P R O C E E D I N G S

Time: 8:04 a.m.

ACTING CHAIRMAN MAISEL: Good morning. I would like to call to order this meeting of the Circulatory System Devices Panel. Today's topic is discussion of a premarket notification for the Cardica Inc. Cardica PAS-Port System, K030434.

I'd like to have Geretta read the conflict of interest statement.

MS. WOOD: Before I read the conflict of interest statement, I would like to remind the panel members that in your folders you will find a survey this morning. I would appreciate it if you would take it with you and fill it out and return it with your reimbursement vouchers. Thank you.

The following announcement addresses conflict of interest issues associated with this meeting, and is made a part of the record so 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 such outweighs the potential conflict of interest involved, is in the best interest of the government.

Therefore, waivers have been granted for Doctors Mitchell Krucoff, John Somberg, and Clyde Yancy for their interests in firms that could potentially be affected by the panel's recommendations.

Dr. Krucoff's waiver involves two consulting interests on unrelated matters for which he receives an annual fee of less than $10,001, with a competing firm, and with a firm that receives funds from a competing firm.

Dr. Somberg's waiver involves a mutual fund with holdings in competing firms. Because of the type of investment, values of the holdings are not known.

Dr. Yancy's waiver involves consulting services for a competing firm on unrelated products for which his fees are less than $10,001 annually.

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-5 of the Parklawn Building.

We would like to note for the record that the agency took into consideration other matters regarding Doctors Krucoff, Yancy, Jeffrey Borer, and Thomas Ferguson. These panelists reported past or current interests involving firms at issues, but in matters that are not related to today's agenda. The agency has determined, therefore, that they may participate fully in these deliberations.

Dr. Eugene Blackstone reported his institution's interest with a firm at issue. Since Dr. Blackstone was not personally involved and there is no continued financial interest, the agency has determined that he may participate fully in the deliberations.

In the event that the discussions involve any other product or firms not already on the agenda for which an FDA participate 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.

ACTING CHAIRMAN MAISEL: Thank you. I would like to have the panel members introduce themselves.

I am William Maisel, a cardiologist at Brigham and Women's Hospital. Dr. Zuckerman?

DR. ZUCKERMAN: Bram Zuckerman, Director, FDA Division of Cardiovascular Devices.

DR. HIRSHFELD: John Hirshfeld. I am an interventional cardiologist at the University of Pennsylvania.

DR. YANCY: Clyde Yancy, heart failure/heart transplant cardiologist at UT Southwestern in Dallas, Texas.

DR. KRUCOFF: Mitch Krucoff, interventional cardiology at Duke University and the Director of the Cardiovascular Devices Unit at the Duke Clinical Research Institute.

DR. NORMAND: Hi. I am Sharon-Lise Normand. I am a Professor of Health Care Policy and Biostatistics at Harvard Medical School and Harvard School of Public Health.

DR. BLACKSTONE: Eugene Blackstone, head of clinical research, thoracic and cardiovascular surgery at the Cleveland Clinic.

DR. BORER: Jeff Borer. I am a cardiologist at Weill Medical College of Cornell University in New York.

DR. FERGUSON: Tom Ferguson, cardiothoracic surgeon emeritus, Washington University, St. Louis.

MS. WOOD: Geretta Wood, Executive Secretary.

DR. SOMBERG: John Somberg, Professor of Medicine and Pharmacology, Rush University, Chicago.

DR. KATO: Norman Kato, cardiothoracic surgery, private practice, Los Angeles, California.

DR. PAGE: Rick Page, head of cardiology, University of Washington, Seattle.

DR. JEEVANANDAM: Val Jeevanandam. I am Chief of Cardiothoracic Surgery, University of Chicago.

MS. MOORE: Deborah Moore, Vice President of Clinical and Regulatory for Proxima Therapeutics.

ACTING CHAIRMAN MAISEL: Thank you. At this point, I would like to invite the FDA to give their brief presentations. The first one will be Dr. Susan Gardner, who is Director of the Office of Surveillance and Biometrics, who will be talking on condition of approval studies.

DR. GARDNER: Thank you. Good morning. I am glad to be your kickoff speaker this morning. I am going to spend a few minutes telling you about a measure of programmatic change in CDRH, and the operational piece of that is the move of the Condition of Approval Studies Program to my office, which is the Office of Surveillance and Biometrics.

Briefly, if you are familiar with CDRH, you think of OSB as being the seat of the post-market piece, but in fact we play a major role in premarket review also, by virtue of the fact that the statisticians are in our office, and you hear from the statisticians at all of your panel meetings.

We also have the epidemiologists in our office, and as you will hear, they are going to start playing a larger role in the premarket piece.

Also in the Office of Surveillance and Biometrics, we do signal detection by use of the various tools we have to monitor the post-market arena. That includes all the medical device reports that come into our office, the MedSun program and other tools that we use.

We are also responsible for characterization of the various risks by virtue of reviewing post-market data, and we coordinate a center of response to post-market issues, if that is sort of a risk communication response. In addition to that, we are responsible for interpretation of the medical device reporting regulation.

A regulatory authority for condition of approval is in 21 CFR 814.82, which says that post-market requirements can include continuing evaluation and periodic reporting on the safety, effectiveness, and reliability of the device for its intended use.

A sort of kickoff event to this major change started a couple of years ago when CDRH took a hard look at our condition of approvals program. The initial look was actually to go and evaluate the quality of the studies that we have been ordering for condition of approval, but as we started to walk this evaluation, we found a much bigger problem.

We looked at all the studies, PMA studies, that -- I'm sorry, all the PMAs that were approved between 1998 and the year 2000, and we found 127 PMAs. Of those, 45 had orders for condition of approval studies.

Unfortunately, we were really unable to do a very thorough evaluation of how those studies were done, because we found that we really didn't have a standardized tracking process in the Center, and we essentially couldn't find the studies -- all the studies.

We turned to our lead reviewers and found, as one might expect, that during the last couple of years many of them had moved on to different positions. So they also were not a way that we could track these studies. So that was one big message.

We also found that we -- we think, although again the tracking was poor, that we hadn't received results for about 22 percent of the studies, and some of them probably had not been started at all. So it was not a very good message, and it was certainly a wake-up call to us to try and improve this program.

So the strategy for change: First of all, what is our goal? I think the goal is somewhat easy. We want to get good post-market information as the device enters the market, to continue to assure the safety and effectiveness of the device as it moves mostly from clinical trials into real world use.

That should help us better characterize the risk/benefit profile that we looked at in the premarket trials and, obviously, add to our ability to make good scientific decisions.

The strategy for change: As I say, the big operational piece was to move the program from ODE to OSB, and that official move was on January 1st, although this has been preceded by a pilot that has gone on now for almost three years, working with one of the divisions to sort of iron out some of these procedures.

Again, another obvious first step was to develop an automated tracking system so we would be able to find these studies, but this also will allow us to acknowledge to industry that we received the reports and also to follow up with them if the reports have not been received.

The second big change is that we have added an epidemiologist to the PMA teams when we anticipate that there is going to be a condition of approval study. The role of the epidemiologist on the team is to develop the post-marketing plan during the premarket review, and this is sort of to step back and take a good look at all the post-marketing tools that we have in place and, again, think about a coordinated plan to monitor the product once it goes into the market.

The epidemiologist will also have a lead in developing a well formulated post-market question, and that is extremely important. We really want to focus on thinking about what we are asking these companies to do in the post-market period.

They will have the lead in the design of the condition of approval study protocol and the lead in the evaluation of the study progress and the results after approval. I want to emphasize that they will continue to work in the PMA team and have close contact with their colleagues in ODE and other parts of the office.

So then the question is: Given the fact that it looks like we have a fairly poor track record in getting these studies done, what do we think now is going to be the motivation for compliance?

Well, the first thing is, if we do a good job, I think, in formulating our post-market hypotheses and asking the right questions, I think that industry is going to be more motivated to carry out the studies, because they are going to be getting information that is going to be important to us.

Second of all, I think the acknowledgment and feedback to industry when they do perform their studies is, obviously, an important piece of the plan. It is not particularly rewarding to do work and then to have it in and have no response, but we are going to be very involved in giving feedback and interacting on what we are finding on these post-market studies.

Third, we are going to be posting the status of these studies on the CDRH website. Hopefully, this will be a positive thing for the folks that are doing it well, but if they are not doing it well, it won't be positive.

Also, we have the ability in CDRH to mandate post-market studies under Section 522. This also allows us to apply penalties for studies when they are not done by leveling civil money penalties and misbranding.

So if we have formulated a good post-market question when the product moves onto the market, if industry does not comply, then being able to implement 522 will be possible.

So what is the impact on the Advisory Panel? The first thing is that during the approval process -- and this will vary a lot depending on what product we are looking at. But we want to attempt to lay out some of the important post-approval public health questions and, certainly, have you help us in giving us feedback on formulating some of those questions, and at times the possible approach for panel consideration.

I think some of the reasons that we do post-market studies are fairly obvious. We obviously look for rare events, events that have not been well characterized premarket. For implants, we want to do long term follow-up. We worry about user error as these products move from the clinical trials into the community practice.

So as those issues become apparent as we have discussions about products, discussion about those, and again helping us to sort of characterize this and think about a post-market approach, if that is appropriate. It is going to be very helpful to us as we move into the condition of approval protocol considerations.

Then the last piece but, I think, an extremely important piece is we want to be able to give feedback to you, and either we will do that as FDA or we will ask industry to do that. So after you have approved a product or not approved a product, but after a product goes to market and we have ordered a condition of approval study, we think that coming back to you a year or two years or whatever after that and giving you feedback on how the study is going, what we've found out, is really an important piece of this.

So that's it, if you have any questions. Thanks.

ACTING CHAIRMAN MAISEL: Thank you, Susan. Our next speaker this morning will be Dr. David Buckles, also from FDA. He is the Branch Chief of the Peripheral Vascular Devices Branch, and will be giving us an update.

DR. BUCKLES: Good morning. I would like to give you an update on a panel meeting that we had this past January.

MS. WOOD: Pull your microphone just a little bit closer. Thank you.

DR. BUCKLES: Okay, I'll try to use my adult voice.

At the panel meeting in January, we discussed the Gore Thoracic Aortic Aneurysm Stent-Graft, which is commonly known as a Gore TAG Device. For this device, the original PMA was submitted to FDA on October 4, 2004. The PMA was granted expedited review status, and the device was also the first of its kind. So we had a somewhat compressed schedule in reviewing the device and getting the device to panel.

The Cardiovascular Devices Panel met on January 13th and voted eight to two to recommend approval of the device. The application was, in fact, approved on March 23rd of this year.

The indication that was approved for the device: It is intended for endovascular repair of aneurysms of the descending thoracic aorta. This indication was modified slightly from the original application in that we added some anatomic specifications to the indication. This is consistent with the advice we received from the panel.

This is a good topic to follow onto Susan Gardner's presentation, because this device did, in fact, have a condition of approval study. The panel recommended a post-approval study with an appropriate number of patients out to five years. The endpoint should include mortality, paraplegia, aneurysm rupture, and the endpoint should -- or recommended by the panel to be consistent with IDE studies.

The panel also recommended a training program, the details of which should be worked out between Gore and the FDA; and the panel's final recommendation was that the labeling should include inclusion and exclusion criteria as well as the anatomic criteria from ongoing studies.

The anatomic criteria, as I just mentioned, were included in the indications for use that were approved for the device.

The FDA approval order included some conditions of approval. The post-approval study was required to be at least 150 patients out to five years, consistent with the panel's recommendation. Endpoints included all cause mortality, aneurysm related mortality, morbidity and device related events that were designed to be consistent with the IDE studies.

In addition, Gore has developed a training program which we discussed at the panel meeting. One of the objectives of the post-approval study is to assess the effectiveness of the training program.

Other conditions of approval included that the company follow all IDE study subjects as well as all post-approval study subjects out to five years, as I mentioned. The company will also provide a clinical update to physician users at least annually, including information from the five-year follow-up and a summary of all available patient data.

So in summary, the final disposition of the PMA application as consistent with the panel recommendation to approve the application. The panel voted to recommend approval, and the device was in fact approved.

There were three conditions of approval that the panel recommended, and all of the conditions were included in the approval order, either in the indications for use or in the conditions of approval that were part of the approval order.

Consistent with the strategy for change that Susan Gardner elucidated, the post-approval study provides for continuing evaluation and periodic reporting on the safety, effectiveness and the reliability of the device for its intended use, as well as providing information on the effectiveness of the company's training program.

Thank you.

ACTING CHAIRMAN MAISEL: Thank you, David. At this point, I would like to move on to the open public hearing. Both the Food and Drug Administration and the public believe in a transparent process for information gathering and decision making. To ensure such transparency at 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 known, its direct competitors. For example, this financial information may include the sponsor's payment of your travel, lodging or other expenses in connection with your attendance at this meeting.

Likewise, FDA encourages you at the beginning of your statement to advise the committee if you do not have any 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.

The first speaker this morning at the open public session is Terry Vanden Hoek.

Any other people wish to address the panel this morning in the open public hearing? Seeing none, we will close the open public hearing portion of the meeting.

We will move on to the sponsor's presentation.

DR. HAUSEN: Panel members, ladies and gentlemen, members of the FDA review team. My name is Bernard Hausen, and I am the President of Cardica. I would like to thank everyone for taking the time to join us here today to review the application from Cardica on the PAS-Port Proximal anastomosis System.

We are aware that the abundance in material that was submitted as part of the panel package has put a significant burden on you over the past few weeks, and we sincerely thank you for providing this service.

Cardica will be presenting information on the clinical need in some detail in order for you to answer the questions the FDA as phrased, give you a product information description, give you data from our clinical trials, a brief overview of approved products, and end our presentation with the clinical perspective.

This is the list of presenters. With me today are Dr. Harringer. He is the Chairman of the Department of Cardiothoracic and Vascular Surgery in Braunschweig, Germany. He was an investigator in our first trial and principal investigator in the second trial.

Also Professor Dan Bloch, who is a full Professor of Biostatistics at Stanford University, and serves as Cardica's Chief Statistical Advisor. And many of you know Dr. Michael Mack, who is the Director of the Cardiothoracic Surgery Associates of North Texas and a very active and well known practicing cardiac surgeon. Dr. Mack serves on many committees of the three main cardiac surgery societies in the world, EACTS, STS and AATS.

This is Cardica's world headquarters. We were founded in 1997, and this is located in Redwood City, which is about 30 miles south of San Francisco.

The company's position is to develop proprietary, automated, anastomotic systems. While Cardica was one of the first companies to engage in research and development in this challenging area, we were one of the last to enter clinical evaluation. Actually, the PAS-Port product has been designed before the symmetry device.

Our perspective of the products that would be needed to make this enabling technology was quite different than that of other companies. With my background as a cardiac surgeon, I stress that any technology we develop would have to be in agreement with the basic principles of surgery, and particularly those pertaining to performing vascular anastomosis.

In the later presentation I will go into more detail, but I do want to stress that the company put forth an enormous effort and substantial resources to address these needs, adhere to these principles in the development of all its products, which may explain some of the good results we have obtained to date.

The company's vision is provide technology that will expedite procedures and reduce associated morbidity and mortality. Currently, Cardica has two approved products in Europe. The PAS-Port product has been approved more than two years ago, the C-Port or Distal Anastomosis System more than a year ago.

The PAS-Port product received approval in Japan more than a year ago, within this year has found tremendous acceptance from Japanese surgeons. Today 25 percent of all vein grafts used in cardiac surgery in Japan are used in conjunction with this product. Beating heart surgery is used very commonly in Japan, and our customers are very devoted to reducing neurologic injury.

Both products have been submitted for clearance to the FDA.

This gory slide -- excuse for the early time -- allows me to describe what we believe is the primary clinical need for a proximal anastomosis system such as the PAS-Port, a system which allows the surgeon to perform a proximal anastomosis without having to clamp.

What you see here is the intimal surface of a diseased human aorta from a 60-year-old female. These are ulcerating atheromatas that can be severed from the aortic wall and released into the blood stream through manipulation of the aorta, thus potentially causing cerebral emboli.

Neurological injury compromises the most serious complications following CABG surgery. The most severe form are strokes, which occur at one to three percent of our patients and are responsible for 21 percent of our operative mortality. Strokes add significantly to intensive care and length of hospital stay. Besides the terrible suffering of these patients, there is an enormous cost to our society.

There are a number of risk factors responsible for strokes, many of which you know and of which the extent of proximal aortic atherosclerosis, as I have shown you in the first slide of this presentation, comprises the absolutely highest risk.

This slide shows a linear association between the incidence of ascending aortic atherosclerosis and patient age with approximately 60 percent of patients 70 years in age presenting a significant degree of atherosclerosis. In light of the fact that patients are getting older and older, this is becoming a more imminent problem.

The mechanism of cerebral injury as a result of proximal aortic atherosclerosis during CABG surgery is described best in this chart from Barbut. The fraction of total embolic signals in the carotid artery is displayed as a function of various manipulations of the aorta, with the most significant release of emboli noticed during cross-clamp or side-biting clamp release.

Alternative strategies are -- Surgeons have been faced with the dilemma of stroke for many years, and have started to develop alternative strategies to reduce the amount of athero-embolization. Today we would like you to consider an additional treatment modality to potentially reduce the incidence of athero-embolization.

For those present here today who are not as familiar with coronary bypass surgery, a conduit such as the patient's own saphenous vein is harvested and then attached to the aorta, which is termed the proximal anastomosis, and to the coronary artery, which is termed the distal anastomosis, thereby bypassing a significant lesion in a coronary artery.

The PAS-Port produce we are discussing today replaces a hand-sewn anastomosis of the proximal anastomosis.

The PAS-Port produce is an integrated device that performs an aortotomy and implant deployment in one action by turning the knob at the end of the tool. This can be performed quickly, and the tool has been designed to be easy and intuitive to use.

One of the most important features is that this tool allows the completion of the proximal anastomosis without placement of a clamp, a process that, as I have shown before, carries with it the risk of embolization. The product has been designed to minimize metal exposure to blood and to provide a firm attachment of the graft to the target vessel.

The package consists of a tray that serves as a loading platform to facilitate conduit loading, delivery tool, cartridge, an implant, a poke-through tool, and a pull-through tool.

This animation shows how the system is deployed. After conduit loading, the tool is placed on the aorta without the need for any preparation of the target vessel, and with a simple rotation of the knob at the end of the tool, first the aortotomy is created and then the wall and then the implant inserted through the aortotomy and deployed and secured. All this occurs within a few seconds without the need for any placement of clamps.

This is a video showing the inside of the human aorta that is hooked up to a water pressure system with an endoscope on the other ends, and you see in the picture and picture view the same deployment videoed from the outside.

For deployment, the surgeon places the tool on the aorta, rotates the knob, which creates the aortotomy, inserts the implant, seats the implant, and then the tool is pulled off.

As you know, there are many factors that impact graft patency. These include the technical quality of the anastomosis, comobidities such as the presence of diabetes or hypolipidemia, vein graft in an arterial circulation. Also, graft patency is affected by graft quality and target vessel selection. It is affected by the amount of trauma to the graft during harvest and loading, dependent on the amount of blood exposed on the endothelial surface, the size of the effective orifice of the conduit at the anastomotic site, and finally graft kinking and long term anticoagulation have been proposed as important variables that impact graft patency.

In the following slides, I will focus on factors that are specific to the use of anastomotic technology. And because the symmetry device was the subject of so much discussion at the last panel and will be addressed in the FDA's presentation, we felt it was important to show you differences between our device and the symmetry device.

This video shows the amount of vein endothelial trauma during loading deployment as the comparison between symmetry device and PAS-Port. This is the symmetry device where the conduit is loaded over a transfer sheet, exposing the entire endothelial surface of the conduit to a metallic surface.

After the conduit is attached to the implant and deployed in the aortic wall, a rod used in the deployment is again pulled through the lumen of the graft, potentially damaging the endothelial surface of the conduit.

In contrast, in the PAS-Port system the section of the conduit pulled through the implant is cut off and discarded. You see here the conduit pulled through the cartridge, exiting at the implant end. The section we just used to pull the graft through is cut off. The surgeon everts the graft over a reduced diameter of the implant, and a poke-through tool with a little membrane pushes the graft onto the implant, thereby allowing the tines to pierce.

This section of the graft that was touched by forceps will end up in the aortic wall, will not be in touch with the endothelial with blood. I think this is a very important fact.

This slide depicts a PAS-Port implant next to a symmetry implant on the left, both deployed in the same human cadaver aorta, both using segments of the same human vein. The amount of blood exposed non-endothelial surface with the symmetry device is approximately 350 percent more than the PAS-Port implant.

More importantly, while there is no metal inside the orifice of the PAS-Port Device, there is significant amount of metal protruding into the lumen of the graft with the symmetry implant. The amount of blood exposed non-endothelial surface of the PAS-Port implants are similar to those of hand-sewn anastomosis.

What we see here is the difference in effective orifice between the two products for the same conduit where especially for small vein grafts, the orifice of the PAS-Port implant is approximately 150 percent larger than the symmetry orifice. This is due to the fact that in the symmetry device, much of the potential orifice of the graft is used up as a flange to connect the graft to the device, shown here. This is graft lumen that gets lost as effective orifice after the deployment has been completed.

The simple eversion of the conduit over the PAS-Port implant avoids this effective loss of graft orifice.

As a last major difference between the two products, I would like to spend a few minutes on the potential for graft kinking. This has been an issue that was raised in the FDA -- by the FDA.

Design of all proximal anastomotic connectors theoretically allow for the creation of a right angle take-off of the graft from the target vessel with the potential of graft kinking at the anastomotic site.

One of the design goals of the PAS-Port implant was to reduce the probability of kinking by minimizing profile height of the implant. This image shows a comparison of profile height of the symmetry device and the PAS-Port implant.

You see that the profile height, which is basically going from here to the top of what you saw protruding into the lumen, is approximately twice as high as with the PAS-Port. For that length, the graft is constrained and can only go in one direction. It cannot take off at an angle.

The video shows here two PAS-Port devices -- one, two -- next to a hand-sewn anastomosis, and there is very little difference in take-off angle between the hand-sewn and the proximal anastomosis developed by the PAS-Port.

Based on early experiences from symmetry, we instructed our users through the instructions for use where to place proximal anastomosis, depending on which area of the heart is planned to revascularized. For grafts going to the left side, we recommend the left or concave surface of the ascending aorta, for grafts to the right, on the ascending part.

So to summarize, the key differences between the two devices that could explain the differences in outcome, as we will show you later, we believe, are the amount of endothelial trauma experienced with the PAS-Port device is minimal, the amount of blood exposed non-endothelial tissue is minimal. The effective orifice area is large, and any incidence of graft kinking due to low profile height of the implant is low, and the occurrence of any negative late sequelae negligible.

One last comment on the sequence of performing the proximal or distal anastomosis first, as this is a question again raised by the FDA. There is no necessarily correct order for performing a proximal or distal anastomosis. Most surgeons do perform the distal anastomosis first, but with the event of beating heart surgery, there are a number of advantages in doing the proximal anastomosis first.

One of the most compelling arguments is that, if you do your proximal anastomosis first, once you have completed the anastomosis and take the bulldog clamp off the graft, you are immediately reperfusing your target vessel, thus reducing the time of temporary ischemia.

The grafts can be distended with physiologic pressure and blood, both of which will optimize endothelial viability. And when patients are operated on pump, repeated cardioplegia through the aorta will provide cardio-protection to grafts already anastomosed.

To conclude this section of my talk, the advantages of the PAS-Port system are that this system allows for rapid deployment, and it is these time savings that will affect total operating times. The system avoids clamping of the aorta, by inference may also affect neurological outcome.

Because this is an automated system, much of the variability between surgeons, intra- and inter-individual variability is eliminated. The design provides surgeons the greater choice in anastomotic sites, making the surgeon less dependent on surgical access.

Performing the proximal anastomosis first has inherent advantages in beating heart surgery, and the PAS-Port Device design addresses, we believe, many of the potential shortcomings of other anastomotic technology.

Before we go into the details of the clinical study we have performed, please allow me to recapture some of the major statements you as the panel had made last year at the Anastomotic Technology Panel.

First of all, the statement was: It is best for patients to undergo as few invasive procedures as possible. Angiography is important, but should be kept to a minimum. Non-invasive methods of patency detection, such as CT or MRI, can be used for intermediate assessment. Clinical trials should assess functional outcomes as well as anatomic outcomes. Primary endpoints should be patency. Secondary endpoints can include major adverse cardiac events and neurological events.

Panel members expressed concern that new guidance from the agency could negatively affect ongoing studies, such as the PAS-Port studies. Dr. Zuckerman stated that the agency would work with the sponsor to ensure that their work to date would not be discarded, and the fact that we are here today, I think, is proof of that.

Cardica completed enrollment in its pivotal trial in 2002, a trial that is referred to in your documents as Cohort I. Based on this data, Cardica submitted the 510(k) for the PAS-Port Device in February 2003, now more than two years ago, using the Symmetry device as a predicate, with the data that was very similar in terms of quantity, length of follow-up, to what has been submitted for the Symmetry approval.

The PAS-Port Device was approved in March of 2003, based on the data from Cohort I for sale in Europe, with the notified body requiring post-market evaluation.

In January 2004 the enrollment for a second prospective trial, a trial that is referred to in your documents as Cohort II, was completed. It is important to note that this was two months before this same panel met to determine objective performance criteria.

Based on the criteria, Cardica discussed pooling of Cohort I and II with the FDA to address this new guidance. The FDA stated that pooling would be acceptable as long as the two studies were comparable and the data was homogenous.

Finally, clinical protocols to Cohort I and II required amendments to allow for the collection of long term data, as the original protocols followed the patients for six months and the amendments allow data collections at 24 months in Cohort I and 12 months in Cohort II.

The intended use for the PAS-Port proximal anastomosis system is to create an everting anastomosis between the aorta and the autologous vein graft. Evaluation of this device was performed, as I said, in two clinical trials. Patients in Cohort I were enrolled in June 2002 and September 2002.

In July 2003 a prospective trial was initiated to test safety and efficacy of the C-Port Distal Anastomosis System we are also developing to perform distal anastomosis.

The clinical trial was designed in such a way as to allow the use of the PAS-Port system in patients requiring more than one vein graft, to prospectively collect data to satisfy European post-market requirements. It is important to note that these two products were never used in the same graft.

Both studies, Cohort I and II, were therefore prospective, non-randomized, multi-center trials, which were approved by the Ethics Committees, had similar selection criteria and the same endpoints for evaluating safety, efficacy and follow-up time points.

Follow-up in both studies consisted of patency evaluations at discharge and at six months and, where these evaluations were, per protocol, included, angiogram, CT and MRI. The studies were designed to compare patency to outcome to historical control.

In the report you have received from the FDA as part of the panel package, the FDA states that these data were collected retrospectively. This is not the case. Data were collected prospectively, primarily to meet European post-market requirements, as I have said, and as the FDA objective performance criteria had not been formulated until after the enrollment of this trial had been completed.

The notified body specifically requested that post-market data be collected using the same study design and endpoints as the pivotal trial.

The C-Port study protocol specifically provided for concurrent use of the PAS-Port Device in patients requiring more than one vein graft. Due to the confidence in both products performing well with low morbidity and mortality, the sponsor and investigators believe that to be an acceptable study design to allow concurrent use of both devices.

Another question raised or implied in the report from the FDA was if there was a target vessel selection bias in favor of PAS-Port index grafts. We have done an analysis, and this flow chart shows the distribution of patients enrolled in the C-Port trial.

One hundred thirty-three patients were enrolled, of which 54 patients received at least one PAS-Port graft. Fifty-two of these patients also received a C-Port vein graft. It is of interest that one patient who could not receive the C-Port due to having a porcelain aorta was successfully implanted with three PAS-Port vein grafts. The decision on which target vessel the respective devices should be used was left to the discretion of the surgeon, and per the criteria set forth in the instructions for use of the device.

This flow chart compares the target vessels, their diameters and native artery stenosis. The C-Port grafts were routed more frequently to branches of the circumflex and to the diagonal vessels than PAS-Port grafts in this subset. Fifty-one percent of the PAS-Port grafts went to the right side, versus 13 percent of the C-Port grafts.

As many of you know, grafts placed to the right side of the heart have historically shown poor patency rates than grafts placed to the left side, such as the circumflex, LAD and diagonal. In addition, the target vessel diameters and the amount of native artery stenosis were very similar.

These data suggest that, if there was a selection bias, it was against the PAS-Port index graft.

Another important question raised was if there had been significant device changes between Cohort I and II which were motivated by the desire to improve the delivery system reliability fall in Cohort I. I believe this is something the public expects a supplier of a product to do, is continue to develop and improve function.

There were a number of small changes made to the deliver system to decrease friction and deployment forces and improve ergonomics. These include changing the lubricant inside the delivery system, making the case halfs differ, increasing clearances within the internal components, adjusting spring forces and making small changes to the torsion shaft. In addition, there was a change to the manufacturing specifications for the cutter speed.

The ergonomics of the delivery system were improved by adding a quarter of a turn to the knob, changing the shape of the knob to improve the ability for the user to grasp it, and by slightly changing the shape of the case to improve visibility.

There was only one minor change to the implant to improve manufacturability during the laser cutting process. You see here the implant that was used in Cohort I, the implant used in Cohort II.

As you can see in this slide, this change consisted of the removal of a chevron feature on the outer flange portion of the implants. This portion only comes into contact with the adventitial surface of the aorta, and as such does not affect the final shape, function or integrity of the implant.

Strain and fatigue performance through shaping and deployment were validated in both versions of the implant, and the two were shown to be equivalent. Physical fatigue testing to 400 million cycles demonstrated adequate fatigue life for 10 years of implantation.

The company made a minor change to the packaging, which facilitated graft loading, and changed the packaging from an outer tray to a pouch configuration.

Finally, a minor change to the pull-through tool, the accessory tool used to pull the graft through the implant and aid in the process of graft loading.

As required, all of these changes were first verified and validated in the laboratory and in Cohort II. These small changes add precision to various components. There have been no functional changes to any component nor changes that affect function of this product.

That concludes my part of the presentation. I would like to ask Dr. Harringer to present our clinical trial data. Thank you.

DR. HARRINGER: Thank you, Dr. Hausen. Good morning, ladies and gentlemen, and thank you for allowing me to present the fundamental design of the studies and their respective results.

I want to disclose that I have been a consultant for Cardica for the last eight years and I have a small equity interest in the company as well as my travel expenses to Washington for this presentation are paid for.

This slide depicts the inclusion and exclusion criteria that were identical for the two cohorts. In a sense, the inclusion and exclusion criteria define a clinically stable patient requiring an elective coronary artery bypass procedure.

The study protocol for patients enrolled in Cohort II allow for enrollment of an older population with more comorbidity and underlying cardiac disease.

The follow-up schedule for the two cohorts is described in this table. I would like to emphasize that the follow-up schedule and clinical assessments at each time point were identical for both cohorts, with the only exception being that the additional long term follow-up differed because enrollment for the first study had been completed a year earlier and, therefore, the follow-up obtained was a year longer.

I would also like to point to the fact that, in contrast to the report by the FDA in panel package which stated that CTs were protocol violations, the use of CT or MRI was allowed as alternative diagnostic method in both study protocols if the patient refused the invasive angiographic procedure.

Again, these protocols were designed and approved by the Ethics Committees well before the objective performance criteria were discussed by this panel a year ago.

The study protocols were initially set up as observational studies with the intention to compare patency rates at six months to historical controls. The results from both cohorts compared favorably in this regard.

Following the panel meeting, the endpoints were redefined. Per FDA guidance, average index graft patency had to be at least 85 percent with a lower confidence bound of 80 percent with 95 percent confidence at six months, using angiography as the primary method to determine graft patency.

We would like to emphasize that angiographic data were assessed by QCA by a core lab, and were available before FDA gave new guidance on patency requirements.

The secondary endpoint was to determine the occurrence and frequency of major adverse cardiac events at one year, which were defined as death, myocardial infarction, and the need for target vessel revascularization. These data were obtained by amending the protocols for long term follow-up.

Core lab analysis was performed by Dr. Alderman from the Department of Cardiology at Stanford University, a physician who has great experience at performing these kind of analysis and who was the core lab cardiologist in important trials such as the BARI trial and the Coalesce and EUCLID trial.

The core lab was blinded to the clinical site evaluation of anastomosis outcome. The assessment was done by quantitative analysis, as mentioned before.

Patency was based on the FitzGibbon grading system which categorizes degrees of stenosis seen at the proximal anastomosis into three groups: FitzGibbon A, less than 50 percent stenosis at the proximal anastomosis; FitzGibbon B, 50 to 99 percent stenosis; and FitzGibbon O, an occlusion of the graft.

In these studies, the centers performed intra-operative heparinization and reversal according to institutional protocols. In order to assess graft patency intra-operatively, graft blood flow was measured with a flow meter, which is an established method to intra-operatively determine graft patency.

Following surgery, the anticoagulation regimen was based on the institutional protocol and not mandated by the study protocol.

Now I would like to describe the study results, starting with the intra-operative and discharge data.

A total of five sites were involved in either or both of the two trials. You can see here one site from Switzerland and four sites from Germany. Two of the sites enrolled patients for both trials.

Demographics are shown for the combined patient population here. Data is depicted as average or percentage, and the respective intervals are shown in parentheses. Demographic data for these studies were comparable to the STS database, as has been published by Dr. Ferguson in the Annals of Thoracic Surgery. The results of the BARI trial as well as a retrospective analysis of more than 150,000 patients performed by Dr. Mack and published recently in the Annals of Thoracic Surgery.

As you can see here, the basic demographic data were very well comparable to the standard elective CABG data in Europe as well as the United States.

In the combined studies, 123 PAS-Port deployments were attempted in 109 patients. A total of 12 PAS-Port deployments were not successful in 10 patients, and I will go into more detail on this point in the next slide. One hundred eleven successful deployments were performed in 99 patients.

Two patients died before discharge. In one patient, the right ventricular infarct had occurred immediately prior to the revascularization procedure. The right ventricle failed during the operation, and the patient was not able to be weaned off cardiopulmonary bypass without the implantation of an ECMO system.

The PAS-Port was used for a double sequential graft that was sequentially placed to the first and second obtuse marginal and posterior descending artery. The patient was reexplored for bleeding on post-operative Day One, and at this time the PAS-Port graft was determined to be patent. The patient died on post-operative Day Two of multi-organ failure. On autopsy, the PAS-Port graft was found to be patent.

The second patient underwent a coronary angiography after the operation that showed a patent PAS-Port graft. Later in his post-operative course, he acquired pneumonia that developed into a full blown ARDS. The patient died on post-operative Day Twenty-six. Neither of these deaths were related to the use of the PAS-Port device.

As I had mentioned, there were a few instances where PAS-Port anastomosis were converted to hand-sewn anastomosis. In the first cohort deployment failures occurred in five instances. In the majority of these failures, device deployment did not occur, and no aortotomy was created. Therefore, the surgeon had only to remove the graft from the device and perform the usual steps of the hand-sewn anastomosis.

Surgical technical failures occurred in an additional five instances. In one instance, the graft was cut too short following successful deployment, and the surgeon had to modify the proximal anastomosis to a T-graft. This is a PAS-Port specific problem, as this can occur whenever a surgeon elects to perform proximal anastomosis first.

In the second cohort, one can observe a significant improvement in procedural success following only minor changes to the delivery device and better surgeon training, as has been mentioned previously by Dr. Hausen.

Note that the two patients that were successfully implanted but died prior to discharge were conservatively considered acute procedure failures, although the causes of death were not considered device related, and both index grafts were found to be patent.

Thus, 109 implants were successfully placed in 97 patients.

The FDA requested that the sponsor collect long term data on the 10 patients that were not implanted and were converted to hand-sewn anastomosis. Since all of these patients were classified as acute procedure failures, they were not included in the six-month efficacy analysis, as the study protocol defined evaluable patients as patients that were discharged with an implant in place. However, nine out of 10 of these patients were reenrolled for the long term follow-up evaluation.

This flow chart describes the data collected at 12 and 24 months. Of these 10 patients, a MACE assessment was possible in nine. One patient who elected not to reenroll was alive and doing well when contacted for long term follow-up.

Resting and stress ECG evaluations were performed in eight and six of the nine patients, respectively. There was no evidence for myocardial infarction or presence of myocardial ischemia, and none of these patients required target vessel revascularization.

So in conclusion, all patients who were not implanted were successfully converted to hand-sewn anastomosis without any compromise to long term outcome.

PAS-Port deployments were attempted in 121 grafts. The majority of these were single grafts. Seventeen percent were single sequentials, and the remaining three percent were double sequentials.

One hundred fifty distal anastomosis were performed with these 121 grafts, 17 of which were placed to the diagonal system, 40 percent to branches of the circumflex, and 43 percent to the right coronary artery tree.

Based on the FDA report, one may have the impression that the PAS-Port patients only received one bypass graft. In fact, we would like to point out that patients actually had an average of 3.1 grafts, 1.8 arterial grafts per patient, an average of two proximal and three distal anastomosis. The surgery times are comparable to those seen in standard CABG surgery.

Protocols required the assessment of index graft patency before discharge. Patients who did not have predischarge patency evaluation either refused the procedure or it was contraindicated for patient health reasons. Eighty-eight patients with 97 PAS-Port implants were evaluated by angiography. Two patients were evaluated by computed tomography.

Using the FDA definition of patency defined as less than 50 percent stenosis, the PAS-Port graft patency at discharge was 99 percent.

I would like to emphasize that the one graft evaluated as FitzGibbon B resolved, and the graft was patent with a stenosis at six months.

This table depicts the different adverse events that were experienced until discharge, and compares the observed rates to those reported in the literature.

In the panel pack we presented revascularization and stenosis occlusion data that were reported for all grafts, including non-index grafts. This appeared to cause confusion regarding the performance of the PAS-Port. Therefore, in this presentation graft revascularization and stenosis occlusion are reported for index grafts only.

In the FDA summary of safety outcomes, each adverse event was counted as occurring in a separate patient. However, some patients had multiple adverse events. Therefore, actually 40 patients had 47 adverse events, and 69 patients or 63 percent had no adverse events compared to a rate of 64 percent cited in the literature.

In addition, the rates of each individual event compare very favorable to those in the literature, as you can see on the slide. The literature analysis was based on thousands of patients. In this slide we have focused on cardiac complications.

Eighty-one percent to 74 percent had no cardiac events. Twenty-eight patients had 31 cardiac complications. Some patients had multiple cardiac events reported. The majority of these complications were arrhythmias, such as atrial fibrillation, and known post-operative adverse event following CABG surgery. Revascularizations were performed in three of these patients. Let's look at these in further detail.

One patient had a clamp injury to the graft, without any evidence of kinking. The injury was more than five centimeters distal to the proximal anastomosis and appeared unrelated. This lesion was treated by PCI. At six months the proximal anastomosis remained patent.

In the second patient, a kink was noted after the first sequential anastomosis, and this graft was surgically revised, yielding excellent short and long term results, including no significant stenosis or occlusion in the six-month angio. This is a technical surgical failure that can also occur in patients with hands-on anastomosis when the length of the graft between the first sequential and the second sequential anastomosis was misjudged by the surgeon. In the third patient a stenosis was seen in the graft body, which was successfully treated by PCI, as you can see here the post-op result -- or the post-interventional result. This graft remained patent by CT at six months.

Recalling the cardiac event summary table, there were three patients with index graft stenosis or occlusions. Let's look at these patients in further detail.

In the first patient, an occlusion of the anastomosis of the sequential vein graft was noted. The first segment of the graft, including the proximal anastomosis, remained patent on angio at six months, whereas the distal bridge from the first to the second sequential anastomosis was occluded.

In the second patient, a kink -- and this is probably difficult to see; it's up here -- was noted close to the proximal anastomosis in the discharge angiogram, and this, interestingly, resolved spontaneously by six months. This graft was considered a FitzGibbon B in the discharge patency results.

In the third patient, a partial thrombus in the body of the vein graft was diagnosed with less than 50 percent stenosis. The patient was not evaluated for patency at six months. However, the patient had a negative stress ECG at three and 12 months. Thrombus was well adhering to the vein wall, and the proximal anastomosis was very far away, with no MACE at three and 12 months.

In summary, of the six index grafts that we have just discussed, the core lab reported patency at discharge in five of these index grafts as FitzGibbon A and one as FitzGibbon B.

Let's look at the clinical data obtained at three, six and 24 months respectively. An important outcome variable in prospective clinical trials is always the completeness of follow-up. We are particularly proud of the follow-up results in these studies, especially in light of the fact that these patients had to be reenrolled for the long term follow-up.

At three months, 97 percent of the patients were followed, at six months 93 percent, and at 12 or 24 months approximately 95 percent of the patients were followed.

Denominator declined between six and 24 months, because four patients died in between. These results compare quite favorably to the published literature for cardiac surgical trials.

Let's look at the results from three month follow-up. Of the 97 patients with PAS-Port's implanted at discharge, 10 patients were not evaluated by stress ECG for several reasons. Of the 87 patients evaluated by stress ECG, 90 percent were negative. For the nine patients with a positive stress ECG, four were noted to have an occluded index graft at six months. The remaining five patients had a patent index graft on the six-month angiogram.

The FDA's report showed concern that these five patients were not followed long term. In fact, all five patients were followed with good long term outcome. This included the stress ECG at 12 or 24 months.

All of these patients were asymptomatic. Four of the five patients had a negative stress ECG. The remaining patients had a similar distribution of ischemia in stress ECG at 12 months as observed in the three-month stress ECG, while the index graft was shown to be patent at six months, and scintigraphy performed at 12 months showed the absence of ischemia in the target region for that index graft.

Now we turn to the six-month patency evaluation. This is a very important slide, as it addresses our primary endpoint. Of the 97 patients discharged with PAS-Port implants, seven patients refused to participate in the clinical follow-up at six months. Three additional patients participated in the follow-up, but declined to undergo a patency evaluation.

Seven of these 10 patients were later evaluated at 12 or 24 months. Eighty-seven patients with 99 PAS-Port implants were evaluated by angiography, MRI or CT, for an overall assessment rate of 90 percent.

At the panel advisory meeting last year, Dr. Zuckerman stated that the paradigm for angiographic follow-up has usually been about 80 percent assessed by an independent core lab. In these studies, 79 percent of patients were evaluated by angiography with core lab analysis. Thus it appears that our angiographic follow-up rate just meets these criteria, in spite of the fact that the study protocol allowed for CT and MRI scans for evaluation of patency additionally.

Eighty-one of the 89 PAS-Port grafts evaluated by angiography were determined to be FitzGibbon A. All 10 PAS-Port grafts evaluated by MRI or CT scan were found to be patent. Therefore, the overall graft patency at six months was 91.9 percent.

At the 2004 annual meeting of the European Association of Cardiothoracic Surgeons, Dr. Mack presented the meta analysis on more than 40 publications in which the patency of more than 28,000 hand-sewn saphenous vein grafts had been determined at various time points.

The blue line depicts the observed patency rate for hand-sewn controls where the definition of patency varies between the individual publications, with most of them defining patency as FitzGibbon O, a complete occlusion only.

In comparison, the PAS-Port graft patency, including FitzGibbon B patients, at discharge was 99 percent, and at six months 91 percent. These are outstanding results, in my opinion.

If you look at the eight PAS-Port grafts that were either occluded or stenosed a six months, the common denominator for seven of these eight grafts was that they were anastomosed to target vessels with a non-critical stenosis. It is underlined in yellow, indicating that perhaps competitive flow was one of the reasons these grafts occluded.

We know that in the FDA report that the primary efficacy data was analyzed in an alternative manner. We show the two analyses side by side. There is a discrepancy, because we used the standard 0.05 significance level in the one-sided test, and the FDA used a 0.025 significance level in a two-side test.

We did our analysis using a one-sided test, because that was how the hypothesis was framed. Nonetheless, the difference in the lower confidence value with the two methods is very small.

When we looked at patients that were analyzed at six months by angiography only, the average patency is 91 percent, with a lower confidence bound of 84.4 percent.

The FDA asked the sponsor to perform a sensitivity analysis, which we will show in the following tables. For this sensitivity analysis, various imputations were performed.

First, patients that had an index graft patency confirmed by MRI or CT were added to the 89 patients with angiographic data. The imputed patency at six months by all patency diagnostic methods in available patients per protocol was 91.9 percent, with a lower confidence bound of 85.9 percent.

Next, 10 additional patients who did not have patency diagnoses at six months were considered. Seven of these 10 implanted patients had 12 or 24 months clinical data with negative stress ECG and absence of angina, indicating that the index graft had a good probability of having been patent at six months.

Three patients in which the data were insufficient were conservatively imputed as having occluded index grafts, for an average patency of 89.9 percent and the lower confidence bound of 83.8 percent.

If we now include data from 12 patients who did not receive a PAS-Port implant but were converted to hand-sewn anastomosis, nine of these patients had sufficient 12 or 24 month clinical data suggesting a strong likelihood that the converted grafts remained patent at six months.

Considering these patients, the average imputed patency at six months were then calculated to be 87 percent with a lower confidence bound of 81 percent.

Finally, in the most conservative estimate of all, we considered all converted hand-sewn grafts, irrespective of actual known patency, to be occluded. In addition, all grafts in the deceased patients were grafts without additional clinical information indicating patency were also considered to be occluded.

In other words, even if an angiogram showed a hand-sewn anastomosis of a converted graft to be patent, the graft was considered to be occluded in this intent to treat analysis and considered a procedural failure.

The average patency in this worst case analysis is then 79,7 percent with a lower confidence bound of 72.8 percent. It is important to say that this is an imputation strategy that leads to a very biased result.

This table depicts the adverse events that occurred from discharge to six months. Again, entries in the table reflect the number of events, not the number of patients. There were 32 patients that had 42 adverse events and 62 patients or 66 percent without an adverse event.

Additionally, it should be noted that the four patients that had ischemia noted at three months and occluded index graft at six months appear on the table twice, as they are reported in both categories, like myocardial ischemia and graft stenosis or occlusion.

Once again, we want to focus on the cardiac complications. Sixty-seven patients or 71 percent have no cardiac complications. Thirty percent of the cardiac adverse events were arrhythmias. As previously noted, eight patients were noted to have myocardial ischemia at three months, of which the index graft was found to be occluded at six months in four.

In the remaining four patients, all grafts were found to be patent by angiography and, therefore, the ischemia was mostly likely a result of progression of native coronary artery disease.

A myocardial infarction was noted in one patient in whom an MRI analysis showed the index graft to be patent. Therefore, the myocardial infarction is not likely to be related to the index graft.

Revascularizations were performed in two patients. Let's look at these in closer detail. A stenosis in the graft at the distal third required dilatation and placement of a stent. This was not considered to be device related, as the location of the stenosis was not in the proximity of the implant, which was up here, which was clearly patent. This patient was followed at 12 months and found to have a negative stress ECG with mo MACE.

In one patient there was a stenosis in proximity of the implant that was successfully dilated. This patient is considered as having significant stenosis in the primary endpoint analysis.

If you recall, in the cardiac complication summary there were 10 stenosis or occlusion associated with an index graft, seven of which are considered as nonpatent index grafts in the primary endpoint analysis. The remaining three index graft occlusion or stenosis are associated with the graft body and the proximal anastomosis was considered to be patent.

Long term follow-up data consisted of 24 month data for Cohort I and 12 month data for Cohort II. Patients were reconsented to participate in long term follow-up, and the data presented here includes only the data collected from six through 12 or 24 months.

Five patients did not reenroll. Of these five patients, three had a six-month angiogram that showed the index graft to be patent. MACE assessment was obtained in 94.8 percent of these patients. Four patients died in the follow-up period with causes unrelated to the device. These include a ruptured infrarenal aortic aneurysm, cerebral tumor, left ventricular heart failure, pulmonary edema and renal failure.

Resting ECG evaluation was completed in 96 percent of the patients and was negative for myocardial infarction in all of these patients. A stress ECG was performed in 89 percent of the patients, with evidence of myocardial ischemia in five patients. There were no index graft revascularizations during the follow-up period.

Let's compare our stress ECG results to those in the literature. A comprehensive review of the available literature and studies that utilized stress ECG for patient follow-up was performed and is depicted in this chart.

In the five studies that reported stress ECG results at three months post-operatively, the number of patients with positive stress ECGs varied between three and 45 percent, with an average of 19 percent. In comparison, the incidence of a positive ECG at three months was 10 percent for the PAS-Port patients.

Two studies reported results of stress ECG at one year follow-up varying between 23 and 35 percent, with an average of 30 percent. By comparison, the incidence of positive stress ECG in Cohort II at one year was nine percent.

At two years, two studies described results of stress ECG evaluation with approximately 35 percent of the patients on average presenting with a positive stress ECG. In comparison, in Cohort I three percent of the patients presented with a positive stress ECG.

Overall, the incidence of positive stress ECG in the PAS-Port study compares very favorably with what has been observed in the published literature.

Throughout the study, the clinical evaluation, including assessment of the Canadian Cardiovascular Society's angina class as well as New York Heart Association heart failure classification was reported. As can be seen in this histogram, a preoperatively large proportion of these patients, as one would expect, was in CCS II through IV.

Following surgery, the vast majority of patients were at CCS 0 or I at the various follow-up time points, as you can see in the blue bars.

The FDA report has requested that we discuss the need for long term antiplatelet therapy, once the device has been cleared to market. This table lists the anticoagulation therapy in patients of these studies. The majority of our study patients were treated with aspirin only, the low line of the table.

One of the most important long term determinants of device safety is the incidence of major adverse cardiac events. In order to respond to issues raised by the FDA, we have added three MACE events. Overall, in this revised analysis there were a total of 21 MACE events, 10 of which were index graft stenosis or occlusions.

We wished to compare these conservative PAS-Port MACE rates to those in the literature. There are very few published studies in coronary artery bypass surgery that provide information on the frequency of MACE at different time points. However, we found one study which was directly applicable.

The ARTS study recently published in Circulation with more than 600 patients followed over three years provide a MACE event rate at six months, 12 months and 24 months, as can be seen on the graph.

The MACE rate at these time points were 9.5, 11.1, and 14.2 percent, respectively. MACE in this ARTS study was defined as cerebral vascular accidents, myocardial infarcts, and target vessel revascularization.

In comparison, the MACE rate in the combined PAS-Port studies was calculated for these same events. As depicted in the blue line, the MACE rates at six, 12 and 24 months were 7.4, 10 percent and 10.8 percent, respectively. Therefore, MACE rates for PAS-Ports are consistent with the literature.

Please allow me to summarize the principal results from our study. First, the acute procedural success improved from 83.3 percent in Cohort I to 93 percent in Cohort II. Patients who did not receive a PAS-Port were converted to hand-sewn without compromise to long term outcome. There were no implant related re-operations for revascularization or bleeding.

Ninety-nine percent of index grafts were patient at discharge. The incidence of peri-operative adverse events compares favorably to the literature.

With respect to our six-month results, clinical follow-up was 93 percent. 79 percent were evaluated by angiography, and 89.7 percent were evaluated by all patency diagnostic methods.

Primary efficacy endpoints demonstrated that 91 percent of PAS-Port index grafts were patent by angiogram, with a lower confidence bound of 84.4 percent. The success criteria in this study were met.

Ninety-two percent of PAS-Port index grafts were patent by angiogram, CT or MRI scan. The success criteria of the study were met.

Eight-nine percent -- 89.9 percent PAS-Port index grafts were patent by clinical imputation or diagnostic procedure. Also these success criteria were met by the study.

Long term results at 12 and 24 months demonstrated excellent patient follow-up of 95 percent. Ninety-four percent of patients evaluated were negative for any evidence of myocardial ischemia. Patients have shown long term stability of their clinical course. The MACE remained low throughout the long term follow-up, with no reports of any index graft related revascularization procedure, no implant related death, no patients symptomatic due to index graft occlusion or stenosis.

The combined results of these studies through one and two years of follow-up demonstrate favorable safety and efficacy of this device for the treatment of patients undergoing coronary artery bypass grafting.

We believe that these data support clearance for the Cardica PAS-Port System in the U.S. Thank you very much.

I would like to hand over for the statistical review to Professor Bloch.

PROF. BLOCH: So I am the consulting biostatistician for Cardica, and I have no equity position in the company.

In the few minutes that I am going to be here, I am going to present evidence that supports the poolability of the data in Cohorts I and II. I am not going to again present the statistical results which you just saw.

Now in thinking about the poolability of data between cohorts, I want to come back to the summary that Dr. Hausen first presented, which had to do with the fact that both studies operated under pre-defined protocols, protocols that were written prior to the studies starting, and the protocols defined very clear methods for data collection. Both had equally stringent and clear methods for data monitoring. The endpoints were identically defined, and all of the analyses for patency were performed by the same CORE lab.

These are important factors to consider in pooling data from cohorts for any study. In the few minutes that I have left, I do want to talk about the similarity of the cohorts with respect to characteristics between the cohorts and also how uniform the patency results were among the five European sites that were studied.

This next slide shows the patency results at the four German sites and the one Swiss site. Notice that the patency results are all high, ranging from 82 percent to 100 percent in Leipzig. The patency results can be statistically compared. They are not statistically different, with a p-value of .5.

These findings across centers that did have some differences, as sites do have differences - - Nevertheless, these findings show an extremely robust finding across the sites.

Now regarding demographic variables as to whether or not they were homogeneous, the 26 demographic baseline variables that were compared used various statistical methods, depending upon if the variable was qualitative or quantitative, and these were standard statistical methods, the two-sample T-test or the Wilcoxon's two-sample rank test, depending upon the nature of the data. The Wilcoxon is an exact test, which is important to use if the data tends not to be normally distributed, and an analogy with qualitative, the Fisher's exact test performs that role, compared to using a chi square test, which is based on normal theory.

In the next two slides are presented these 26 baseline characteristics. For your convenience to be able to quantify the differences between the two cohorts, there is a column on the left of the table which I've put a p-value on top of.

It is important to keep in mind that, when you are looking at 26 different characteristics of cohorts, we are looking at multiple different variables, and these are what statisticians would refer to as nominal p-values or so called P-values. Strictly speaking, they should be adjusted for all the comparisons that are being made.

Nevertheless, there are differences among the variables, and several of them are by design. So, for example, if you look at this table, the very first entry is age. And you will notice that the average age was three years older in Cohort II than in Cohort I. This is not by accident, because the inclusion criteria for the second cohort allowed older patients to be enrolled.

Also, the enrollment criteria allowed patients to have more comorbidity and more cardiac involvement in the second cohort compared to the first, and in this continuation of the table you see that there are several variables that have p-values which are on the low side.

For example, the angina comparison has this nominal p-value of .014, and this is also picked up in the Canadian score of angina severity score, CCS, where that is a variable: Is a patient class III or IV or not -- more severe or not? That p-value is .008, actually the lowest nominal p-value in the group.

So altogether there are some differences. Also, as we have noted, the two cohorts were operating under different protocols, the biggest difference being that the C-Port and PAS-Port were both allowed to be deployed in the second cohort.

Because of this, the FDA asked Cardica to perform a propensity score analysis to compare patency outcomes, having adjusted for the baseline characteristics.

Now let's keep in mind what we are doing when we are adjusting for potentially, let's say, confounding variables when we are assessing an outcome. In order for the analysis to be appropriate, the variables that we choose should have two properties.

One is that they should differ between the two groups that we are comparing, because if they are completely balanced, it doesn't make any difference. The second property of the variable is that it should have an influence on the outcome variable of interest -- for example, patency in our case.

For adjustment procedures to work -- and the most common method that people are used to seeing are analysis of covariates where the variables would be directly controlled for in the light of comparison of groups where the dependent variable is the outcome of interest or in this method, which I will describe briefly in a minute, the propensity score method, it is important for these methods that the variables that have the properties of differing between the two groups and are different -- and have an effect on the outcome, that all such variables be included in the model.

With the propensity score method, what is done is that these variables are related through a model to the propensity or the tendency or the probability, all synonymous with each other, of belonging to either Cohort I or Cohort II. The idea is that, once this model is formed -- so it's a very simple model -- it relates membership in Cohort I or II, with a probability of belonging to Cohort I and II, to these identified variables.

Once a propensity score model is obtained, then each person's set of variables that were used -- age, presence of nonpresence of diabetes or whatever the variables may be -- are then entered into the model, and a probability is then calculated for that patient. All patients have a probability calculated using the model.

The theory of propensity scores tells us that, if the variables in the model have the properties I have described, then patients that have similar propensity score values or similar probabilities from the model have then been balanced with respect to the variables of interest. That's the theory.

Now to go on, I am going to present two analyses using propensity score models. One is contained in your panel pack. So I won't dwell on it very long. But here all 26 variables were considered as potential variables to enter into the model, but the consideration was not based only on whether or not they differed between Cohorts I and II, but first the most important consideration was did they or did they not have an influence on outcome.

In considering the variables, only three had a potential effect on outcome statistically -- that is, with p-values less than 0.15. Those were presence of diabetes, smoking history, and vessel disease. All the other variables had p-values that exceeded 0.15.

So I have to emphasize, all 26 variables were considered. Only three had a possible statistical association with the outcome. Using these three variables in the propensity score model and then adjusting in the way that I will show you with a follow-up of the second method, the p-value for the comparison was 0.22. That is, the rates did not differ.

Now in the FDA's review of the analysis I just presented, they remarked that we may not have controlled for, adjusted for, enough variables. So I did a second propensity score analysis where, again back to the table of 26 variables, identified an additional five variables that differed between the two groups at the p-value of less than 0.15.

Now these variables did not have an association with outcome -- that is, with patency at the 0.15. It was higher than that. But it is legitimate to include such variables if one can justify it, and I think we would all say that, if you are older, there is more chance that you are going to have a bad outcome almost with anything in medicine. So it is a perfectly reasonably thing to include.

So these eight variables were included in a second propensity score model, and the propensity scores are contained in this histogram, just so you can see that the propensity scores varied. Some were as low as 0.1 and some over 80-90 percent.

Again the theory is that, if you are in the same bucket, that your covariates have been balanced. One then takes the propensity scores, and one forms strata, a few strata, and in those strata -- and so the strata are obtained by grouping propensity scores, and in the next slide you will see that there were four strata formed where the propensity scores in the range of zero to 0.3 were all put together, the patients in those ranges of propensity scores into the first low here, the first bucket of patient values.

I formed four strata here. The theory of statistics says that, if you form three to five strata for adjusting for confounders, that if data is sufficiently normally distributed -- that is summary statistics, which is true if you have 50 or more observations per sample -- then you capture over 90 percent of the variability that is obtained by the confounders.

In this case, I took four, because there was no natural division of the propensity scores into, say, three different categories or five. It is clear that, if you don't form many strata, you are not making any adjustment at all. If you just take one bucket, it's just a crude rate. You haven't adjusted for anything.

If you take too many strata, then there may not be enough sample, and that can also be misleading. So you can see here that I formed the buckets to have approximately equal number of patients in the combined cohorts. Next slide.

Now the way it works with propensity scores is, having done all that, the theory now says, well, in the stratum with scores in the zero to 0.3 range, we now look at the actual patency rates. They can be compared in that bucket, because the theory tells us they have been balanced in that bucket with respect to the covariates, and that both rates are 100 percent.

You can go down the line and look in the different strata and compare the patency events. They tend to be lowest in the fourth stratum and most disparate in the fourth stratum.

Also, if you remember the histograms, there was a lack of balance in the numbers of patients in the different buckets, and that is exactly why you want to do an adjustment.

The way the adjustment works is through a statistical test is to combine the evidence across strata, and that's what the Mantel-Haenszel test does here, and the p-value here is 0.57, again showing that the results are highly poolable. Notice that the p-value here was 0.57, in the first analysis was 0.22. both of them point to the same result, and that is that they really are not different.

I just wanted to show you one more slide as to what poolability -- I mean what this method actually does in terms of the interpretation of the unadjusted versus adjusted rates.

So on the first row here, I simply present the results that you have already seen. That is Cohort I patency was 86 percent, Cohort II 96 percent. With the first adjustment, if we use direct standardization as the adjustment procedure, the adjustment for the covariates -- in this case, three of them that were in the model -- results in the Cohort I patency rate being adjusted upward by about two percent, with the second rate not really changing.

In the second propensity score model with all eight covariates entered, the Cohort I rate doesn't change compared to the adjustment that was made with the analysis I, but the second rate changes quite a lot. It goes from 95 percent down to about 90 percent.

Of course, you can see that these results of 88 percent versus 90 percent are pretty close, and that it is not surprising that the p-value of the Mantel-Haenszel test was 0.57.

So in conclusion, I want to again point out that you have seen that the differences between the cohorts are minor, and have little effect on the assessment of the primary endpoint. These two cohorts have been shown to be poolable by two separate propensity score analyses.

I now want to have Dr. Michael Mack come up.

DR. MACK: Good morning. I am Mike Mack, and I am a practicing cardiac surgeon in Dallas, Texas, and I appreciate the opportunity of making this presentation this morning.

My conflict of interest disclosure is: I am here as a consultant to Cardica and, as such, my travel expenses are being reimbursed today. I have no equity interest or financial interest in the firm.

The outline of this presentation will be, first of all, a brief rationale for anastomotic connectors; secondly, the current status of connectors; thirdly, a historical look at saphenous vein graft patency; fourth, a look at current saphenous vein graft patency in sutured anastomosis, which I think will be somewhat eye opening; then looking at PAS-Port comparison to both historical and current sutured outcomes; and lastly, a brief comparison of PAS-Port to the Symmetry (St. Jude) connector.

First of all, what is the rationale behind anastomotic connectors? My initial interest in this whole area and initial interest in the St. Jude Symmetry Device was to facilitate less invasive surgery, and that was from two standpoints: First of all, to facilitate beating heart surgery by both shortening the period of time required at anastomosis, and secondly to increase the accuracy of performing an anastomosis.

There have been concerns raised in beating heart surgery, because the operating conditions aren't as optimal, that perhaps a less accurate anastomosis may be performed. I thought that anastomotic connectors had the potential to address this issue.

Second is to facilitate small incision or limited access surgery. There have been a lot of attempts over the last 10 years to perform totally endoscopic coronary bypass using robotics and using endoscopic suturing techniques. However, the most difficult thing to do endoscopically is to suture.

It appeared to me that, if there was an anastomotic connector that could eliminate suturing, then you could facilitate small incision endoscopic surgery.

Then lastly, it appeared to me that you could create a uniform, reproducible and perhaps superior anastomosis compared to sutures. Sutured anastomosis is what defines the skill of a cardiac surgeon. However, it also is a major variable in terms of reproducibility.

Similar to the way that stent grafts have created -- or stenting has created uniform outcomes with PCI and minimized some of the operator variable, it was thought by me and others that perhaps, if you could do this in a sutured anastomosis, you could minimize surgeon variable in terms of performing the anastomosis.

Indeed, we got involved in this field very early on, and this is a report of our initial experience with proximal anastomosis performed with the St. Judge's Symmetry connector. That was reported in 2003, and we found that we had good initial results with this connector.

So the question is, really, why are we here today? Those have been concerns that were raised about clinical outcomes with the St. Judge Proximal Anastomotic Connector called Symmetry.

Indeed, we were the ones that first raised issues regarding this connector. This is a report from our series in 2004. If we look at the abstract of this, we took 162 patients in which we had performed anastomotic connectors and compared them to the 159 immediately preceding patients that had sutured anastomosis in the year before it, and found that there is a significant increase in early MACE.

Indeed, it was an early presentation of this data at TCT in 2003 that Dr. Sapirstein was at, and he and I had a hallway conversation afterwards in which I told him, yes, I had significant concerns about the connector and this outcomes. I think that that was probably the index event that led to the concerns about the St. Jude Symmetry Device.

So, therefore, the question that needs to be answered is: Are all anastomotic connectors the same or do we have the potential of throwing the baby out with the bath water here?

This was a very hot, fertile field in innovation three and four years ago, with at least 14 companies involved in this, and Cardica is the last one left standing right now.

The question is: Has St. Judge poisoned the well forevermore, and should all anastomotic connectors be painted with the same paint brush, or perhaps is this one different?

So in attempt to answer that, I am going to compare the outcome of the PAS-Port with historical sutured controls, a comparison with current suture outcomes, and then with the Symmetry Device.

The analysis has been referred to already. I pooled all studies that looked at saphenous vein graft patency in the literature in the period between 1979 and 2001, and there were 30 studies that looked at 28,000 grafts.

These seven studies reported early outcomes, less than 30 days, and we can see that in meta analysis the total patency is 87.9 percent. Similarly, at three to six months there has been a total of 10 studies reported, 2200 grafts. The patency rate is 84 percent. At 12 months, you can see the studies here, a total of 11,000 grafts. The patency rate is 82.7 percent.

If you go out to two to five years, there have been three studies that looked at that, a total of 3,000 grafts, and the patency rate of saphenous vein grafts is 74.3 percent.

So if we look at what we are interested in today in a total of these 28,000 grafts examined, the patency rate at three to six months is 84 percent, and at 12 months is 82.7 percent.

Now the next three studies that I am going to talk about, I think, are fairly sobering to cardiac surgeons today. The first is a recent report out of Portland comparing endoscopic versus open and saphenous vein harvest.

The second is the PREVENT IV trial, the database which has just been unlocked recently, and outcomes reported in a third is the PRAGUE-4 trial, and we will look at each of these.

The first is a randomized trial of endoscopic versus open saphenous vein graft for coronary bypass grafting that look at six-month patency rates that was published last month in The Journal of Thoracic and Cardiovascular Surgery.

It was a randomized single center trial. Six-month angiographic patency was the endpoint. A total of 336 grafts were evaluated. The patency rate was 69.1 percent. In arterial grafts in the study, the patency rate was greater than 90 percent. There was no difference between endoscopic and open harvested saphenous vein grafts.

The PREVENT IV trial database was unlocked on March 29th. This was a trial run out of DCRI, the evaluation of gene blocker with E2F to prevent intimal hyperplasia in saphenous vein grafts.

There were a total of 3,000 isolated CABG patients in 84 U.S. centers, a 13-month enrollment period between 2002 and 2003. Eight percent of these patients received angiographic follow-up at one year. 4,557 grafts were evaluated. The patency rate is 64 percent.

There was no difference between the control group and the treatment group. There was no difference whether it was performed on pump or off pump. The arterial graft patency rate in this group, again, was in the 90 percent range.

The third is the PRAGUE-4 trial, which is a single center randomized trial comparing off pump versus on pump CABG. This was published in Circulation in December of 2004. One-year angiographic follow-up was the primary endpoint of this trial. Two hundred fifty-five grafts were evaluated. The patency rate of the saphenous vein grafts was 52.5 percent.

So why should current saphenous vein graft patency be different from historical controls? Well, I think right now is that, with the success of PCI and drug-eluding stents, that the current surgical material that we have to operate on today is that it is typical of diffuse distal disease with multiple risk factors.

Limited proximal disease is now uniformly treated by stents, and with target vessels that are uniformly poor, I think that -- We always like to think that we are doing better than we were before, but I think the reality is, because of worse target vessels, our patency rate is not as good as historical controls.

The second is that there is an increased use of arterial grafts in surgery, and those arterial grafts go to the best vessels. If we put saphenous vein grafts to the left anterior descending, they are going to be better than saphenous vein graft patency to ramus, circumflex or diagonals or to the posterior descending branch of the right coronary artery.

So even the best vessels that we have to bypass right now, we are using arteries for, then veins. So I think, although these results are somewhat sobering, they are not surprising.

So if we look at a summary of this, of current trials, we see that three trials using sutured anastomosis have patency rates ranging between 52 and 69 percent, and in the same time period the Cardica PAS-Port saphenous vein graft patency was 91 percent.

Now let's look at the clinical outcomes of the St. Jude versus the Cardica. This is our series of one-year follow-up that I alluded to earlier.

There are a total of 162 patients that we followed up. We had angina recurrence in a total of 12.3 percent of these patients, and in these patients there were 23 index grafts that got a Symmetry Device, of which 15 were occluded.

In the 109 patients in the Cardica series, only two patients presented with angina at 12 months, and both were found to be in non-index grafts. We studied 38 percent of these patients.

After we began having problems early on, we initiated surveillance angiography, and that was mainly by EBCT, and we were able to study 38 percent of the patients and found the patency rate of the St. Judge grafts, and those were 77 percent.

In this Cardica study, as you've heard already, 91 percent of the patients were restudied, either by cath or by CT or MRI, and the patency rate was 91 percent.

Now I would like to clarify one thing that I have seen in the panel packet here, and that is there has been extensive discussion of MACE. Now the definition of MACE, as everybody knows, is major adverse cardiac events, and those include death, MI, and target revascularization, not graft occlusion or stenosis discovered on surveillance angiography. But if you see, both on the presentations here this morning and in the FDA panel packet, grafts discovered by surveillance angiography are included in MACE data, and I don't think that that is appropriate.

We all know that clinical events occur about half as frequently as angiographic events do, and if we are doing surveillance angiography in everyone, your event rate is going to be twice as high as what your clinical rate is.

So I think, to include graft occlusion discovered on routine surveillance angiography that was not intervened on and did not cause a clinical event really legitimately shouldn't be included as MACE.

So, therefore, in my looking at the results of this Cardica trial of 109 patients, there were 12 patients or 11 percent that had MACE. There were six deaths, of which four were clearly non-cardiac but still need to be included in MACE.

There was one patient that had a myocardial infarction, and there were five patients that underwent target vessel revascularization, but these fell prey to the atherostenotic reflex. They did not have clinical events. They were discovered on routine angiography. Yet still they did have target vessel revascularization.

So in summary, I think anastomotic connectors are enabling technology to facilitate less invasive CABG. All anastomotic connectors are not the same, and I'm the one that suffered through the initial clinical experience with the St. Jude Device and had to face each of those patients that did have a problem with that. Although I have no experience with this device, it certainly looks to me like this is not the same device as the St. Jude device.

The current sutured saphenous vein graft patency is less than the historical standard controls against which -- of which this is the benchmark against which this device is being compared. And PAS-Port saphenous vein graft patency appears to be equivalent to historical controls, and possibly superior to current saphenous vein graft patency. Thank you.

Bernard, are you going to close? That completes our presentation then.

ACTING CHAIRMAN MAISEL: Okay, thank you very much for a very thorough presentation. At this point I would like to invite the panel members to question the sponsor about any burning issues, reminding them, of course, that we will have ample time later in the session to ask questions. Sharon?

DR. NORMAND: I just have one question of clarification. When you report the patency rates at the level of the patient, how are you counting those with multiple implants? I just don't know how you define that as a success.

DR. HAUSEN: Patency is evaluated on a graft basis. So if we have evaluated 89 grafts, index grafts, and 81 were patent, that gives us our --

DR. NORMAND; No, but I was asking -- I'm sorry. Sometimes you report the rate at the level of the patient. So, for example, if there were 109 patients, you report the patency rate for those 109 patients. I was just wondering if somebody had three implants and -- do you need to have them all patent? I mean, that's the question I have.

DR. HAUSEN: We have some patients who have multiple implants, and I am not aware that we are reporting on a per patient basis.

DR. KRUCOFF: Can I just follow on that question. Was any sort of correction done? Was a GEE used for multiple observations in individual patients on the patency rate?

DR. HAUSEN: No. The patency, as we have presented it, is just a percentage, and primarily we want to focus on a per graft basis.

ACTING CHAIRMAN MAISEL: Norm, did have a question?

DR. KATO: A question for the sponsor. I guess what I don't understand kind of running throughout the presentation is that Dr. Mack concluded that your device is different than the predicate device, that all connectors are not the same, and yet you are coming to us saying that you are the same, because you are relying upon the predicate device as the model for getting approval. So can you explain that discrepancy?

DR. HAUSEN: Well, I think it is historical. Basically, when we submitted, everything was still fine with the Symmetry device. Therefore, that was the predicate for our submission, and it did not change as the results turned sour on the Symmetry device. We were still using that same submission.

DR. KATO: And so you weren't allowed to change that halfway through?

DR. HAUSEN: I think we could have withdrawn our submission, which would have started us at zero again right there.

ACTING CHAIRMAN MAISEL: Maybe at this point -- I know the FDA will make their presentation, but maybe Dr. Zuckerman could just clarify the current status of the Symmetry. My understanding is it was a voluntary withdrawal from the market, and so it is still legitimate to use it as a predicate device. Is that accurate?

DR. ZUCKERMAN: That's correct, and in your panel pack you will see that there is a series of predicate devices that are being used by the sponsor to get through the 510(k) flow chart. I think, though, that the primary questions that this panel is asked to address, those of safety and effectiveness for the current device, remain independent of what the status is of the St. Jude device.

ACTING CHAIRMAN MAISEL: Thank you. John?

DR. HIRSHFELD: This is a question for both the clinical presenters. The rationale that you presented for why this device is important is in order to prevent the complications of aortic cross-clamping.

Yet it appears to me from the data in the panel pack that virtually all of the operations in your series were done on bypass. So these patients all had aortic cross-clamping. So I have two questions.

First is, how well can you generalize the applicability of this device to off-pump surgery when all of the surgery was done on-pump? And despite the fact that the surgery was done on-pump, you had a fairly good neurological outcome in this series. So I wonder whether you may be overstating the value of this device in preventing aortic cross-clamp complications.

DR. HAUSEN: Those are two very valid comments. I was careful to present as inference, because the studies we would have to do to show that this has an impact on stroke would range to 3000 patients, just to bring it down by half a percent, and that there is a significant variability in the incidence of stroke in coronary bypass surgery that can vary from one percent or even less that we saw in our study to three or four percent.

I tried in my initial presentation to show that this is an area that I think there is a very common understanding among surgeons that the aorta is the problem, is the most important problem and source for stroke.

So if by inference you use a device or you don't touch the aorta at all, you will probably affect the incidence of stroke. In addition, Dr. Mack pointed out that creating anastomosis in a very standardized form is probably beneficial. If we point to intestinal anastomosis that are 95 percent of which are now done by staplers, and as a young surgeon I wouldn't believe that and always thought as a surgeon we're always the best -- we don't think to replace what we do. But it turned out that, yes, this is the most widespread use is through staplers.

I think this is -- If we are given the opportunity, this technology can drive -- get to that point, too.

ACTING CHAIRMAN MAISEL: Jeff?

DR. BORER: I have actually the same question as John, which you have now answered. But there is a second part to it. You suggested in the presentation that one of the important advantages of this tool is that it provides a greater selection of implantation sites than if one were doing this free-hand. I'm not sure I understand that.

You know, if you are doing the procedure off-pump, and these were not done off-pump, but if you were doing them off-pump, I suppose you would choose your site the same way with the tool as you would without, because you have nothing -- you can't look inside. But how is it that you have more sites that you can choose from if you use the tool than if you don't?

DR. HAUSEN: Another good question. The tool does not require any clamping. So as a surgeon I am basically, through an open sternotomy, limited to the ascending aorta and maybe the proximal arch of the aorta. With this device you can get to the descending aorta.

You can get to various parts that you can just barely get through -- just have visibility through an open chest procedure, because all you do is place the tool and rotate the knob and pull the tool off. There is no actual manipulation or proximity to the aorta needed.

DR. MACK: I think, Jeff, in point of fact you can see inside the aorta, and that is what we routinely do. Everybody nowadays gets at the aortic scanning at the time to define disease in the ascending aorta, and that is where we thought that the Symmetry device would have its benefit, because you could pick -- you didn't have to put a clamp on. You could pick a non-diseased spot or a minimally diseased spot in the aorta and put the graft there.

Now that we don't have that device to use, what we do is we still do the epiaortic scanning. We use a device called Heartstring, which is a little umbrella that fits in through an aortotomy, and then we suture that anastomosis. It is cumbersome but still, we feel, saves adverse neurological outcomes.

So you can see where it is, and it is an appropriate valid potential use of this.

ACTING CHAIRMAN MAISEL: Dr. Blackstone?

DR. BLACKSTONE: I have a couple of questions, but could I carry on this particular conversation.

What Dr. Mack has just said and what you have just said is that this may allow you more flexibility in where you can put your graft than if you were doing it conventionally. And yet your very first contraindication for use of this device is use in sites that you would not use with clamping.

That seems to be a contradiction.

DR. HAUSEN: I think this is how our instructions for use are starting. I think what we need is experience to gain more data to support what I have just said. I think, in theory and what has been used in practice in some sites over the world is that, yes, you do have access to other sites.

I think, for the instructions for use in the U.S., we want to be conservative and reduce the risk in the use of this product and for the patient, and be very careful in what we select and use this product in.

MR. BLACKSTONE: And following on with that, at least the surgeons at Cleveland Clinic tell me -- and according to our book, Cardiac Surgery -- that most of the proximal anastomosis tend to be an oblique, not an inside anastomosis, to the aorta, and maybe some of the surgeons will help us with this.

It seems that that obliqueness, in a way, helps with this problem of kinking and how long or redundant a graft can be.

You have minimized that, and I'm just interested in -- and yet you highlight it multiple times about the kinking. It seems like this actually potentially introduces more potential for kinking than a hand-sewn anastomosis.

DR. HAUSEN: In our study it did not, and what we asked the CORE lab to do is determine the takeoff angles in discharge angiograms of grafts that were hand-sewn, of which there were over 150 in these two studies -- vein grafts -- to the PAS-Port grafts, and determine what the takeoff angle was.

I have a slide that I won't find as quickly, but it shows a distribution from very flat angle, by 10 degrees, 15 degrees to 90 degrees, almost identical in both cohorts. I think that has to do with the low profile of the implant.

While theoretically the implant would require or make the graft takeoff in 90 degrees, in practice it doesn't. It takes off at various angles. In addition, all the graft kinks we saw -- and there were some, and we saw them in hand-sewns, too. So just making the anastomosis oblique doesn't get rid of the potential for kinking.

I think the primary factor that makes kinks happen is graft length determination or inadequacy in routing the graft and fixating at various points. That is irrespective of how you do your proximal anastomosis.

DR. BLACKSTONE: So final question on that. So it means that you have looked at these takeoff angles at the hand-sewn and the other. That means you have also looked at the patency of your hand-sewn grafts.

DR. HAUSEN: That's right.

DR. BLACKSTONE: And yet you have brought Dr. Mack here to tell us what some other studies and what historical controls are, but you have in your back pocket just now told us you actually know this in the hand-sewn, but we don't that.

DR. HAUSEN: We used the hand-sewns with the Sequent device. So it is a different product, and we didn't think that was appropriate to present this, plus it is a completely different variable using an anastomotic technology on the distal end. But I can tell you what those numbers are.

DR. BLACKSTONE: How about Cohort I, though?

DR. HAUSEN; Cohort I had some, and I think the average patency was in the 82 percent hand-sewns.

DR. BLACKSTONE: Okay. So it is more like the historic controls than the picture that has just been painted.

DR. HAUSEN: In this instance, it is, yes.

ACTING CHAIRMAN MAISEL: Jeff?

DR. BORER: I would like to follow on that question, and I don't want to reinvent the wheel, and this is the second panel, and I wasn't on the first. So just tell me if this has all been run through.

You know, we are talking about evaluating a device for its potential for allowing graft patency -- enabling graft patency and preventing the complications of graft occlusion. Those events, the clinical events, as we have seen from this database, are relatively few.

It would be unrealistic to expect you or anyone doing device studies to do large randomized controlled trials to look at these relatively uncommon outcomes. So although that would have been wonderful, that's okay, as long as -- I think our greater concern is stable point estimates of the actual frequency of the adverse outcomes, however we choose to define them, so that we can make reasonable judgments as to whether they are the same as what would expect from historical controls.

The problem that I have here then is the extrapolability of these data to comparisons with the historical controls. You know, Dr. Mack's presentation was -- For me, it was very compelling. I mean, I think that was very informative. But when I look at the large group of historical controls, 1979 through 2002, we are talking about a period during which techniques changed, patency rates undoubtedly changed, and the severity of disease in the target population may have been very different from that in the group that was studied here.

This is a very small group of patients that was studied. I mean, even when you add on Cohort II, we are talking about very small populations. So that the point estimates we would draw are inherently relatively unstable.

So I'm wondering why, just to assure us a little bit more, because you couldn't do it statistically, but to assure us that there's comparability between the hand-sewn sort of contemporaneous control group that you really have or could have had in Cohort I -- why you didn't randomize initially to the C-Port device versus hand-sewn. Why was that decision made?

I think, if it hadn't been made, one would have an easier time being confident about extrapolating these data to comparison with historical controls.

DR. HAUSEN: I can elaborate on that. It's basically again a timing problem. As I tried to point out, we designed these studies knowing or believing to know what the FDA required on the predicate device for approval. All the decisions, all the protocols, all the ethics committee's approval were completed.

The enrollment was completed before this panel convened a year ago. Then after the fact, objective performance criteria were presented to us that we now had to meet with studies we designed before. So that was a very difficult situation for us.

We had to then be creative. We had to sit down with Dr. Zuckerman and his team and discuss how can we get the power into our sample size, and the only way we could do that is by pooling the two datas, if we were going to -- if the FDA was going to honor what the panel suggested, which was make it -- don't make it impossible for those companies who have spent millions and millions of dollars to develop this technology, and already well underway in this process.

I agreed. I mean, a randomized trial would be perfect for this, absolutely perfect. Randomizing between samples -- that was something that was discussed at the panel. It is very complicated to do, because you have to not only randomize which target vessels you are using, you have to randomize the runoff of that target vessel, the vein graft size. There's so many factors that influence that, that you probably need sample sizes in the 500-600.

ACTING CHAIRMAN MAISEL: I will -- The Panel met about a year ago to discuss these issues, and we spent an entire day debating these issues, and some of the points that were brought up, randomizing -- you know, do you randomize within a patient, right coronary artery versus circumflex, the differences between those arteries, the difficulty in randomizing to no device for patients who would be forced to get the angiograms that they won't otherwise require.

The upshot of that Panel meeting was that the FDA decided on an 85 percent patency rate with a 95 percent lower confidence interval of 80 percent, which is the bar we have been given today.

We will hear from the FDA later, and they can comment on it, and maybe we can talk about the patency issues after their presentation later. Clyde?

DR. YANCY: One question to follow up on the issue of study design, understanding that the design predated the new FDA requirement. I still am curious as to whatever deliberations took place to determine your sample size.

It seems quite small in comparison to either Dr. Mack's single center experience or the larger trial that was done by DCRI. So even in the absence of the FDA perspective, what was the thought process about a relatively small sample size?

DR. HAUSEN: The primary predicate devices performed observational studies, and that's what we did, too. So there wasn't initially a formal hypothesis. It was determined from studying what others had done, and even with the very early discussion with the FDA back in 1999 that one could do this as an observational study, and it then got elevated to being hypothesis driven within OPC.

DR. YANCY: So let me follow up with one additional question. Since the predicate or precedent had been established with technology that is no longer extant, was there any thought process on the sponsor's part in view of the Symmetry experience to redesign a larger prospective, even if it is observational series?

DR. HAUSEN: What we did is we started the registry. I mean, it is within our financial means. The biggest market for these products are the U.S. That revenue stream would then trigger studies like that, absolutely. I think it is important. We have performed surveys and registries on over 2000 implants so far, sold about 500, and looked at MACE events and looked at how many patients, for example, came back with angina. Is it close to the 12.3 percent, because I have to be honest, that was something we were concerned about.

You look at the implants. At first glance, you think, well, that's very similar. And from this survey, we know now that there is a one percent -- that's one percent, six out of 550 patients, came back with angina where there was index graft occlusion or stenosis. So it is a very low rate, and probably full in agreement with what you would expect in this patient cohort.

ACTING CHAIRMAN MAISEL: John and then Rick.

DR. SOMBERG: Three, hopefully, brief questions. One is: You mentioned that the fatigue rate of the device had been tested with the evolution up to ten years. Is there any test -- that would be tantamount to 10 years of wear on the device. Is there anything beyond that? Does that matter? Is the healing process such that after a period of time, the device integrity is not as critical because the anastomosis heals together or something? Is that right?

DR. HAUSEN: Yes. That's a good question, yes. I would say within three months the graft is completely healed into the aorta, that if we could design an implant that dissolves, it would probably be perfectly fine.

DR. SOMBERG: Okay. The last question is: It seems to me that most of the action is at the very proximal part of the venous graft with the aorta.

DR. HAUSEN: Yes.

DR. SOMBERG So what happens distal, at the other end, may be progression of disease, etcetera, of the organism state. But what is related to your device seems to be what is proximal. Do you have a breakdown of proximal anastomosis reocclusion in the device in terms of this Cohort I/Cohort II? I seem to only find one at the proximal area, the Cohort I/Cohort II, and then for your overall registry at this point Europe and Japan?

DR. HAUSEN: Right. We have -- In Cohort I we had a discharge, one that had a FitzGibbon B grading, that Dr. Harringer pointed out became patent with a FitzGibbon A at six months, and we had at six months one graft that had a stenosis right at the proximal -- we showed the angio -- that was successfully dilated.

So the incidence of problems directly at the proximal anastomosis was very low. We cannot say from the six occlusions we had in Cohort I and the one occlusion in Cohort II where that originated. That is basically impossible unless you saw slipping of contrast retrograde from the coronary during the angiogram, which we did not see.

DR. SOMBERG: And the registry?

DR. HAUSEN: In the registry, the incidence of occlusions -- again, except for two reports where we -- or four reports where saw a stenosis at the proximal implant, they were flush occlusion or they were complete occlusions -- I shouldn't say flush occlusions -- and it is impossible to say where that would happen.

Just to clarify, you expect some degree of stenosis with hand-sewn anastomosis. We actually saw that in our study, too. So it is not a specific finding to an implant. It's just a default that you see a problem. It is by default assumed to be the implant. I mean, that's kind of the experience I have had working with my colleagues.

ACTING CHAIRMAN MAISEL: Rick.

DR. PAGE: Thank you. I was impressed and indeed sobered by Dr. Mack's presentation of saphenous vein patency, and was struck by the contrast between the data he presented and your data for this device. It makes me wonder the difference in populations studied.

Specifically, in the European sites where these studies were performed, was there as aggressive use of percutaneous intervention and arterial conduit use that would be comparable to United States practice, because indeed with the high patency rates we are seeing there, I wonder whether the device is being adequately tested in the sorts of vessels that in the United States receive a saphenous vein conduit; because as was mentioned, the saphenous vein grafting is placed in the worst candidates and in the most difficult vessels, and that would not be consistent with the patency data that we saw today.

DR. HAUSEN: Yes. Dr. Harringer.

DR. HARRINGER: If I may briefly comment on this. Well, as far as we can observe the data and evaluate the data, they seem to be similar with regard to distal runoff. At least, that is what the information was that came from the CORE lab. We actually got the information back that Dr. Alderman was surprised how small the vessels were that we were actually bypassing, that somebody in the U.S. might not be attacking such small vessels.

So there seems to be a similar comparable distal runoff in these vessels. However, I do agree with you that probably the aggressiveness of PCI in Europe is probably not quite the same as in the U.S., although -- I mean, you know your colleagues as well as I do. They are catching up in the territory.

ACTING CHAIRMAN MAISEL: Val?

DR. JEEVANANDAM: I have a couple of questions. You know, the device failures that turned into hand-sewn anastomosis is an interesting group, and I think that's a group that you probably need to take out of the total evaluation, because those were not -- the device was not implemented in those. But with those 10 patients, it is interesting that all 10 of those grafts were patent, I see, through the analysis.

DR. HAUSEN: Yes.

DR. JEEVANANDAM: So I think you can't really include those in terms of your overall patency, because those were device failures.

DR. HAUSEN: You are speaking from my heart. That was not our intention. This is something the FDA asked us to do, to do a intent-to-treat analysis. They are part of the intent-to-treat, and therefore, we did this sensitivity analysis where we took you through from angiographically only evaluated. Then we added the CT patients. Then we added patients where we clinically assessed the status with PAS-Ports and finally we used these 10 patients, and they were part of the intent-to-treat analysis. That was mandated.

DR. JEEVANANDAM: The other interesting thing is a lot of your occlusions were in vessels that did not have significant stenoses. While that disturbs me, because then when you say that patients don't have MACE and you are trying to -- you know, we are trying to equate angina with MACE -- right? -- as part of MACE, and whether that is a surrogate for graft occlusion.

Well, if you have a graft occlusion in a vessel that has a 50 percent stenosis, you are not going to have MACE, because you probably won't have angina from that vessel.

DR. HAUSEN: Right, but those we did then detect with surveillance angio. I mean, that's the advantage of having that system. I think, if we would just rely on MACE as your final outcome, that would probably be very true.

These anastomosis to target vessels with so little stenosis were the absolute exception. That's why we pointed it out. That is probably competitive flow in some of them, perhaps, the cause for occlusion, but that is completely speculative.

DR. JEEVANANDAM: And my last question is: With final labeling -- I mean, if we are predicating this as a comparison to the St. Jude device, do you now have to label this as being equivalent to the St. Jude device when you market it?

DR. HAUSEN: I'll let Dr. Zuckerman answer that question.

DR. ZUCKERMAN: The answer would be no. I would remind the Panel members that the purpose of this initial questioning period is just to answer any burning questions. There will be adequate time this afternoon for the Panel to thoroughly review the data in the submission and the intended labeling.

ACTING CHAIRMAN MAISEL: With that comment, I think we will move on and take our break. We will reconvene in 15 minutes at 10:40.

(Whereupon, the foregoing matter went off the record at 10:23 a.m. and went back on the record at 10:41 a.m.)

ACTING CHAIRMAN MAISEL: Why don't we get started. I would like to invite the FDA to give their presentation.

DR. ENYINNA: Good morning, Panel members and guests. This portion will be the FDA portion, in which we will present our review of the PAS-Port anastomotic system.

My name is Kachi Enyinna. I am a Scientific Reviewer from the Circulatory Support and Prosthetics Branch of Office of Device Evaluation, and also on my review team will be Barbara Krasnicka who is a statistician and will be presenting the statistics portion of our review, as well as Wolf Sapirstein, Medical Officer, who will be presenting the clinical portion of the review.

I will also add that I have no equity or financial interest in the firm.

The indications for use for the PAS-Port, Cardica's proposed indication for use, is as follows: The PAS-Port system is intended to create an everting anastomosis (end-to-side) between the aorta and an autologous vein graft. That word kills me, autologous. The predicate is the St. Jude Medical Symmetry Aortic Clip.

The PAS-Port consists of the implant, a stainless steel tubing used to connect veins, the conduit, to the vascular vessels in an end-to-side fashion. The system also includes a delivery system used to attach the conduit.

The implant is laser cut from medical grade 316L stainless steel tubing. It contains nine barbed prongs designed to penetrate through the vein graft after it has been inverted. The implant is initially configured in a compressed state and expands as it is deployed.

So as you see on the slide there, the first image is the pre-shaped implant, the stainless steel tubing and then the middle is the expanded state of the implant before deployment, and the actual implant after deployment where the top region separates from the bottom -- the rest of the skeleton of the stainless steel tubing. That is what is left inside the patient, the image on the right, bottom right.

Here we have what the delivery system looks like packaged with the cartridge and the deployment -- delivery tool and deployment system, the guide clip, etcetera, etcetera.

The process of deployment: As Bernard has gone through already -- I will just do a quick discussion of the operation of the device, but as it ties into the performance testing, which I will talk about shortly.

In Figure 1, 2, and 3, you see that the implant vein graft attached is introduced into the target vessel. The body of the implant expands, and the nine tines penetrate through the everted vessel to form an inner or intimal flange.

In Figure 3 you see that the implant inner flange is now deployed. It is during this phase the implant is expanded and the lumen is increased.

In Figures 4, 5 and 6 the inner flange is now seated against the intimal surface of the target vessel. In Figures 4 and 5 the implant is further deployed to form an outer or adventitial flange.

The outer flange is now seated against the adventitial surface of the target vessel, and the implant is released from the delivery tool. The two flanges then compress the target vessel wall, attaching the implant.

Comparison to the predicate: The PAS-Port was designed to be similar to the St. Jude Symmetry. Both systems deliver and implant a metallic star shaped connector for hemostatic anastomosis between the aorta and saphenous vein during CABG surgery.

There is no new technological characteristics differentiating the two systems. So it performs similar to the St. Jude Symmetry, except with some tweaks. However, the PAS-Port does incorporate designs that may address some of the problems seen with the predicate only after widespread use that led to eventual withdrawal from the market.

These include system integration with arteriotomy creation and capture and implant deployment in one step; absence of conduit cannulation; reduced profile -- outer tines are flush with the aorta outer wall; and reduced amount of metal in blood path.

Some of the controversial features of the predicate do persist, however, with the PAS-Port. Exposure of the sub-intimal tissue and metal to bloodstream is reduced, but still present.

A non-compliant annular containment of the anastomosis exists at the circular orifice, created with a stent-like scaffolding that is conducive to kinking.

The performance of the device was evaluated in a series of tests prior to deploying the device and during and after deployment. The tests are representative of actual clinical use. Below summarizes the tests conducted, the performance criteria and reliability and confidence and results attained.

Functional test summary includes formation of anastomosis. The deployment system will deploy and correctly form the implant, creating an anastomosis between the human vein and aorta. Leakage at site of anastomosis: The anastomosis between a saphenous vein and human aorta using water as test fluid will not leak below 200 millimeters mercury.

Poke-through test: The poke-through tool will poke through at least one feature of all nine implant tines without damage to the implant or vein.

Pull-through test: The vein will pull through the deployment system without dislodging from the pull-through tool with no more than one re-attempt.

The aortic plug capture: The cutter will consistently capture the test plug.

Implant pull-out test: The implant will withstand 10 ounces of tensile forces.

Hard Stop: The hard stop should be felt by the user when the knob has been completely turned and deployment is complete.

Pre-fire testing includes knob and switch. Where the knob and switch are locked down and prior to loading cartridge in the delivery tool.

The Cartridge Lock: The cartridge locks into place. Safety Switch: The device cannot fire prior to unlocking the safety switch.

According to preclinical data submitted by the sponsor, the test samples met and passed the set minimum performance criteria.

Safety testing summary includes biocompatibility. The PAS-Port proximal anastomosis system implant and all of the component parts have been tested, according to ISO 109993 in order to establish device safety. Device passed cytotoxicity, sensitization, acute intracutaneous reactivity, acute systemic toxicity, material mediated pyrogen, and hemolysis testing.

Testing also included chemical characterization where a traceable implant sample is analyzed to confirm that it was composed of stainless steel. Wave-length X-ray fluorescence confirmed the purity and that the stainless steel was used in the implant assembly.

Corrosion testing: Testing on the stainless steel implant portion was performed by a third party laboratory. Testing demonstrated that the ligation clips have good corrosion resistance in vivo, and clips will have an adequate resistance in pitting corrosion and crevice corrosion in the proposed use environment.

Testing also included animal testing, preclinical animal testing, where two chronic animal studies were conducted using an ovine model. Finally, the clinical testing which Wolf Sapirstein will go into detail after my presentation.

History of this device with the FDA: While there were cleared devices, FDA convened an Advisory Panel one year ago to clarify clearance requirements for the anastomotic devices. Since that time, the St. Jude Symmetry device has been withdrawn from the market. The Bypass Maakafim CorLink AAD, another predicate, has not been commercialized, and the Coalescent U-Clip is marketed more as a suture.

Device Modifications: The sponsor conducted a finite element analysis on the device implant. The analysis modeled stresses and strains produced in a device during manufacture, surgical placement, and continued performance during its expected lifetime as an element of vascular anastomosis.

The analysis predicts that the device should be able to withstand the stresses placed on it during surgical placement and during expected lifetime.

Finite element analysis was performed only on the preclinical device implant. The sponsor stated that minor changes were made to the delivery system and implant in response to mechanical issues noted with the delivery system during the clinical study that resulted in five of the device failures.

These modifications involve changes in material, tolerances and manufacturing procedures in the delivery system, and minor changes in the pre-shaped implant geometry and implant shaping procedures.

The changes to the system were only tested in human tissue model using an aorta and saphenous vein with no actual deployments. Use of human tissue model to assess the effectiveness of these changes is not considered adequate to address the device malfunction.

Devices at the preclinical, clinical and post-clinical stages differ in design. The numbers in the center indicate the areas where -- The number in the center indicates the area where the changes -- The numbers in the center indicate the areas that were changed and are summarized in the following.

The part -- Modifications from clinical to current implant: The part edges were shifted up by 0.0007 inches. The large barbs in the outer flange first seen in the preclinical implant and then replaced with a chevron feature is now replaced with three small barbs, seen in area 2.

Small barbs added along the internal edges of the outer flanges are now removed. The discard section was changed to a full mesh. Sponsor's intent was to provide more structure and rigidity during deployment. Also the discard section was lengthened to compensate as the mesh shortened and expands.

So in conclusion, the sponsor has not performed the finite element analysis and fatigue testing on the final version. Fatigue lifetime determination should not be based on earlier models but on the final product.

It is required that the sponsor provide documented evidence that the PAS-Port can withstand accelerated fatigue testing of one-year equivalent to use under simulated physiological conditions on the final product before any type of clearance can be granted.

This ends my -- the engineering portion of my presentation. Wolf Sapirstein will now continue with the clinical trial. Thank you.

I'm sorry. Barbara will now do the statistical.

DR. KRASNICKA: Good morning. I am the statistician reviewer for this submission. In my presentation I will focus mainly on the problems connected with the study design and statistical analysis performed.

The objective of the sponsor's submission was to present effectiveness and safety of the PAS-Port anastomosis system in patients requiring CABG. The statistical inferences under consideration are really based on four studies.

The studies are: The pivotal study conducted under the investigational plan CP2001-01; a prospective, nonrandomized, multi-center follow-up study carried out to evaluate long term health status of patients from study CP2001-01; The C-Port study conducted under the investigational plan CP2002-02; and a prospective, nonrandomized, multi-center follow-up study to evaluate long term health status of C-Port patients.

In the presentation some studies will be shortly introduced, and then statistical analysis performed, and the problems will be discussed.

Let me introduce the design of the pivotal study. The study was a nonrandomized, prospective, one-arm clinical trial conducted in two German and one Swiss clinical sites. It was carried out without IDE. The objective of this study was to assess ease of the PAS-Port device use, identify any procedural failure, and evaluate safety and effectiveness of the PAS-Port device. However, hypotheses were not defined.

According to the study plan, all patients should be evaluated pre-operatively, at the time of discharge from hospital, and at three and six months after operation. Coronary angiography was to be performed pre- and post-procedure and in the six months after operation.

In the process of the study evaluation, FDA gave the sponsor some post hoc recommendations related to the effectiveness and safety outcomes. According to these recommendations, the primary endpoint should be defined as the patency of the PAS-Port graft at six months, as assessed by angiography. The patency was to be defined as less than 50 percent stenosis in the proximal anastomosis related to a PAS-Port graft.

The PAS-Port graft patency rate at six months should be compared with the fixed OPC equal 80 percent. This number was recommended during the Panel discussion at a meeting on March 18, 2004, and was based on the historical control patency rates.

The second additional primary effectiveness endpoint was the patency of a proximal anastomosis of a PAS-Port graft at 12 months, as assessed by stress ECG test. Occurrence and frequency of MACE were to be assessed at one year, and the MACE frequency rate was to be treated as one of the effectiveness endpoints.

No specific recommendations with regard to adverse event was imposed. However, safety endpoint was defined as any adverse event that occurred during the study.

Now let me discuss the study course and its results. As already mentioned, the prospective, one-arm pivotal study was carried out at three sites. Altogether, 60 patients signed consent forms. However, due to intraoperative screening, five patients were not enrolled into the study. Therefore, the ITT population consisted of 55 patients with 60 intended or attempted PAS-Port grafts.

Implantation of the PAS-Port device were successful in 47 patients. IN eight patients, 15 percent, devices were not implanted successfully. During the pivotal study, no further information was collected for these eight patients. However, two years after operation seven of these patients were reenrolled into the long term study.

Please notice that the six month evaluation was performed for 80 percent of enrolled patients, and the patency was assessed by angiography with CORE lab analyses in only 71 percent of patients.

The results of the statistical analysis for the primary effectiveness endpoint are as follows: The observed percentage of patent PAS-Port grafts assessed by angiography at six months is 0.86 percent with 95 percent confidence interval, 71 percent to 95 percent.

Out of 60 grafts, only 42 grafts, 70 percent, were evaluated by the angiography at six months visit. Therefore, sensitivity analysis with respect to missing outcomes was needed.

The sponsored performed an imputation of missing patency statuses based on MRI or CT or stress ECG results, and clinical observations. Five patients were evaluated by MRI diagnostic methods. Their PAS-Port graft status was imputed as patent.

Under this imputation method, the patency rate estimate is 0.87 with 95 percent confidence interval, 0.74 to 0.95.

If we would assume for the ITT population the occlusion of grafts for all technical failures, all grafts in deceased patients and grafts without additional clinical information, then the ITT patency rate estimate would be equal, 0.72 with 95 percent confidence interval, 0.59 to 0.83.

In summary, lower confidence limits of the 95 percent confidence interval for the patency rates for different methods of assessment are all below the recommended 80 percent level. This means the FDA's recommended criterium of 80 percent was not met for the pivotal study.

One of the primary endpoints of the pivotal study was the MACE frequency rate during the 12 or 24 months period. The Kaplan-Meier estimate of the MACE frequency rates are shown in this table prepared by the sponsor.

Based on this table, the first impression is that the MACE rate at six months was very low. For instance, for the per-protocol population, the rate was 4.4 percent. However, this impression is misleading, because the MACE ratio at six months plus 15 days was 13 percent. Only additional 15 days caused essential difference in estimates. Moreover, information on the MACE for some months after the last pivotal study visit is not available, since no question relating to the history of adverse event occurrences was stated in so called 12 months case report forms.

The last pivotal study visit could be a visit at the discharge or at three months or at six months. As I mentioned before, the seven patients who withdrew at the baseline were reenrolled for the two-year evaluation, but only four of them received the stress ECG test. Despite that, the sponsor qualified all seven patients as MACE-free during two years after procedure.

Main results for the safety outcomes are summarized in this table. According to the sponsor, most of the adverse events listed in this table were not connected with the use of the device. At the discharge, 12 patients experienced arrhythmias, and there was one stenosis. This means at least 22 percent of patients experienced one cardiac adverse event.

The problems of adverse events will be discussed in more detail by Dr. Sapirstein.

It is worth noting that two patients died during the follow-up period between the sixth and the 24th month visits. Ten technical failures of device use were not included in the sponsor's adverse event summary.

As I mentioned before, the effectiveness endpoint was not met either for the ITT or for per-protocol or for the observed populations. Moreover, the point estimates and limits of confidence intervals may be biased, because the analyses were performed post hoc. We are dealing with one-arm study carried out only in three hospitals, and about 30 percent of grafts were not evaluated by angiography at six months.

The study was carried without IDE and formal interaction with the FDA. Additionally, it is impossible to compare 12 months MACE rates in the pivotal study with CABG historical data. One of the reasons is that in the historical studies each graft within the patient was included into a corresponding analysis, while in the sponsor's study only PAS-Port grafts were taken into analysis.

Target vessels implanted grafts were special, and they had to reveal special intraoperative inclusion criteria.

As said before, the original pivotal study was a very small one, and it had no power to demonstrate the safety and effectiveness of the device. Therefore, the sponsor proposed to increase the study power by adding to the pivotal study a subset of study CP2002-02.

I will refer to the study as study 2. Study 2 was a prospective, non-randomized, five-center study with the objective to assess the safety and effectiveness of another device, namely the C-Port Distal Anastomosis System. However, for some patients with multiple vein grafts, study 2 protocol allowed parallel employment of other devices.

The PAS-Port and C-Port devices were never utilized in the same graft. What is very important, the PAS-Port placement was based on the surgeons' discretion and was determined by aorta disease state and preferred grafting sequence.

I want to stress again that study 2 main endpoints reflected only evaluation of the C-Port System, and an assessment of the PAS-Port System was performed beyond the original scope of the study.

A subset of the CP2002-02 data, called further Cohort 2, should be considered only as a complementary dataset for the pivotal study, but not as an independent clinical study.

This Cohort 2 dataset on the PAS-Port system use was created retrospectively without stringent clinical rules normally imposed on device clinical studies. I am speaking about PAS-Port device study. Data extracted from a broader dataset collected for other purposes. It is impossible to estimate biases embedded in the data.

Evaluation of the PAS-Port system in Cohort 2 was performed retrospectively in conjunction with evaluation of another not cleared by the FDA anastomosis system. What is very important, the PAS-Port used in the pivotal study and in Study 2 were not exactly the same.

The PAS-Port System was improved during and after the pivotal study. Moreover, populations of two studies were different with respect to patients' pre- and intra-operative covariates.

Based on this table, we can conclude that there were differences between two cohorts in such clinical characteristics of patients as age, angina, CCS, and aspirin use within five days before operations.

Additionally, two cohorts were different with respect to such intra-operative characteristics like operation time, number of artery grafts and number of proximal anastomosis.

Therefore, the sponsor tried to justify pooling of two datasets by applying the propensity score method. Correctly employed propensity score method could be used as a diagnostic tool to show, to some extent, comparability between the pivotal study cohort and Cohort 2.

If two groups would overlap well enough in terms of propensity scores, then it would be possible to check the influence of cohort effect on the outcome variable, adjusting for baseline differences.

In our case, the propensity score for a patient is defined as a conditional probability of patient being assigned to Cohort 1, given the patient's covariates. The propensity score can be used to approximately balance the covariate differences of two groups, but can be seriously degraded if important covariates have not been collected or not taken into account in the analysis.

Therefore, a proper propensity score analysis should include as many covariates as possible.

Sponsor's propensity score analysis was performed in two steps. In the first step, the sponsor established main predictors of patients' graft patency: Diabetes, smoking history, vessel disease, and then calculated the propensity scores for each patient using predictors found in the first step.

Finally, the sponsor grouped patients into three subgroups and compared the outcome, adjusting for baseline differences, by using three subgroups of propensity scores. Based on such a procedure, the sponsor claims that the six months angiographic patency rates are comparable for two cohorts.

Distributions of sponsor's propensity scores are shown in this figure. Based on this figure, one could conclude that distributions for two cohorts do overlap. However, the sponsor's propensity analysis could not lead to meaningful justification of pooling of two cohorts' datasets, because propensity scores were based only on three covariates, which were predictors of patency. Some important covariates were not included in the analysis, and the dataset was very small.

I cannot discuss the second sponsor's propensity model, because it was not submitted to the FDA.

Due to the flaws of the sponsor's analysis, the propensity scores analysis was performed by the FDA. The propensity score model was built in a few steps. A logistic regression model with a stepwise selection was utilized.

The following covariates were included in the final model: Age, angina, CCS, gender, hyperlipidemia, vessel disease, NYHA or number of proximal anastomosis, and use of aspirin within five days of operation.

The entire population was divided into propensity score tertiles, with 36 patients in each tertile. One patient was excluded, because she/he did not have CCS, and most patients from Cohort 2 were in the first tertile.

The propensity score distributions for the final model are shown in this figure. Based on this figure, one can immediately conclude that the distributions for the pivotal study and cohort 2 do not overlap sufficiently, and is consistent with observed differences between groups. Therefore, it is impossible to check the cohort effect on outcomes.

The FDA conclusion is that the sponsor's original result about the pooling of data for two cohorts is simply invalid.

Due to lack of statistical support for combining two datasets, it did not make sense to perform statistical analysis for the combined dataset. Additionally, I want to stress again that the subset of study 2, cohort 2, should not be treated as a separate clinical study.

As mentioned previously, the pivotal study alone did not supply sufficient evidence of effectiveness and safety of the PAS-Port System. Point estimates for the effectiveness and safety endpoints may be biased due to post hoc analyses, a lot of missing information, and the questionable imputation of patency for many missing data.

Thank you very much for your attention. Now Dr. Sapirstein will present the clinical review.

DR. SAPIRSTEIN: Good morning. I undertook the clinical review of this device. In a former life, I was a cardiothoracic surgeon in the golden age of that discipline when catheters were used to direct us to the vessels to treat, not to treat the vessel itself.

The device for your review is used to create an anastomosis between the aorta and the autologous saphenous vein graft conduit for CABG or those studied entirely to create the proximal aortic anastomosis on the ascending aorta for a CABG conduit, clearance is sought for generic use of a saphenous vein anastomosis to any part of the aorta.

At the onset I should clarify that for a 510(k) clearance application, the predicate device can be used to demonstrate equivalence in construct and function, not to provide data for comparison to the device under consideration.

Because a CABG procedure has such critical impact on patient wellbeing and indeed survival, the FDA has stipulated very specific clearance criteria for devices employed in their construction. These are being developed with input from this Panel at a meeting on March 18th last year, as well as with a three-decade of history for the "gold standard" sutured CABG anastomosis.

Now I do not question much of the data presented by the sponsor. I only have some questions on the interpretation of that data in some areas, and I will try not to be too repetitive in my presentation, but try and present these.

Although the historical data which Dr. Mack has already introduced may not reflect present day clinical setting for CABG which now occurs in older patients and with anatomy unsuited for percutaneous interventions, and Dr. Page has already mentioned this, study inclusion for these cases for study of an anastomotic device do enroll samples which tend to be comparable to patients of the pre-drug-eluding stent era, and Dr. Page has questioned the sponsor on this particular subject.

These are the data that we -- criteria we formulated at the previous panel meeting to discuss these anastomotic devices. Next slide, please.

This is the trial -- the pivotal study number 1. Let me at the onset also indicate that the FDA had no input on the protocols of either this study or the second cohort. We only suggested after these studies had been undertaken some of the criteria that we would require for clearance of devices.

For instance, we had no discussion about method of demonstrating patency, whether computerized tomography or MRI would be acceptable, and this study, designated Study 1 by the sponsor, is considered the pivotal study.

As is summarized in this slide, the demographics of the trial is extremely curtailed. Next slide, please.

Because of the failure to comply with a stipulated angiography, the sponsor undertook a sensitivity analysis of patients which they claim assures absence of selection bias for cases which underwent angiograms, the requirement for demonstration of patency by the FDA, and those not imaged and for implantation failures.

Imputation of patency for CABG in various patients, not image, was made with several surrogates. The sponsor has adopted these assessments to generate the several analyses presented here, including an intention to treat analysis, none of which meet the protocol's hypothesis for effectiveness in terms of lower confidence bounds, and these, incidentally, were determined for 95 percent confidence intervals. Next slide, please.

The post hoc addition of a one-year clinical follow-up to the protocol resulted in patients needing to be re-consented, and the evaluation extended over a 12 to 24 month period.

The attribution of an adverse event to the device is often based on a judgment that the electrocardiographic regional abnormality is not consistent with that of the index graft's region of supply.

The hospital and death rates are similar to those of the Society of Thoracic Surgeons' database and to other historical literature sources. However, several device modifications were made to address failed implantations in eight patients. These failures were essentially eliminated from any analysis of device adverse events or even procedural failures, but were retained for intent-to-treat analysis of effectiveness.

Additional suture control of bleeding was required in 12 cases. Kinking of the graft was eliminated with additional sutures placed in six cases, and it is not clear how many of the later required reexploration for an angiographic finding after chest closure. Next slide, please.

This cohort of patients was recruited to supplement the essentially failed small pivotal study. Now while we agree that this was a prospectively designed study, this was for the distal anastomotic device, and it is our understanding from many interactions with the sponsor that patients were recruited for supplementing the Study 1 after the device had been -- after the study had been undertaken.

Patients thus recruited had been enrolled, and the proximal anastomosis used in a graft that did not incorporate the distal anastomosis. Retrospectively, this resulted in the stratification of 54 such patients for this cohort.

The selection bias inherent with this recruitment process is compounded by the PAS-Port use being determined not so much by protocol but by the surgeon's decision, whether it is because of his preference for an order in which anastomosis were to be constructed or the available conduit that he could work with. Next slide.

As similar methodology was used for these analyses are patency, as for Study 1, and the improvement in effectiveness must be viewed with some caution in light of the recruitment process, the design changes to the device, and the learning curve effect created by two of the four study centers for this Cohort 2 who enrolled about 50 percent of the 55 patient sample, having also participated in Study 1. Next slide.

The adverse event rate was higher in this cohort than in Study 1, and this may possibly be related to concomitant surgery permitted. The reduction to two conversions to sutured anastomosis from the eight that occurred in Study 1 may reflect design changes to the device since the Study 1. The sites for these two deployment failures were not available to estimate a possible learning curve factor.

There are several caveats to the uncritical acceptance of these outcomes which I will discuss later. Next slide, please.

The poolability of these two datasets claimed by the sponsor as based on their propensity analysis has been reviewed by Dr. Krasnicka. A summary of this pooled data is provided in this slide, 111 implants in 99 patients of whom 77 with 89 grafts had six-month angiography with patency seen in 81 grafts. Stress testing was performed at the 12 to 24 months period in 78 patients. Next slide, please.

The effectiveness for this sample combination in a different analysis undertaken with imputed patency was determined by the sponsor as shown here. The observed patency meets the lower confidence bound for 80 percent for the angiographic study grafts, with exclusion of those patients not studied in the denominator.

I should comment on Dr. Mack's remarks about the importance of clinical outcome as opposed to lumenology or anatomical patency. From our point of view, the replacement of the "gold standard" of a suture anastomosis, which has prevailed for three decades must be demonstrated with an anatomical patency, and we cannot rely on the vagaries of a clinical outcome. Next slide, please.

This is how the patency was imputed by the sponsor. In Study 1 angiography was performed on 39 patients. This constituted 72 percent of all 50 implants. Magnetic resonance imaging was used in five patients, and a stress ECG in two patients.

The suture conversions in eight patients were evaluated with stress ECG in five patients, a resting ECG in one patient, and with a Canadian Cardiac Society's functional scale in one patient.

Now we realize that angiography is an invasive evaluation which is difficult to perform one year -- to recruit patients to perform one year after a procedure, particularly if they are well. Therefore, we insisted on the stress ECG as a screening test, and any abnormality of the stress test was to be evaluated further with angiography. Next slide, please.

In Cohort 2 with 50 patients, angio was studied in 38 patients with 45 grafts. A multiple slide computerized tomography was used in five patients, stress ECG in three patients, and the Canadian scale in two patients. Suture conversion in two patients was imputed as patent with stress ECG. Next slide.

We have these limitations with the submission for this clearance. The pivotal study with prospective design and relative freedom from bias, despite need for protocol modification, failed to meet the study effectiveness hypothesis.

The poolability of Study 1 and Cohort 2 has not been satisfactorily established to compensate for failure of the pivotal study. The surrogates employed to impute patency was not evaluated with angiography.

The problem with the pooling included these items, and also what was neglected to be entered into the propensity were important factors, such as the modifications to the device, the use of anticholesterol therapy, and anticoagulants and antiplatelet medication. Next slide, please.

Our concerns for the evaluation of safety and effectiveness relate to the low incidence of angiography carried out for follow-up evaluations, the imputation of patency of 20 percent of the study implanted patients and 23.8 percent of the intent-to-treat patients, and with imputation performed on these rather self-criteria in the combined groups. Next slide, please.

The stress ECG was only performed in 36 of 45 survivors in Study 1 and 42 of 48 survivors in Cohort 2, and with ischemia demonstrated in five of these cases.

There were two episodes of thrombo-embolism identified, and these were not considered to be related to the device, even though the propensity for thrombogenicity in steel devices exposed to the blood stream is well known.

Stenosis or occlusion in adverse events listed occurred in 11 cases, and nine cases of ischemia were reported and one myocardial infarction. The stenosis are difficult to evaluate in relation to the index graft in the absence of radiographic details regarding their location and extent.

The revascularization procedures undertaken in 13 cases were not described in relation to the device graft except in two treated for stenotic lesions within two and five centimeters of the device, and the stenting of a 90 percent stenosis described in a graft body in a third case.

Review of nine cases of electrocardiographic ischemia fails to reject the possibility of compromise to the index graft in five of these cases. The single myocardial infarction involved in the posterior wall of the recipient of a right posterior descending graft was characterized as not device related, as was an area of hyperkinesia that was noted in a region of index graft supply in another patient.

The index graft supplied 16 to 22 percent of left ventricular mass which, without rigorous angiographic study, might make assessment of MACE difficult. Perhaps, Dr. Mack, we could remove the "C" and just call it MAEs and make it more acceptable. Next slide, please.

These are our study conclusions. Study 1 did not meet the pre-specified effectiveness hypothesis, and the poolability of Study 1 and Cohort 2 is problematic from our point of view.

The angiographic follow-up of patients is incomplete, and the sponsor relied heavily on imputations with non-invasive technologies to assess this patency.

The attribution of several adverse events as not device related is questionable and open to some dispute. Thank you very much.

ACTING CHAIRMAN MAISEL: I would like to thank the FDA for their presentation, and at this point open the discussion up to the Panel to directly question the FDA on any issues. Mitch.

DR. KRUCOFF: I would like to ask, since we are not all at Sharon-Lise's level, just a couple of statistical clarification questions, as may relate to this discussion.

One is in the term "intention to treat." I'm sorry if I missed it, but we obviously are going to have some discussion around patients who were enrolled in the study in whom the deployment of the device failed and whose outcomes ultimately were gathered in some sort of way or other.

So can you help me understand where statistically that would follow your presentation? Are these patients more appropriately -- Are their outcomes included or separately from safety and efficacy assessment of the device?

DR. KRASNICKA: In the ITT analysis these patients were included. The patients --

ACTING CHAIRMAN MAISEL: Can you speak into the microphone, please.

DR. KRASNICKA: Okay. The patients signed consent forms and were enrolled in the study.

DR. KRUCOFF: So we have consensus on that?

DR. KRASNICKA: Yes.

DR. KRUCOFF: Okay. Thank you. The second question is: In a setting where we are looking at multiple grafts in some patients and single grafts in others, where at least again at a lower level statistical tools like GE corrections, my understanding has been if you have three grafts in one patient and one graft in three other patients, and all three grafts in one patient do great and each of the one graft in the other three patients do poorly, that it may be because that one patient has a characteristic that is different and that there are ways of approaching that.

So is that a relevant discussion?

DR. KRASNICKA: No, because in the pivotal study is 55 patients with 60 grafts. But I am speaking about only PAS-Port graft, because in the analysis of the graft it was in patients who were not included. So, therefore, I didn't see really since --

DR. KRUCOFF: Okay. And again sounds like a consensus with the sponsor, because they also were not presenting any sort of correction for that. So is that a consensus?

DR. KRASNICKA: But in 109 patients I don't remember how many patients got more than --

ACTING CHAIRMAN MAISEL: It's safe to say that there were some patients who got multiple PAS-Ports and no correction was done for that for the fact that multiple patients got -- multiple PAS-Ports were implanted in --

DR. KRUCOFF: So it may be such a small number that it doesn't matter.

DR. KRASNICKA: It's a very small number. So it doesn't matter.

DR. KRUCOFF: It's a small number of a small number. Okay.

Lastly is with regard to the use of propensity scores, which at least I would like some help in clarifying. There are two ways that I could envision propensity scores might be of interest here. One is, obviously, to assess whether Cohort 1 and Cohort 2 are poolable, and we heard an extensive presentation earlier, and I think we have heard some about that as well.

The other is with regard to where these patients might fit relative to historical controls or as OUS patients might be comparable to patients who we would be likely to encounter in the practice of medicine in the United States.

I'm not sure that I am hearing both of those from you all or from the sponsor. So I've heard a lot about the use of propensity scores to understand whether or not these are poolable sets. Do you all have anything to say on the use of a propensity profile of the enrolled patients in one or both of these cohorts relative to historical controls or patients in the United Stats?

DR. KRASNICKA: No.

DR. KRUCOFF: Okay. Last question, honestly. This is just to help me in the understanding of how you pick variables for a propensity score, and I'm going to ask them a little bit more about this later, but just one quick question.

Obviously, you said the omission of key variables will undermine the value of a propensity score. What about selection bias? If two different studies actually had the same variables, same descriptors, but actually in the way the patients were selected there was, as we have here, two different studies where the surgeon's decision about where to put a graft or who to include or exclude from a study, is subject to different kinds of thinking and, therefore, potentially different kinds of bias.

Is there anything in a propensity score that captures that or that characterizes that?

DR. KRASNICKA: I believe, yes, in some point, a propensity score can catch to some extent the bias, the selection. But the studies were very small ones. So it is very difficult to --

DR. YUE: No. That variable was not included in the propensity score modeling.

MS. WOOD: Lily, please state your name.

DR. YUE: My name is Lily Yue, Branch Chief of Cardiovascular and Ophthalmic Devices Branch.

DR. KRUCOFF: So again, at a layperson's level would it be fair to say that preexistent bias, selection bias, like that might be like any other important variable that is left out of a propensity score, that it would undermine the value of the propensity score?

DR. YUE: You are right. I am not sure whether that variable was measured or not. Was it captured?

DR. KRASNICKA: Vessel disease into cohorts. So an adjustment was based on the vessel disease, too. So it was to some extent, it was captured.

DR. KRUCOFF: Okay.

ACTING CHAIRMAN MAISEL: Sharon, and then Tom.

DR. NORMAND: Just for clarification, I guess, I would state, if there is an important variable that is not measured, the propensity score analysis doesn't eliminate that particular bias.

I just have a quick point of clarification about the construction of the 95 percent confidence intervals. Were they indeed 95 percent and two-sided?

DR. KRASNICKA: Ninety-five percent? Yes, two-sided.

DR. NORMAND: So .025 in both tails.

DR. KRASNICKA: Yes.

DR. NORMAND: Thank you.

ACTING CHAIRMAN MAISEL: Tom.

DR. YUE: I thought the FDA presented the issue, it was two-sided. But I am not sure it's one-sided or two-sided.

DR. NORMAND: My question was to the FDA.

DR. YUE: FDA, that's right.

ACTING CHAIRMAN MAISEL: Tom.

DR. FERGUSON: Like Dr. Krucoff, I'm a layperson, but I am -- From what I've seen and what I have heard at other meetings, I am appalled by the difference in the results of the two propensity scores, not so much the fact that one overlaps and the other doesn't, but the fact that they are so widely apart.

My question, I guess, is maybe to you and then also to the sponsor later. How can this be, if you are using your -- you are picking your variables from the same group. Admittedly, you might not take variables from -- All of you might not take the same variables, but I think -- I just don't understand, because this is, to me, the critical issue for the day.

If, in fact, we support the FDA analysis that the two are not poolable, then we can't use the data from the second study.

ACTING CHAIRMAN MAISEL: Jeff.

DR. BORER: I have a question about the quality of the data that were analyzed. The FDA probably can answer this question, but if not, maybe we will want some additional information from the sponsor.

A primary issue here is drawing firm inferences from very small numbers, and that's a problem. To try to resolve some of that difference, as the FDA pointed out and as the sponsor presented, surrogates were used when angiography wasn't available.

A surrogate like exercise electrocardiography might be acceptable in a randomized trial with large populations, so that you could directly compare -- so that you could eliminate other biases and the populations are large enough so that you can draw some reasonable conclusions.

In small populations, the exercise electrocardiogram breaks down, because of a lack of positive and negative predictive value, as was demonstrated in the data that were presented to us. But there is another issue that I want to get to.

That is, how to interpret the positivity or negativity of either the exercise electrocardiogram or the symptom of angina in a situation where we haven't heard about the pre-operative finding for that descriptor.

From the table that was presented to us, one-third -- more than a third of the patients in Cohort 1, in Study 1, didn't have angina pre-op. So whether they had it post-op or not, is -- You know, it's difficult to interpret unambiguously the finding that they didn't have it post-op as a surrogate for something.

The same thing is true of the exercise electrocardiogram. If we don't know if it was positive or negative pre-op, it's hard for me, at any rate, to interpret what the negativity might mean post-operatively.

So I'd like some comments from the FDA about the quality of the data that were being analyzed. Given that there are some concerns about surrogates in general, what about this issue with these surrogates in these studies?

DR. SAPIRSTEIN: Yes. We agree with you about that, and the problem arose because it was a feeling that a six-month evaluation of patency was the best that we could arrive at. Knowing what is happening with the predicate device, we felt that this was not a sufficient determinant for something as important as this.

The public health impact on about 200-300,000 people having this procedure done annually was enormous. So we compromised in having the stress ECG at one year, irrespective of what it was pre-operatively; and if it was abnormal, then it should be followed up with a repeat angiogram, and that, we felt, would take care of that problem of what was before and what came after.

DR. BORER: I would accept that as a reasonable compromise, but I was really more concerned about the six-month data where angiograms weren't available for everybody and, as you pointed out in your presentation, Dr. Sapirstein, we had -- I think it was a total of seven electrocardiograms that were used out of a very small number of patients, and one symptom classification out of a very small number of patients as the surrogate for angiography.

ACTING CHAIRMAN MAISEL: John.

DR. SOMBERG: A couple of questions, and I will try to be brief. The first: In discussion of propensity score, I am taken by the large number of variables the FDA chose, and the more focused approach of the sponsor. It is my understanding that the more you bin a small study, the more likely you are going to find it not comparable, because you are just going to keep picking variables.

So can you defend your choice of -- I think it was seven, and in one slide it was eight, variables in the propensity analysis which you then claim makes the two cohorts noncomparable?

I think that addresses some of my earlier colleagues' problems, because why are they so disparate, and the reason is you are using seven or eight and they are using three or five.

DR. KRASNICKA: I tried to use -- From the study point of view, this means I use stepwise analysis, and I look at confidence interval, and I look at what is going on with the whole model when I add more covariates or drop them.

I found not only one model but few models, and they support what I saw on this model. And --

DR. SOMBERG: A few models in terms of the same number of variates or were using less variates when you use three or four or five?

DR. KRASNICKA: No, around six, seven, eight. In some case, I use only covariates -- intraoperative covariates, only based on really covariates which can collectivize patients before operation, and this model give me something between what sponsor presented and what I got from intra- and intraoperative covariates.

Additionally, I look at C index, and I see the -- of my model.

ACTING CHAIRMAN MAISEL: I would also just comment that I don't agree that the poolability or propensity score analyses are all that different. If you look at the sponsor's poolability, Slide Number 11, where they break up the strata, in the lower strata there are three times more patients in Cohort 2 than in Cohort 1, and in the upper strata there's two to seven times more patients in Cohort 1 than Cohort 2.

So it appears that there is a clear difference.

DR. SOMBERG: But, Bill, in their slide there was a lot of overlap. In the FDA's final slide there was no overlap and statistics showing that they were diametrically noncomparable. So I think it was night and day.

DR. NORMAND: If I could interject, I don't think that's the conclusion I would make from a comparison between those two slides. Pretend I had a variable that was an exclusion criterion, that people that had to have -- everybody had to have single vessel disease in one cohort, and in the other cohort they had to have multi-vessel disease.

Now if I included that covariate in my analysis, they could be completely separated, because it's a perfect predictor. So if you are going to pick variables that -- So I wouldn't say it doesn't matter how many variables they put in. You say, gee, it looks like they are very, very different. It's because I decided to put in variables that actually are measured and make sense.

So -- Just let me say one more thing. So the issue about the number of variables for the propensity score analysis is a very different thought process than looking at worrying about sort of Type I error in terms of looking at treatment effectiveness.

So the idea in the propensity score analysis is trying to balance as many as possible. You don't care that you have -- you are going to look at the coefficients of age and weight. You don't care about that. You are just trying to get overall the comparability between the two cohorts.

So I think the thought process that you are applying in terms of the number of variables -- I know it doesn't apply in this situation.

So just to emphasize, in terms of looking at the balance between the two cohorts, you want to try and include as many variables as possible. It's very different than worrying about a multiple comparison problem.

Moreover, if one cohort purposely excluded males, and they are the other one included -- you know, only had females -- they are completely different. So indeed they would be separate and, if I didn't include that gender in the model, they might overlap a lot. So I think you have to think about the variables that are included and sort of the reason why they are included or excluded.

ACTING CHAIRMAN MAISEL: Val.

DR. JEEVANANDAM: I have a question to --

ACTING CHAIRMAN MAISEL: Can you speak into the microphone, please.

DR. JEEVANANDAM: I have a question to Mr. Enyinna on the engineering component of it.

At the end, do you say that both these devices were actually quite different, and the last device had actually not undergone the testing that you would have required to get it approved?

MR. ENYINNA: That's absolutely correct, yes.

DR. JEEVANANDAM: Okay. Because it appears, if you look at Cohort 1 and Cohort 2, the patency difference is pretty significant between those two studies, and one wonders whether it has to do with the differences in the design of the product, because they are quite different in what is actually implanted into the patient.

MR. ENYINNA: Right. Whether that affects -- improves patency or not is one thing, but FDA puts a lot of value on engineering and preclinical data before trials occur, and it is important that they are able to validate any kind of changes made to a device and that the final device that they intend to market has been put through required performance testing, such as finite element analysis and fatigue testing.

Changing from one model to another, from one generation to another or improving is one -- that's good. We like improvements, but it needs to be validated before we can release that to market.

It is unknown whether these changes has any kind of failure modes or any kinds of strains or stresses that it might be weakened by those improvements. But in order for us to determine that, we need some kind of documented evidence that, despite the changes, the device performs and functions as intended.

DR. JEEVANANDAM: And we have had a lot of these discussions about combining cohort 1 and Cohort 2 in propensity analysis, but are we combining two different devices, and is that something that we should allow them to do?

DR. ZUCKERMAN: Okay. Again, in a general construct it would be helpful if this Advisory Panel would leave the, say, engineering analysis to us. We just mentioned that to show that we need to do due diligence.

There is a component of the engineering analysis that hasn't been completed, but let's assume that it passes muster. Given the clinical data and what was presented, we are asking your comment about the clinical acceptability of pooling.

DR. JEEVANANDAM: But if the two -- In Cohort 1 and Cohort 2, if the rates of patency are different, can they actually be pooled as one device?

DR. ZUCKERMAN: Again, it comes back to the methodology that is being discussed here, what you think of the selection of patients for Cohorts 1 and 2, and all these clinical issues.

ACTING CHAIRMAN MAISEL: John.

DR. HIRSHFELD: I think in trying to understand this, what we are really trying to figure out is how reliably this device does what it is intended to do and how durably it does this.

I think it would be useful for some historical guidance here from the people who have been through the past, because what we are dealing with is basically a surrogate endpoint for a device failure and then surrogates for the surrogate endpoint, as far as what we actually have.

I think I, in particular, and I think probably a lot of the Panel, are uneasy about the fact that Symmetry got through and then was discovered only in post-market surveillance to have its problems.

So the question I would have from people who went through the Symmetry approval process, if we had some of those people here, was: Do we have data now that help us be more confident about how well this device performs than the people who evaluated Symmetry did when they made that judgment?

ACTING CHAIRMAN MAISEL: Dr. Sapirstein, could you provide us a little perspective on Symmetry and where we are with this device?

DR. SAPIRSTEIN: Well, with the Symmetry we only had short term data, six-month data, and in fact -- so even more limited data than we have for this device. And the result -- and the clearance of that device and its commercialization indicated to us quite clearly, as Dr. Mack first suggested, that six-month data per se isn't adequate and that you need something out to one year or even longer.

Well, there is a limit to what we can ask for. We have accepted the one-year evidence based on stress testing, risks residing on the six-month angiogram, and on a follow-up repeat angiogram, if necessary.

We don't think that MRI or computed technology at this stage is adequate to demonstrate significant obstruction to an anastomotic device. It may illustrate filling of a graft, a conduit, and we know the value of that from the Vineburg operation. So that is -- and it's largely because of a failure of the St. Jude history that we insisted on this.

ACTING CHAIRMAN MAISEL: Bram?

DR. ZUCKERMAN: So to fully respond to Dr. Hirshfeld, number one, we have learned a lot from our prior device experience in this arena and have attempted to convey that to the industry.

Two, if there is a concern that post-market, as you said, it would be the same story rather than a different story regarding the agency's ability to look at potential post-market problems with this class of device, as noted by Dr. Gardner this morning, the agency is really pursuing a different approach, and we would consider with the Panel potentially that type of approach with this class of devices.

So that the ability to surveil these anastomotic devices, both in pre- and post-market, is very different from the landscape of a few years ago.

ACTING CHAIRMAN MAISEL: John.

DR. SOMBERG: Could I have clarified the -- We have a denominator of about 100 with the PAS-Port device. In the Symmetry device -- and what was the denominator, and did you have any follow-up beyond six months, because here you have follow-up for at least 35-40 patients for two years, and we are talking about for about another 35 or some 40 patients one-year follow-up.

That's why I understand the FDA asked for the MACE data, because there was built in to have angiography, but you couldn't keep doing angiography. So you asked for MACE.

So let's say, if you had this type of data, MACE on 100 patients at one year, would you have found out what you have with Symmetry -- the problem you had discovered later on with Symmetry?

DR. SAPIRSTEIN: If we would have relied on stress testing at one year, we would have found out what was going on with --

DR. SOMBERG: That's not MACE, though. Stress testing is not MACE, because the endpoint in this study that PAS-Port is presenting is MACE data, which is myocardial infarction or active angina, to my understanding.

DR. SAPIRSTEIN: But the primary --

DR. SOMBERG: Is that not right?

DR. SAPIRSTEIN: The primary effective endpoint was patency, and we just collected the additional secondary data on MACE and adverse events, but the primary effectiveness was anatomical patency.

DR. SOMBERG: No, I understand that. My question was: The MACE data that is presented with PAS-Port at one year -- let's just say at one year in approximately 100 patients -- would that have given you the signal with Symmetry if you had MACE data at one year?

DR. SAPIRSTEIN: It probably would have, yes.

DR. SOMBERG: And then I would ask the follow-up question. With the FDA's analysis, have you seen a signal approaching that scene with Symmetry with the PAS-Port System on those 100 patients for one-year follow-up?

DR. SAPIRSTEIN: I think what we would have to compare to would be studies such as those that were performed by Dr. Mack and his group on the St. Jude Symmetry, and based on that data I think -- and this may approximate it. But it is difficult to compare. I am not sure.

The data that Dr. Mack published was based on clinical -- wide clinical use, and this data is provided -- This submission contains brief and circumscribed data for a single device in a patient on very circumscribed criteria. For instance, as was suggested, it was supplying an area of about 16 to 20 percent of ventricular mass, not very large areas.

So it is difficult to -- I think they are not very easily compared, but I think they were comparable.

ACTING CHAIRMAN MAISEL: Dr. Blackstone.

DR. BLACKSTONE: Since I am not sure that you have answered the question, let me pursue it a little bit.

First of all, Dr. Mack would say that he could pick up -- and he did tell us that he could pick up a difference with Symmetry when just looking at MACE. The problem I am having with looking at MACE in all of these presentations is that everyone has presented a different definition of MACE.

So for example, if you look back at your slide that says problems with MACE evaluations, you are not looking at MACE. You are looking at MACE plus the occluded graft event data. That isn't MACE data from any definition that any of us have ever heard before.

One of its major problems is that, when you present these both in the submission and in the slides, for example, as Kaplan-Meier curves, the data on occlusions is not points of time. It is angiograms taken at various points in time. So it's interval- censored data. There is nowhere in here that anyone has looked at any interval-censoring data.

Then you have combined that interval-censored data with clinical MACE data. So I think none of us have any good idea of what the MACE really is for this. So that I think you are right, Wolf. You can't answer the question that was asked, because no one has looked at MACE like all of us look at MACE and know whether, from one year to two years, six months, whatever, we see a falling pattern like we did in Symmetry.

DR. SAPIRSTEIN: And I used the data that was presented by the sponsor for MACE, and that is death, myocardial infarction, and revascularization. So there is sort of a consistency in our two presentations.

DR. BLACKSTONE: But they have plenty of curves that they call MACE that include the graft occlusion data. In fact, that is the primary focus of the MACE data.

ACTING CHAIRMAN MAISEL: Clyde.

DR. YANCY: I respect what Dr. Zuckerman said regarding limiting our discussion on the engineering issues, but I share Val's concerns about the comparison between the iterative developments of this device.

During the presentation, you indicated that the change in mesh density may have provided more rigidity to make the application easier, and that would be an important concept, if that's true.

With regard to the other changes and the slide that you shared with us, do you have similar concepts or ideas that these changes may have indeed been instrumental in changing outcomes, because the changes -- I'll allude to the engineering perspective, but if you think there was a change that could have had a difference in the usability of the technology, I'd like for you to develop that just a bit more.

DR. ENYINNA: I can't say honestly if those changes improved the outcomes of the study, the second study. But from an engineering perspective, we need to look at that, regardless of whether -- before clinical trials begin, and we look at these things before it begins and determine whether the device is safe enough to be implanted into the patient.

So, unfortunately, they are giving us data after it has been implanted into the patient and using that to sort of confirm the validity of their FEA, final element analysis performed prior to clinical implantation.

DR. YANCY: So would be extended to say that you don't see any overt advantage of at least several of the changes, with the exception of the one that you commented on? You are relatively neutral?

MR. ENYINNA: I'm sorry. Could you repeat that again?

DR. YANCY: Do you see any overt advantage of several of the changes that were made, changing the time design, etcetera? You indicated one advantage, for instance.

MR. ENYINNA: My point was that I can't tell whether there is an advantage or not, because the proper procedure for performance testing was not applied to the new generation of the implant. So for us it needs to -- For me to make that kind of decision or assessment, I need to see -- we need to see the testing performed on the device on the final product.

DR. SAPIRSTEIN: Dr. Yancy, from a clinical point of view, there is so much noise associated with it, but there seems to be a clearcut advantage to using the new device over the previous device, whatever little changes were made. There were less failures to implant, and the ease of implantation must impact on the performance of a CABG procedure.

DR. YANCY; Well, I mean, that's part of the direction of the question, which obviously we can develop more in the afternoon. But were there changes in outcomes? If it's clearcut or not, I'm not sure. Was that due to design changes in the platform? Was it due to learning curve? Was it due to different patient selection issues? So I think those are the things that we need to flesh out.

ACTING CHAIRMAN MAISEL: Linda?

MS. MOTTLE: Along the same line on engineering, just a point for clarification, we went from 10 device failures in Cohort 1 down to two failures after the modifications in Cohort 2, and that is two failures out of 61 device implantations. Is that a fairly acceptable rate of failure with the deploytation on that, and that, of course, was the major intent stated for the modifications?

DR. YANCY: There is an improvement in the failure rate, and for me to say that it is based on the engineering or the design, it's hard for me to do that because I didn't see. We don't have the engineering. From a clinical perspective, you might say that, yes, but as an engineer I can't say that, no.

ACTING CHAIRMAN MAISEL: And whether or not it is an acceptable failure rate will be discussed this afternoon. I think that is one of the reasons the Panel is convened. Mitch, and then Jeff.

DR. KRUCOFF: I think this has been said, but I just want to be absolutely clear about when the first conversations between the sponsor and the FDA occurred. Is it accurate to say that both cohorts had completed enrollment when the first discussion with FDA occurred, or is that -- Can you just help me focus on where in the timeline conversations with FDA -- Neither one of these were IDEs. Right?

DR. SAPIRSTEIN: Neither were IDEs. Both studies were undertaken without any input from us. Study 1 was completed, I think, when the first discussions occurred with FDA, and it was my understanding that Cohort 2 -- the study for the distal anastomotic device had already been commenced--

DR. KRUCOFF: Had begun enrollment.

DR. SAPIRSTEIN: -- at that time.

DR. KRUCOFF: Thank you.

ACTING CHAIRMAN MAISEL: Jeff.

DR. BORER: John raised the issue of durability, and I want to ask a question about that, make a comment and ask a question, because I don't see it in the questions for later, and I think it is important.

When I look at this device and see these tines that are sharp enough to impale the everted edge of a vein, I wonder what they could do long-term to an aorta if they were pointed in the wrong direction at some point.

Now we are not going to learn that prior to approval, if the device is approved, and that's okay. I think, though, that that speaks to the issue that was raised earlier about mandatory follow-up. But I wonder if we have any information of a longer term nature about adverse events that could be related to the mechanics of the device with the Symmetry device. I mean, is this something that has shown up, because there's several years of potential follow-up available now?

Having said that, I want to make another point about durability. That is that I agree completely with Dr. Mack about the occulostenotic reflex and its inappropriateness. That is why really you have to look at events and whatever. But the truth is that the patency of the graft short term -- that is, within a year to two years -- is the best information we have from which to draw prognostic inferences about the durability of the treatment as opposed to the device.

So I think it is useful to have that information. I just make that as a comment. But the question is with regard to the potential for injuring the aorta and for the manifestation of that to occur later.

We are not going to know anything about that, I think, from the data on this particular device, but can we draw any inferences from Symmetry? Do we have some reports that would help us there?

DR. SAPIRSTEIN: No, we don't have any reports of failure of the device causing inadvertent injury to some structures. We do have reports -- for instance, patient suffering a chest injury or CPR events and dislodging a device very shortly after -- within a month or so of its implantation.

There were no instances of false aneurysms or of aortic disruption.

ACTING CHAIRMAN MAISEL: Deborah.

MS. MOORE: Yes. I had another question about the modifications and a clarification. It is not clear the timing of what those modifications were with respect to each of the clinical trials.

The current design that people are talking about not having the fatigue testing that needs to be done, which it sounds like that would be done as part of the approval process -- was that specific device used in Cohort 2 or are those changes that were implemented after Cohort 2 was completed?

MR. ENYINNA: To the best of my knowledge, I believe they were implemented before Cohort 2 and after Study 1. There is an image in one of the sponsor's -- I don't know if it was displayed on the screen or not, but they show modifications to the implant itself from Study 1 to the one used in Cohort 2 in which the chevron was removed from the tines.

ACTING CHAIRMAN MAISEL: Any other questions for FDA at this point. So seeing none, why don't we break for lunch and reconvene at 1:30.

(Whereupon, the foregoing matter went off the record at 12:25 p.m.)

- - -

A F T E R N O O N S E S S I O N

Time: 1:32 p.m.

ACTING CHAIRMAN MAISEL: Good afternoon. We will start this afternoon's session with the primary reviewers and, as soon as Dr. Krucoff gets his computer in gear, we will have him begin his review.

DR. KRUCOFF: Should we start? Do we have the sponsor here? Should we proceed?

ACTING CHAIRMAN MAISEL: It's 1:35. So --

DR. KRUCOFF: Okay. Well, my motion is -- ACTING CHAIRMAN MAISEL: Mitch, why don't you go ahead?

DR. KRUCOFF: Okay. I actually wanted to start with a question for Dr. Mack.

ACTING CHAIRMAN MAISEL: Actually, if I could ask you to use the podium, because we like to leave that free so that the panelists can question either the FDA or the sponsor.

DR. KRUCOFF: Just a quick question, and I think it will be -- I hope it will be obvious why. You enrolled patients in the St. Jude Symmetry experience. Right?

DR. MACK: Correct.

DR. KRUCOFF: And if this isn't fair, don't answer. I just would be very interested to know: When you were actively enrolling patients in that study, where did you think you were heading with that device? Were you enthusiastic? Were you excited about it?

DR. MACK: Yes, I was very enthusiastic about it, and it was until we had a significant number of patients and it started to get eight, nine months down the line and people came back with angina. We just had never seen that before.

Our cardiologists had kind of said, wait a minute, you guys may have a problem with this device. And it wasn't until then it became obvious. I did -- perhaps we can put it up a little later -- take a table out from our paper that showed that we looked at historical controls versus St. Jude. The lines were identical until six months, and then six months began separating out to 18 months.

So it became clear from that standpoint. So in regard to the question of, if St. Jude had been followed from the start, would this problem have been picked up on a clinical basis? The answer is unequivocally yes.

DR. KRUCOFF: Thanks, because I think for my understanding of where we are with the Panel of the product we are discussing today, it's going to be real important for us to separate out what is concept, and very interesting concept, and potential benefit to patients in reduction of suffering concept from what data.

I think all of us have learned in one fashion or another the same kind of lesson, that what really appear to be uniquely engineered approaches to concepts of what would make processes better when we actually put them in human beings and use them frequently are more complicated than we expect.

DR. MACK: And like I said at the beginning, my whole long term view of this was to facilitate minimal access surgery. It was appropriate to start in beating heart surgery, but at the end of the day that's not where I saw it ending up.

DR. KRUCOFF: Thanks. Okay, I think it is important to start as we sort through this with areas that are going to need to be kept unambiguous. One of those areas is that this is a 510(k) application and, not to abuse Dr. Zuckerman's comment earlier this morning, technically that is not our issue, but conceptually equivalence, substantial equivalence to what is, and whether it is to a preexisting sutureless anastomotic device that ultimately was enthusiastically explored but ultimately has been voluntarily withdrawn or whether we are asking the question is this equivalent to a suture technique, that is sort of an established gold standard are separate questions.

I think, if we touch on issues related to one or the other, we need to be careful that they are separate.

I think, conceptually, the notion that a quicker way to attach a durable, reliable anastomosis to the aorta without cross-clamping has potential to reduce embolic events, as long as it is, in fact, durable in the short term and in the long -- short term safe, doesn't thrombose, in the long term doesn't restenose or do other bad things, that I don't think anybody has trouble seeing the concept.

The ingeniousness of the engineering, I think -- again, we don't have to get into details. We have two devices, but even the first version was a pretty ingenious piece of engineering.

I think fundamentally, to me, the question is do we have data about this device in human beings where we can understand whether the device is safe and effective?

One of the issues that I would sort out, as Mike presented, and some of the data were done at my home base, is why grafts fail. So let me haul you up here one more time, Dr. Mack, and ask: If you were to take a best educated guess, what percentage of graft failure would you say comes from the proximal anastomosis?

DR. MACK: A low percentage. I would say it's not the predominant factor. If I had to pick a number, I'd say 10 to 20 percent. I think that the target vessel being bypassed, the amount of distal disease, and then other body-of-the-graft issues are much more important than the proximal anastomosis.

DR. KRUCOFF: And I think probably we will all agree that it's probably a low number, if it's compared to a hand-sewn anastomosis. So we are talking about small numbers of events in small numbers of patients, and that's, of course, where things get ambiguous.

The other issue that, to my mind, in this particular device is an unresolved one, is whether events that are observed five centimeters distal to the proximal anastomosis are events that might involve thrombotic and embolic issues as opposed to local.

I just wondered if anybody from the sponsor would care to comment on whether, in fact, seeing a new lesion in a vein graft five centimeters distal to the proximal anastomosis is, in fact, unrelated to the device.

DR. HARRINGER: Well, I have demonstrated to you the angiograms and how the stenosis looked like. If you expect an embolic event -- If you expect small embolic events that would occur from the proximal anastomotic device, you probably would see them more distally. You probably would see target vessel occlusions, actually, less than stenosis within the graft.

These five centimeter lesions in the grafts are not quite uncommon, because that's where you sometimes place your bulldog clamps, where you place your clamps in order to occlude the grafts, and that might cause endothelial damage, too.

DR. KRUCOFF: Understood, and to me the issue is do we really know. And having seen, for instance, embolic material catch at a point of a reversed venous valve in a vein graft, I'm just not sure that we are always going to be able to know, either angiographically or etiologically, whether some of or what percentage of the events that we can angiographically see five centimeters distally are unrelated to the way the proximal anastomosis is constructed.

DR. HARRINGER: It is a small percentage, and we do see those in hand-sewn anastomosis, that within the body of the vein graft there is some kind of thrombus. These numbers are actually small, but they look usually different. If you see a thrombus formed within the vein graft, they look more like the graft demonstrated at the 50 percent stenosis where you have adherence to the vessel wall and the kind of suture-like fashion inside the lumen.

DR. KRUCOFF: Okay. Well, I am going to keep going, with all respect, to saying I'm not sure how much we can tell from angiograms.

I think the other area that we have to be very clear about, and it's been brought up a couple of times but rather haphazardly, is that to understand how this device behaves in the human, we have anatomic features and physiologic features and clinical features.

So whether we have in-stent-like restenosis is really an anatomic feature ocular, stenotically recognized, or otherwise, but it does tell us about the behavior of tissue around the device that impales vein graft and creates, certainly, inflammation. I haven't seen much in the way of animal data, but I think you could expect there probably some inflammatory aspects.

Ultimately, whether or not that generates ischemia or has physiologic consequences is a separate question, and then whether patients come to harm, have myocardial infarctions or death, is a -- They are all related but independent questions.

I think that is particularly important when we think about the methods of assessing each of those issues. So angiography is invasive. Patients don't like it if they are feeling well, but it is about the only way, or certainly one of the best ways, to understand anatomically how much tissue reaction there is in a human being to this novel technique.

Stress tests or symptoms may tell us more physiologically about whether there is ischemia and ultimately clinical events about outcomes.

I think this is where, in these very small numbers of patients, some of the issues arise as to how reasonable is it to combine these different assessment modalities and impute information about the behavior of the device in this small population.

So my question for whoever from the sponsor is: I found no information on patients who actually had multiple methods of assessment. So do you all have any data you can share with us on patients who had an angiogram and a stress test and a symptomatic assessment for us to help understand -- The way you have imputed these, by and large, is that they are concordant. So do you have any data that would actually help us understand in this patient population whether these measures are indeed concordant, CT, MRI, stress testing, angiography, clinical outcomes and symptoms?

DR. HAUSEN: I will answer that. We have very limited information, because -- Let me pick the ones where we had positive stress EKGs and then they were angiogrammed at -- stress EKG at three months and angiogram at six months. That's one indication.

Half of the patients had occlusions. The other half were patent. That's an example. We did not angio routinely after six months. So that's kind of the only time we could compare two pieces of information, concordant information.

DR. KRUCOFF: Do you have any numbers, for instance, of patients who angiographically had compromised vessels and who had normal stress tests or freedom from symptoms?

DR. HAUSEN: Yes. It is again half/half. The sensitivity and specificity in this setting early post-op in stress EKGs, I would say, is fairly poor.

DR. KRUCOFF: Okay. So then I am -- and I may have missed it. So I'll just ask. What I didn't get out of the information provided or today's presentations is whether any sort of correction for that relative concordance or discordance is made in any of the imputations of any of the sensitivity analyses that were presented.

DR. HAUSEN: It wasn't, because the sample -- We didn't discuss that, because the sample sizes are so small that the types of corrections just wouldn't hold.

DR. KRUCOFF: Okay. So on average what we are really saying is the mosaic of endpoints that are constructed to try and recover the sample size include endpoints that are roughly 50/50 concordant with one another, which would fit clinical practice, but which, I think, would again -- would have an impact on some of the statistical conclusions of the reasonableness of the way these endpoints have been imputed. Is that 50/50?

DR. HAUSEN: I can state something on that. Fifty/fifty, but if we recall from Dr. Mack's presentation, 12.5 percent of his patients came back symptomatic, and none in our patient population. So I think that is a variable that seems to separate in patients with Symmetry devices versus ones with PAS-Port devices, if patients -- if occlusions occur.

DR. KRUCOFF: Symptomatic?

DR. HAUSEN: Symptomatic, yes. It's a big difference.

DR. KRUCOFF: The primary endpoint was re-narrowing. The primary endpoint was patency.

DR. HAUSEN: For us, yes.

DR. KRUCOFF: Okay. Again, whoever from the company can help me. One of the concerns that came up briefly is how generalizable would any of these data be in 300,000 patients undergoing bypass surgery in the United States.

So can somebody help us understand that a little bit?

DR. MACK: I think the answer to that is I quoted the PREVENT IV trial this morning, which was a 3,000 patient U.S. trial, and the inclusion -- exclusion criteria of that trial are very similar to what was included here.