look at, but that's what we're looking for here.
The last line of questioning and then I'm complete has to do with the question of constrictive pericarditis. I think that this, in my mind, an unresolved area because we all appreciate this is a time dependent phenomenon, and so the first question is: how soon in the surveillance process did you see changes consistent with restrictive physiology, and what would you anticipate the long-term issues would be?
Because I don't think we've seen sufficient data to know if a significant cohort of this group won't end up with pretty symptomatic constrictive pericarditis in another five to ten years. If you equate it to radiation injury, it can take one or two decades before it appears.
DR. KUBO: Yes, Spencer Kubo again.
The issue about restrictive pericarditis and constricted physiology is very important, and the time dependency of this issue is also very important to follow.
We have evaluated over 1,300 echoes during this period of time, and the echoes could be obtained at three, six, 12, 18, 24 months, and there have been isolated echoes throughout that period. So there doesn't seem to be any predilection for early or late.
The issue though, and as was discussed already by Dr. Jessup, we found nothing to indicate an association with those echoes with adverse events, with hospitalizations, with other adverse outcomes. The issue though about following patients for a longer period of time, that this may be something that won't occur within the first year, but later on, is very important, and because of that, we have continued, we will continue our surveillance with the echoes with the same reading by the echo core lab using the exact same criteria that we've had in the trial, continuous out for five years in the trial because it's very important that they have the ability to use this and interpret it in the same context.
DR. YANCY: I would assume there are no animal data available on long-term issues like that.
DR. KUBO: The animal data are more difficult to obtain long term because of the expense associated with keeping an animal alive for a longer period of time.
However, we do have the opportunity to do other experiments in the animals, and if you could put Slide 65 on.
These are data from Dr. Sabbah's (phonetic) model of microembolization in which we create heart failure due to episodic administration of emboli through the LAD and the circumflex arteries. And in this model we are able to provide a volume load, giving 400 mLs of Destran 40 acutely, and I think this actually refers to a question that Dr. Pina said, that the CorCap would take away one of the compensatory mechanisms of the heart that they couldn't dilate.
But, in point of fact, the ventricle is able to adapt to this volume load with an increase in size both at three months and six months, as well as an associated increase in ejection fraction. So the compliance of the CorCap which is intended to reduce the amount of chronic filling is able to allow the acute increase that is the more normal response to a volume load.
This is an important physiologic point, but again measured in animals, not in patients.
I think Dr. Mann had something else he wanted.
DR. YANCY: Well, let me just ask one other thing. Before you leave the podium, particularly, because I think you're probably best suited to address this, several times during the presentations this morning we heard the comment of missing data, and I just think it merits a response from the sponsor and the investigators.
Was this due to patient acuity, a tracking issue, change in the CRFs? Give me some sense of why we hear that qualification on so many data subsets.
DR. KUBO: Yes, I'm going to let Dr. Mann speak to that.
I'd like though you to think about the difference in missing data with respect to the core lab baseline NYHA, which we've discussed at some length and will continue to discuss, with the missing data from the exercise test which we have discussed and will discuss some more, but actually the high degree of compliance and the very good follow-up that we have had in all the other endpoints, and Dr. Mann will describe that to you.
DR. MANN: This is Doug Mann again.
I just want to follow on with a question with the constrictive physiology. It's important to emphasize that about 80 percent of the constrictive physiology was episodic. Okay? We know that post surgery you can get episodic constrictive physiology, but not true cardiac compression or true cardiac constriction.
In the cases that we followed it was episodic. It was a single event. It wasn't evident on a repeat analysis. So we're in an era now where we have this exquisitely sensitive way to detect physiology, which may have no clinical significance, and in fact, 15 percent of the control patients actually had episodic constrictive physiology as well.
So in answer to your question, and we will have long-term follow-up.
Could I have the first slide on please?
This slide shows the patient follow-up status for the patients who were randomized to a CorCap at the control arm. As you can see here, there's only one patient who is alive whose clinical status is unknown in the CorCap group and in the control group there was only one patient whose vital status was unknown.
So our overall follow-up of the cohort was really quite good.
Next slide, please.
This slide shows the visit attendance and the compliance with visit attendance, and as you can see here, the overall compliance with visit attendance was quite high, ranging somewhere between 94 to 90 -- about 90 to 95 percent over the main part of the trial. So overall compliance with visits.
In some instances there were missing data because the patients were too sick to exercise, but the overall attendance at the visits was quite high.
Next slide, please.
And this slide here just shows you the test compliance overall by major endpoints/secondary endpoints. So you can see that the compliance for the people who completed the blinded assessment of the New York Heart was 92 percent. We had 93 percent compliance with echo, 93 percent compliance with SF-36, and 94 percent compliance with the Minnesota Living with Heart Failure questionnaire.
I think if you take episodic missing pieces of data, you can come up with very high numbers, but I think that they're spurious. As you can see, visit attendance was quite good. Our follow-up on vital status was quite good, and as you can see here, the overall assessment or the ability to collect the relevant data was quite good.
Did I answer your question?
DR. YANCY: Sure. Since you're still at the mic, let me have you answer the question that Mariell answered in her presentation, what do you tell your patients, because in a contemporary context with patients on ACE, beta blockers, aldo antagonists, device therapy in place, especially this patient population, we've already gone over what their identifying characteristics were.
What do you tell them the benefit of this intervention would be based on your data?
DR. MANN: That's a great question.
You know, I as a clinician, the way I view this as somebody who has put the device in patients and who has worked with the patients who have the device, I think that this actually stabilizes the disease process.
What's my evidence for saying that? Well, we reduce the need for major cardiac procedures, and there was an improvement in quality of life.
So what I tell a patient is that despite optimal medical management and despite optimal device therapy, there is a chance that you're going to progress or you have progressed on optimal therapy.
I have nothing for you at this point in time. If you get sick or you may qualify for a VAD or you may qualify for a transplant, but right now you're feeling poorly, and I don't have anything for you.
We have a device that we think will stabilize your disease process and leave you feeling better for a longer period of time and hopefully forestall the need for transplant or aggressive device therapy with a ventricle or support device. That's what I would tell a patient.
DR. YANCY: And I guess that stabilized the disease process part that I really want to continue to think on this afternoon because we don't have the mortality benefit. We don't have the hospitalization benefit, and most of the cardiac procedure interventions, including transplantation, involves some subjective decision making, and that is compromised a bit when you have an unblinded study.
So that one phrase you just mentioned, stabilization, is what is still a friction point for us.
DR. MANN: Yes, I guess I would agree with you that implantation of a biventricular pacemaker would be subjective, and we've shown you the data that the primary endpoint is robust even after subtracting that out.
I would disagree that the implantation of a VAD or a transplant is purely subjective and that most of us just don't go and walk in the clinic one day and say, "You need a VAD," or, "You need a transplant." This is something that after a long period of time of trying every medical therapy we have, realizing that we're not doing well, we then go to the patient and say, "Okay. We think you should be listed for transplantation. Let me refer you to a transplant specialist."
I don't think it's quite as subjective as you think. I think it's really driven by lack of alternative therapy that will improve the quality of life.
CHAIRPERSON MAISEL: Okay. At this point we'll give each of the panel members an opportunity speak. Given the large size of our panel, I'd ask you to keep your comments to about five minutes, and why don't we start with Judah?
DR. WEINBERGER: Okay. I actually will try to be very succinct. I'll try to, anyway.
Can somebody from the sponsor tell us what's the frequency of inadequate echo cardiograms? Nondiagnostic, nonevaluable, inadequate, technically difficult echo cardiograms.
DR. KUBO: For which endpoint?
DR. WEINBERGER: Post implantation of the Acorn device. We're making decisions about diastolic volumes, about constriction, about cardiac physiology based upon these echoes. My assumption is since there has been a lack of a statement, that we're getting 100 percent information on these endpoints; is that correct?
DR. KUBO: Correct. We have the specific numbers. I can get you those. For the endpoints of end diastolic volumes and dimensions and so forth, the number of missing patients are missing data because of inadequate echoes is very small.
For the Doppler evidence of constrictive physiology because it's a much more complicated set of measures that would include a respirometer, the percentage is higher, and it's about 15 percent, I believe, but I can get you the specific number on that.
DR. WEINBERGER: So that's 15 percent of the echoes or 15 percent of the patients?
DR. KUBO: Fifteen percent of the echoes.
DR. WEINBERGER: So we don't really know what the instance of constriction is because that's the only readout for constriction.
DR. KUBO: Correct. There are patients who will have, among six echoes, might have five that would be interpretable and one which is not. So it does happen that way.
DR. WEINBERGER: All right. So presumably there's a subset of patients which are technically difficult which don't have any Doppler studies at all. No? That's wrong?
DR. KUBO: No. The studies that we showed you include patients who have at least one evaluable echo.
DR. WEINBERGER: Okay, all right. Have there been any PCIs in patients who got the Acorn device?
DR. KUBO: By percutaneous coronary interventions?
DR. WEINBERGER: Right.
DR. KUBO: No, there have not been.
DR. WEINBERGER: Okay. So we don't know anything about the natural history of Acorn on top of ischemic disease.
DR. KUBO: That's correct, and I should mention the reason why that is so low, and I think that's a question that Dr. Borer asked, is one of the requirements at the entrance at trial was if the patient had ischemic heart disease, they were to have an angiogram, an angiogram to determine their likelihood of a need for coronary bypass surgery. If they needed it, they were obviously excluded from the trial or if, on the basis of that angiogram, they thought that a coronary intervention might be necessary within the foreseeable future, those patients were eliminated because of the issue.
DR. WEINBERGER: Let me flip it around. Has anyone who had a previous stent gotten the Acorn device?
DR. KUBO: I believe in our --
PARTICIPANT: (Speaking from an unmicked location.)
DR. WEINBERGER: No, I want to know whether you know anything about the natural history of putting an Acorn on top of a stent that has already been implanted.
DR. KUBO: Yeah, we'll get you the exact numbers of patients without coronary interventions. We have done angiograms in patients who have had the CorCap placed, and there has been no compromise of the coronary vasculature. We don't anticipate any issues, but our experience is very limited, but I can get you the exact numbers of patients who have had stents prior.
DR. WEINBERGER: Okay, and I'd just like to drill in a little bit more on some surgical issues. I mean, the descriptions of the surgeons as were reiterated by the FDA are rather impressive. If you have to go back into this chest, is just getting in safely an issue? Is that or is the issue at a subsequent point?
In other words, do you have to be careful about slicing into the right ventricle? Is everything of that matter down?
DR. KUBO: I'd like Dr. Acker to respond to that.
DR. ACKER: In any redo operation, I'm concerned about injuring the heart, okay, whether it's a coronary bypass operation, a valve or VAD or whatever. So we always have that in our mind, and this is a redistrinotomy the RV can be stuck up to the sternum.
Having said that, you know, today, you know, if you're a transplant surgeon, half of the transplants you do are for re-ops. So we're very familiar with how to open the sternum to minimize the chance of a catastrophic RV laceration which did not occur in this study at all.
As far as the other concerns of reoperations, I think the important point being the right atrium, the great vessels are not involved in the CorCap, and in my experience, having done reoperations on this patient group, that can be easily dissected out very similar to a more standard re-op and you can get on bypass safely, and then you can take your time dissecting out the ventricle.
So for a transplant it is not an issue. For a valve it is not an issue. For an aortic valve it's not an issue. Would it be an issue for coronary bypass? Yeah, we've already said that, but for other issues it's not.
DR. WEINBERGER: So in your practice you would not recommend this for somebody who has an underlying coronary artery disease that you think the natural history of that patient might come to revascularization at some point?
DR. ACKER: Well, in my practice, and certainly in the study we do not look at that.
Can you put on the back-up slides regarding coronary bypass surgery?
There is an experience with that, and there is at least 45 patients in Europe and Australia that have had one previous to that perhaps, have had ‑- slide on, 132.
As this slide shows, there's 45 patients who have received CABG and CorCap. So if I had a patient with ischemic myopathy, for instance, who had needed revascularization and this was an approved therapy, I would have no qualms about putting a CorCap on.
Again, a patient with heart failure, we know what their natural history is. They're going to die within, you know, several years, and revascularize them at the same time.
MS. WOOD: Excuse me, Dr. Acker, but I believe that I'm being told by the review team that this information was not part of the submission and, therefore, cannot be considered by this panel.
Is that correct, Ms. Fleischer?
CHAIRPERSON MAISEL: I think it would be helpful for -- they're only showing it in direct response to a question from a panel member.
MS. WOOD: All right.
CHAIRPERSON MAISEL: So you can go ahead and continue.
DR. ACKER: I was just trying to respond to the direct question.
And what I would be reluctant to do is three years later reoperate and do surgery revascularization. However, with the advances in endovascular revascularization, stent therapy and complete the original revascularization, especially in arterial graphs, I would find that unlikely to be an issue and would outlive the heart failure syndrome.
DR. WEINBERGER: That's music to my ears as an invasive cardiologist.
DR. ACKER: Okay.
DR. WEINBERGER: I'll turn it over.
DR. ANDERSON: Excuse me. Steve Anderson, Vice President of Regulatory.
I just want to point out that the information on concomitant CABG was, indeed, in the PMA submission.
CHAIRPERSON MAISEL: thank you very much.
DR. KATO: A question for the sponsor. There was a slide that was put up that showed that you had something like a 92 percent compliance on SF-36 and peak VO2. I think that was one of the slides that we just saw.
But, on the other hand, in the presentation this morning, it said missing follow-up data. The numbers were much worse than that.
Can you explain that discrepancy?
DR. KUBO: Yes. Dr. Brown would like to answer that question.
DR. BROWN: FDA cited several numbers on missing data in their presentation. I believe one of the ones you may be referring to is an analysis that there's an indicator that 47 percent of patients were missing a test at some point.
DR. KATO: Yes.
DR. BROWN: They did cite that number. What that number refers to is -- and we haven't precisely confirmed this with FDA -- but we believe what that analysis is is for a given patient among any of these tests, as well as additional tests which were not analyzed but were done up, for example, blood work and X-ray, among any test at any visit that patient might have attended, as many as five or six visits, if they missed one of those tests at some point in time, they were counted as having a missed test by FDA.
So that's where that 47 percent number comes from. One of those potentially 20, 25, 30 tests was missed at some point. These numbers indicate on a test-by-test basis what the compliance was, and you can see the numbers are in the low to mid-'90s.
DR. KATO: Because at least according to this graph on page 5 of 29, I think, of the document that we have from you is that the numbers are a lot worse. Again, this graph is not enclosed, at least on my document here, but for example, the SF-36, I guess, you're probably down in the 85 percent range. The MLHF, it looks like you're down also in the 80 to 85 percent range also.
So I don't understand why those numbers saying 93 percent, but these numbers are saying --
DR. ACKER: I'll get the document to which you are referring. I thought one of the slides from our presentation is at our panel pack.
DR. KATO: Yeah, it's on page 5 of 29 on the document that came from you.
DR. ACKER: Okay. I'll take a look at that. There are a number of reasons why a test could have been missed, including, for instance, patient death, intervening procedures which prevented a visit, visit missed attendance, and then, of course, if you did attend a visit you may have missed the test even so.
But I'll get a look at that page and see what the discrepancy might be.
DR. KATO: Okay.
DR. ACKER: Oh, I'm sorry. I'm being told that's FDA's results again.
DR. KATO: Oh, yes. I'm sorry. This is the FDA summary memorandum, but I think we may have seen this. I'd have to look through.
DR. ACKER: I believe I can tell you what those numbers are. If I'm looking at the right numbers, what FDA has done is to look at the number of available tests at the 12-month visit as the numerator, and the denominator is going to be the number of patients who were enrolled in the trial.
So the fractions of missing data you're looking at there include patients who, for example, died. Yeah, that's correct. This is based on 12-month available data for the entire patient population.
So, for example, any patient who died three months after enrollment is being counted as missing data for the purpose of this FDA analysis. That actually accounts for probably half of the missing data in both the Minnesota Living and SF-36 and a smaller proportion of the exercise data.
CHAIRPERSON MAISEL: Norm, would you like for the FDA to comment on that?
DR. KATO: Yeah, Clarence, would the FDA comment on that, please?
CHAIRPERSON MAISEL: Dina, could you have someone comment, please, on the graph on page 5 of 22 and why the missing data appears to not agree with what was just shown on the slide?
DR. PINA: I just thought that, yes, that is correct. It's any test missing in any of the patients. So if a patient is missing a test at six months or missing at 12 months, that would have counted. So when we say here missing test occurred in 47 percent of the patients, it can be a test at any time of follow-up, whether it's at six months or 12 months.
And there were patients, for example, who had a six-month test but didn't have a 12-month, and there were some patients who had a 12-month but didn't have a six months.
DR. KATO: Well, then does this graph exaggerate what we just saw up on the screen from the sponsor?
DR. PINA: No, I don't believe it exaggerates. I can ask Laura if she has the n's for each of these because this is the percent of missing follow-up data, but there are n's that go with this.
Laura, do you have those numbers? Let me look at my back-up slides here. I may have the answer to those.
DR. THOMPSON: I don't have them at hand, no.
CHAIRPERSON MAISEL: So it sounds like, in short, we're looking at similar but slightly different calculations.
DR. KATO: Well, I mean, obviously this is important because we have to -- you know, one of the major criticisms is this lack of data, and if, on the other hand, the sponsor is saying that 93 percent of the time the data is there, then you know, is the data there or not and are we able to then extract, you know, use this data in order to make our decision today?
DR. SOMBERG: But aren't we talking about different data? I mean, these are not the tests -- this is not the New York Heart Association classification which is the critical --
DR. KATO: No, no. This is the SF-36.
DR. SOMBERG: Yeah, I understand, but what cam out earlier is that the most critical data, you know, that they are fairly complete in some areas, but they are at least half the time incomplete in other areas.
MS. WOOD: A reminder to the panel. Please wait to be recognized by the Chair if you wish to speak.
DR. SOMBERG: I'm sorry.
CHAIRPERSON MAISEL: I retroactively acknowledge Dr. Somberg.
DR. SOMBERG: And I apologize for speaking out of turn.
DR. PINA: Yeah, I think we're referring to the slide that we have, missing tests occurred in X number of patients, 47 percent of the patients. So there are patients who had, say, 15 percent of patients may have had an SF-36 at baseline, but then not at six months, and then another one at 12 months. So it was very difficult to always see the progression with spotty, here and there data.
So I think we're looking at the same thing just from two different viewpoints. I don't think it's that dissimilar.
DR. KATO: Okay.
DR. THOMPSON: I just want to add one point. For some of these secondary endpoints, there may be some small discrepancies in percentages between ours and the sponsor's. We haven't had time to communicate up to the last minute with respect to all of the slides, and the reason being is because we received a data set from the sponsor, but I don't have all of the same programming code that they used, and so it may be that there are some, you know, cases where some of the switches that were used were not exactly the same.
So especially in Dr. Pina's reports I haven't looked at everything to make sure it matches the sponsor's, but that may take care of some of some small discrepancies.
But with regard to the primary endpoint, I believe that we have the same missing percentages in that regard.
CHAIRPERSON MAISEL: And, Norm, just a final point of clarification, the FDA graph has peak VO2 and six-minute walk, which obviously had a lot of missing data, and that was not on the slide that we just looked at from the sponsor.
DR. KATO: Yeah, more of that, right.
One question to Dr. Acker. What happens to the bypass graphs when you do a simultaneous CABG and Acorn procedure. I'm a little concerned that, you know, if all of the surgeons -- and I have to respect to Dr. McCarthy and Dr. Trento who I've heard of the reputations before. If they describe these are the worst adhesions they've ever seen in their whole lives, quote, unquote, one, I have to be concerned about simultaneous CABG with the Acorn procedure, and number two is that obviously not everybody is going to be able to have the skill set to redo procedures on people with this device.
CHAIRPERSON MAISEL: So maybe you can confine your comments to the second question. We're not here to consider a simultaneous CABG and Acorn placement.
DR. KATO: Okay.
CHAIRPERSON MAISEL: So if you want to comment on secondary operations, that's fine.
MS. WOOD: I would also like to remind the panel that we cannot refer to the doctors' names in the op reports. Those are not publicly releasable information.
DR. ACKER: Okay. You know, it's part of what heart surgeons do these says, is redo operations, and you know, this isn't just the heart surgeons from the top centers in the country. These are 28 centers across the country that did face the need to do a redo operation after a CorCap.
I go back to my presentation from the morning. You can Slide 105 on. The operations were done safely. Adverse events were not increased, bleeding was not increased. These are dense adhesions. I mean when I have a BiVAD, I know I'm going to have dense adhesions, and what makes that a much more difficult operation in a VAD is because you know that those canulars are actually in the right atrium and in the great vessels. The key structures that need to be identified earlier on and dissected out safely. With a CorCap you don't have that worry.
There are dense adhesions. We know about it. We know how to handle it. The data suggests that this is done safely and I feel with very good dissemination and mentoring that this device could be disseminated across the country and done safely by cardiac surgeons as a rule across the country.
DR. KATO: One final question to the sponsor. You know, we know that there is some type of remodeling that occurs in the face of both drugs and devices, but the remodeling that occurs is really a secondary event of the heart reacting to either after load or pre-load reduction or volume control with the device.
I'm a little bit unsure of how you get from the fact that you can put a mechanical structure around the hear to prevent to dilate in order to do a basically a primary reduction and have it work in the same way, and I'm still struggling with that in light of the fact that we have had other -- you know, there are other operations out there where we take out parts of the heart, and most of the times it really doesn't work.
DR. MANN: So I guess I don't understand the latter part of the question. The first part of the question I understood. I guess you're having difficulty with the mechanism, how unloading would lead to reverse remodeling or --
DR. KATO: Oh, no, no.
DR. MANN: -- how CorCap leads to unloading?
DR. KATO: I actually understand quite well how unloading with drugs or mechanical means leads to remodeling the heart beats. The heart then shrinks down.
But then the heart is functioning as a secondary -- this is a secondary mechanism, that the heart shrinks because of another primary event. Here on the other hand you're taking the heart and mechanically shrinking it as the primary event, and then hoping that it remodels as a result of that.
I mean, in effect, you could, I guess, in theory take a piece out, you know.
DR. MANN: Yes, but this raises the question of how ACE inhibitors, beta blockers and biventricular pacers work, whether remodeling is the primary effect whereby they're working. We have an article in Circulation this month entitled the biomechanical model, which just begins to look at heart failure as a combination of function and remodeling, but that that's what's really important in heart failure and drives it forward.
So the question is whether it's a secondary or primary event. We would argue it's a primary event and that one of the ways in which biventricular pacing works is by making the ventricle smaller, unloading the ventricle and thereby improving contractile performance. I mean for all of the reasons I articulated in my presentation. Big hearts have high wall stress and makes it harder for the heart to work and activates a variety of stress activated and maladapted genes.
When you unload the heart that whole process reverses, strictly by reversing it, in addition to which, you know, bigger is not better even if you're from the great State of Texas. You know, bigger is not necessarily better that you have these large, remodeled ventricles that are mechanically and energetically inefficient, in addition to be leaky. They sort of develop mitral regurgitation.
So that's the first answer to your question, is I don't view what happens after beta blockers being secondary. I view it as primary. Again, maybe that's philosophical.
Now, in answer to your question, the CorCap doesn't shrink wrap the heart. Okay? It applies a gentle, two millimeters of external pressure, and this has been measured in experimental studies, to the heart. What goes into the calculation of wall stress is the transmural pressure which is the left ventricular end diastolic pressure minus the intrathoracic pressure, which is traditionally about minus two millimeters.
So take an example of a heart failure patient with a left ventricular end diastolic pressure of 20, minus intrathoracic pressure of minus two. The distending pressure there is 22. Okay? That goes into the calculation of wall stress or the amount of work that the ventricle does just to empty each beat.
Okay. Now put on the CorCap, gentle too millimeters of external pressure. So now it's 20 minus/plus two or 18. So the transmural pressure, the distending pressure that goes into the wall stress calculation has gone down just by this gentle external pressure.
This is also, parenthetically, the way that if you give CPAP to people and you increase intrathoracic pressure, their ventricles get smaller for the same reason. You're increasing intrathoracic pressure and decreasing transmural pressure.
I'd be happy to provide you with all of the preclinical data that showed that this applies it. So we're not shrink wrapping the heart. You're basically unloading the heart. It's not quite analogous to an LVAD, but you're unloading it. In response to the unloading there's a reversal of the whole field gene program. There's a reversal of the hypertrophic gene program, and it goes all the way down and then transmits into a reverse remodeling process.
Did I make myself clear? Did I talk too ‑-
DR. KATO: Yeah, that's fine.
CHAIRPERSON MAISEL: George.
DR. NETROVEC: Well, I think it has been a really interesting and well presented series of data. I have a couple of maybe minor points, but if I heard it right this morning, I heard the comment that the echo follow-ups were blindly read, and I find it hard to believe that on the echo you can't tell that these people have had this operation. Am I missing something?
DR. KUBO: Spencer Kubo.
The CorCap is not visible by echo, and so the treatment allocation is completely blinded to the reviewer. The ability of the patient to have a mitral valve surgery is not blinded. So one could tell what stratum they were based on the echo, but it is completely blinded, and Dr. Oh (phonetic) is in the audience who was our echo core lab, who could confirm that.
DR. NETROVEC: The second comment I have is there's something a little peculiar about, at least as I think about it, we want to see patients get better, and so in this trial 33 percent of the patients in the control group improved their New York Heart Association classification.
One has to assume that doesn't happen by chance, but through better medical management, and yet on the other hand if a patient -- that was presumably adjusted because the patient wasn't doing well at some point in time, and on the other hand, if the patient wasn't doing well, quote, unquote, they got a BiV pacemaker, and that counted negatively, and yet at the end of that the patient may have been better.
So if one looks at the worsened group and assume that maybe half of the worsened group in the control population actually ultimately got better because of it, then I'm confused about how the totals are.
Can anyone make a comment about that? It looks like you have two different categories. If you had to have a procedure, that was a fault or a foul, but if you got better medical treatment, that was okay.
DR. KUBO: I'm not sure I'm entirely following your question. Patients could be worsened by one of three ways: death, major cardiac procedure, or the NYHA got worse by at least one NYHA class.
The maximum medical therapy in the control group may have made some of the control patients get better. Compliance in a control trial is almost always better than in a free living patient population, and of course, two-thirds of the patients have got mitral valve surgery, which also presumably made them better.
Just through some facts. I'm not sure I was following exactly.
DR. NETROVEC: I was just having trouble figuring out how a third of the control patients could get better and we just select out a certain group that may have gotten better because of a treatment they got and we call that a foul that they didn't improve.
DR. KUBO: Right. Well, yes, control patients get better. That's obviously an important reason why we have a controlled trial, because we don't want to necessarily just assess the effect of the better follow-up.
DR. NETROVEC: All right. Well, while you're there then, the other question I have has to do with Slides 57 and 58 that were shown, and I guess what I really am having trouble with is figuring out what an SAE is in this because if you look at 57 and then 58, actually the control patients do better.
DR. KUBO: Right.
DR. NETROVEC: And that somehow doesn't fit everything else about the trial.
DR. KUBO: Yes. I could take a start at that.
Slide on, please, on 57.
An SAE, a serious adverse event, is defined as any adverse event that either caused death, caused or prolonged a hospitalization, or resulted in permanent disability. So a pretty standard definition for a serious adverse event.
And it's important, of course, to combine serious adverse events and death because if you had a higher death rate, you would reduce the number of SAEs.
In this trial, in a Kaplan Meier analysis, we look at the number of patients in the control and treatment groups and see no difference. There are a large number of patients who get SAEs because of the effect of surgery, which is something well known to all cardiac surgeons, but there's no incremental risk of adding the CorCap on top of patients with the mitral valve surgery. The safety is very clear here.
Now, the second point though on the next slide, 58, is that there is a difference in the no mitral surgery stratum in which the treatment group has a much higher degree of SAEs, especially early. This is almost certainly related to the effects of the surgery per se, not necessarily the CorCap because this line, this blue line is almost superimposable on the two lines from the preceding curve, indicating that this risk of surgery is well known, well defined, and equal no matter what kind of surgery you're getting, but it's not necessarily the CorCap per se.
Now, the control group does not get surgery. So they do not have that early event of AEs, but you see they're getting a lot of AEs as they go further. We believe that that increase AEs that occur after that initial phase is related to their progression of heart failure and their worsening condition, and that there is this catch-up phenomenon.
Does that --
DR. NETROVEC: Yes, thank you.
CHAIRPERSON MAISEL: Gene.
DR. BLACKSTONE: Five comments, one a little long, but let me start with one that relates to Slide 58 that you just had on.
You say that the difference in adverse events is not related to the CorCap, but it is entirely related to the CorCap because the only thing that these patients had was the CorCap. Therefore, this represents the SAE for isolated CorCap.
Now, second, I am having a lot of trouble with the primary endpoint for several reasons. First of all, you are combining the additional surgery for heart failure for which there is some hope that that will improve the patient; with death, for which there is little hope for improvement so these are not the same kind of events; but both of those are time related point processes.
You're combining that though with the status of the patient, which is not a point process. It is the status of the patient across the time for which you have taken multiple snapshots across time. So it's a longitudinal data as opposed to event data, and now you're taking those three and combining them into an endpoint.
I probably have analyzed more cardiac surgery data than anyone in the room and probably anyone in the world. I would have no idea how to do an analysis that combines this kind of information.
So I have trouble with even thinking about how you would put such disparate things together.
Now, furthermore, there's one other thing that is being done, and that is not only are these equally weighted, but whether you go for a New York Heart Class IV to Heart Class I or from III to II, you get the same weighting for this, and so you're only looking at these changes.
I'm not sure how responsible an n point. You say it's very clinically meaningful, but I have no idea how to interpret that data. So that's just a comment on the primary endpoint. I have no idea what it means.
Okay. Now, the second thing, we've talked a lot about missing data here and there, and so on. I'm not so terribly worried about the missing baseline New York Heart functional class data if, in fact, 85 percent of patients from the core lab are in the same class to begin with. You know, they're all Class IIIs. So let's just say they're on all Class III.
But what bothered me both with the sponsor and with the presentation by Dr. Thompson from the FDA on I think it's their Slide 20, and that is the wastefulness of the data. What you have done is to very expensively get data across time. The coordinators here have followed these patients at two week intervals, at monthly intervals and so on. So there's an enormous amount of data, and yet it seems for most analyses formal methods of longitudinal data analysis with looking at time as a fixed effect and then looking at interactions with time to see if there's a change in the rate across time, I don't see any of that data being used. It just seems like a small amount of data.
There's a couple of slides that say just focus on the data at 12 months and you have all of this sea of data, and we're not seeing all of that information that is put together.
Then for the secondary endpoints as well as some of the primary endpoints, we're only looking at change scores. There's a lot of literature about change scores, and I think we have seen one slide where we find some of these absolute numbers of where people are. It seems to me that the change scores introduce much less efficient type of analysis than could have been. So, again, I'm not sure I'm seeing everything that could have been gleaned from the data.
And finally, a totally different point, and that is in the animal studies that were shown very early here, it looks like the CorCap was being used rather acutely to prevent the kind of dilatation that is now being used for patients after they already have the remodeling.
So it looks like the animal studies and clinical studies are really addressing something rather different than the human study that we have had presented today. Those are the concerns that I have.
CHAIRPERSON MAISEL: Thank you.
At this point -- did you want the sponsor to respond?
DR. BLACKSTONE: I'd be happy for the response. What I don't know is how on earth did the statisticians think about how they were going to analyze this.
DR. KUBO: Thank you very much.
I wrote down or jotted down at least four different points, and I'd like to respond to each one rather quickly if I can.
The first question you have, and I'll have Dr. Jessup come up to address this, is what about the time dependency, and you're absolutely correct. We have a combination of an endpoint that includes events that occur from any time from enrollment to the common closing date, as well as a part of the component, the NYHA which was done at specific time points, and that was meant to try to capture as many of the adverse events like deaths and major cardiac procedures and extend the time as possible.
We have also performed the analysis though with the primary endpoint at 12 months, and I'll have Dr. Jessup describe that on this slide on, please.
DR. JESSUP: I agree that it's a challenge to assess what happens to heart failure patients. They're a mixed group of people with advanced disease, and many things go wrong with them. Some of them are related to their heart failure and some of them are related to their co-morbid conditions.
And so as you well know, in many heart failure trials it has been considered important to try and get an overall view of what happens to that kind of patient, especially if you don't have a huge patient population, a huge power to do a mortality study alone.
And so as we've said, our primary endpoint was trying to encompass all the things that can go wrong, and you're absolutely right. Some of them are very finite and some of them can happen across a time course. Nevertheless, this composite endpoint has been useful in many, many heart failure trials, not necessarily device trials, but certainly it has been a way that has been commonly accepted in other drug therapies.
Let me just show you what happens to the primary endpoint at 12 months as opposed to the primary endpoint at a common closing date to try and emphasize that although the individual numbers may change and the individual patients, depending on how long they've been followed in the common closing date, the magnitude of change is still the same with respect to the odds ratio, and it's still a significant value.
So the graph shows the primary endpoint at 12 months, where almost all of the patients were followed, and the primary endpoint in the common closing date where we had a variety of follow-up.
If I could have Slide 33 on, please.
This is another way of looking at the primary endpoint at various time points, again, trying to address one of your concerns, and although obviously the confidence intervals change somewhat, the relative odds ratio continues to be the same across 12 months, 18 months, and 24 months.
Our understanding about the heart failure syndrome, as you saw from Dr. Mann's presentation in the beginning, is that this is a progressive disorder, and so even if we had done nothing, we would expect that a variety of assessments of the patient would change.
And finally, I just want to make one point as far as the control population. There have been a variety of very large heart failure trials that have been confounded by control improvement, the most recent of which was VAL HEFT, and despite stable medical therapy, the control arm continued to improve and show reverse remodeling as well. So hopefully that addresses some of your concerns.
DR. KUBO: Thank you, Dr. Jessup.
Another point that you raised was sort of the wastefulness of the data, and I think that that does represent the differences that we have with the agency in terms of analyzing by ANOVA (phonetic), which is at a fixed time point, like 12 months, or longitudinal regression, which will take advantage of all the data collected over time. I'd like Dr. Brown to address that issue.
DR. BROWN: Slide on, please.
This is Scott Brown.
The point was made briefly about the use of data at specific time points as opposed to, as it was pointed out, data is being collected all throughout this trial at the various study centers. For precisely this reason, the secondary endpoints that you see that we presented this morning and that you see on the board right now are, in fact, longitudinal analyses making use of all available data throughout the various time points in the trial.
This does contrast the agency's presentation by and large focused on 12-month change scores, but the data you see here is the more powerful longitudinal analysis making use of all available data.
Dr. Blackstone, does that help?
DR. BLACKSTONE: Let me ask you one detail about that. Specifically, did you look at the interaction of time and group? Because my bet is that your treatment difference is looking at an intercept difference.
DR. KUBO: The treatment difference in this analysis is going to be the time points were looked at as a factor variable as opposed to a continuous variable. So rather than an intercept it's an average treatment difference across the various time points.
CHAIRPERSON MAISEL: Did you have any additional unresolved problems? Gene, are you all set?
DR. BLACKSTONE: Yes.
CHAIRPERSON MAISEL: Okay. Thank you very much.
DR. NAFTEL: Can I make one comment?
MS. WOOD: Please identify yourself.
DR. NAFTEL: My name is David Naftel. I'm a Professor of Biostatistics and Surgery at the University of Alabama at Birmingham.
Certainly I appreciated Dr. Blackstone's remarks and I think they're very appropriate. A composite endpoint is always so difficult, and it's really a result of a committee of clinicians and statisticians trying to come up with a single number that measures how well a patient is doing, and in the end, I think often the clinicians are not quite happy with it, and the statisticians aren't. But I believe it does work to come up with a single number that says how does a patient do.
Now, certainly the time relatedness is absolutely critical. In this study everyone had to have at least one year of follow-up. So that helps a little bit, but certainly when we looked at the components, when we can really get into where we're comfortable and statistically and even clinically all of those analyses of time to death, those are all time related. There's a time connected to when you die. The adverse events, the major cardiac procedures, all of those do have time tied to them, and those analyses I think were probably done the way most statisticians would do them.
So I just want to point out this is always a problem, and what you're seeing is a combination, a joint effort from the clinical and the statistical side.
CHAIRPERSON MAISEL: Thank you.
At this point I'd like to invite Dr. Page to make his comments. I'll preface his remarks by saying that I would --
DR. THOMPSON: Pardon me. Laura Thompson.
Just in response to the previous question, the sponsor did prespecify to do longitudinal analyses as well as the ANOVA at six months and at 12 months for most of the secondary endpoints, and for some of the components of the primary endpoint, in particular, NYHA.
And I believe they also presented interaction with p values as well. I don't quite remember what they were off the top of my head, but both of those types of analyses were prespecified by the sponsor.
CHAIRPERSON MAISEL: Thank you.
I would advise you to listen very carefully to Dr. Page not only because he's always very insightful, but because he has laryngitis.
DR. PAGE: Excuse me. Can you hear me?
DR. PAGE: Does it work like this?
CHAIRPERSON MAISEL: Yeah, that works.
DR. PAGE: Okay. Now that I've gotten your attention, my experience in surgery has been in pacemakers, and there's a device called "Personnet Pouch," which is a device that's really designed to create adhesions around a pacemaker, and I hear what Dr. Acker is saying about -- I guess he said it was not an issue about the adhesions, but having read the operative reports that the audience isn't privy to, and I won't name names nor institutions, but these are really unusual operative modes, like good people commenting on the difficulty. So whether or not Dr. Acker would be concerned about doing a re-op, I would be concerned about sending my patient to a surgeon to do a re-op where it may take an extra two hours to do the procedure. So I still consider that an unresolved issue.
But I do have a question for Dr. Acker, and that is occasionally a patient will require an epicardial pacing lead for biventricular pacing. I understand that a CABG would not be practical, but could you place an epicardial pacing lead after the CorCap has been in place for a year or two?
CHAIRPERSON MAISEL: So just so I can repeat the question in case the audience didn't hear, could an epicardial pacing lead for biventricular pacing be placed after the CorCap is placed.
DR. ACKER: I don't believe it was needed to be done, but there would be theoretical reason or surgical reasons why it couldn't be done through a small thoracotomy and place an epicardial lead in that fashion.
DR. PAGE: And would you take the set through the device to then get onto the epicardium to place basically and would that disrupt the effect of the CorCap?
DR. ACKER: No, it would not. You know, again, you would make a small incision in the pericardium, dissect down to the device, to the jacket, and then place the lead through that.
As far as your first comment goes, it's not that I'm not concerned about doing reoperations. It's just something that we are quite -- we do all the time and we're quite good at it,a nd you know, a third of our patients, for instance, at the University of Pennsylvania had 50 transplants. Twenty to 25 a year are patients with VADs,a nd I can tell you it takes an extra one to two hours to do those operations.
DR. PAGE: I'm well aware of the VAD issue, but having looked at the op notes for, say, a repeat mitral valve surgery five years after CorCap valve, I'm convinced based on those notes that there is a higher level of operative technique required to perform that surgery. So as a cardiologist I've worked that into my own thought process if I were sending a patient to surgery.
DR. ACKER: One of those op notes are mine. So I am familiar with the reoperation after a CorCap, and there is certainly a -- it is harder than doing a patient who had five years ago a simple mitral valve.
DR. PAGE: Exactly.
DR. ACKER: Yeah, but not as hard as other operations.
DR. PAGE: I'll stop there.
DR. KUBO: If I could just make --
CHAIRPERSON MAISEL: Please.
DR. KUBO: -- one comment about the repeat mitral valve surgeries because that would be another potentially important issue. That is, we do have experience with three mitral valve surgeries in this trial, as well as a couple of patients who have required repeat mitral valve surgeries from our safety studies in Europe, and those have all been done very successfully we think probably because of the approach through the atria, not through the ventricals.
We think the biggest reoperation risk is with transplant, and that's why we've showed you data on transplant. The issues about bypass surgery are also discussed, but the other operations seem to be done very safely.
CHAIRPERSON MAISEL: Thank you.
DR. WHITE: Thank you, Mr. Chairman.
I've been paying attention, and I don't have any unresolved issues or questions to ask that haven't already been asked. So I will pass.
CHAIRPERSON MAISEL: Thank you.
DR. NORMAND: Thank you.
Not surprisingly, I'm not going to beat this one to death, but I want to talk about the missing data again, and I guess I want to raise the issue because of what Dr. Blackstone said because I think I'm a little confused or perhaps other people are as well.
It seems to me that, Dr. Blackstone, I think you had said something about 85 percent didn't change. They worsened or something. You weren't worried about the missing data.
DR. BLACKSTONE: The initial NYHA, 85 percent were in a single class.
DR. NORMAND: Okay. I guess the point I wanted to make is we don't know if 85 percent were in the same class because I think, unless my calculations are wrong, almost 60 percent were missing baseline data. So that can't make sense.
Now, I realize you had to impute the data. In fact, I'm a big fan of imputation. I know there's at least one other panel member that indicated that he was, but my concern -- I just want to state it formally -- relates to the following. With that much missing data, you're really in a non-ignorable missing data set. So anything you say about that model, your inferences really depend on the model they used to impute the data. That's a fact. That's not disputable. That's a fact.
And unfortunately in our panel packets we weren't given any data about the imputation model. The first time I saw it was on the screen when someone listed some of the variables in the model, but that clearly when you have missing data to that extent, one really does want to see the imputation because, in fact, you're in a non-ignorable setting, and what does that mean? Why am I using these terms?
That means that the plausibility of the statistical assumptions relate to the richness of the data that are used in the model, which we didn't see, and moreover, it also relates to the complexity of the model that you're imputing, and those are two key concepts for anybody who deals with missing data that you need to see, and we didn't see it.
With that said, my understanding is that based on your primary endpoint, that indeed you didn't have to use the imputed data because of this algorithm you had for assigning patients in terms of improved or didn't improved, but I couldn't figure out the math to figure out, well, okay, how many did you actually have to use the data for, and one doesn't like to have to try and figure that out.
So, again, I am concerned about the missing data, but I do want to admit, I guess, my understanding is that the imputed data, all of the imputed data wasn't necessarily used in the primary endpoint, but that was completely unclear, I think.
So I just will state that about the missing data and that in the future would be nice to have if you're going to input the missing data. We need it. Not everybody will read it, but surely I will.
And then I guess I do have a question about the missing data mechanism given you do have so much missing data. Typically statisticians would use tighter mixture models, and I'm wondering what drove your decision to use a model that I didn't see, but apparently it had a lot of variables in it.
Did you use a pattern mixture model? What kind of model did you use for your missing data I guess is the --
DR. BROWN: I'm sorry. This isn't the slide I want. Let me address the first question. Yeah, let's go with this. Slide on, please.
I think what's being asked for probably is a little bit more detailed on the imputation. It is a big deal here. We contend that the amount of missing data is not as great as is being perceived due to the fact that it is restricted to one corner of one of the three components of the primary endpoint, and in fact, that it's the more benign corner. It is the baseline data which one expects to be balanced just due to randomization.
Having said that, there are a lot of data missing in that particular corner of the primary. So a little bit of discussion of what exactly was done in the imputation is appropriate.
One thing we did do, and Dr. Thompson made reference to this in her presentation, is that when you have a statistical model with a lot of importance, it behooves you to test the assumptions and the basis for your modeling.
So that's true of any statistical method. One test of reliability, as we mentioned on this slide, is to alter the imputation model and examine the impact on the statistical inference that you've drawn, and for this particular trial, we built three different imputation models, the first of them submitted in the PMA and the results of which are those you have in front of you at various points.
It did use a multi-variate normal assumption with rounding, as Dr. Thompson mentioned, and a set of predictor variables which were selected in a step-wise fashion from a list of baseline covariates specified by the steering committee, and if Dr. Normand is interested, I can provide a list of what those variables were.
There were additionally two more models built to try to make sure that we weren't making specific assumptions about the data that were driving the endpoint here, and this is the important thing. We don't want to have our results specifically tied to one and only one side of assumptions in a model.
So the two validation models, and these were built in discussion with FDA, both used an ordinal data assumption, which is quite different from the multi-variate normal, both using a broader set of predictor variables, each slightly different from the other, but a broader set of predictor variables that did include some follow-up information as well.
And the point that we want to make about all of this is that when you look at these three models, you get very similar results in terms of the statistical inference with regard to the primary. The three p values are .024, .029, and .033, respectively.
So this gives us some good feeling that at least the specific form of the imputation is now what's driving significance in the primary.
Now, having said that, we can also get to the same results in this trial without using imputation at all, and this was made reference in the main presentation, and let me see here. Could I get main Slide 111 please? No, I'm sorry. One, twelve, 112. Yes, slide on, please.
This is the analysis that was presented at the tail end of this morning's presentation by the sponsor looking at the status of the patient at the end of the trial only. This, I would mention, it notes on the slide here, but this was an a priori analysis. This was specified in the protocol as a confirmatory analysis for the primary for just this purpose, that if we run into problems, being able to think about the imputation. Let's have something on hand that gets you to the same place without having to worry about statistical modeling to the same degree.
So what this does is it takes the various pieces of the primary endpoint and reorders them in a fashion which is clinically reasonable, and then tests the question does the treatment group perform better on a whole than the control group.
And so what we've done is we've ordered the four NYHA classes at the final visit, stacked major cardiac procedures indicative of worsening heart failure in with the Class IV patients, and then categorized death as the worst possible outcome.
And you can see here that this recombination of the pieces of the primary endpoint gives you a significant result, p value of .042, without using any imputation whatsoever.
I should mention as an aside the particular choice of putting major cardiac procedure with the Class IVs instead of categorizing it differently, that was made on a clinical basis. It could be discussed further by the clinicians. It doesn't actually matter with the conclusions if you ordered it a little differently.
If you put major cardiac procedure below Class IV, for instance, you get a slightly different p value, but the same significant result. So it's not dependent on that particular assumption.
DR. NORMAND: And the question about the pattern mixture models?
DR. BROWN: Right.
DR. NORMAND: Again, the issue is you could show lots of imputation models, but the fact is you can't test for nonignorability, as you know.
DR. BROWN: Right.
DR. NORMAND: So I guess just quickly, did you try pattern mixture models which were appropriate when you have this degree of missing data?
DR. BROWN: That was not attempted.
DR. NORMAND: Okay. Thank you.
DR. KUBO: If I may, Spencer Kubo.
Just to provide one other commentary on this, the issues about imputing data and a large amount of data is important, and we've consulted with many different experts in the field, including Dr. Brown, Dr. Naftel, Dr. Kinley Lawrence, in an effort to make sure we would do this the right way in an appropriate way that would be acceptable to all audiences, and realizing that there may be differences with certain techniques and certain opinions, but that this was done according to the way that would be appropriate and acceptable by many experts.
Having said that, for the clinicians among us though, we want to say is this just some sort of statistical test that proves a point. How can we be sure that we're actually seeing the clinical benefit that is meaningful to our patients?
And that's why we did the analysis using the data without any imputation, and the first Slide 115, please, is -- slide on please.
This is in the subset of patients of 121 patients who were randomized, enrolled in the trial after we implemented the core lab so that they'd have the exact same technique at before and the exact same technique after. There's no imputation at all.
And with this we find the exact same proportional odds ratio of 1.75. Now, we've lost a lot of power because we're dealing with 121 of 300 patients, but I submit to you had we extended this at the same pattern extended to the 300 patients, we'd have the same or the significant p value.
The other way of doing this analysis without imputation as Dr. Brown just mentioned in the next slide, please, is to look at the follow-up data at the end of the trial without -- again, it's assuming everybody is starting at the same baseline, and if you do this, you have the exact same pattern or results. The results of our trial are not dependent on imputation. The use of imputation allows us to have a more comprehensive assessment of all the patients in the trial and using as much of the data as possible, but it is not dependent on the imputation itself.
DR. NORMAND: Yes. Thank you very much for that.
I have just two more types of questions. The first one is very specific. I think Dr. Acker said there were 28 centers involved in the trial. Were patients randomized within center or how did you -- just a quick yes, no or --
DR. BROWN: The quick answer is yes. Randomized, promoted blocks by center.
DR. NORMAND: Within center. Okay.
DR. BROWN: That's good.
DR. NORMAND: I just have a few questions about the post market study, and because it was nice to see it in there, and I wanted to ask a few questions about it. And one related to I think some of your inclusion criteria for the continuous access protocol is different, and can you explain why the inclusion criteria is different and justified that.
So maybe they're clinically not meaningful changes, but you've got no hepatic function requirement, no pulmonary function requirement, et cetera.
DR. BROWN: The basic contents of our continued address protocol and our post market surveillance protocol in terms of patient selection will be quite similar, and in fact, we designed it to be similar.
Some of the requirements which we did drop though were requirements for some of the exercise testing, like the six-minute walk test or rather the peak VO2 and so forth because those we are not going to be following the peak VO2 test as an example on long term.
But by and large, the inclusion and exclusion of arterial will be quite similar.
DR. NORMAND: Well, I'd highlight it in green and maybe a clinician could look at it and tell me at some point whether it would make a difference or not because when I read it, because I'm compulsive, I looked at both columns, and they are different. So I just want to understand why you would do that, and maybe it's not clinically meaningful.
DR. KUBO: When we come to the questions about the post market surveillance design we can discuss that further
DR. NORMAND: Okay, and then I had another -- just two more questions related to that, and that is now did I read you're going to use the Kansas City? Did I make that up? Are you measuring using the Kansas city -- no? So I made it up. Sorry. Because I thought I saw that somewhere.
And I guess the question I have with regard to the last thing to do with the post market study is the lack of a control group, and I think you could probably easily use some matching using some administrative data somehow. I mean, I think that would be fairly straightforward to do.
But I feel that not having a control group is a big wasted opportunity.
CHAIRPERSON MAISEL: Thank you, Sharon.
DR. FERGUSON: Well, first I want to echo Dr. Yancy's comments. I've seldom heard a more lucid, clearly displayed presentation at these meetings,a nd I thank you for that.
My questions are related to the extent to which CorCap will be used in other situations. Being a simple surgeon, it would have been great for me to see a study where you looked at the medical treatment of these patients versus CorCap, period, because what I see here is a device which fulfills a need that is not there in the market, a very good need, and for several reasons I would not like to see it diluted at least at the beginning.
and to me, as I go through the protocol, that's what the mitral valve inclusion has done. You've excluded coronary bypass by necessity, and I can understand that very well from being where Dr. Acker has been, but I want some assurance, I guess, that in the use of the device it will not include statements such as this device can be used in any patient who needs cardiac remodeling, which is different from a patient who is in heart failure.
Because I think that dilutes the value of it very much so. I still have a hard time, but we won't dwell on that, but I still have a hard time putting mitral valve surgery -- and I understand the permutations that go with all kinds of heart failure and what things are going to happen. It's not that, but my reading of the material, and I'd like to be assured that this is correct; my reading of the material and the inclusion criteria are that thee were patients who had heart failure from some cause, and that cause turned out to be idiopathic for reasons that you've already mentioned.
But it also excluded patients at the time of entrance who had mitral valve disease. Is that wrong or right? That's wrong. Is that wrong?
DR. KUBO: Yes, right.
DR. FERGUSON: Yes, I'm wrong.
DR. FERGUSON: Okay. So you included from the beginning the patients who had mitral valve disease with the assumption that the mitral valve disease is the cause of their problem, and I have a little problem with that.
DR. KUBO: The inclusion criteria for the mitral valve arm was a mixture of patients who had anatomically normal mitral valves, but incompetent because of the ventricular dilatation process.
In addition, there were a few patients who had structural abnormalities with the mitral valve.
DR. FERGUSON: Well, that's the problem I have right there, of course, because I can understand the first group of patients where you have mitral valve, which is normal but a dilated annulus and, therefore, have MR.
The same question is applied, and see if I'm wrong here again, that patients who you thought were on the horizon are going to need transplant were excluded from this study; is that right or wrong?
DR. KUBO: That's correct.
DR. FERGUSON: That's correct. Okay, because when a patient needs, as Dr. Acker put it very well, when a patient needs a cardiac transplant, he needs a cardiac transplant. No question about that.
So to summarize, that's all I have to ask you. The summary that I have is that I would like to see the labeling -- and we'll talk about this in the panel later -- but the labeling should be fairly specific because I don't think that it does you a service or the public a service to put a device out there that you say, "Well, my patient has mitral valve regurgitation. Just put this on at the end of the operation and this will be helpful in addition."
That's all I have.
DR. ACKER: May I make one response to Dr. Ferguson's comments?
DR. FERGUSON: Sure.
DR. ACKER: The patients that all had one thing in common, and they all had heart failure, Class III or Class IV heart failure of ischemic or nonischemic etiologies.
The other thing they had in common is a depressed EF and a severely dilated ventricle. Some of those patients had valvular disease as the primary problem, but with, let's say, an EF of 20 percent, a situation that many surgeons would not address or maybe cardiologists would not refer them for anything. So it was our feeling it was reasonable then to include those patients, randomize them all to getting mitral valve surgery, but the additional treatment group has been CorCapped to see if we can further impact on what is a difficult situation with a decreased natural history.
DR. FERGUSON: Could I respond to that just briefly?
CHAIRPERSON MAISEL: Please.
DR. FERGUSON: Yes, I agree 100 percent with that, and we all understand in this room, but what I'm talking about is when the device gets out into the marketplace you don't want to see it misapplied, and it could be misapplied very easily in my view when it gets out there and where people get it in their idea that they can just take a patient who's in minimal heart failure with mitral valve disease, a large heart, and say, "Well, we'll put this on in addition."
DR. ACKER: I couldn't agree with you more. Specifically for heart failure patients.
DR. FERGUSON: All right.
CHAIRPERSON MAISEL: thank you.
DR. CALIFF: Well, I'll try to be brief. You said five minutes. I'll try to see if I can come out under that. There's a lot of --
CHAIRPERSON MAISEL: A Washington, D.C. five minutes.
DR. CALIFF: You notice I've got on the outfit for a filibuster here, but I'll --
DR. CALIFF: -- I'll avoid that.
I'm going to stick to the primary outcomes and just try to make sure I've got down what you've thinking was and how you dealt with modifications that were made in trying to get this study done. This is a hard study. I'm getting the sense there were many discussions and debates about the trial design before it went on. I guess a couple of my questions are probably procedural, and a couple are more clinical trials questions.
I guess the first one I'll pick is this issue of bias and measurement of functional status, and there's one procedural question and one clinical trials question.
The procedural one is I'm really confused by this extraordinary delay in resolving the issue of independent adjudication, and Dr. Pina said there was a prolonged negotiation that went out. I would have taken it for granted that in any study that can't be blinded that you would need to have an independent adjudication of a primary endpoint.
I mean if I was teaching Clinical Trials 101 in our K-30 program and you're talking about a protocol without doing that, I would send it back and say, you know, "Go read the textbook and come back and let's discuss it again."
But obviously you had some very experienced investigators and they must not have believed that this was necessary.
My clinical trials question, I'd sort of like to hear from the FDA over why this is allowed to happen and there's not a guidance on this already, but the clinical trials question is it's intriguing that if you assume that the investigators would have been biased in favor of the device, at least as I look at page 8 of the FDA report, it looks like more people, not significantly more, but it sort of trends against the device if you just go with the investigator reading of New York Heart Association class.
So I don't know what to make of New York Heart Association class altogether with these two things. Maybe to boil it down to a simple question: did the investigators really believe when they started the study that it's good to do device trials without having an independent, blinded adjudication of a subjective symptom?
DR. KUBO: That's a very important question, and if I could provide some historical context for this, when the trial is first designed, the primary endpoint was an LV dimension endpoint, which we felt was dealing with a fundamental intricacy feature of the device, which was to cause reverse remodeling and make the ventricle smaller.
The challenge that we got in return for that was that the agency was interested in endpoints that the patient can see and feel, that the patient wouldn't know that their ventricle was smaller, that the patient wouldn't know that their ejection fraction was larger or smaller. And so we only had to deal with endpoints that the patient could see and feel.
Now, we had an additional problem of having a trial that could not be blinded because in the no MVR stratum, you had patients who were undergoing surgery in the treatment group and patients in the control group who were not.
So we have a situation which I think will come up more and more often in devices in that you have an inability to blind fully like we normally like to do, but you're dealing with or an insistence on endpoints that the patient can see and feel. So you're going to have that conflict at all times, and we sort of dealt with that.
We had the issue then of finding an endpoint that the patient could see and feel and was clinically relevant. We thought that mortality was an important endpoint, but I should mention that this trial was not designed as a mortality endpoint, and we were never going to find a mortality endpoint.
In point of fact, given our patient trial of 300 with our baseline mortality rate, if you could put this slide on because I think this point has come up several times, for us to see a mortality benefit in this trial, given a 300 --
DR. CALIFF: Wait a minute, wait a minute. Now you're filibustering. We all know this.
DR. KUBO: Sorry.
DR. CALIFF: I think everybody on the panel is familiar with power.
DR. KUBO: Okay. I'll get back on task. Thank you for reminding me to get back on task.
DR. CALIFF: You're using my five minutes. That's a well known tactic.
DR. KUBO: I'll refund you that five minutes.
DR. CALIFF: So if I've got it right, the investigators didn't believe it was the right thing to do, but they did it anyway until halfway into the trial.
DR. KUBO: Right. We had -- clearly, the investigators wanted information that was going to be as bias-proof as possible in way that could have the most robust method of determination and would have preferred the volume change or the dimension change, but knowing that we needed to deal with an acceptance and congruence of interest between the agency and the sponsor, we needed to come to a compromise, which we did.
DR. CALIFF: And last question on this issue. There was a complete firewall between the company and the investigators and the database because as I'm looking at the investigators' call, it wouldn't have looked good for you. You wouldn't have ever made it on the primary endpoint based on what the doctors were saying.
DR. KUBO: That's correct.
DR. CALIFF: So there was absolutely no contamination.
DR. KUBO: That's correct.
DR. CALIFF: Then my next question, and I've only got two more, obviously, the -- well, I guess my statement about this would be that the trial rides on its primary endpoint. Everything else that you've shown, I've been there before. You can pick things that look good and ignore things that look back. You don't like to talk about the BNP. I understand that, and I'm sure you believe what you're doing, but to me you design an experiment, you signify the primary endpoint, and it rises or falls based on that and the believability of the primary endpoint.
So we've got this outcome of avoiding major cardiac procedures, and I accept that you got a committee. That was a good tactic to use, but fundamentally, the question that I think Rick is addressing is if everything else about two patients was exactly the same and one had an Acorn device in nd the other didn't, is there an equal chance that a surgeon or a cardiologist would send the patient for this second procedure?
And if I'm understanding what you're saying, you really believe there's an equal chance, that there would be no bias. Because you've only really dealt with the people that had the procedures, not the ones that would have met the same criteria that weren't referred for the procedure.
DR. MANN: I won't filibuster and I won't borrow your jacket. This is Doug Mann.
I think for elective procedures, Rick's concerns would weigh into my decision making. If I have no other choice, if I have lifesaving therapy for a patient and I know it's more difficult, I still recommend lifesaving surgery.
So in terms of a VAD or a transplant where you thought that was the only option to keep the patient alive, then the degree of difficulty wouldn't enter into my decision making.
I can't speak for the other investigators, but I can speak for myself as a clinician, and I think if you put yourself in the clinician's shoes, you also would opt for a lifesaving surgery, albeit with more difficulty, and you have to inform the patient of that risk, but you still go ahead and do it, and it really doesn't influence your final decision.
DR. ACKER: Can I follow up on that? Can I have Slide 129 on?
I think this speaks to whether or not the patients were really sick at the time of transplant, whether things are being withheld, whether perhaps the control patients were going to transplant early, and what this shows is that the vast majority of patients at the time of transplant were on 1(a) or 1(b). They were really sick. They were on inotropes.
DR. CALIFF: Right, but pardon me. I'm not arguing about the people that had the procedure. I'm questioning whether people who didn't have the procedure who otherwise would have met criteria, and it's hard. I think we're left with a subjective impression here. I'm just bringing this out as something for people to think about.
DR. ACKER: I guess you have to go back to the interdependence of the primary endpoint with New York Heart Association class getting worse or dying if the transplant was being withheld.
DR. CALIFF: So my final --
CHAIRPERSON MAISEL: So, Dina, did you want to comment?
DR. FLEISCHER: Dr. Califf, I just wanted to bring up two points. One is I wanted to sort of also echo what Dr. Kubo had said, but, you know, some of the discussions that were ongoing with the company and with FDA, one being that what was going to be the primary endpoint for this trial, and what the company did propose in the beginning was sort of a change in the left ventricular end diastolic volume as sort of the primary endpoint for this trial, and that was one of the discussions that we had, and one was what were going to be something that were clinically relevant for this patient population, and that was sort of what was the lengthy discussions that we had is what would be the composite primary effectiveness endpoint for this patient population, and that was, as we know, what we ended up with in this trial.
And because of the New York Heart Association class core lab assessment was decided upon after 58 percent of the patients had already been enrolled in the trial. So that is really how that had happened when we saw we had our final agreement. So that's how that had come about.
So I know that answers that question, while the study was ongoing.
The other question is that there are some deficiencies that are still outstanding that we had brought up to the company, and one was that we had that same concern, and perhaps that there could be -- I mean it may not be feasible, but we did bring up the concern that there could be an independent assessment perhaps of all the patients that were enrolled in the trial and sort of looking at those events independently and see if there was sort of an independent assessment, exactly what you're talking about. So we did ask that question.
CHAIRPERSON MAISEL: Thank you.
Would the sponsor like to respond to the FDA comments?
Okay. Dr. Califf.
DR. CALIFF: Just really, I guess, to express two opinions and see if there's disagreement about it. I mean, first of all, I commend the FDA actually for trying to get outcomes instead of biological measurements.
And one of the big reasons is because of this missing data on the other end, for which there's no way to compensate when you look at things like diastolic dimensions, if the patients did or didn't come back. You have no way of knowing what the answer would have been, and there is no statistical way to correct for that that I know of.
So as opposed to Dr. Blackstone, actually I like this endpoint if good methodology was used to collect it because every patient at least is accounted for in some way.
And secondly, I just got out of control with Dr. Yancy because I do not think it's appropriate to do within group comparisons when the between group comparison is neither statistically significant, nor is there an interaction between treatment and the outcome because all you're really saying is that one group sort of made it below .05 and the other didn't, but if you ask the question is the response of the two groups different, the answer would be it's not.
So that slide actually didn't help me at all except to just show that, you know, if you look at things enough ways, you can often find p values less than .05.
So I'm actually happy with the endpoint. My concern is that when you add up the uncertainty due to the imputation and the uncertainty due to the fundamental issues in the New York Heart Association class and the second surgeries, it's just a question of everything trends in the right direction for the device. The question is: how much uncertainty does all of that add up to relative to a primary endpoint of just slightly less than .05 to begin with.
I guess that's what we'll discuss at the end.
CHAIRPERSON MAISEL: Dr. Borer.
DR. BORER: I'm going to withhold any comments until I have to defend myself during the voting because I tend to be a little more optimistic than some of the things I've heard, but I do have four questions that I need to have clarified. They're simple.
First of all, with regard to the concerns about pericarditis and constriction, if I understood the data correctly, patients were followed for as long as three years in the large trial, and in addition, you have observational data from Charite in Berlin that among people who were operated on even longer ago than that, even though perhaps they haven't been followed as rigorously over time.
So, number one, do you have albeit in a small number of patients, and you can tell us how many were followed at least two and a half or three years, has anyone developed constrictive pericarditis?
Okay, and how many patients is that between those two studies, approximately?
DR. KUBO: You're asking a question about how many patients have been followed after three and a half years. There are over 330 CorCap implants performed worldwide since 1999. Exactly how many patients have gotten beyond that three-year follow-up point I'd have to get for you. It's about 12 in the Charite safety studies. It's I don't know how many in the U.S. trial.
DR. BORER: Okay. I mean, it might be fair to say it's something north of 20?
DR. KUBO: Yes.
DR. BORER: Yeah, and we haven't had a problem. That's question number one.
Number two, perhaps less important, again, among the approximately 30 patients who had coronary disease and were included in this trial, 15 of them approximately had to have a CorCap. Did any of those patients require subsequent coronary procedures?
No. Okay. Next, about eight patients were refused surgery. Do we know anything about what happened to them? They're randomized in one way, but they actually were treated in another. Do we know what their outcome was?
DR. BROWN: Right. So the question is this trial was conducted under intent to treat principles, and as a surgical trial there was a small group of patients who were randomized to the CorCap either with or without MVR surgery concomitantly and did not get it.
I should note that the reverse did not happen. There were no patients randomized to control who crossed over to the CorCap device. So we have a group of nine patients.
Slide on, please.
Thee are the nine patients randomized to the CorCap who did not receive it surgically for one reason or another, and you can see three of them simply opted out of the trial before that. Four of them refused the surgery at the time it was due, and then two of them died prior to their scheduled surgical implant date.
And the question is: how does this impact the results?
So the first thing I would note, of course, is all of these patients were followed. Even those who withdrew were followed to the best of our ability, and we do have information on the clinical course of these nine patients.
Three of them have died as of the extended follow-up date of April 15th, 2005, and one additional patient from that group received a biventricular pacemaker which was adjudicated by the committee as worsening heart failure. So that is out of the nine patients.
Again, all of those results do count against the CorCap in terms of the intent to treat analysis.
Slide on, please.
This would represent the protocol analysis of the primary endpoint, excluding those patients who did not receive their assigned treatment, and as you can see, the results are very comparable to the primary endpoint we've seen previously.
DR. BORER: thank you. That's what I need to know.
CHAIRPERSON MAISEL: Tom.
DR. VASSILIADES: I have just a couple of questions. One is we haven't heard any information, either preclinical or clinical, on the effect the CorCap device has on the right ventricle. Do you have any animal studies or clinical electrocardiographic studies to demonstrate what the effects are on the right ventricle?
That's question number one. Number two is you mentioned histopathology on the device in animals. Do we have some histology information in the patient that underwent transplantation after getting the CorCap in terms of was there any degree of invasion within the myocardium, et cetera?
Then in some of the information that we received, it says that bench testing reveals that the device applies a pressure of five. Clinically what is the pressure?
You mentioned two a little while ago, and I'm just curious as to what the actual number really is.
The rest are comments that I'll save for later.
DR. KUBO: Three very important topics which we have not talked about. The first is the effect on the right ventricle. This is an important device that's different from other interventions because this one involves both the right and the left ventricle combined.
The data that Dr. Sabbah has done in terms of right ventricular performance, as well as following the echocardiographic, demonstrates a similar effect in reducing the RV size, as well as the left ventricular size, that being much more difficult to measure through echocardiography, of course, given the shape of the right ventricle.
DR. VASSILIADES: Do you have pressure volume loops for the right ventricle in animals?
DR. KUBO: Dr. Sabbah is here who has done other analyses with Dr. Kass. That is involving the left ventricular pressure volume loops.
We have pressure volume loops from patients from the safety study. Those are also in the panel pack demonstrating a leftward shift of the end systolic pressure volume relationship, but all on the left ventricle side.
DR. VASSILIADES: You had a question about histology. The histology has been very carefully examined in all of the animal models as one demonstration slide showing no invasion of the fibrotic response.
Slide on, please.
Just to refresh people's memory, no invasion of the fibrotic response into the epicardium, being a clear plane, and the absence notable of any inflammatory cells so that the process of fibrotic reaction at least in the animal model is present by three months and appears to be stable at a very short period of time.
The amount of histology that we have in the patients who have been transplanted is limited and sporadic. So it is not all patients had evidence or had a -- we don't have data on all of the histologic examinations, but the ones that we do have did not show any invasive properties of the CorCap into the heart.
The last comment is regarding the amount of pressure applied by the CorCap. This has been measured by a number of experimental methods, including the Tibert pillow (phonetic) and other message, and it's somewhere between two and five millimeters of mercury, correct.
CHAIRPERSON MAISEL: Tom, did you have anything else?
I just had a couple of quick questions. I didn't see it anywhere. Maybe you provided it and I missed it, but can you comment on how many patients were screened to be included in this study and how many refused randomization that were eligible for this study just so we can get a sense of generalizability?
DR. KUBO: We don't have those numbers, but we do have certain centers that are doing logs of potential in patients, but this was not kept consistently throughout all the centers.
CHAIRPERSON MAISEL: I also wondered if you could clarify the number of procedures that occurred that were adjudicated as not due to heart failure.
DR. KUBO: Yes. This is an important point. If you could put slide on, Slide 6 here.
This is the results of the CERC review of the major cardiac procedures. There were a total of 41 major cardiac procedures that were adjudicated by the CERC committee. Thirty-two of these were deemed to be associated with heart failure and those 32 were counted in the primary endpoint.
Nine of the 41 procedures were not associated with heart failure. These were not counted towards the primary endpoint, and these were much more in the control group.
So, in fact, the committee did what we were thinking they might do. If there was a lower threshold for a major cardiac procedure in the control group, these were eliminated through the use of this committee.
CHAIRPERSON MAISEL: Okay. Thank you.
One of my comments just has to do with the primary endpoint, and I guess in contest to Dr. Califf, I'm not a big fan of the primary endpoint. Let's say you had 100 patients who underwent the CorCap and had cardiac surgery and had no change in the New York Heart Association class, and then 100 patients who got biventricular implants. I know that wasn't the case here, but had improvement in their New York Heart Association class, all of those 100, quote, control patients would be counted as worsened because of their procedure, and the endpoint is meaningless.
So I have trouble interpreting the major cardiac procedure for heart failure as part of the endpoint. I might have picked a second cardiac procedure as their endpoint because if some patient in the control group needed a first one, now we're even. The second one would be the one that would be more concerning to me.
Perhaps you can comment.
DR. KUBO: Yes. Another important point, how do you account for lifesaving procedures in a group of patients like this? We had to make a decision that we would allow lifesaving procedures like transplant. So if someone gets worse, is in extremis, gets a cardiac transplant and is now functional Class I, it's not really fair to the device to call that patient New York Heart Class I because they've had an additional procedure.
CHAIRPERSON MAISEL: I certainly agree and don't dispute the transplant and LVAD issue. It's more the cardiac surgery, the valve replacement, the biventricular device.
DR. KUBO: The pacemaker, yeah. We've had lots of discussion about the issue about BiV pacers because the threshold is so low.
And if you could put on Slide 29, please.
We specifically looked at this and said, "You know what? Let's forget about all the BiV pacers. Don't even consider the CERC committee. Just assume all of the BiV pacemakers are noise."
And so if you exclude all of the BiV pacemakers from this trial, you have the results as shown here. We've shown this before. The odds ratio is about the same, and the p value is significant.
So there is some noise relating to this, but it's not causing the primary results that we see.
CHAIRPERSON MAISEL: Okay. Any other panel members have any questions?
MS. MOTTLE: Thank you, Dr. Maisel.
I have not heard anything dealing with patient education or physician education. Would the sponsor please address that?
DR. KUBO: Yes. Patient education is very important. As part of the trial, we talked to them about their heart failure and this being a part of their heart failure regimens, including all of the recommendations that physicians give regarding medications, diet, exercise, and so forth.
So this is meant to be a part of their ongoing heart failure education.
For physicians, we have an extensive training program that we implemented for the trial.
Slide on, please.
This just demonstrates at least the first phase of this where we're training not just the surgeon, but also the entire team. So we want them to be working as a triumvirate, if you will, the cardiologist, the heart failure, nursing, and the cardiac surgeon working together, and so we train them together on many things, including the mechanisms of action, the ability to identify the appropriate patients, understanding the patient education material, the implant technique, and the ongoing management, both peri and postoperatively.
And that training is supplemented by on-site training in the next slide on, please, in which we go to the sites, train the people who would be involved in the implant procedure, including personnel in the ICU, the anesthesiologist, the echocardiographer, and rehab people as well.
And then, of course, there's ongoing training for physicians and surgeons after that, too.
MS. MOTTLE: Can I also trust that depending on the final labeling indications, contraindications, warnings, that the sample patient education material that we have had presented will be updated --
DR. KUBO: Absolutely.
MS. MOTTLE: -- for grammatical errors, completeness of risks, potential contraindications, future CABG, et cetera?
DR. KUBO: Absolutely.
MS. MOTTLE: Thanks.
DR. KUBO: Thank you.
CHAIRPERSON MAISEL: Okay. At this point we'll take a 15 minute break. I have 3:17. We'll reconvene at 3:35.
(Whereupon, the foregoing matter went off the record at 3:19 p.m. and went back on the record at 3:38 p.m.)
CHAIRPERSON MAISEL: Before we begin with the FDA questions, I've been asked to remind the audience to sign in at the registration desk at some point before you leave today if you have not already done so.
At this point we will begin with the FDA questions, and I'll ask Geretta to read them. Can we do one at a time?
MS. WOOD: Mike, could I get you to project the questions for me?
We just want to remind you that today FDA must determine the reasonable assurance of safety and effectiveness. The safety and effectiveness are to be determined with respect to patients for whose use the device is intended, conditions of use prescribed, recommended or suggested, probable benefit versus probably injury.
The first three questions will deal with safety. The patient operative reports indicate that performing a subsequent operation on CorCap patients was complicated by the presence of dense adhesions.
Number one, please comment on whether placement of the CorCap device and the resulting increased difficulty for follow-on surgery, especially coronary bypass operations, will compromise patient safety during a subsequent operation.
CHAIRPERSON MAISEL: For some of these questions I will try to summarize points that we have belabored or at least spoken about quite a bit.
For this one we certainly expressed concerns about adhesions. Some of the descriptions were notable about how severe the adhesions were. The sponsor seems to have gone to great lengths to exclude patients who would be candidates for coronary artery bypass operations. Certainly that would be relevant for any labeling of the device.
I don't know that we have any evidence that it compromises patient safety. There was only a handful of patients that were required to undergo repeat operations. Certainly the operation time seemed to be a little bit longer. I don't know that we have data to say that it compromises patient safety, but it is a concern.
Do other people want to add to anything I just said.
CHAIRPERSON MAISEL: Yes, Tom.
DR. VASSILIADES: As a cardiac surgeon, I did not hear anything to make me gravely concerned about this device with respect to reoperations, other than for coronary bypass surgery, and I agree with what has already been said in terms of reoperations causing adhesions, and I honestly don't think that this device places the patient at any risk for safety in subsequent operations beyond what a normal reoperative operation should require.
CHAIRPERSON MAISEL: Would some of our other cardiac surgeons like to comment? Norm?
DR. KATO: Well, in general, when you're looking at your risk stratification for cardiac operations, just doing a redo basically doubles your mortality risk.
I think that given the description by many of the surgeons, who will remain unnamed and many of them are quite talented and leaders at their institutions, which also will remain unnamed, I have to take their comments seriously, which describe the adhesions as the worst they've ever seen, and these guys are doing, you know, high volumes of procedures.
And perhaps, you know, during a clinical trial you're dealing with obviously talented technicians, and you're right. We don't have any clinical data right now, but perhaps if the device goes to market general application might see an increase in mortality risk.
CHAIRPERSON MAISEL: Tom.
DR. FERGUSON: I would agree with Tom that I don't think this poses a greatly increased risk over what all cardiac surgeons see in the operating room today. The reason I wanted to comment was that I think there's a kind of either/or situation developing here with this device that makes it important. It's being put in for heart failure, but I don't understand yet what we're going to do with that body of patients who have ischemic heart failure, and clearly many of those are going to have minimum or maybe moderate degrees of heart failure when you decide to send them for coronary bypass surgery and/or invasive cardiology techniques.
I do think, and I want to go on record for this, that if you have this device placed for that indication, that is, heart failure, that you have removed essentially -- and I'd like to be corrected if this is wrong -- but you have removed essentially any possibility of doing an external procedure on the heart, i.e., bypass grafting.
Would you agree with that?
CHAIRPERSON MAISEL: This is noninteractive time period.
DR. FERGUSON: Yeah, okay.
CHAIRPERSON MAISEL: So any other comments?
DR. FERGUSON: He nodded. That's okay.
CHAIRPERSON MAISEL: Any other comments regarding Question 1?
DR. VASSILIADES: I have one more. I mean, I think that's kind of a long-range statement that pretty much lays to rest the possibility of any learning curve being accomplished with this device or the possibility of being able to learn what you can do at the beginning of the operation at the time of implant to mitigate the effects of reoperation.
So I'm confident that there will be surgical techniques in the future, and things that will be learned to mitigate some of the problems with subsequent operations.
So I am not as certain about the fact that you have laid to rest the ability to revascularize the patient.
CHAIRPERSON MAISEL: I should also comment in fairness to the sponsor they are advocating this device for people who are not candidates for revascularization.
DR. VASSILIADES: Yes.
CHAIRPERSON MAISEL: So it's not really in the panel's purview. Certainly we can talk about it and worry about it, but the patients who should be getting this device will not be candidates for revascularization.
John? Okay, Gene.
DR. BLACKSTONE: The patients in this study had an average age of 55 years. Assuming that they have a long life, how can you say that they would not be candidates later for an operation of coronary disease, which these days is usually done at much later age?
CHAIRPERSON MAISEL: Jeff.
DR. BORER: Can I try to respond to that? Because I think, as usual, Dr. Blackstone, you've raised a very important point. And I'll refer back to something Dr. Acker said, but he said it very quickly, and it flew by and nobody commented on it. I think it deserves emphasis.
These are very sick people. They've had heart failure in this study for an average of five years, and they're 55. The likelihood of their making it not a much older age group is rather slim with the best that we can do.
I think that this technique holds promise for improving quality of life. No one has suggested it improves length of life, although there may be tendencies in that direction from the April 2005 data, but the important point is that I don't think we can or should judge this against what might be done 20 years down the road because it's unlikely that any of the people who receive this would be living unless we have a dramatic improvement in our therapeutic armamentarium, and if we do, that may affect coronaries, too.
Obviously we still have endovascular therapy for the coronaries, and that could be applied. So I take the point. I think it's important, but I think we have to remember for whom this therapy is intended.
CHAIRPERSON MAISEL: Dr. Somberg.
DR. SOMBERG: With that said, I think it is fair to say that half the heart failure population has been excluded by this evaluation and, therefore, whatever this committee would consider recommending has to be cognizant of the fat that coronary bypass surgery at the time of or after the use of this device is truly in the unknown.
So why patients with Class III/IV heart failure have a very poor prognosis, certainly we're all aware that there may be a TOLT (phonetic) ischemic etiologies here that interventions may change things, and therefore, whatever we do has to be cognizant of the fact that ischemia and, as our Chairman has pointed out, ischemia has to really be excluded because we cannot predict what new surgical techniques will permit. We can't predict what may occur with populations.
All we can say is that it hasn't been studied, and I think, you know, from my point, and this goes beyond this question, but that is a tremendous disservice to the heart failure population by so limiting the initial valuation.
CHAIRPERSON MAISEL: Clyde.
DR. YANCY: Just to modify the context a bit, I can't let a comment pass that the prognosis for these patients is so dismal that longevity is not a possibility.
We are just entering an era where we have appropriate multi-drug treatment strategies and devices, and so there is no reason to think that a significant proportion of people that are Class III that have some measurements that are suggestive of at least a moderate prognosis won't continue to do reasonably well.
So even though it's tangential, the question about future surgical interventions is not entirely inappropriate to be considered.
CHAIRPERSON MAISEL: George.
DR. NETROVEC: I think the question is an excellent one, and I think based on the data we have, it would be a tough process.
On the other hand, surgeons are pretty clever people as best I can see, and if this really becomes a problem in the future and this device continues to provide benefit for patients so that we continue to use it, we'll figure out how to solve that, I think.
So we may be getting way off track.
CHAIRPERSON MAISEL: Bram.
DR. ZUCKERMAN: So, Dr. Maisel, after hearing this rich discussion, how would you summarize the answer to one? Does it go back to your initial point is if these patients can be picked carefully, the answer is it can be done safely and to leave the perhaps alteration of the indications question to a different question.
CHAIRPERSON MAISEL: I think what we've heard from some of our cardiac surgeons is that the risk of subsequent transplant surgery, which might be of great concern to these patients obviously, is not -- that transplant surgery is still possible in these patients. The coronary artery bypass operation, I don't think we have any data to comment one way or the other, other than the sponsor has told us that that planned or expectation of coronary bypass surgery is not to be expected.
DR. ZUCKERMAN: With the caveat that this is a carefully selected population.
CHAIRPERSON MAISEL: Of course.
Geretta, why don't we move on to Question 2, please?
MS. WOOD: Forty-two patients had evidence of constrictive physiology on at least one echocardiogram after CorCap implantation. Please comment on whether there is adequate information to assess whether placement of the CorCap CSD causes clinically significant pericardial constriction.
CHAIRPERSON MAISEL: I think we had a number of questions related to this topic. Certainly, there appears to be some echo data that there could be some, quote, constrictive physiology. It appears that there is not evidence of, quote, clinically significant pericardial constriction acknowledging that it could be many years before that type of physiology might develop.
Does anyone disagree?
CHAIRPERSON MAISEL: Why don't we move on to Question 3, please?
MS. WOOD: The regulations under --
DR. CALIFF: I'm surprised you didn't, I mean -- being a noted epidemiologist you didn't note that 150 cases can't tell you much of anything about what real risks are, or do you think it can?
CHAIRPERSON MAISEL: I was waiting for you to chime in.
DR. CALIFF: Okay. So my answer to the question would be who knows, but it could be up to five or seven percent and you might not see it at this point.
CHAIRPERSON MAISEL: I mean, I think because of the time course that Dr. Yancy discussed this is the type of thing that would need to be followed out for many years. It would be the perfect topic for a post market study or for certainly excellent echo data on any subsequent clinical trials.
DR. ZUCKERMAN: Yes, I just had one point of clarification on Dr. Califf's comments, "who knows." The sponsor today at their presentation said in order to get us to better know what is the appropriate answer, they were going to do their post approval echo reading via core lab as opposed to what is written right now in your panel pack book, which is site readings.
Was that an appropriate response by the sponsor that will help us to get to the root of Dr. Califf's concerns?
DR. CALIFF: Well, I had a choice between ten times as many patients with admissions for volume overload versus one-tenth as many patients with delicate echo core lab measurements. I'd take ten times as many patients, but if we can't get that, the core lab I think is a good thing.
CHAIRPERSON MAISEL: Jeffrey.
DR. BORER: You know, we're talking about constrictive pericarditis here, not an echocardiographic finding. Constrictive pericarditis is a clinical syndrome, and that does have to be looked for, but it won't be looked for effectively with an echo cardiogram.
I would certainly agree with Rob that if we had an adequate number of patients so that we could make a reasonably stable point estimate, that would be great, but we don't. What we do have is some information that suggests that hasn't been a problem yet.
Does that mean there won't be? Of course not. It doesn't mean that at all. We don't know that, but we don't have a problem yet, and there has been some looking, and that at least begins to reassure one that it's reasonable to keep on looking rather than to stop the process here.
So that's what I would say about it.
CHAIRPERSON MAISEL: Okay. Geretta.
MS. WOOD: The regulations under which a premarket application is assessed for marketing approval require demonstration of a reasonable assurance of safety. Please comment on whether the information provided in the panel pack and presented at this meeting is sufficient to demonstrate a reasonable assurance of safety for the CorCap device system.
CHAIRPERSON MAISEL: I'm going to call on Dr. Somberg to summarize the panel's thoughts and provide your own.
DR. SOMBERG: Well, I think I'm going to fall into the same problem you are in summarizing these things. I think the panel has a very diverse range of opinions, and the FDA is going to have to take note of that. I think it goes from some people who feel that not adequate numbers have been studied, and this may as we heard in the previous discussion not have picked up an appropriate signal to others who have a recognition that tin device development this has to be considered a large study.
The issue that I raised and I think some other shave touched upon as well is that at least half of the population with ischemic disease has not been evaluated, and therefore we have a different population that we're looking at here, and there are some very interesting findings in this population that suggest that it may behave differently, such as the robust response to medical therapy more than interventions, as compared to some other studies.
So with that said, I'm not sure we've adequately evaluated the safety in all populations which should include ischemic heart disease in patients who were end stage and who would need no further revascularization to adequately assess things.
But looking at the subpopulation, which has 60 percent idiopathic etiology, I think this is a fair assessment, not seeing any highly significant disturbing safety problem that has come to the fore to this date.
CHAIRPERSON MAISEL: So in a yes or no --
CHAIRPERSON MAISEL: -- do you feel that the information is sufficient to demonstrate reasonable assurance of safety?
John, do you have -- I'm not trying to put you on the spot, but I just want to clarify your position.
DR. SOMBERG: My position would be the safety issue has to be taken in context with the demonstration of efficacy, and you know, it's a balance, but you asked me to summarize the opinions I heard here, and I thought there was a range, you know.
So I would say possibly yes, possibly no.
CHAIRPERSON MAISEL: I'd actually like to hear from each panel member on this issue. So we'll just kind of go around the table. Judah, can you comment on your views on the safety issue?
DR. WEINBERGER: I think that the way for me to think about the safety is I'm going to ignore the ischemia breakdown. Of course, clearly the sponsor is not interested in having this be put out there for ischemic cardiomyopathy. So assuming that ultimately after much negotiation this will be put out there for dilated idiopathic cardiomyopathy, the question is: is it safe for that population? That is the question in my mind that I would like to slightly modify.
And to answer that question, you then have to dichotomize the population to those people who need their chests opened otherwise and those people who don't need their chests opened otherwise.
And if you look at that question, do the patients need mitral valve surgery or not need mitral valve surgery, the ones who don't need mitral valve surgery, I don't think this is a safe device at least because of the early surgical deaths.
The people who need mitral valve and who need their chests opened anyway, it appears to be a safe procedure for them.
So I can't give this a blanket safety backing, but I can say that there is a population that has been found here in the trial where it appears to be safe.
CHAIRPERSON MAISEL: Norm.
DR. KATO: I think that with those qualifications, I think what Judah just said is appropriate. I think there is a very small subset here where safety can be shown. And I think there's a large population where safety is really unclear. So I think that also leads you into, you know, later on for the labeling. The labeling will just have to be very precise.
CHAIRPERSON MAISEL: George.
DR. NETROVEC: Yeah, I would support that. I think if you're already going to operate on the patient, I think it has got safety. I think to do it otherwise there's some question.
CHAIRPERSON MAISEL: Gene.
DR. BLACKSTONE: I agree.
CHAIRPERSON MAISEL: Rick.
DR. PAGE: I agree in terms of the acute situation, but I also have concerns in terms of long term because you're leaving a device in place which may become an issue with further surgery.
That being said, if the efficacy were adequate, it could overwhelm the concerns that I have for safety, but it becomes a balance. But my perspective is I remain concerned.
CHAIRPERSON MAISEL: Thank you.
DR. WHITE: I have reservations about the safety of this device, both acutely and long term. I think the issue of constrictive physiology may become a bigger issue than we've given it now, and five years from now I'd be concerned about looking at that data.
So I'm concerned about it long and short term.
CHAIRPERSON MAISEL: Sharon, could you speak into the microphone, please? We couldn't hear Chris very well. Could you please speak into the microphone?
DR. NORMAND: Is there a button?
So actually I don't have any concerns about safety based on the data that was presented.
CHAIRPERSON MAISEL: Tom.
DR. FERGUSON: Yes. I think one of the strong issues for the device is its safety in the short term. None of us can predicate what's going to happen long term. So let's not get into that.
CHAIRPERSON MAISEL: Clyde.
DR. YANCY: I am concerned about safety in the immediate perioperative period, acknowledging that it may get better over time as surgeons get more comfortable with the technology, but based on what we've seen so far, I am concerned about safety.
CHAIRPERSON MAISEL: Rob.
DR. CALIFF: Yeah, I guess the way I would put it is relative to most devices that I've seen I would say there's an out in the clause about "relatively" or something like that. You know, I would say the risks are explainable and understandable. As long as it was labeled appropriately, you're going to take a two to ten percent chance of being dead almost immediately in order for a benefit which I guess we'll talk about later.
I would also point out that in practice outside of clinical trials everything we've seen is going to show a higher risk of mortality and surgical complications.
CHAIRPERSON MAISEL: Jeff.
DR. BORER: I agree with Rob. I think that for the patients for whom I believe this device is intended, which is people with very severe congestive heart failure who are on maximum medical therapy, plus or minus a biventricular pacemaker, that it is acceptably safe for the intended use as long as the patients are highly unlikely to imminently require coronary bypass grafting, and I would put that caveat in.
But with that caveat, I think it is acceptably safe.
CHAIRPERSON MAISEL: Tom.
DR. VASSILIADES: I agree with Tom and Bob and Jeff, that I believe the device is safe, and I also think that those panel members that do not feel it's safe are really using a different definition of safety because if you compare it with medical therapy, then by that definition no device will ever be implanted by means of the fact that it's not safe. So I think we have to rely on the relative safety, relative to a surgical procedure, and I don't think there are issues.
I think that the study has supported it fairly well actually.
CHAIRPERSON MAISEL: Certainly we'll need to consider the safety and effectiveness balance which we'll discuss in a few moments, but it sounds like most of the panel members can identify at least a small population or a subpopulation that is feeling safe, and not everyone in sum feel it's safe in the whole group.
DR. ZUCKERMAN: Yes. Dr. Maisel, we've had a very good discussion of the global safety issues, but if I could ask Dr. Yancy to clarify some of the comments on the acute safety issues.
Were you referring to things that perhaps could be optimized in a post approval study, such as the way the operation is done a little bit more investigation to Frank Starling mechanisms. There seem to be hemodynamic compromise acutely. Those are minor points compared to the global safety issue, but what acute safety things were you just referring to?
DR. YANCY: Well, I mean, to be candid, I think death is a concern. Now, the investigators appropriately reassessed where things were going and took interventions, and's why I suggested that over time, I think the learning curve would lead to a lesser risk, but the question is fairly precise, the information provided in this panel pack, and I can't walk away from the fact that there appears to be an early perioperative concern and that it's in the context of efficacy, et cetera. So that's specifically what I had mentioned, too, and it is possible that in the post marketing survey or with a physician education program that's very rigorous, adjustments can be made to bring that down, but that's an undefined target right now.
CHAIRPERSON MAISEL: Dr. Somberg.
DR. SOMBERG: I just wanted to interject that the adjustment that was being, I think, talked about to deal with the hemodynamic stability in the bypass, and we all know that there's a trend towards a lot of procedures to be done off pump to try to avoid the debits of bypass. So when we're trying to mitigate one problem we may be exposing another series of problems to this patient population as well.
So I don't think we can blindly say that maybe things could be made better here or there, and you might change mortality, but you might change, you know, cerebral dysfunction, et cetera. So I'm concerned, and those things all have to be weighed, and that's why, you know, I'm going to defend myself. That's why there was a, you know, maybe yes, maybe no, because there's no clear answer here. It has to be in a risk-benefit context when you evaluate this device.
CHAIRPERSON MAISEL: I would also just comment that I think it is worth pointing out that the no MVR group that had the early postoperative mortality did have equal mortality at 12 months, and this panel has quite frequently seen devices implanted that have early complications. So certainly it's a balance between the safety and the effectiveness.
Why don't we move on to the next question?
MS. WOOD: Questions four through ten will deal with the effectiveness.
As the protocol evolved during the course of the trial, FDA requested that NYHA class be assessed by a blinded observer rather than by the site investigator. Blinded assessment of the NYHA class at entry was not available for more than half of the patients. Their NYHA class for the purpose of the statistical analysis of the composite primary effectiveness endpoint was imputed using a statistical model.
Please comment on whether the imputation of NYHA class compromises the analysis of the composite primary effectiveness endpoint, which includes NYHA class as one element.
CHAIRPERSON MAISEL: Well, we certainly had a lot of discussion on this topic and saw several different imputation models as well as models that did not impute any data. Maybe I'll ask my favorite statistician on the panel, Sharon, to comment on her views.
DR. NORMAND: I'm also the only statistician on the panel. That's why.
I'd like to clarify that I don't think the issue is imputation. It's the fact that there's so much missing data. So I would comment by saying that I think that the amount of missing data does compromise the analysis. It's not necessarily imputation procedure, but the amount of missing data.
CHAIRPERSON MAISEL: Clyde, what are your thoughts on that topic?
DR. YANCY: I agree with Sharon.
CHAIRPERSON MAISEL: Does anyone disagree with Sharon and Clyde? Does anyone feel that the data on NYHA class is satisfactory?
DR. CALIFF: Well, so this is like everything else we've been discussing. The question is how much, and I don't think any card carrying biostatistician or ambulatory clinician statistician would say that when you have to impute with large amounts of data that it's as good as actually having the data.
So the real question is, you know, when you have a primary endpoint of .04 whatever it is and then you have the uncertainty or loss of confidence provided by imputation, you know, how much? Could I ask Sharon just to comment on a scale of how much?
DR. NORMAND: Typically in the statistical field, if you have -- this is typical -- if you have more than 50 percent missing data, you're in trouble, and although I can give you an example of where you have five percent missing data it would be a problem, but when you have more than 50 percent missing data, you are actually in a situation that statisticians would characterize as nonignorable missingness, and it does compromise your analysis if you use a model that assumes missing at random.
So I think it's safe to say with more than 50 percent of missing data you're actually -- your statistical model should drive your results, and so that's the bottom line answer.
That's not to say that less than that amount wouldn't cause you problems, but certainly that amount raises a big red flag and says that, indeed, this is a lot of missing data and this compromises your analysis.
CHAIRPERSON MAISEL: Gene.
DR. BLACKSTONE: I'm a little disappointed that the sponsor didn't then go back and at least retrospectively have the panel that was adjudicating the NYHA look at the first half of the group. Even if they flagged those, it would give some information, more than we have here.
CHAIRPERSON MAISEL: John.
DR. SOMBERG: But when you say that you can make the argument why not take the site selected data, and that goes, if you will, the other way, and that would take a large component of the composite primary endpoint out.
So I hear you, but you can see why someone like myself have major problems with the endpoint because of this issue. So I think for some of us this is really -- you know, if there was another endpoint chosen another way that went that way but it was biased, okay, and this one has problems, but you know, we sort of take that.
Maybe that's the clinician amateur/very amateur, but when one goes one way and the other goes the other way and this is really compromised because of numbers, then I'm very concerned.
CHAIRPERSON MAISEL: Jeff.
DR. BORER: I think Rob raised the important point. The question isn't worded that way, but you know, I'm not a statistician. There's obviously a large proportion of the data missing, and that's hard to draw a firm conclusion about the strength of the analysis, and I would defer to Sharon certainly about that.
But I think there's another aspect to this. These are the data we have, and yes, in addition, I would wish that the decision had been made by the investigators to include heart failure hospitalizations in the endpoint. That would have helped.
But they thought about it and came to the conclusion they didn't want to, and, you know, they're good investigators and it's reasonable for them to have that opinion.
However, despite those issues, I am persuaded by the fact that no matter how you look at the data, except for the subset of patients who had MVR and the site definition of New York Heart Association functional class as the outcome, except for that subset because the entire population when you looked at site selection, the site selected New York Heart tended slightly in the direction of favorability for the CorCap; except for that subset, no matter how you look at the data, they seem to go the same way, and that is that the CorCap is superior to no CorCap, and the odds ratio is approximately the same, again, no matter how you look at the data.
So you know, as a non-statistician, I look at the concordance of all these analyses, and I say, "Gee, it looks like there's something there." So I would echo Rob's comment. I mean I don't know how everyone would come out on that if you thought about that, but I would echo Rob's comment that it's not that there isn't concern because of the missing data. Of course there's concern. The question is how much concern in the context of judging the relation of benefit to risk for this device.
CHAIRPERSON MAISEL: Any other comments? Sharon.
DR. NORMAND: I guess the way I'm thinking about it is that the composite is really driven by not very much by the New York, the missing data. Death has nothing to do with it. So we've got these two endpoints, and although we saw some analyses where it didn't matter which model you used, I can bet you a dollar I can come up with a model where it would matter, and that's the issue.
The issue is that because there's so much missing data for that endpoint, again, in a clinical trial that's a primary endpoint, there has been lots of analyses done. The issue is that certainly I can show you more models where it would matter, and the plausibility of which is the right model is up in the air. And that's where the issue is.
So that's why I have trouble with the composite. If they were going to use a different measure and pull that out, maybe they'd have a different answer, but what I was presented today is a composite measure where there's more than half of the data are missing. It has been cut lots of ways. I could cut it a lot of different ways, and I guarantee you I could make you, depending on my multiple imputation model.
So that's the issue. The issue is the plausibility of the model for the missing data when there's so much missing data, and it makes no sense to look at a model where you exclude the people who had the missing data.
So that's statistical theory and protocol that's been advocated and you may feel comforted by seeing that, but I would tell you that I'd caution you against that comfort level.
DR. BORER: I don't want to lose my dollar.
CHAIRPERSON MAISEL: Why don't we move on to the next question, Geretta?
MS. WOOD: The primary determination of device effectiveness was assessed by a composite of mortality, NYHA score, and the necessity of reoperation for worsening heart failure. As individual measures, neither mortality nor NYHA class were statistically significantly improved in device as compared to control patient.
However, CorCap patients experienced significantly fewer reoperations to treat worsening heart failure than did patients in the control group.
Please comment on the extent to which possible physician treatment bias, reluctance to refer forward or to perform reoperation on CorCap patients might have affected the disparity in the number of devices versus control patients who received a reoperation, major cardiac procedure for worsening heart failure.
CHAIRPERSON MAISEL: Why don't we do number six at the same time? Can you read that, too, please?
MS. WOOD: If it is your opinion that physician treatment bias did affect the number of reoperations in device patients, please indicate whether you think such bias affected the outcome of the primary effectiveness endpoint. If so, to what degree?
CHAIRPERSON MAISEL: Well, we also discussed this at length. I think we all agree that there's the potential for physician treatment bias. Obviously physicians and patients were not blinded by placement of their CorCap. It seems unlikely that that would have a major effect on transplant or LVAD placement. Certainly it could affect biventricular device placement or -- I'm sorry -- I don't think it would have really affected biventricular device placement because it's a pretty benign procedure that wouldn't be affected by the CorCap.
DR. CALIFF: So you're shocking me multiple times this afternoon already. I mean, given what we know about the capriciousness of transplant and ANOVA implantation, you can look at race. You can look at geographic location. There are huge -- you know it's not as if every patient that meets some criteria gets this procedure done. Why are you so confident it wouldn't have an effect when the difference is already so small between the two groups?
CHAIRPERSON MAISEL: I do think that physician treatment bias has potential to affect the outcome, but I think I've been convinced by the sponsor that those patients would be captured in other ways, that a patient that didn't get a transplant because they weren't considered a good risk would have had worsening heart failure and might have been picked up by the New York Heart Association Class.
DR. CALIFF: But the sites actually saw that the people without the device were doing better, more improved at least. I mean, when they were looking at the patients they were saying the patients didn't get the device or at least it's good, if not better.
So I'm a little less certain than you seem to be even without these two major endpoints.
CHAIRPERSON MAISEL: John.
DR. SOMBERG: Well, I just may be a contrarian thinker, but I think because the CRT is equally distributed in the treatment versus the control group, and that's the one intervention which would not be biased by whether you should or should not open the chest again, I think it suggests that there was a bias and that the other devices which are not equally placed in both groups, much more in the control than the treatment group, is because people were reluctant to open the chest.
And, therefore, I think there is a suggestion of bias, and that becomes important when two of the three, to put all of the questions together, when two of the three factors in the composite endpoint are not composite if you disqualify the New York Heart the way it was done. Then you have to say the one that is moving, the endpoint is potentially biased, and the endpoint has to be considered questionable.
So I'm leaning more on that it is an important question of bias than there is less of one, but maybe if you summarize our panel, it would say yes/no, again.
CHAIRPERSON MAISEL: Other comments? Judah.
DR. WEINBERGER: Yeah, I think that possibly one could explain some of the reasons for lower reoperation rates in the intervention group because anyone who needed mitral valve surgery got it at the time they got their CorCap. People who might have not needed mitral valve surgery until later on in the course, who randomized them into the control group, then had an event later on, and that added to the control column operations.
So I think that it's not at all clear cut that there was a bias or if anything, it could have been that, you know, randomizing to getting the device lowered the threshold for adding mitral valve surgery. I don't know that we have a way of deciding that.
CHAIRPERSON MAISEL: Tom.
DR. FERGUSON: I think it's dangerous for us as a panel to impugn bias when we have no proof for that any more than we -- I know what our job is here, and so on and so on. But I don't think that well meaning physicians are going to deny a treatment modality to a patient if, number one, they feel that the treatment modality is reliable and if it poses a little bit of a risk for the patient, then so be it. That's the way we take care of patients as doctors. I don't understand this bias talk.
CHAIRPERSON MAISEL: Jeff.
DR. BORER: I agree with Tom. I must say I looked at these data a little bit differently than perhaps what we've heard, and I agree with the sponsor's analysis. I can't say that there was no bias. I certainly cannot. I mean, we didn't ask; we didn't look in a way that would allow us to come to a rigorous conclusion about that.
But I don't think, first of all, that the biventricular paces were equally distributed. There were 50 percent more in the control group.
The fact is that even though there was no significant difference in mortality, and I'm not going to suggest that there's a mortality benefit here, as time went on there tended to be fewer deaths as a proportion in the patients who received the CorCap and the patients who didn't, and even though there are a lot of missing data, and, you know, maybe everything is thrown off because the imputation model that was used, et cetera, et cetera, at least what we've seen, does not suggest that there was an increase in the severity of heart failure selectively in the people who received the CorCap.
So I don't think there was even a real basis for a bias, even if a bias might have existed. I just don't think that it's likely that that, as you said, Bill, that it's likely that that had an impact on the data set that we see here.
CHAIRPERSON MAISEL: Other comments?
I always like to end with a comment that it agreed with me, and so we'll move on to Number 7.
MS. WOOD: FDA notes that the incremental effectiveness of the CorCap when used together with mitral valve replacement appears to be less than that of the CorCap alone. Please comment on this apparent difference in effectiveness between the two groups.
CHAIRPERSON MAISEL: Judah, you were commenting on the safety between the different groups. Do you want to tackle this one?
DR. WEINBERGER: I think that the model of the statistical uncertainties with interpreting the primary endpoint, the amount of effectiveness or confidence in effectiveness in patients who are getting mitral valve replacement is certainly lower, even based upon these statistics than it is in the patients who are not getting mitral valve replacements.
So I know I'm not being very clear. Let me try and just summarize and say that the people who it appears to be safer to do this to don't appear to be getting incremental benefit to the best that I can determine.
CHAIRPERSON MAISEL: George.
DR. NETROVEC: I don't know the breakdown because it wasn't presented as far as I can see, but remember the inclusion criteria allowed a patient getting mitral valve surgery to have a 45 percent ejection fraction, whereas the patients in the non-mitral valve had to be 30 percent or less.
And so the possibility is that if most of the patients getting mitral valve replacement had a 45 percent ejection fraction, it's quite conceivable over 18 months' follow-up you wouldn't see much benefit. I think that may have more to do with it than anything else.
CHAIRPERSON MAISEL: Jeff.
DR. BORER: Yeah, I think that George's point is well taken. There were also functional Class II patients involved in that group which would make it harder to see an improvement in functional class, but I think that the key point that we have to remember is that the study was not really powered to see, to be able to be likely to see a difference between the two groups even if it was really there and there was no interaction term.
You know, it sort of looks like there was less of an impact on the people who had mitral valve surgery, but of course, mitral valve surgery by itself has a positive impact presumably.
So I think that these data are too few and potentially confounded not by intent or not because of poor study design so that it's hard for us to really draw a firm conclusion about whether there is, indeed, a true difference in the impact of the CorCap and the people who got an MVR and the people who didn't get an MVR.
Certainly it's plausible, and maybe it happened, but I don't know that we can draw a firm conclusion about it.
CHAIRPERSON MAISEL: Sharon.
DR. NORMAND: I agree with the comments that were just made, and unless I'm mistaken, I thought that the sponsor did include an interaction. It was nonsignificant. Now maybe they might not have been powered to find that interaction, but indeed, they fit an interaction term. It was nonsignificant, and hence, I'm not concerned about -- I don't have a concern about this.
CHAIRPERSON MAISEL: Rob.
DR. CALIFF: Two comments. First, I can't resist just saying this because it's intriguing. We've been filming doctors and patients talking about risk and benefits of statins in actual practice, and it's amazing to see how the doctors stumble over it. I'm trying to imagine based on this panel's discussion of subgroups and multiple factors how on earth a reasonable discussion would occur between a doctor and a patient, but I guess we'll get to that later.
But this just points out there are 300-patient trials for the fastest growing component of the leading causes of death in our country leaves us with huge numbers of unanswered questions, and I hope that we can get beyond this view that you can only do these tiny studies because, in essence, we're left guessing, is my perspective on this whole thing.
As Sharon said, the interaction test is not significant. Well, what do you expect when you have so few events and so few patients? It's not powered to tell it, and we're left as far as I'm concerned with a pure guess.
CHAIRPERSON MAISEL: So are there other panel members who feel that the data is sufficient to make a distinction between the two populations, the MVR and the no MVR?
DR. ZUCKERMAN: Yeah, Dr. Normand, can you clarify Dr. Califf's comments? I'm referring to page 5 of the FDA slides, Dr. Thompson's slides, and with the numbers and the widths of the confidence interviews would you agree with him? Was there anything else to be made out of that?
CHAIRPERSON MAISEL: Where are you looking?
DR. ZUCKERMAN: It's the fourth one, page 5 of Dr. Thompson's.
DR. NORMAND: Oh, do you me the page 5 of Slide 5?
Well, I think you asked two questions. One is did I agree with Dr. Califf. I mean, I agree with him in terms of the point made that the interaction was nonsignificant. I queried as to whether or not they were powered enough to find it, I guess, but I can't do that calculation that quickly in my head, but I suspect they weren't powered. So I agree with that point.
And hence, when one looks at the confidence intervals that Dr. Thompson did, they do certainly overlap. So, you know, even looking at that, they don't look that different because the intervals overlap.
So, again, but it's just based on whether there's enough data or sample size to differentiate the odds ratios in those two groups, and I don't know if they were powered enough to find it, but based on what we see, we cannot say that those intervals are different.
CHAIRPERSON MAISEL: Any other comments on the MVR/no MVR subgroups?
CHAIRPERSON MAISEL: Number 8.
MS. WOOD: Please comment on whether the results of this clinical trial are sufficient to demonstrate a reasonable assurance of effectiveness for the CorCap CSD.
CHAIRPERSON MAISEL: Who would like to tackle this one? Dr. Califf.
DR. CALIFF: Reasonable. Can I ask for -- I haven't done devices for a while. It seems like the standards have gotten higher than they used to be, but can you define reasonable?
DR. ZUCKERMAN: The agency is not looking for absolute assurance, but we are looking for a situation with a medical device where if it is appropriately labeled for a specific intended patient population, we believe that we have developed a data set where there is a positive risk-benefit profile and the device should be approved.
So I think it's very similar to the usual, thoughtful clinician judgment that's used in making many decisions with patients, hence the need to bring this device to an Advisory Panel.
DR. CALIFF: And so I guess what I would say is that if the standard was an interpolated p value of .05 in favor of a primary endpoint of effectiveness, you know, I'd say it doesn't quite meet that standard because it's just below .05, and we've got the uncertainties that we've discussed, particularly weighing in my mind the need to interpolate data that's not there.
And I think in the questioning it has been clear that there has to be some penalty for having to do that. It's unspecified what the penalty ought to be, but it wouldn't take much to get from .04 to .07.
If you ask me relative to other devices that are on the market as a criteria for reasonableness, you know, the trends are almost all in the right direction, and so, you know, I guess we haven't gotten to the balance question yet. So I would say it all depends on your definition of reasonable, and it's right on that borderline, and I think the tradeoff could be described to a patient and a doctor in a way that would allow a reasonable decision to be made, which I think is the way Bram asked the question.
CHAIRPERSON MAISEL: Clyde, what are your thoughts?
DR. YANCY: I think one has to acknowledge that based on the agreed upon primary composite endpoint by the investigators, they did reach that, and so if the question is restricted to a reasonable assurance of effectiveness from this clinical trial, I think we have to embrace that with varying degrees of enthusiasm, and the ultimately issue will be the balance of that with risk and indication.
CHAIRPERSON MAISEL: John.
DR. SOMBERG: Well, I must say from my review I felt that the composite endpoint was very muddled and iffy. You're saying it goes in that direction, but it's really a rehash of many different ways of looking at that endpoint, and when you look at mortality, when you look at hospitalization, when you look at exercise performance, six-minute walk, I don't see a benefit, and we're asking people to risk their lives for anywhere you say from a two to ten percent incidence of up front mortality and what is the very hard benefit you're getting?
And in this day and age where there are a lot of alternatives, sure, you could throw it to a patient and possibly offer that in an individual case, but you're talking more of a humanitarian potential here as opposed to based on a hard, randomized controlled study.
So I think this study has unfortunately chosen the wrong endpoints and not reached a conclusive determination.
CHAIRPERSON MAISEL: Other comments. Tom.
DR. FERGUSON: Well, I just take the opposite view then. They've done a good job of taking a disease process for which there is really no effective surgical treatment. I emphasize the surgical, and I would even include the pacemaker. That's yet to be determined, and have given us data which show that they can turn these patients around.
Now, I understand all of the things that I've heard today and you've heard today, but I guess what I'm saying the bottom line for me is that they have done their best to give us reasonable assurance and I'm assured.
CHAIRPERSON MAISEL: Rick, did you want to?
DR. PAGE: I agree with Dr. Califf that this barely but I think just falls short of demonstrating efficacy. My concern is the efficacy is fairly minor, and this is, indeed, a high risk population.
So, yes, I agree we don't have good surgical options for this high risk population, but I'm not convinced that this is a surgical disease we're talking about right now.
Then you find we have to balance the risk versus what I see is maybe a modest benefit, and that will come down through the final vote.
CHAIRPERSON MAISEL: Chris.
DR. WHITE: I agree with Dr. Somberg, and I think that there's not much evidence of benefit here at all, and I think that virtually every part of the endpoint has been in question of whether they met or didn't meet benefit. I think that whether we can measure the actual risk or not in this small population is uncertain.
So I would feel that I don't feel that we've met a reasonable assurance of efficacy given the uncertainty in the endpoint.
CHAIRPERSON MAISEL: Any other comments on effectiveness before we move on?
DR. WEINBERGER: I think I'm going to incorporate Bram's comments here about reasonableness. If I were a reviewer reviewing this data for publication and somebody presented data where in the endpoint the entire p value of the endpoint depended upon having less than 50 percent of the data, I don't know that I would accept that paper for publication. I'd have a real hard time, and by that criteria, if that's going to be my criteria, this doesn't meet my reasonable effectiveness criteria.
CHAIRPERSON MAISEL: Gene.
DR. BLACKSTONE: On the other hand, as you know, I've already expressed an opinion about the primary endpoint. In terms of whether the device is effective, the fact that there's a difference in end diastolic volume says that it is doing something. That doesn't mean that it translates into clinical effectiveness, but in terms of device effectiveness, it seems to be doing what it is supposed to be doing.
CHAIRPERSON MAISEL: Jeff.
DR. BORER: Yeah, I'm going to say this only so that it can be on the record because I had intended to hold my comments for Number 10, but I believe without going into details that there is sufficient evidence to demonstrate a reasonable assurance of effectiveness for the device for what it's intended to do, and what it is intended to do is something we're going to have to talk about a little bit.
CHAIRPERSON MAISEL: Bram, did you want to comment?
DR. ZUCKERMAN: Yeah. I just wanted to clarify the definition of effectiveness. When you get to Question 10, as Dr. Borer was just saying in response to Dr. Blackstone's comment, the regulatory definition of effectiveness is not that it's just a mechanical restraint, as he indicated. There must be a clinical utility also.
CHAIRPERSON MAISEL: Okay. Thank you.
Why don't we do Question 9 now, please?
MS. WOOD: Secondary measures of device effectiveness included measures of ventricular geometry, left ventricular dimensions, and patient quality of life, MLHF and SF-36, and functional status, six-minute walk and MVO-2. Please comment whether some or all of the secondary measures of devise effectiveness can be used to assess the effectiveness of the CorCap CSD.
If your answer is yes, please explain which measures are useful and how you interpret them.
CHAIRPERSON MAISEL: Well, we also spent an awful lot of time talking about this. Certainly MOHF of 36 or potentially subject to bias in patients and physicians that are unblinded. It would have been very nice to see physiologic improvement in these patients with six-minute walk in MVO-2, and I am biased towards using those preferentially over the others, although they are certainly all useful, as are some of the geometry and EF and ventricular size measurements.
Does anyone have additional comments? Rob?
DR. CALIFF: Well, I'll ask you. What do you make of the absence of benefit for New York Heart from the bias of the -- since there appears to be no bias there, would that influence you to think that there was lack of bias in these other secondary outcomes?
There definitely was a bias in favor of the device and the site assessment of New York Heart class. It went --
CHAIRPERSON MAISEL: I would say there's a lot of bias in the New York Heart Association classification, as evidenced by the core lab versus the site.
DR. CALIFF: Which one do you think is more accurate?
CHAIRPERSON MAISEL: I think blinded analysis is more accurate.
Any other comments?
CHAIRPERSON MAISEL: So let's do Number 10.
MS. WOOD: Please comment on whether the benefits of CorCap CSD therapy outweigh its risks.
CHAIRPERSON MAISEL: Why don't we go around the table on this one? And I'll spare Judah and we'll start at the other end, and I'll ask Tom to comment first.
DR. VASSILIADES: I think looking at the study and weighing all of the information, in my opinion, the benefits outweigh the risks.
CHAIRPERSON MAISEL: Jeffrey.
DR. BORER: I agree with Tom. I think the benefits outweigh the risks for a population that has severe heart failures on maximum medical therapy plus or minus a biventricular pacemaker, if that's indicated and appropriate for the person and with the caveat that the patient doesn't require or doesn't appear to require, isn't likely to imminently require a bypass crafting procedure.
With those ground rules, I think the benefits outweigh the risks.
CHAIRPERSON MAISEL: Rob.
DR. CALIFF: Well, the truth is I don't know based on the data we have. I guess what I would say is that for some patients who have certain values there's a range of benefit and risk describable in the population that could lead some individuals to choose this treatment, even given the uncertainties that exist.
So would I say this device is safe? Would I say it's effective? Neither.
Would I say that there's enough data here relative to other devices for an individual to make a choice? I'd say yes.
CHAIRPERSON MAISEL: John.
DR. SOMBERG: I think the device is promising. I think it does demonstrate a decrease in the volume, which is useful, and it has a hurdle that pharmaceutical interventions don't have, and that is the implantation of it, and we ask more for a device possibly than we ask for a pharmaceutical.
I know pharmaceuticals haven't got side effects, and that's certainly the case, as well. That can be limiting.
With that said, I think there's promise here, but we have not had a cleaner demonstrate of benefit as to add to the risk of the implantation, and thus, I am very concerned about an approval, but I do think it's certainly not a condemnation of the device, that it doesn't work.
CHAIRPERSON MAISEL: Clyde.
DR. YANCY: I think about Question 9 and preceded it. As I mentioned earlier, I'm interested in seeing consistency. I think the composite is fairly soft and modest at best in terms of benefit, but it would reassure me to see some of the other markers going in a direction that is more robust, and I don't see that.
I suspect that, indeed, there are individuals in highly selected situations where this could be effective, maybe even dramatic. I'm not sure if based on these data we yet know who those persons are. There would be every reason to try to have this available, but I don't know how we can truly do that in the context of the data.
So being forced to answer the question, I think I'd have to say that in aggregate, looking at benefits defined both by primary and secondary, I don't think they outweigh the short term and the still uncertain long-term risk, specifically constrictive pericarditis and any risks related to reoperations.
CHAIRPERSON MAISEL: Tom.
DR. FERGUSON: I agree with Tom.
CHAIRPERSON MAISEL: Sharon.
DR. NORMAND: Given the data that were presented to us and having to answer the question, I would say that I cannot say that the benefits outweigh the risks.
CHAIRPERSON MAISEL: Chris.
DR. WHITE: I think the uncertainty makes me say that the benefits don't outweigh the risks.
CHAIRPERSON MAISEL: Rick.
DR. PAGE: I agree that I don't believe the benefits outweigh the risk here.
CHAIRPERSON MAISEL: Gene.
DR. BLACKSTONE: I agree with Rob and Sharon on that, too.
CHAIRPERSON MAISEL: George.
DR. NETROVEC: I would concur with the position that the effectiveness does not outweigh an unknown risk.
DR. BORER: Norm.
DR. KATO: I will also agree that the benefits do not outweigh the risks here.
CHAIRPERSON MAISEL: And Judah.
DR. WEINBERGER: First I'd like to thank the Chairman for being such a wise judge of body language and starting from the other side, but I have to agree that across the population as defined in these studies, the benefit do not outweigh the risks.
CHAIRPERSON MAISEL: Bram, did you want any other additional comments on that question?
Okay. Let's move on to labeling, please.
MS. WOOD: One aspect of the premarket evaluation of a new product is the review of its labeling. The labeling must objectively describe which patients are appropriate for treatment, indicate potential adverse events associated with the use of the device, and explain how the product should be used to maximize clinical benefit and minimize adverse events.
If you recommend approval of the device, please address the following questions regarding product labeling.
Number 11, with references to the proposed labeling, the majority of patients studied had dilated cardiomyopathy and yet the number one cause of heart failure in the U.S. is ischemic cardiomyopathy. Please comment on whether the results of this study can be extrapolated to an ischemic cardiomyopathic population.
CHAIRPERSON MAISEL: John.
DR. SOMBERG: Well, I think we may -- maybe I'm guessing the body language of the panel differently, but I would say that if there was a favorable recommendation, it would have to state that there is little information on ischemic myopathy patients, and the device should not be indicated in those patients unless all ischemic improvement measures have already been taken.
So, in other words, those type of patients, like the protocol would have to be excluded as was done in the protocol.
DR. BORER: I agree with John. I would say it a slightly different way, but I think John made the point that I would. I don't think there's anything in the data that suggests that patients with ischemia as the reality of the heart failure didn't respond or don't respond.
Ten percent of the population, indeed, presumably had an ischemic etiology, and you know, we didn't hear anything to suggest they responded differently from the others. It's a small group, and so it's hard to draw a firm conclusion about that subpopulation specifically.
But I think the key point, and John made it, is that those people did not include anyone who are likely imminently to require bypass grafting or another coronary revascularization procedure. So I think that we cannot extrapolate to that group. We have no data about that group, and we can't extrapolate to them.
May I add to that comment a little bit? Because I thought that this labeling issue -- I didn't realize the questions were written just exactly as they were. Question 9, you know, I think Clyde made a very important point about the lack of consistency there. I would take it one step further.
As I read the label that was in here, it talks about reverse remodeling, and I would have to infer from the data we've seen that we don't have very much information specifically about reverse remodeling. It may be occurring, but we don't really have the information that would allow us to say that.
And it was in response to the items that were measured relevant to Question 9 that we could have looked for something, and I don't think that it was there in strength.
So I think that when you look at this label, the references to reverse remodeling and improved modeling of the heart and whatever might have to be revised.
CHAIRPERSON MAISEL: So just getting back to the ischemic population again, do people object to ischemic patients; assuming that the device were approved, do people object to ischemic patients getting it if they need the inclusion and exclusion criteria from the clinical trial? Meaning that in the clinical trial they describe exclusion criteria such as candidates for surgical revascularization, et cetera.
So for patients who are ischemic but have no planned surgical revascularization in the future?
DR. CALIFF: I would take the position that, you know, it was ten percent of the patients. There was no striking signal there, and as long as there was really agreement between the doctor and the patient that surgical revascularization was not in the future, that wouldn't bother me. A reasonable person could make that decision.
CHAIRPERSON MAISEL: Does anyone object to that? Okay.
DR. SOMBERG: I wouldn't object to that, but I would say it was not just surgical. It is interventional as well because that has not been looked at, and I'm understanding that there may be some concerns in that regard as well. With the development of -- you know, I say they're not shrink wrapping the heart, but for all intent and purposes it is shrink wrapping the heart after a period of time. How much pressure is put on is just unknown at this point.
So it may even preclude -- I doubt it, but it may preclude or make difficult coronary, and I would defer to coronary interventionists who would like to comment on that.
CHAIRPERSON MAISEL: Getting back to one of Jeff's prior points about the use of the term "remodeling," it states in the labeling "associated with a reverse remodeling effect." Does anyone -- Jeff said he was uncomfortable with the term "remodeling." Is anyone okay with the term "remodeling"?
DR. CALIFF: If you asked 100 doctors what "remodeling" is as it relates to left ventricular function, I would bet that most of them would get it wrong.
So I don't really even know why this would even be in a label. It doesn't describe a patient benefit. It describes a biological phenomenon.
CHAIRPERSON MAISEL: For a label to state "remodeling," what would you expect to see? Pathology? What would the label have to show in order for it to be able to claim a remodeling effect?
DR. BORER: Well, I think, you know, there would be a few characteristics you'd like to see met. I mean, it would be nice to see a clear reduction in left ventricular size, including mass, which was not clear-cut here. It would have been nice, and I think Illeana raised this point. It would have been nice, although it wasn't done and that's okay, to have seen some histological data from the hearts that were explanted. That could have been helpful. Maybe it wouldn't have been.
There are some cellular and molecular markers that would have been consistent with a potentially beneficial myocardial alteration. So, I mean, there are a variety of characteristics. Several were measured here, left ventricular size, left ventricular mass. They just weren't -- as Clyde said, they didn't really go -- they weren't consistently in the right direction by a magnitude that would cause one to be able to draw firm conclusions that the processes we're looking for actually are happening in a meaningful way.
CHAIRPERSON MAISEL: Bram.
DR. ZUCKERMAN: Dr. Borer, could I ask you to provide a little bit more clarification for here we're dealing with a device that seeks the claim of ventricular remodeling, and by design it's going to reduce volumes if the bench testing is predictive? So could you help us?
Where would you place more emphasis, on all of those variables? How would you separate device remodeling from drug remodeling, especially given the fact that right now for devices, this concept of remodeling may be a surrogate in that we don't have a lot of evidence that says remodeling translates into reduction in mortality or I should say it's not a proven surrogate yet.
DR. BORER: Yeah, I mean, I think there are several comments I would make. First of all, I agree 100 percent with what you said. It's not a surrogate. I mean, or rather, it's not an acceptable surrogate, even if we could define it. I don't think that there are sufficient data so that we could look at it as a surrogate for clinical benefit. I think there are very few such things right at the present time.
Number two, I'm not really the proper one to answer the question you've asked. We have two experts in the room here who are far more erudite about all of this than I am. One is Illeana, and the other is Doug Mann, and they could answer your question better than I.
However, having said that, I think that you have to make a distinction between a device that limits physically the capacity of the left ventricle to expand and the inference that that device has caused a fundamental structural change in the myocardium, and I'm not suggesting that that doesn't happen. It may well happen, and there was some animal data that suggested it did, but I just don't think that those data were consistently supported by the clinical data, and that's why I have this concern.
CHAIRPERSON MAISEL: John.
DR. WHITE: Well, I can see Jeff's concern, and I would just caution about the use of remodeling, which is charged and makes physicians think of certain very specific concerns. I would be more inclined if it came to that to talk about reduction in left ventricular volume, reduction if there was left ventricular mass. If there wasn't, that should be stated and to state what its effects on ejection fraction are, and then let the physician and the patient -- and when the physician explains to the patient what those mean.
But when you talk about remodeling, it's a very charged term, and we all like to -- there's electrical remodeling and mechanical remodeling. There's all sorts of things, and you know, it's very nice for meetings, but it's a charged term, and if you will, it may be even bordering on advertising as opposed to being substantive.
So I would stick to the physiologic markers that they're actually discussing.
CHAIRPERSON MAISEL: George.
DR. NETROVEC: To maybe come back to Dr. Zuckerman's comment, it seems to me that there's a small change. We've argued about how significant it is. I guess I would feel better that the device was really having a long-term impact if you showed at three months it looked X and at nine months it looked better, but that there was actually some progressive, persistent change in volume contraction and so forth rather than kind of just one absolute change.
Then I would feel like the device was doing something just like a drug would do over time.
DR. CALIFF: They did show that, but with, you know, the caveat of missing data even for the biological measures. There were multiple measures.
CHAIRPERSON MAISEL: Jeff.
DR. BORER: Well, I don't want to put words in George's mouth, but if I can, I agree with them.
DR. BORER: It would have been more compelling; the data would have been more compelling than they are if the patients who had the device implanted, the reduction in left ventricular size, for example, among the patients who have the device implanted was greater than or was progressively greater than the reduction in left ventricular size in the control group, whatever that control was.
The two curves seem to fall in parallel. I could understand in the mitral valve replacement or repair group why they might both fall, and I could even understand how they might fall in the control group treated optimally with drugs, but the fact that they fell in parallel after an initial reduction in size after the device was put in makes one wonder about what the device was doing besides merely putting a cap on the capacity of the left ventricle to dilate.
You know, I don't know that it's not doing something more than that, but I don't think we were shown enough to allow us to firmly conclude that it is doing something more than that.
CHAIRPERSON MAISEL: Why don't we move on to the next question, 11(b)?
MS. WOOD: With reference to the proposed labeling, do the proposed indications for use adequately define the patient population studied and for which the device will be marketed?
CHAIRPERSON MAISEL: I think, in short, the indications in the labeling are not detailed enough and should include many of the indications in exclusions from the clinical trial.
Are there any other specific comments that we haven't already touched on that someone would like to make?
Okay. Let's go on to C.
DR. ZUCKERMAN: Well, can we just stop there for a moment? Usually in our labeling we say simpler is better, and then in our clinical trial section we fully describe the clinical trial, which is -- and there's no clinical trial section in this current label. So we have some work to do.
But with the caveat that we'd like to identify the patient population and what the device is intended to do, is there anything that people would add to the indications statement, which is in the patient labeling on page 1?
For example, Dr. Borer, you've made the point that this is a selected patient population where one tries to reduce the up front surgical risk. There is an extremely small proportion of Class IV patients in this trial, four percent. Does there need to be something short and sweet that better describes who can be selected?
DR. BORER: Yes. Well, I think you have to say something about the severity of the heart failure of the patients you're treating. Fundamentally these were Class III and some Class IV patients who were being treated, except for that small subset with MR, with functional Class II.
It makes it a little hard, but my understanding here is that we were dealing with people with very severe heart failure, not so severe that it seemed highly unlikely they could ever go to the operating room.
So I think you do have to describe the population a little bit, with a little bit more detail. Very severe, but still operable. I don't know how you say that in the label, and that may be too hard to do here on a committee, but I think you do have to say something more.
CHAIRPERSON MAISEL: I mean, another way to do that might be to describe optimal heart failure medical management, you know, a statement such as where optimal medical management is defined as, you know, treatment with an ACE inhibitor or ARB, beta blocker, diuretic, et cetera.
DR. ZUCKERMAN: Right. I don't want to beat this to death, but I'm going back to page 22 of the sponsor's panel pack baseline characteristics, and even though these are very sick patients, you know, when you look at the site assessed, it's actually New York Heart Class IV. It's only 4.6 percent. The majority are II and III, as show in the baseline demographics.
In other words, similar to certain other medical device therapies, is there enough data for a Class IV patient?
CHAIRPERSON MAISEL: Jeff.
DR. BORER: Yes. My understanding, and you know, you have to correct me if I've got this wrong, was that except for protocol violators, the only patients in functional Class II in this trial were patients who were going to have a mitral valve procedure. So that's a separate population, and I think that that population has to be described fairly specifically within the label.
But for patients who don't have mitral regurgitation, all of them were in functional Class III, and some were in functional Class IV. I would be very much happier if we have more information about patients in functional Class IV, but there were some limitations on who was acceptable to take to the operating room if you weren't going to do a transplant.
And so, you know, I don't mean this as a way of evading answering your question. You asked if there is enough information about Class IV, and I would like more, but at least the information we have about people who don't meet the exclusion criteria, that is, they cannot go to the operating room, it doesn't appear that the people in functional Class IV did differently than the people in functional Class III, and so I would accept that as a basis for -- I would accept that group as being acceptable to receive the device.
But I do think we have to define fairly specifically up front who shouldn't get it, too, and those are the people who were defined in this protocol as being too sick to go to the operating room.
CHAIRPERSON MAISEL: Rob.
DR. CALIFF: Yeah. I mean I summarized my feelings on this, Bram, as I said before. This is such a small study with so few endpoints that I think a label should reflect the astounding amount of uncertainty that exists around any decision made about the use of this device and should get people fairly simple summaries of the risks and benefits that were observed in the trial, should include the exclusion criteria and the inclusion criteria that were used in the trial.
And I would just comment that if ever there was a case that this got out for coverage under protocol, I'm assuming that most of these, almost all of the patients eligible for this device would be either Medicaid or Medicare covered. If ever there was a case for coverage under protocol, whoever is paying for these devices if it's not the patient is going to have a great interest.
PARTICIPANT: Dr. Maisel.
CHAIRPERSON MAISEL: Yes, I will preempt Mr. Morton's comment just to say that the panel is not here to discuss payment or cost effectiveness.
DR. CALIFF: I didn't mention cost effectiveness. I just said --
CHAIRPERSON MAISEL: Okay. Or cost.
DR. ZUCKERMAN: But I think what's useful ‑-
DR. CALIFF: Conditions of use.
DR. ZUCKERMAN: -- in the two recent comments, as well as Dr. Borer's schema comments, was that in a post approval study to concentrate on these particular patient populations where the data looks encouraging, but we don't have an overwhelming amount of data might be useful, and we'll get to those questions.
CHAIRPERSON MAISEL: Burning comments. Gene.
DR. BLACKSTONE: Just remember the comment about no need for coronary bypass grafting.
CHAIRPERSON MAISEL: Of course. Sharon.
DR. NORMAND: I'm just perplexed about the conversation to do with the New York Heart Association, New York Heart measure, just because (a) I thought there was an agreement. I may have misunderstood, but that the site assessed one was biased. So we can't use that, and so I don't know why we would use that information if we've already said it was biased.
And moreover, the core assessed is only based on, you know, 126 people. So I'm not sure how we used this information, given the very long discussion we've had about this.
DR. CALIFF: But I think we have agreement, Bill, that the entry criteria of the trial would be what goes in the --
CHAIRPERSON MAISEL: Right. I think the challenge would be to construct a brief, accurate label that reflects all the comments of the panel, but specifically the inclusion criteria and defining the patient population for whom the device is intended.
DR. BORER: But just with regard to Sharon's point because it's an important one, I think the concern about bias was at the other end. After you had treated the patient and you knew what you had treated them with, that might affect how you reclassified them in the post treatment analysis. The entry analysis was just to get them into the study, and I don't think there was great concern about bias there.
CHAIRPERSON MAISEL: Why don't we move on?
DR. FERGUSON: I had a comment.
CHAIRPERSON MAISEL: Tom
DR. FERGUSON: The baseline characteristics only include four patients that are Class IV, and I'm not sure how we can make a comment on that. Why not just say Class III? Bram?
DR. ZUCKERMAN: Because it gets back to Dr. Norman's point. That was the site assessed. The core lab assessed had a higher percentage. So there are lots of question marks. It's certainly something we can look at in a little bit more detail with the sponsor.
But I think the general nature of the discussion has been these are sick patients that we can describe a bit better, but not to make such a critical line between three and four.
CHAIRPERSON MAISEL: I think we're having the trouble drawing the line between MVR and no MVR, and they were prespecified, let alone making a line between the different New York Heart Association classes.
Let's move on to the next question, Geretta.
MS. WOOD: Are there any additional warnings, precautions, or contraindications that you think should be included in the labeling to assist practitioners?
CHAIRPERSON MAISEL: The one comment I would like to make is that I didn't feel the label accurately reflected the issues related to whether or not the patient should receive bypass surgery if they're not undergoing MVR.
I think obviously the sponsor and the investigators have a lot of experience and showed a remarkable reduction in operative mortality throughout the trial, and somehow that needs to be reflected both in the label and in a training program.
Any other labeling comments?
CHAIRPERSON MAISEL: Okay. Let's move on.
MS. WOOD: Training. Selection of patients to receive heart failure therapy with and surgical replacement of the CorCap CSD requires the referring cardiologist and the surgeon to assess multiple factors, some of which are unique to the device. The sponsor has not included a physician training program as part of the proposed labeling.
Please comment on whether a physician training program is necessary for this device system. If your answer is yes, please comment regarding the specific elements that should be included in the training program.
CHAIRPERSON MAISEL: Who would like to tackle this one? Training program.
DR. CALIFF: I mean, it's hard to imagine not having a training program. I'd like to see the consent. It should say, "I've never done one of these things before. Why don't you let me try it out on you?"
CHAIRPERSON MAISEL: I mean it seems like a training program would have to include heart failure specialists, cardiac surgeons, involve, you know, observing or participating in surgery, maybe an animal models.
DR. FERGUSON: Didn't we see a slide that showed this?
CHAIRPERSON MAISEL: So we saw a slide from the sponsor that showed an investigator training program. There was nothing in the panel packet about a training program.
DR. SOMBERG: I think it's clear cut there has to be a training program for the surgeons involved in this study. I took it as more generic, the question would there be a training program for referring doctors, and I don't really think that would be indicated. I think you just have to have a clear specifications here, but certainly, I mean, everything that was said in the sponsor's showing how we mitigated it is presupposed on a training program. Otherwise all of that information would be lost.
CHAIRPERSON MAISEL: Tom.
DR. VASSILIADES: Well, I would agree with that. I think perhaps maybe anesthesiologists should be included in the operative management as a training program.
But I'm also wondering whether or not the echocardiographer needs to be involved or not in terms of training for that, whether it makes any difference afterwards in terms of getting a good study.
CHAIRPERSON MAISEL: I think it becomes a slippery slope if we start including -- in other words, we could end up with a system where you only an interpret these echoes if you're, quote, trained, and obviously I don't think that would be beneficial, but undoubtedly the surgeons should be included, and I think the heart failure physicians and anesthesiologists make sense.
DR. CALIFF: I mean, I think the discussion has been good on this. There's a lot in it for the sponsor obviously to have a good training program because outcomes are being measured so much. I think you had a good description.
CHAIRPERSON MAISEL: Let's move on to the post approval study.
MS. WOOD: The sponsor has proposed a post approval study to include up to 348 patients who will be followed for five years, the intent being to follow the long-term performance of the CorCap CSC in the general patient population.
Given the concerns expressed above, please comment as to the suitability of the proposed post approval study. If additional elements should be included, what are they?
CHAIRPERSON MAISEL: Can you read the next one, too, please?
MS. WOOD: Please comment on an appropriate comparison group to evaluate the safety and effectiveness of the CorCap CSD against based on the results of surveillance during a five-year period.
CHAIRPERSON MAISEL: Sharon.
DR. NORMAND: In looking at what was in our panel packets, I had some comments, first of all about the extended follow-up phase. I think the sponsor proposed five years of annual follow-up including clinical assessment, and there they say site assessment of New York Heart Association functional classification, and I would want to warn them. Wouldn't they want to have the same -- you might have ascertainment bias if you're going back and forth between a core labs assessment and all of the discussion.
So I would not use a site assessment of that particular endpoint. If you wanted to combine the data over all years, I think that's a pretty basic notion that you should introduce this as a bias.
So I had a comment about that. I did have a question. I don't know if anybody has had a chance to look at it, and that is actually the very thoughtful plan to look at extended follow-up and then a continued access protocol, and in that protocol, as I mentioned earlier, there seems to be some differences in the inclusion criteria, and not being the clinician, I have no idea whether or not some of the differences between the pivotal trial and the continued access protocol are significant.
So, for example, no hepatic function required, no pulmonary function required, no serum creatinine requirement. The clinicians just need to comment whether or not changing the inclusion criteria that I just described is worrisome. I don't know.
CHAIRPERSON MAISEL: John.
DR. SOMBERG: I think we're late in the afternoon. So I just want to make the comment that I think this is a reasonable trial if you thought that the database showed efficacy and safety and there was a favorable balance. If you don't feel that, you would advocate, you know, another trial for another attempt at a PMA using different endpoints and different type of follow-up.
So, you know, I'm of the latter opinion. So, therefore, I don't think this is inappropriate to talk about, but if you do feel that efficacy and safety has been demonstrated, this would be appropriate. I wou